xref: /linux/fs/ext4/super.c (revision c75c5ab575af7db707689cdbb5a5c458e9a034bb)
1 /*
2  *  linux/fs/ext4/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18 
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
43 
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
46 
47 #include "ext4.h"
48 #include "ext4_extents.h"	/* Needed for trace points definition */
49 #include "ext4_jbd2.h"
50 #include "xattr.h"
51 #include "acl.h"
52 #include "mballoc.h"
53 
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
56 
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62 
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 			     unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 					struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 				   struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 		       const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 
85 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
86 static struct file_system_type ext2_fs_type = {
87 	.owner		= THIS_MODULE,
88 	.name		= "ext2",
89 	.mount		= ext4_mount,
90 	.kill_sb	= kill_block_super,
91 	.fs_flags	= FS_REQUIRES_DEV,
92 };
93 MODULE_ALIAS_FS("ext2");
94 MODULE_ALIAS("ext2");
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96 #else
97 #define IS_EXT2_SB(sb) (0)
98 #endif
99 
100 
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 	.owner		= THIS_MODULE,
104 	.name		= "ext3",
105 	.mount		= ext4_mount,
106 	.kill_sb	= kill_block_super,
107 	.fs_flags	= FS_REQUIRES_DEV,
108 };
109 MODULE_ALIAS_FS("ext3");
110 MODULE_ALIAS("ext3");
111 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
112 #else
113 #define IS_EXT3_SB(sb) (0)
114 #endif
115 
116 static int ext4_verify_csum_type(struct super_block *sb,
117 				 struct ext4_super_block *es)
118 {
119 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
120 					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
121 		return 1;
122 
123 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
124 }
125 
126 static __le32 ext4_superblock_csum(struct super_block *sb,
127 				   struct ext4_super_block *es)
128 {
129 	struct ext4_sb_info *sbi = EXT4_SB(sb);
130 	int offset = offsetof(struct ext4_super_block, s_checksum);
131 	__u32 csum;
132 
133 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
134 
135 	return cpu_to_le32(csum);
136 }
137 
138 int ext4_superblock_csum_verify(struct super_block *sb,
139 				struct ext4_super_block *es)
140 {
141 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
142 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
143 		return 1;
144 
145 	return es->s_checksum == ext4_superblock_csum(sb, es);
146 }
147 
148 void ext4_superblock_csum_set(struct super_block *sb)
149 {
150 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
151 
152 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
153 		EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
154 		return;
155 
156 	es->s_checksum = ext4_superblock_csum(sb, es);
157 }
158 
159 void *ext4_kvmalloc(size_t size, gfp_t flags)
160 {
161 	void *ret;
162 
163 	ret = kmalloc(size, flags);
164 	if (!ret)
165 		ret = __vmalloc(size, flags, PAGE_KERNEL);
166 	return ret;
167 }
168 
169 void *ext4_kvzalloc(size_t size, gfp_t flags)
170 {
171 	void *ret;
172 
173 	ret = kzalloc(size, flags);
174 	if (!ret)
175 		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176 	return ret;
177 }
178 
179 void ext4_kvfree(void *ptr)
180 {
181 	if (is_vmalloc_addr(ptr))
182 		vfree(ptr);
183 	else
184 		kfree(ptr);
185 
186 }
187 
188 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
189 			       struct ext4_group_desc *bg)
190 {
191 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
192 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
193 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
194 }
195 
196 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
197 			       struct ext4_group_desc *bg)
198 {
199 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
200 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
201 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
202 }
203 
204 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
205 			      struct ext4_group_desc *bg)
206 {
207 	return le32_to_cpu(bg->bg_inode_table_lo) |
208 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
209 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
210 }
211 
212 __u32 ext4_free_group_clusters(struct super_block *sb,
213 			       struct ext4_group_desc *bg)
214 {
215 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
216 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
217 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
218 }
219 
220 __u32 ext4_free_inodes_count(struct super_block *sb,
221 			      struct ext4_group_desc *bg)
222 {
223 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
224 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
225 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
226 }
227 
228 __u32 ext4_used_dirs_count(struct super_block *sb,
229 			      struct ext4_group_desc *bg)
230 {
231 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
232 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
233 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
234 }
235 
236 __u32 ext4_itable_unused_count(struct super_block *sb,
237 			      struct ext4_group_desc *bg)
238 {
239 	return le16_to_cpu(bg->bg_itable_unused_lo) |
240 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
241 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
242 }
243 
244 void ext4_block_bitmap_set(struct super_block *sb,
245 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
246 {
247 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
248 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
249 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
250 }
251 
252 void ext4_inode_bitmap_set(struct super_block *sb,
253 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
254 {
255 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
256 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
257 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
258 }
259 
260 void ext4_inode_table_set(struct super_block *sb,
261 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
262 {
263 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
264 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
265 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
266 }
267 
268 void ext4_free_group_clusters_set(struct super_block *sb,
269 				  struct ext4_group_desc *bg, __u32 count)
270 {
271 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
272 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
273 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
274 }
275 
276 void ext4_free_inodes_set(struct super_block *sb,
277 			  struct ext4_group_desc *bg, __u32 count)
278 {
279 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
280 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
281 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
282 }
283 
284 void ext4_used_dirs_set(struct super_block *sb,
285 			  struct ext4_group_desc *bg, __u32 count)
286 {
287 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
288 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
289 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
290 }
291 
292 void ext4_itable_unused_set(struct super_block *sb,
293 			  struct ext4_group_desc *bg, __u32 count)
294 {
295 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
296 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
297 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
298 }
299 
300 
301 static void __save_error_info(struct super_block *sb, const char *func,
302 			    unsigned int line)
303 {
304 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
305 
306 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
307 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
308 	es->s_last_error_time = cpu_to_le32(get_seconds());
309 	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
310 	es->s_last_error_line = cpu_to_le32(line);
311 	if (!es->s_first_error_time) {
312 		es->s_first_error_time = es->s_last_error_time;
313 		strncpy(es->s_first_error_func, func,
314 			sizeof(es->s_first_error_func));
315 		es->s_first_error_line = cpu_to_le32(line);
316 		es->s_first_error_ino = es->s_last_error_ino;
317 		es->s_first_error_block = es->s_last_error_block;
318 	}
319 	/*
320 	 * Start the daily error reporting function if it hasn't been
321 	 * started already
322 	 */
323 	if (!es->s_error_count)
324 		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
325 	le32_add_cpu(&es->s_error_count, 1);
326 }
327 
328 static void save_error_info(struct super_block *sb, const char *func,
329 			    unsigned int line)
330 {
331 	__save_error_info(sb, func, line);
332 	ext4_commit_super(sb, 1);
333 }
334 
335 /*
336  * The del_gendisk() function uninitializes the disk-specific data
337  * structures, including the bdi structure, without telling anyone
338  * else.  Once this happens, any attempt to call mark_buffer_dirty()
339  * (for example, by ext4_commit_super), will cause a kernel OOPS.
340  * This is a kludge to prevent these oops until we can put in a proper
341  * hook in del_gendisk() to inform the VFS and file system layers.
342  */
343 static int block_device_ejected(struct super_block *sb)
344 {
345 	struct inode *bd_inode = sb->s_bdev->bd_inode;
346 	struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
347 
348 	return bdi->dev == NULL;
349 }
350 
351 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
352 {
353 	struct super_block		*sb = journal->j_private;
354 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
355 	int				error = is_journal_aborted(journal);
356 	struct ext4_journal_cb_entry	*jce, *tmp;
357 
358 	spin_lock(&sbi->s_md_lock);
359 	list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
360 		list_del_init(&jce->jce_list);
361 		spin_unlock(&sbi->s_md_lock);
362 		jce->jce_func(sb, jce, error);
363 		spin_lock(&sbi->s_md_lock);
364 	}
365 	spin_unlock(&sbi->s_md_lock);
366 }
367 
368 /* Deal with the reporting of failure conditions on a filesystem such as
369  * inconsistencies detected or read IO failures.
370  *
371  * On ext2, we can store the error state of the filesystem in the
372  * superblock.  That is not possible on ext4, because we may have other
373  * write ordering constraints on the superblock which prevent us from
374  * writing it out straight away; and given that the journal is about to
375  * be aborted, we can't rely on the current, or future, transactions to
376  * write out the superblock safely.
377  *
378  * We'll just use the jbd2_journal_abort() error code to record an error in
379  * the journal instead.  On recovery, the journal will complain about
380  * that error until we've noted it down and cleared it.
381  */
382 
383 static void ext4_handle_error(struct super_block *sb)
384 {
385 	if (sb->s_flags & MS_RDONLY)
386 		return;
387 
388 	if (!test_opt(sb, ERRORS_CONT)) {
389 		journal_t *journal = EXT4_SB(sb)->s_journal;
390 
391 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
392 		if (journal)
393 			jbd2_journal_abort(journal, -EIO);
394 	}
395 	if (test_opt(sb, ERRORS_RO)) {
396 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
397 		sb->s_flags |= MS_RDONLY;
398 	}
399 	if (test_opt(sb, ERRORS_PANIC))
400 		panic("EXT4-fs (device %s): panic forced after error\n",
401 			sb->s_id);
402 }
403 
404 void __ext4_error(struct super_block *sb, const char *function,
405 		  unsigned int line, const char *fmt, ...)
406 {
407 	struct va_format vaf;
408 	va_list args;
409 
410 	va_start(args, fmt);
411 	vaf.fmt = fmt;
412 	vaf.va = &args;
413 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
414 	       sb->s_id, function, line, current->comm, &vaf);
415 	va_end(args);
416 	save_error_info(sb, function, line);
417 
418 	ext4_handle_error(sb);
419 }
420 
421 void ext4_error_inode(struct inode *inode, const char *function,
422 		      unsigned int line, ext4_fsblk_t block,
423 		      const char *fmt, ...)
424 {
425 	va_list args;
426 	struct va_format vaf;
427 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
428 
429 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
430 	es->s_last_error_block = cpu_to_le64(block);
431 	save_error_info(inode->i_sb, function, line);
432 	va_start(args, fmt);
433 	vaf.fmt = fmt;
434 	vaf.va = &args;
435 	if (block)
436 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
437 		       "inode #%lu: block %llu: comm %s: %pV\n",
438 		       inode->i_sb->s_id, function, line, inode->i_ino,
439 		       block, current->comm, &vaf);
440 	else
441 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
442 		       "inode #%lu: comm %s: %pV\n",
443 		       inode->i_sb->s_id, function, line, inode->i_ino,
444 		       current->comm, &vaf);
445 	va_end(args);
446 
447 	ext4_handle_error(inode->i_sb);
448 }
449 
450 void ext4_error_file(struct file *file, const char *function,
451 		     unsigned int line, ext4_fsblk_t block,
452 		     const char *fmt, ...)
453 {
454 	va_list args;
455 	struct va_format vaf;
456 	struct ext4_super_block *es;
457 	struct inode *inode = file_inode(file);
458 	char pathname[80], *path;
459 
460 	es = EXT4_SB(inode->i_sb)->s_es;
461 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
462 	save_error_info(inode->i_sb, function, line);
463 	path = d_path(&(file->f_path), pathname, sizeof(pathname));
464 	if (IS_ERR(path))
465 		path = "(unknown)";
466 	va_start(args, fmt);
467 	vaf.fmt = fmt;
468 	vaf.va = &args;
469 	if (block)
470 		printk(KERN_CRIT
471 		       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
472 		       "block %llu: comm %s: path %s: %pV\n",
473 		       inode->i_sb->s_id, function, line, inode->i_ino,
474 		       block, current->comm, path, &vaf);
475 	else
476 		printk(KERN_CRIT
477 		       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 		       "comm %s: path %s: %pV\n",
479 		       inode->i_sb->s_id, function, line, inode->i_ino,
480 		       current->comm, path, &vaf);
481 	va_end(args);
482 
483 	ext4_handle_error(inode->i_sb);
484 }
485 
486 const char *ext4_decode_error(struct super_block *sb, int errno,
487 			      char nbuf[16])
488 {
489 	char *errstr = NULL;
490 
491 	switch (errno) {
492 	case -EIO:
493 		errstr = "IO failure";
494 		break;
495 	case -ENOMEM:
496 		errstr = "Out of memory";
497 		break;
498 	case -EROFS:
499 		if (!sb || (EXT4_SB(sb)->s_journal &&
500 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
501 			errstr = "Journal has aborted";
502 		else
503 			errstr = "Readonly filesystem";
504 		break;
505 	default:
506 		/* If the caller passed in an extra buffer for unknown
507 		 * errors, textualise them now.  Else we just return
508 		 * NULL. */
509 		if (nbuf) {
510 			/* Check for truncated error codes... */
511 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
512 				errstr = nbuf;
513 		}
514 		break;
515 	}
516 
517 	return errstr;
518 }
519 
520 /* __ext4_std_error decodes expected errors from journaling functions
521  * automatically and invokes the appropriate error response.  */
522 
523 void __ext4_std_error(struct super_block *sb, const char *function,
524 		      unsigned int line, int errno)
525 {
526 	char nbuf[16];
527 	const char *errstr;
528 
529 	/* Special case: if the error is EROFS, and we're not already
530 	 * inside a transaction, then there's really no point in logging
531 	 * an error. */
532 	if (errno == -EROFS && journal_current_handle() == NULL &&
533 	    (sb->s_flags & MS_RDONLY))
534 		return;
535 
536 	errstr = ext4_decode_error(sb, errno, nbuf);
537 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
538 	       sb->s_id, function, line, errstr);
539 	save_error_info(sb, function, line);
540 
541 	ext4_handle_error(sb);
542 }
543 
544 /*
545  * ext4_abort is a much stronger failure handler than ext4_error.  The
546  * abort function may be used to deal with unrecoverable failures such
547  * as journal IO errors or ENOMEM at a critical moment in log management.
548  *
549  * We unconditionally force the filesystem into an ABORT|READONLY state,
550  * unless the error response on the fs has been set to panic in which
551  * case we take the easy way out and panic immediately.
552  */
553 
554 void __ext4_abort(struct super_block *sb, const char *function,
555 		unsigned int line, const char *fmt, ...)
556 {
557 	va_list args;
558 
559 	save_error_info(sb, function, line);
560 	va_start(args, fmt);
561 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
562 	       function, line);
563 	vprintk(fmt, args);
564 	printk("\n");
565 	va_end(args);
566 
567 	if ((sb->s_flags & MS_RDONLY) == 0) {
568 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
569 		sb->s_flags |= MS_RDONLY;
570 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
571 		if (EXT4_SB(sb)->s_journal)
572 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
573 		save_error_info(sb, function, line);
574 	}
575 	if (test_opt(sb, ERRORS_PANIC))
576 		panic("EXT4-fs panic from previous error\n");
577 }
578 
579 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
580 {
581 	struct va_format vaf;
582 	va_list args;
583 
584 	va_start(args, fmt);
585 	vaf.fmt = fmt;
586 	vaf.va = &args;
587 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
588 	va_end(args);
589 }
590 
591 void __ext4_warning(struct super_block *sb, const char *function,
592 		    unsigned int line, const char *fmt, ...)
593 {
594 	struct va_format vaf;
595 	va_list args;
596 
597 	va_start(args, fmt);
598 	vaf.fmt = fmt;
599 	vaf.va = &args;
600 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
601 	       sb->s_id, function, line, &vaf);
602 	va_end(args);
603 }
604 
605 void __ext4_grp_locked_error(const char *function, unsigned int line,
606 			     struct super_block *sb, ext4_group_t grp,
607 			     unsigned long ino, ext4_fsblk_t block,
608 			     const char *fmt, ...)
609 __releases(bitlock)
610 __acquires(bitlock)
611 {
612 	struct va_format vaf;
613 	va_list args;
614 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
615 
616 	es->s_last_error_ino = cpu_to_le32(ino);
617 	es->s_last_error_block = cpu_to_le64(block);
618 	__save_error_info(sb, function, line);
619 
620 	va_start(args, fmt);
621 
622 	vaf.fmt = fmt;
623 	vaf.va = &args;
624 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
625 	       sb->s_id, function, line, grp);
626 	if (ino)
627 		printk(KERN_CONT "inode %lu: ", ino);
628 	if (block)
629 		printk(KERN_CONT "block %llu:", (unsigned long long) block);
630 	printk(KERN_CONT "%pV\n", &vaf);
631 	va_end(args);
632 
633 	if (test_opt(sb, ERRORS_CONT)) {
634 		ext4_commit_super(sb, 0);
635 		return;
636 	}
637 
638 	ext4_unlock_group(sb, grp);
639 	ext4_handle_error(sb);
640 	/*
641 	 * We only get here in the ERRORS_RO case; relocking the group
642 	 * may be dangerous, but nothing bad will happen since the
643 	 * filesystem will have already been marked read/only and the
644 	 * journal has been aborted.  We return 1 as a hint to callers
645 	 * who might what to use the return value from
646 	 * ext4_grp_locked_error() to distinguish between the
647 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
648 	 * aggressively from the ext4 function in question, with a
649 	 * more appropriate error code.
650 	 */
651 	ext4_lock_group(sb, grp);
652 	return;
653 }
654 
655 void ext4_update_dynamic_rev(struct super_block *sb)
656 {
657 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
658 
659 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
660 		return;
661 
662 	ext4_warning(sb,
663 		     "updating to rev %d because of new feature flag, "
664 		     "running e2fsck is recommended",
665 		     EXT4_DYNAMIC_REV);
666 
667 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
668 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
669 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
670 	/* leave es->s_feature_*compat flags alone */
671 	/* es->s_uuid will be set by e2fsck if empty */
672 
673 	/*
674 	 * The rest of the superblock fields should be zero, and if not it
675 	 * means they are likely already in use, so leave them alone.  We
676 	 * can leave it up to e2fsck to clean up any inconsistencies there.
677 	 */
678 }
679 
680 /*
681  * Open the external journal device
682  */
683 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
684 {
685 	struct block_device *bdev;
686 	char b[BDEVNAME_SIZE];
687 
688 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
689 	if (IS_ERR(bdev))
690 		goto fail;
691 	return bdev;
692 
693 fail:
694 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
695 			__bdevname(dev, b), PTR_ERR(bdev));
696 	return NULL;
697 }
698 
699 /*
700  * Release the journal device
701  */
702 static int ext4_blkdev_put(struct block_device *bdev)
703 {
704 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
705 }
706 
707 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
708 {
709 	struct block_device *bdev;
710 	int ret = -ENODEV;
711 
712 	bdev = sbi->journal_bdev;
713 	if (bdev) {
714 		ret = ext4_blkdev_put(bdev);
715 		sbi->journal_bdev = NULL;
716 	}
717 	return ret;
718 }
719 
720 static inline struct inode *orphan_list_entry(struct list_head *l)
721 {
722 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
723 }
724 
725 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
726 {
727 	struct list_head *l;
728 
729 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
730 		 le32_to_cpu(sbi->s_es->s_last_orphan));
731 
732 	printk(KERN_ERR "sb_info orphan list:\n");
733 	list_for_each(l, &sbi->s_orphan) {
734 		struct inode *inode = orphan_list_entry(l);
735 		printk(KERN_ERR "  "
736 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
737 		       inode->i_sb->s_id, inode->i_ino, inode,
738 		       inode->i_mode, inode->i_nlink,
739 		       NEXT_ORPHAN(inode));
740 	}
741 }
742 
743 static void ext4_put_super(struct super_block *sb)
744 {
745 	struct ext4_sb_info *sbi = EXT4_SB(sb);
746 	struct ext4_super_block *es = sbi->s_es;
747 	int i, err;
748 
749 	ext4_unregister_li_request(sb);
750 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
751 
752 	flush_workqueue(sbi->dio_unwritten_wq);
753 	destroy_workqueue(sbi->dio_unwritten_wq);
754 
755 	if (sbi->s_journal) {
756 		err = jbd2_journal_destroy(sbi->s_journal);
757 		sbi->s_journal = NULL;
758 		if (err < 0)
759 			ext4_abort(sb, "Couldn't clean up the journal");
760 	}
761 
762 	ext4_es_unregister_shrinker(sb);
763 	del_timer(&sbi->s_err_report);
764 	ext4_release_system_zone(sb);
765 	ext4_mb_release(sb);
766 	ext4_ext_release(sb);
767 	ext4_xattr_put_super(sb);
768 
769 	if (!(sb->s_flags & MS_RDONLY)) {
770 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
771 		es->s_state = cpu_to_le16(sbi->s_mount_state);
772 	}
773 	if (!(sb->s_flags & MS_RDONLY))
774 		ext4_commit_super(sb, 1);
775 
776 	if (sbi->s_proc) {
777 		remove_proc_entry("options", sbi->s_proc);
778 		remove_proc_entry(sb->s_id, ext4_proc_root);
779 	}
780 	kobject_del(&sbi->s_kobj);
781 
782 	for (i = 0; i < sbi->s_gdb_count; i++)
783 		brelse(sbi->s_group_desc[i]);
784 	ext4_kvfree(sbi->s_group_desc);
785 	ext4_kvfree(sbi->s_flex_groups);
786 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
787 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
788 	percpu_counter_destroy(&sbi->s_dirs_counter);
789 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
790 	percpu_counter_destroy(&sbi->s_extent_cache_cnt);
791 	brelse(sbi->s_sbh);
792 #ifdef CONFIG_QUOTA
793 	for (i = 0; i < MAXQUOTAS; i++)
794 		kfree(sbi->s_qf_names[i]);
795 #endif
796 
797 	/* Debugging code just in case the in-memory inode orphan list
798 	 * isn't empty.  The on-disk one can be non-empty if we've
799 	 * detected an error and taken the fs readonly, but the
800 	 * in-memory list had better be clean by this point. */
801 	if (!list_empty(&sbi->s_orphan))
802 		dump_orphan_list(sb, sbi);
803 	J_ASSERT(list_empty(&sbi->s_orphan));
804 
805 	invalidate_bdev(sb->s_bdev);
806 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
807 		/*
808 		 * Invalidate the journal device's buffers.  We don't want them
809 		 * floating about in memory - the physical journal device may
810 		 * hotswapped, and it breaks the `ro-after' testing code.
811 		 */
812 		sync_blockdev(sbi->journal_bdev);
813 		invalidate_bdev(sbi->journal_bdev);
814 		ext4_blkdev_remove(sbi);
815 	}
816 	if (sbi->s_mmp_tsk)
817 		kthread_stop(sbi->s_mmp_tsk);
818 	sb->s_fs_info = NULL;
819 	/*
820 	 * Now that we are completely done shutting down the
821 	 * superblock, we need to actually destroy the kobject.
822 	 */
823 	kobject_put(&sbi->s_kobj);
824 	wait_for_completion(&sbi->s_kobj_unregister);
825 	if (sbi->s_chksum_driver)
826 		crypto_free_shash(sbi->s_chksum_driver);
827 	kfree(sbi->s_blockgroup_lock);
828 	kfree(sbi);
829 }
830 
831 static struct kmem_cache *ext4_inode_cachep;
832 
833 /*
834  * Called inside transaction, so use GFP_NOFS
835  */
836 static struct inode *ext4_alloc_inode(struct super_block *sb)
837 {
838 	struct ext4_inode_info *ei;
839 
840 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
841 	if (!ei)
842 		return NULL;
843 
844 	ei->vfs_inode.i_version = 1;
845 	INIT_LIST_HEAD(&ei->i_prealloc_list);
846 	spin_lock_init(&ei->i_prealloc_lock);
847 	ext4_es_init_tree(&ei->i_es_tree);
848 	rwlock_init(&ei->i_es_lock);
849 	INIT_LIST_HEAD(&ei->i_es_lru);
850 	ei->i_es_lru_nr = 0;
851 	ei->i_reserved_data_blocks = 0;
852 	ei->i_reserved_meta_blocks = 0;
853 	ei->i_allocated_meta_blocks = 0;
854 	ei->i_da_metadata_calc_len = 0;
855 	ei->i_da_metadata_calc_last_lblock = 0;
856 	spin_lock_init(&(ei->i_block_reservation_lock));
857 #ifdef CONFIG_QUOTA
858 	ei->i_reserved_quota = 0;
859 #endif
860 	ei->jinode = NULL;
861 	INIT_LIST_HEAD(&ei->i_completed_io_list);
862 	spin_lock_init(&ei->i_completed_io_lock);
863 	ei->i_sync_tid = 0;
864 	ei->i_datasync_tid = 0;
865 	atomic_set(&ei->i_ioend_count, 0);
866 	atomic_set(&ei->i_unwritten, 0);
867 	INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
868 
869 	return &ei->vfs_inode;
870 }
871 
872 static int ext4_drop_inode(struct inode *inode)
873 {
874 	int drop = generic_drop_inode(inode);
875 
876 	trace_ext4_drop_inode(inode, drop);
877 	return drop;
878 }
879 
880 static void ext4_i_callback(struct rcu_head *head)
881 {
882 	struct inode *inode = container_of(head, struct inode, i_rcu);
883 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
884 }
885 
886 static void ext4_destroy_inode(struct inode *inode)
887 {
888 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
889 		ext4_msg(inode->i_sb, KERN_ERR,
890 			 "Inode %lu (%p): orphan list check failed!",
891 			 inode->i_ino, EXT4_I(inode));
892 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
893 				EXT4_I(inode), sizeof(struct ext4_inode_info),
894 				true);
895 		dump_stack();
896 	}
897 	call_rcu(&inode->i_rcu, ext4_i_callback);
898 }
899 
900 static void init_once(void *foo)
901 {
902 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
903 
904 	INIT_LIST_HEAD(&ei->i_orphan);
905 	init_rwsem(&ei->xattr_sem);
906 	init_rwsem(&ei->i_data_sem);
907 	inode_init_once(&ei->vfs_inode);
908 }
909 
910 static int init_inodecache(void)
911 {
912 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
913 					     sizeof(struct ext4_inode_info),
914 					     0, (SLAB_RECLAIM_ACCOUNT|
915 						SLAB_MEM_SPREAD),
916 					     init_once);
917 	if (ext4_inode_cachep == NULL)
918 		return -ENOMEM;
919 	return 0;
920 }
921 
922 static void destroy_inodecache(void)
923 {
924 	/*
925 	 * Make sure all delayed rcu free inodes are flushed before we
926 	 * destroy cache.
927 	 */
928 	rcu_barrier();
929 	kmem_cache_destroy(ext4_inode_cachep);
930 }
931 
932 void ext4_clear_inode(struct inode *inode)
933 {
934 	invalidate_inode_buffers(inode);
935 	clear_inode(inode);
936 	dquot_drop(inode);
937 	ext4_discard_preallocations(inode);
938 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
939 	ext4_es_lru_del(inode);
940 	if (EXT4_I(inode)->jinode) {
941 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
942 					       EXT4_I(inode)->jinode);
943 		jbd2_free_inode(EXT4_I(inode)->jinode);
944 		EXT4_I(inode)->jinode = NULL;
945 	}
946 }
947 
948 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
949 					u64 ino, u32 generation)
950 {
951 	struct inode *inode;
952 
953 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
954 		return ERR_PTR(-ESTALE);
955 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
956 		return ERR_PTR(-ESTALE);
957 
958 	/* iget isn't really right if the inode is currently unallocated!!
959 	 *
960 	 * ext4_read_inode will return a bad_inode if the inode had been
961 	 * deleted, so we should be safe.
962 	 *
963 	 * Currently we don't know the generation for parent directory, so
964 	 * a generation of 0 means "accept any"
965 	 */
966 	inode = ext4_iget(sb, ino);
967 	if (IS_ERR(inode))
968 		return ERR_CAST(inode);
969 	if (generation && inode->i_generation != generation) {
970 		iput(inode);
971 		return ERR_PTR(-ESTALE);
972 	}
973 
974 	return inode;
975 }
976 
977 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
978 					int fh_len, int fh_type)
979 {
980 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
981 				    ext4_nfs_get_inode);
982 }
983 
984 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
985 					int fh_len, int fh_type)
986 {
987 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
988 				    ext4_nfs_get_inode);
989 }
990 
991 /*
992  * Try to release metadata pages (indirect blocks, directories) which are
993  * mapped via the block device.  Since these pages could have journal heads
994  * which would prevent try_to_free_buffers() from freeing them, we must use
995  * jbd2 layer's try_to_free_buffers() function to release them.
996  */
997 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
998 				 gfp_t wait)
999 {
1000 	journal_t *journal = EXT4_SB(sb)->s_journal;
1001 
1002 	WARN_ON(PageChecked(page));
1003 	if (!page_has_buffers(page))
1004 		return 0;
1005 	if (journal)
1006 		return jbd2_journal_try_to_free_buffers(journal, page,
1007 							wait & ~__GFP_WAIT);
1008 	return try_to_free_buffers(page);
1009 }
1010 
1011 #ifdef CONFIG_QUOTA
1012 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1013 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1014 
1015 static int ext4_write_dquot(struct dquot *dquot);
1016 static int ext4_acquire_dquot(struct dquot *dquot);
1017 static int ext4_release_dquot(struct dquot *dquot);
1018 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1019 static int ext4_write_info(struct super_block *sb, int type);
1020 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1021 			 struct path *path);
1022 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1023 				 int format_id);
1024 static int ext4_quota_off(struct super_block *sb, int type);
1025 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1026 static int ext4_quota_on_mount(struct super_block *sb, int type);
1027 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1028 			       size_t len, loff_t off);
1029 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1030 				const char *data, size_t len, loff_t off);
1031 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1032 			     unsigned int flags);
1033 static int ext4_enable_quotas(struct super_block *sb);
1034 
1035 static const struct dquot_operations ext4_quota_operations = {
1036 	.get_reserved_space = ext4_get_reserved_space,
1037 	.write_dquot	= ext4_write_dquot,
1038 	.acquire_dquot	= ext4_acquire_dquot,
1039 	.release_dquot	= ext4_release_dquot,
1040 	.mark_dirty	= ext4_mark_dquot_dirty,
1041 	.write_info	= ext4_write_info,
1042 	.alloc_dquot	= dquot_alloc,
1043 	.destroy_dquot	= dquot_destroy,
1044 };
1045 
1046 static const struct quotactl_ops ext4_qctl_operations = {
1047 	.quota_on	= ext4_quota_on,
1048 	.quota_off	= ext4_quota_off,
1049 	.quota_sync	= dquot_quota_sync,
1050 	.get_info	= dquot_get_dqinfo,
1051 	.set_info	= dquot_set_dqinfo,
1052 	.get_dqblk	= dquot_get_dqblk,
1053 	.set_dqblk	= dquot_set_dqblk
1054 };
1055 
1056 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1057 	.quota_on_meta	= ext4_quota_on_sysfile,
1058 	.quota_off	= ext4_quota_off_sysfile,
1059 	.quota_sync	= dquot_quota_sync,
1060 	.get_info	= dquot_get_dqinfo,
1061 	.set_info	= dquot_set_dqinfo,
1062 	.get_dqblk	= dquot_get_dqblk,
1063 	.set_dqblk	= dquot_set_dqblk
1064 };
1065 #endif
1066 
1067 static const struct super_operations ext4_sops = {
1068 	.alloc_inode	= ext4_alloc_inode,
1069 	.destroy_inode	= ext4_destroy_inode,
1070 	.write_inode	= ext4_write_inode,
1071 	.dirty_inode	= ext4_dirty_inode,
1072 	.drop_inode	= ext4_drop_inode,
1073 	.evict_inode	= ext4_evict_inode,
1074 	.put_super	= ext4_put_super,
1075 	.sync_fs	= ext4_sync_fs,
1076 	.freeze_fs	= ext4_freeze,
1077 	.unfreeze_fs	= ext4_unfreeze,
1078 	.statfs		= ext4_statfs,
1079 	.remount_fs	= ext4_remount,
1080 	.show_options	= ext4_show_options,
1081 #ifdef CONFIG_QUOTA
1082 	.quota_read	= ext4_quota_read,
1083 	.quota_write	= ext4_quota_write,
1084 #endif
1085 	.bdev_try_to_free_page = bdev_try_to_free_page,
1086 };
1087 
1088 static const struct super_operations ext4_nojournal_sops = {
1089 	.alloc_inode	= ext4_alloc_inode,
1090 	.destroy_inode	= ext4_destroy_inode,
1091 	.write_inode	= ext4_write_inode,
1092 	.dirty_inode	= ext4_dirty_inode,
1093 	.drop_inode	= ext4_drop_inode,
1094 	.evict_inode	= ext4_evict_inode,
1095 	.put_super	= ext4_put_super,
1096 	.statfs		= ext4_statfs,
1097 	.remount_fs	= ext4_remount,
1098 	.show_options	= ext4_show_options,
1099 #ifdef CONFIG_QUOTA
1100 	.quota_read	= ext4_quota_read,
1101 	.quota_write	= ext4_quota_write,
1102 #endif
1103 	.bdev_try_to_free_page = bdev_try_to_free_page,
1104 };
1105 
1106 static const struct export_operations ext4_export_ops = {
1107 	.fh_to_dentry = ext4_fh_to_dentry,
1108 	.fh_to_parent = ext4_fh_to_parent,
1109 	.get_parent = ext4_get_parent,
1110 };
1111 
1112 enum {
1113 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1114 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1115 	Opt_nouid32, Opt_debug, Opt_removed,
1116 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1117 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1118 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1119 	Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1120 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1121 	Opt_data_err_abort, Opt_data_err_ignore,
1122 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1123 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1124 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1125 	Opt_usrquota, Opt_grpquota, Opt_i_version,
1126 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1127 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1128 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1129 	Opt_dioread_nolock, Opt_dioread_lock,
1130 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1131 	Opt_max_dir_size_kb,
1132 };
1133 
1134 static const match_table_t tokens = {
1135 	{Opt_bsd_df, "bsddf"},
1136 	{Opt_minix_df, "minixdf"},
1137 	{Opt_grpid, "grpid"},
1138 	{Opt_grpid, "bsdgroups"},
1139 	{Opt_nogrpid, "nogrpid"},
1140 	{Opt_nogrpid, "sysvgroups"},
1141 	{Opt_resgid, "resgid=%u"},
1142 	{Opt_resuid, "resuid=%u"},
1143 	{Opt_sb, "sb=%u"},
1144 	{Opt_err_cont, "errors=continue"},
1145 	{Opt_err_panic, "errors=panic"},
1146 	{Opt_err_ro, "errors=remount-ro"},
1147 	{Opt_nouid32, "nouid32"},
1148 	{Opt_debug, "debug"},
1149 	{Opt_removed, "oldalloc"},
1150 	{Opt_removed, "orlov"},
1151 	{Opt_user_xattr, "user_xattr"},
1152 	{Opt_nouser_xattr, "nouser_xattr"},
1153 	{Opt_acl, "acl"},
1154 	{Opt_noacl, "noacl"},
1155 	{Opt_noload, "norecovery"},
1156 	{Opt_noload, "noload"},
1157 	{Opt_removed, "nobh"},
1158 	{Opt_removed, "bh"},
1159 	{Opt_commit, "commit=%u"},
1160 	{Opt_min_batch_time, "min_batch_time=%u"},
1161 	{Opt_max_batch_time, "max_batch_time=%u"},
1162 	{Opt_journal_dev, "journal_dev=%u"},
1163 	{Opt_journal_checksum, "journal_checksum"},
1164 	{Opt_journal_async_commit, "journal_async_commit"},
1165 	{Opt_abort, "abort"},
1166 	{Opt_data_journal, "data=journal"},
1167 	{Opt_data_ordered, "data=ordered"},
1168 	{Opt_data_writeback, "data=writeback"},
1169 	{Opt_data_err_abort, "data_err=abort"},
1170 	{Opt_data_err_ignore, "data_err=ignore"},
1171 	{Opt_offusrjquota, "usrjquota="},
1172 	{Opt_usrjquota, "usrjquota=%s"},
1173 	{Opt_offgrpjquota, "grpjquota="},
1174 	{Opt_grpjquota, "grpjquota=%s"},
1175 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1176 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1177 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1178 	{Opt_grpquota, "grpquota"},
1179 	{Opt_noquota, "noquota"},
1180 	{Opt_quota, "quota"},
1181 	{Opt_usrquota, "usrquota"},
1182 	{Opt_barrier, "barrier=%u"},
1183 	{Opt_barrier, "barrier"},
1184 	{Opt_nobarrier, "nobarrier"},
1185 	{Opt_i_version, "i_version"},
1186 	{Opt_stripe, "stripe=%u"},
1187 	{Opt_delalloc, "delalloc"},
1188 	{Opt_nodelalloc, "nodelalloc"},
1189 	{Opt_removed, "mblk_io_submit"},
1190 	{Opt_removed, "nomblk_io_submit"},
1191 	{Opt_block_validity, "block_validity"},
1192 	{Opt_noblock_validity, "noblock_validity"},
1193 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1194 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1195 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1196 	{Opt_auto_da_alloc, "auto_da_alloc"},
1197 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1198 	{Opt_dioread_nolock, "dioread_nolock"},
1199 	{Opt_dioread_lock, "dioread_lock"},
1200 	{Opt_discard, "discard"},
1201 	{Opt_nodiscard, "nodiscard"},
1202 	{Opt_init_itable, "init_itable=%u"},
1203 	{Opt_init_itable, "init_itable"},
1204 	{Opt_noinit_itable, "noinit_itable"},
1205 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1206 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1207 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1208 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1209 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1210 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1211 	{Opt_err, NULL},
1212 };
1213 
1214 static ext4_fsblk_t get_sb_block(void **data)
1215 {
1216 	ext4_fsblk_t	sb_block;
1217 	char		*options = (char *) *data;
1218 
1219 	if (!options || strncmp(options, "sb=", 3) != 0)
1220 		return 1;	/* Default location */
1221 
1222 	options += 3;
1223 	/* TODO: use simple_strtoll with >32bit ext4 */
1224 	sb_block = simple_strtoul(options, &options, 0);
1225 	if (*options && *options != ',') {
1226 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1227 		       (char *) *data);
1228 		return 1;
1229 	}
1230 	if (*options == ',')
1231 		options++;
1232 	*data = (void *) options;
1233 
1234 	return sb_block;
1235 }
1236 
1237 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1238 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1239 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1240 
1241 #ifdef CONFIG_QUOTA
1242 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1243 {
1244 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1245 	char *qname;
1246 	int ret = -1;
1247 
1248 	if (sb_any_quota_loaded(sb) &&
1249 		!sbi->s_qf_names[qtype]) {
1250 		ext4_msg(sb, KERN_ERR,
1251 			"Cannot change journaled "
1252 			"quota options when quota turned on");
1253 		return -1;
1254 	}
1255 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1256 		ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1257 			 "when QUOTA feature is enabled");
1258 		return -1;
1259 	}
1260 	qname = match_strdup(args);
1261 	if (!qname) {
1262 		ext4_msg(sb, KERN_ERR,
1263 			"Not enough memory for storing quotafile name");
1264 		return -1;
1265 	}
1266 	if (sbi->s_qf_names[qtype]) {
1267 		if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1268 			ret = 1;
1269 		else
1270 			ext4_msg(sb, KERN_ERR,
1271 				 "%s quota file already specified",
1272 				 QTYPE2NAME(qtype));
1273 		goto errout;
1274 	}
1275 	if (strchr(qname, '/')) {
1276 		ext4_msg(sb, KERN_ERR,
1277 			"quotafile must be on filesystem root");
1278 		goto errout;
1279 	}
1280 	sbi->s_qf_names[qtype] = qname;
1281 	set_opt(sb, QUOTA);
1282 	return 1;
1283 errout:
1284 	kfree(qname);
1285 	return ret;
1286 }
1287 
1288 static int clear_qf_name(struct super_block *sb, int qtype)
1289 {
1290 
1291 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1292 
1293 	if (sb_any_quota_loaded(sb) &&
1294 		sbi->s_qf_names[qtype]) {
1295 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1296 			" when quota turned on");
1297 		return -1;
1298 	}
1299 	kfree(sbi->s_qf_names[qtype]);
1300 	sbi->s_qf_names[qtype] = NULL;
1301 	return 1;
1302 }
1303 #endif
1304 
1305 #define MOPT_SET	0x0001
1306 #define MOPT_CLEAR	0x0002
1307 #define MOPT_NOSUPPORT	0x0004
1308 #define MOPT_EXPLICIT	0x0008
1309 #define MOPT_CLEAR_ERR	0x0010
1310 #define MOPT_GTE0	0x0020
1311 #ifdef CONFIG_QUOTA
1312 #define MOPT_Q		0
1313 #define MOPT_QFMT	0x0040
1314 #else
1315 #define MOPT_Q		MOPT_NOSUPPORT
1316 #define MOPT_QFMT	MOPT_NOSUPPORT
1317 #endif
1318 #define MOPT_DATAJ	0x0080
1319 #define MOPT_NO_EXT2	0x0100
1320 #define MOPT_NO_EXT3	0x0200
1321 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1322 
1323 static const struct mount_opts {
1324 	int	token;
1325 	int	mount_opt;
1326 	int	flags;
1327 } ext4_mount_opts[] = {
1328 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1329 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1330 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1331 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1332 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1333 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1334 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1335 	 MOPT_EXT4_ONLY | MOPT_SET},
1336 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1337 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1338 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1339 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1340 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1341 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1342 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1343 	 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1344 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1345 	 MOPT_EXT4_ONLY | MOPT_SET},
1346 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1347 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1348 	 MOPT_EXT4_ONLY | MOPT_SET},
1349 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1350 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1351 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1352 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1353 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1354 	 MOPT_NO_EXT2 | MOPT_SET},
1355 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1356 	 MOPT_NO_EXT2 | MOPT_CLEAR},
1357 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1358 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1359 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1360 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1361 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1362 	{Opt_commit, 0, MOPT_GTE0},
1363 	{Opt_max_batch_time, 0, MOPT_GTE0},
1364 	{Opt_min_batch_time, 0, MOPT_GTE0},
1365 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1366 	{Opt_init_itable, 0, MOPT_GTE0},
1367 	{Opt_stripe, 0, MOPT_GTE0},
1368 	{Opt_resuid, 0, MOPT_GTE0},
1369 	{Opt_resgid, 0, MOPT_GTE0},
1370 	{Opt_journal_dev, 0, MOPT_GTE0},
1371 	{Opt_journal_ioprio, 0, MOPT_GTE0},
1372 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1373 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1374 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1375 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1376 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1377 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1378 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1379 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1380 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1381 #else
1382 	{Opt_acl, 0, MOPT_NOSUPPORT},
1383 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1384 #endif
1385 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1386 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1387 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1388 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1389 							MOPT_SET | MOPT_Q},
1390 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1391 							MOPT_SET | MOPT_Q},
1392 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1393 		       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1394 	{Opt_usrjquota, 0, MOPT_Q},
1395 	{Opt_grpjquota, 0, MOPT_Q},
1396 	{Opt_offusrjquota, 0, MOPT_Q},
1397 	{Opt_offgrpjquota, 0, MOPT_Q},
1398 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1399 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1400 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1401 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1402 	{Opt_err, 0, 0}
1403 };
1404 
1405 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1406 			    substring_t *args, unsigned long *journal_devnum,
1407 			    unsigned int *journal_ioprio, int is_remount)
1408 {
1409 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1410 	const struct mount_opts *m;
1411 	kuid_t uid;
1412 	kgid_t gid;
1413 	int arg = 0;
1414 
1415 #ifdef CONFIG_QUOTA
1416 	if (token == Opt_usrjquota)
1417 		return set_qf_name(sb, USRQUOTA, &args[0]);
1418 	else if (token == Opt_grpjquota)
1419 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1420 	else if (token == Opt_offusrjquota)
1421 		return clear_qf_name(sb, USRQUOTA);
1422 	else if (token == Opt_offgrpjquota)
1423 		return clear_qf_name(sb, GRPQUOTA);
1424 #endif
1425 	switch (token) {
1426 	case Opt_noacl:
1427 	case Opt_nouser_xattr:
1428 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1429 		break;
1430 	case Opt_sb:
1431 		return 1;	/* handled by get_sb_block() */
1432 	case Opt_removed:
1433 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1434 		return 1;
1435 	case Opt_abort:
1436 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1437 		return 1;
1438 	case Opt_i_version:
1439 		sb->s_flags |= MS_I_VERSION;
1440 		return 1;
1441 	}
1442 
1443 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1444 		if (token == m->token)
1445 			break;
1446 
1447 	if (m->token == Opt_err) {
1448 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1449 			 "or missing value", opt);
1450 		return -1;
1451 	}
1452 
1453 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1454 		ext4_msg(sb, KERN_ERR,
1455 			 "Mount option \"%s\" incompatible with ext2", opt);
1456 		return -1;
1457 	}
1458 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1459 		ext4_msg(sb, KERN_ERR,
1460 			 "Mount option \"%s\" incompatible with ext3", opt);
1461 		return -1;
1462 	}
1463 
1464 	if (args->from && match_int(args, &arg))
1465 		return -1;
1466 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1467 		return -1;
1468 	if (m->flags & MOPT_EXPLICIT)
1469 		set_opt2(sb, EXPLICIT_DELALLOC);
1470 	if (m->flags & MOPT_CLEAR_ERR)
1471 		clear_opt(sb, ERRORS_MASK);
1472 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1473 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1474 			 "options when quota turned on");
1475 		return -1;
1476 	}
1477 
1478 	if (m->flags & MOPT_NOSUPPORT) {
1479 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1480 	} else if (token == Opt_commit) {
1481 		if (arg == 0)
1482 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1483 		sbi->s_commit_interval = HZ * arg;
1484 	} else if (token == Opt_max_batch_time) {
1485 		if (arg == 0)
1486 			arg = EXT4_DEF_MAX_BATCH_TIME;
1487 		sbi->s_max_batch_time = arg;
1488 	} else if (token == Opt_min_batch_time) {
1489 		sbi->s_min_batch_time = arg;
1490 	} else if (token == Opt_inode_readahead_blks) {
1491 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1492 			ext4_msg(sb, KERN_ERR,
1493 				 "EXT4-fs: inode_readahead_blks must be "
1494 				 "0 or a power of 2 smaller than 2^31");
1495 			return -1;
1496 		}
1497 		sbi->s_inode_readahead_blks = arg;
1498 	} else if (token == Opt_init_itable) {
1499 		set_opt(sb, INIT_INODE_TABLE);
1500 		if (!args->from)
1501 			arg = EXT4_DEF_LI_WAIT_MULT;
1502 		sbi->s_li_wait_mult = arg;
1503 	} else if (token == Opt_max_dir_size_kb) {
1504 		sbi->s_max_dir_size_kb = arg;
1505 	} else if (token == Opt_stripe) {
1506 		sbi->s_stripe = arg;
1507 	} else if (token == Opt_resuid) {
1508 		uid = make_kuid(current_user_ns(), arg);
1509 		if (!uid_valid(uid)) {
1510 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1511 			return -1;
1512 		}
1513 		sbi->s_resuid = uid;
1514 	} else if (token == Opt_resgid) {
1515 		gid = make_kgid(current_user_ns(), arg);
1516 		if (!gid_valid(gid)) {
1517 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1518 			return -1;
1519 		}
1520 		sbi->s_resgid = gid;
1521 	} else if (token == Opt_journal_dev) {
1522 		if (is_remount) {
1523 			ext4_msg(sb, KERN_ERR,
1524 				 "Cannot specify journal on remount");
1525 			return -1;
1526 		}
1527 		*journal_devnum = arg;
1528 	} else if (token == Opt_journal_ioprio) {
1529 		if (arg > 7) {
1530 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1531 				 " (must be 0-7)");
1532 			return -1;
1533 		}
1534 		*journal_ioprio =
1535 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1536 	} else if (m->flags & MOPT_DATAJ) {
1537 		if (is_remount) {
1538 			if (!sbi->s_journal)
1539 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1540 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1541 				ext4_msg(sb, KERN_ERR,
1542 					 "Cannot change data mode on remount");
1543 				return -1;
1544 			}
1545 		} else {
1546 			clear_opt(sb, DATA_FLAGS);
1547 			sbi->s_mount_opt |= m->mount_opt;
1548 		}
1549 #ifdef CONFIG_QUOTA
1550 	} else if (m->flags & MOPT_QFMT) {
1551 		if (sb_any_quota_loaded(sb) &&
1552 		    sbi->s_jquota_fmt != m->mount_opt) {
1553 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1554 				 "quota options when quota turned on");
1555 			return -1;
1556 		}
1557 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1558 					       EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1559 			ext4_msg(sb, KERN_ERR,
1560 				 "Cannot set journaled quota options "
1561 				 "when QUOTA feature is enabled");
1562 			return -1;
1563 		}
1564 		sbi->s_jquota_fmt = m->mount_opt;
1565 #endif
1566 	} else {
1567 		if (!args->from)
1568 			arg = 1;
1569 		if (m->flags & MOPT_CLEAR)
1570 			arg = !arg;
1571 		else if (unlikely(!(m->flags & MOPT_SET))) {
1572 			ext4_msg(sb, KERN_WARNING,
1573 				 "buggy handling of option %s", opt);
1574 			WARN_ON(1);
1575 			return -1;
1576 		}
1577 		if (arg != 0)
1578 			sbi->s_mount_opt |= m->mount_opt;
1579 		else
1580 			sbi->s_mount_opt &= ~m->mount_opt;
1581 	}
1582 	return 1;
1583 }
1584 
1585 static int parse_options(char *options, struct super_block *sb,
1586 			 unsigned long *journal_devnum,
1587 			 unsigned int *journal_ioprio,
1588 			 int is_remount)
1589 {
1590 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1591 	char *p;
1592 	substring_t args[MAX_OPT_ARGS];
1593 	int token;
1594 
1595 	if (!options)
1596 		return 1;
1597 
1598 	while ((p = strsep(&options, ",")) != NULL) {
1599 		if (!*p)
1600 			continue;
1601 		/*
1602 		 * Initialize args struct so we know whether arg was
1603 		 * found; some options take optional arguments.
1604 		 */
1605 		args[0].to = args[0].from = NULL;
1606 		token = match_token(p, tokens, args);
1607 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1608 				     journal_ioprio, is_remount) < 0)
1609 			return 0;
1610 	}
1611 #ifdef CONFIG_QUOTA
1612 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1613 	    (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1614 		ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1615 			 "feature is enabled");
1616 		return 0;
1617 	}
1618 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1619 		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1620 			clear_opt(sb, USRQUOTA);
1621 
1622 		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1623 			clear_opt(sb, GRPQUOTA);
1624 
1625 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1626 			ext4_msg(sb, KERN_ERR, "old and new quota "
1627 					"format mixing");
1628 			return 0;
1629 		}
1630 
1631 		if (!sbi->s_jquota_fmt) {
1632 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1633 					"not specified");
1634 			return 0;
1635 		}
1636 	} else {
1637 		if (sbi->s_jquota_fmt) {
1638 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1639 					"specified with no journaling "
1640 					"enabled");
1641 			return 0;
1642 		}
1643 	}
1644 #endif
1645 	if (test_opt(sb, DIOREAD_NOLOCK)) {
1646 		int blocksize =
1647 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1648 
1649 		if (blocksize < PAGE_CACHE_SIZE) {
1650 			ext4_msg(sb, KERN_ERR, "can't mount with "
1651 				 "dioread_nolock if block size != PAGE_SIZE");
1652 			return 0;
1653 		}
1654 	}
1655 	return 1;
1656 }
1657 
1658 static inline void ext4_show_quota_options(struct seq_file *seq,
1659 					   struct super_block *sb)
1660 {
1661 #if defined(CONFIG_QUOTA)
1662 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1663 
1664 	if (sbi->s_jquota_fmt) {
1665 		char *fmtname = "";
1666 
1667 		switch (sbi->s_jquota_fmt) {
1668 		case QFMT_VFS_OLD:
1669 			fmtname = "vfsold";
1670 			break;
1671 		case QFMT_VFS_V0:
1672 			fmtname = "vfsv0";
1673 			break;
1674 		case QFMT_VFS_V1:
1675 			fmtname = "vfsv1";
1676 			break;
1677 		}
1678 		seq_printf(seq, ",jqfmt=%s", fmtname);
1679 	}
1680 
1681 	if (sbi->s_qf_names[USRQUOTA])
1682 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1683 
1684 	if (sbi->s_qf_names[GRPQUOTA])
1685 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1686 
1687 	if (test_opt(sb, USRQUOTA))
1688 		seq_puts(seq, ",usrquota");
1689 
1690 	if (test_opt(sb, GRPQUOTA))
1691 		seq_puts(seq, ",grpquota");
1692 #endif
1693 }
1694 
1695 static const char *token2str(int token)
1696 {
1697 	const struct match_token *t;
1698 
1699 	for (t = tokens; t->token != Opt_err; t++)
1700 		if (t->token == token && !strchr(t->pattern, '='))
1701 			break;
1702 	return t->pattern;
1703 }
1704 
1705 /*
1706  * Show an option if
1707  *  - it's set to a non-default value OR
1708  *  - if the per-sb default is different from the global default
1709  */
1710 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1711 			      int nodefs)
1712 {
1713 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1714 	struct ext4_super_block *es = sbi->s_es;
1715 	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1716 	const struct mount_opts *m;
1717 	char sep = nodefs ? '\n' : ',';
1718 
1719 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1720 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1721 
1722 	if (sbi->s_sb_block != 1)
1723 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1724 
1725 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1726 		int want_set = m->flags & MOPT_SET;
1727 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1728 		    (m->flags & MOPT_CLEAR_ERR))
1729 			continue;
1730 		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1731 			continue; /* skip if same as the default */
1732 		if ((want_set &&
1733 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1734 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1735 			continue; /* select Opt_noFoo vs Opt_Foo */
1736 		SEQ_OPTS_PRINT("%s", token2str(m->token));
1737 	}
1738 
1739 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1740 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1741 		SEQ_OPTS_PRINT("resuid=%u",
1742 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
1743 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1744 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1745 		SEQ_OPTS_PRINT("resgid=%u",
1746 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
1747 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1748 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1749 		SEQ_OPTS_PUTS("errors=remount-ro");
1750 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1751 		SEQ_OPTS_PUTS("errors=continue");
1752 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1753 		SEQ_OPTS_PUTS("errors=panic");
1754 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1755 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1756 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1757 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1758 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1759 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1760 	if (sb->s_flags & MS_I_VERSION)
1761 		SEQ_OPTS_PUTS("i_version");
1762 	if (nodefs || sbi->s_stripe)
1763 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1764 	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1765 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1766 			SEQ_OPTS_PUTS("data=journal");
1767 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1768 			SEQ_OPTS_PUTS("data=ordered");
1769 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1770 			SEQ_OPTS_PUTS("data=writeback");
1771 	}
1772 	if (nodefs ||
1773 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1774 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1775 			       sbi->s_inode_readahead_blks);
1776 
1777 	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1778 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1779 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1780 	if (nodefs || sbi->s_max_dir_size_kb)
1781 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1782 
1783 	ext4_show_quota_options(seq, sb);
1784 	return 0;
1785 }
1786 
1787 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1788 {
1789 	return _ext4_show_options(seq, root->d_sb, 0);
1790 }
1791 
1792 static int options_seq_show(struct seq_file *seq, void *offset)
1793 {
1794 	struct super_block *sb = seq->private;
1795 	int rc;
1796 
1797 	seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1798 	rc = _ext4_show_options(seq, sb, 1);
1799 	seq_puts(seq, "\n");
1800 	return rc;
1801 }
1802 
1803 static int options_open_fs(struct inode *inode, struct file *file)
1804 {
1805 	return single_open(file, options_seq_show, PDE(inode)->data);
1806 }
1807 
1808 static const struct file_operations ext4_seq_options_fops = {
1809 	.owner = THIS_MODULE,
1810 	.open = options_open_fs,
1811 	.read = seq_read,
1812 	.llseek = seq_lseek,
1813 	.release = single_release,
1814 };
1815 
1816 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1817 			    int read_only)
1818 {
1819 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1820 	int res = 0;
1821 
1822 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1823 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1824 			 "forcing read-only mode");
1825 		res = MS_RDONLY;
1826 	}
1827 	if (read_only)
1828 		goto done;
1829 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1830 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1831 			 "running e2fsck is recommended");
1832 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1833 		ext4_msg(sb, KERN_WARNING,
1834 			 "warning: mounting fs with errors, "
1835 			 "running e2fsck is recommended");
1836 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1837 		 le16_to_cpu(es->s_mnt_count) >=
1838 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1839 		ext4_msg(sb, KERN_WARNING,
1840 			 "warning: maximal mount count reached, "
1841 			 "running e2fsck is recommended");
1842 	else if (le32_to_cpu(es->s_checkinterval) &&
1843 		(le32_to_cpu(es->s_lastcheck) +
1844 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1845 		ext4_msg(sb, KERN_WARNING,
1846 			 "warning: checktime reached, "
1847 			 "running e2fsck is recommended");
1848 	if (!sbi->s_journal)
1849 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1850 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1851 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1852 	le16_add_cpu(&es->s_mnt_count, 1);
1853 	es->s_mtime = cpu_to_le32(get_seconds());
1854 	ext4_update_dynamic_rev(sb);
1855 	if (sbi->s_journal)
1856 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1857 
1858 	ext4_commit_super(sb, 1);
1859 done:
1860 	if (test_opt(sb, DEBUG))
1861 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1862 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1863 			sb->s_blocksize,
1864 			sbi->s_groups_count,
1865 			EXT4_BLOCKS_PER_GROUP(sb),
1866 			EXT4_INODES_PER_GROUP(sb),
1867 			sbi->s_mount_opt, sbi->s_mount_opt2);
1868 
1869 	cleancache_init_fs(sb);
1870 	return res;
1871 }
1872 
1873 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1874 {
1875 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1876 	struct flex_groups *new_groups;
1877 	int size;
1878 
1879 	if (!sbi->s_log_groups_per_flex)
1880 		return 0;
1881 
1882 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
1883 	if (size <= sbi->s_flex_groups_allocated)
1884 		return 0;
1885 
1886 	size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1887 	new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1888 	if (!new_groups) {
1889 		ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1890 			 size / (int) sizeof(struct flex_groups));
1891 		return -ENOMEM;
1892 	}
1893 
1894 	if (sbi->s_flex_groups) {
1895 		memcpy(new_groups, sbi->s_flex_groups,
1896 		       (sbi->s_flex_groups_allocated *
1897 			sizeof(struct flex_groups)));
1898 		ext4_kvfree(sbi->s_flex_groups);
1899 	}
1900 	sbi->s_flex_groups = new_groups;
1901 	sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1902 	return 0;
1903 }
1904 
1905 static int ext4_fill_flex_info(struct super_block *sb)
1906 {
1907 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1908 	struct ext4_group_desc *gdp = NULL;
1909 	ext4_group_t flex_group;
1910 	unsigned int groups_per_flex = 0;
1911 	int i, err;
1912 
1913 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1914 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1915 		sbi->s_log_groups_per_flex = 0;
1916 		return 1;
1917 	}
1918 	groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1919 
1920 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1921 	if (err)
1922 		goto failed;
1923 
1924 	for (i = 0; i < sbi->s_groups_count; i++) {
1925 		gdp = ext4_get_group_desc(sb, i, NULL);
1926 
1927 		flex_group = ext4_flex_group(sbi, i);
1928 		atomic_add(ext4_free_inodes_count(sb, gdp),
1929 			   &sbi->s_flex_groups[flex_group].free_inodes);
1930 		atomic64_add(ext4_free_group_clusters(sb, gdp),
1931 			     &sbi->s_flex_groups[flex_group].free_clusters);
1932 		atomic_add(ext4_used_dirs_count(sb, gdp),
1933 			   &sbi->s_flex_groups[flex_group].used_dirs);
1934 	}
1935 
1936 	return 1;
1937 failed:
1938 	return 0;
1939 }
1940 
1941 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1942 				   struct ext4_group_desc *gdp)
1943 {
1944 	int offset;
1945 	__u16 crc = 0;
1946 	__le32 le_group = cpu_to_le32(block_group);
1947 
1948 	if ((sbi->s_es->s_feature_ro_compat &
1949 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1950 		/* Use new metadata_csum algorithm */
1951 		__u16 old_csum;
1952 		__u32 csum32;
1953 
1954 		old_csum = gdp->bg_checksum;
1955 		gdp->bg_checksum = 0;
1956 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1957 				     sizeof(le_group));
1958 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1959 				     sbi->s_desc_size);
1960 		gdp->bg_checksum = old_csum;
1961 
1962 		crc = csum32 & 0xFFFF;
1963 		goto out;
1964 	}
1965 
1966 	/* old crc16 code */
1967 	offset = offsetof(struct ext4_group_desc, bg_checksum);
1968 
1969 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1970 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1971 	crc = crc16(crc, (__u8 *)gdp, offset);
1972 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
1973 	/* for checksum of struct ext4_group_desc do the rest...*/
1974 	if ((sbi->s_es->s_feature_incompat &
1975 	     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1976 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1977 		crc = crc16(crc, (__u8 *)gdp + offset,
1978 			    le16_to_cpu(sbi->s_es->s_desc_size) -
1979 				offset);
1980 
1981 out:
1982 	return cpu_to_le16(crc);
1983 }
1984 
1985 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1986 				struct ext4_group_desc *gdp)
1987 {
1988 	if (ext4_has_group_desc_csum(sb) &&
1989 	    (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1990 						      block_group, gdp)))
1991 		return 0;
1992 
1993 	return 1;
1994 }
1995 
1996 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1997 			      struct ext4_group_desc *gdp)
1998 {
1999 	if (!ext4_has_group_desc_csum(sb))
2000 		return;
2001 	gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2002 }
2003 
2004 /* Called at mount-time, super-block is locked */
2005 static int ext4_check_descriptors(struct super_block *sb,
2006 				  ext4_group_t *first_not_zeroed)
2007 {
2008 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2009 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2010 	ext4_fsblk_t last_block;
2011 	ext4_fsblk_t block_bitmap;
2012 	ext4_fsblk_t inode_bitmap;
2013 	ext4_fsblk_t inode_table;
2014 	int flexbg_flag = 0;
2015 	ext4_group_t i, grp = sbi->s_groups_count;
2016 
2017 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2018 		flexbg_flag = 1;
2019 
2020 	ext4_debug("Checking group descriptors");
2021 
2022 	for (i = 0; i < sbi->s_groups_count; i++) {
2023 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2024 
2025 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2026 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2027 		else
2028 			last_block = first_block +
2029 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2030 
2031 		if ((grp == sbi->s_groups_count) &&
2032 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2033 			grp = i;
2034 
2035 		block_bitmap = ext4_block_bitmap(sb, gdp);
2036 		if (block_bitmap < first_block || block_bitmap > last_block) {
2037 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2038 			       "Block bitmap for group %u not in group "
2039 			       "(block %llu)!", i, block_bitmap);
2040 			return 0;
2041 		}
2042 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2043 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2044 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2045 			       "Inode bitmap for group %u not in group "
2046 			       "(block %llu)!", i, inode_bitmap);
2047 			return 0;
2048 		}
2049 		inode_table = ext4_inode_table(sb, gdp);
2050 		if (inode_table < first_block ||
2051 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2052 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2053 			       "Inode table for group %u not in group "
2054 			       "(block %llu)!", i, inode_table);
2055 			return 0;
2056 		}
2057 		ext4_lock_group(sb, i);
2058 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2059 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2060 				 "Checksum for group %u failed (%u!=%u)",
2061 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2062 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2063 			if (!(sb->s_flags & MS_RDONLY)) {
2064 				ext4_unlock_group(sb, i);
2065 				return 0;
2066 			}
2067 		}
2068 		ext4_unlock_group(sb, i);
2069 		if (!flexbg_flag)
2070 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2071 	}
2072 	if (NULL != first_not_zeroed)
2073 		*first_not_zeroed = grp;
2074 
2075 	ext4_free_blocks_count_set(sbi->s_es,
2076 				   EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2077 	sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2078 	return 1;
2079 }
2080 
2081 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2082  * the superblock) which were deleted from all directories, but held open by
2083  * a process at the time of a crash.  We walk the list and try to delete these
2084  * inodes at recovery time (only with a read-write filesystem).
2085  *
2086  * In order to keep the orphan inode chain consistent during traversal (in
2087  * case of crash during recovery), we link each inode into the superblock
2088  * orphan list_head and handle it the same way as an inode deletion during
2089  * normal operation (which journals the operations for us).
2090  *
2091  * We only do an iget() and an iput() on each inode, which is very safe if we
2092  * accidentally point at an in-use or already deleted inode.  The worst that
2093  * can happen in this case is that we get a "bit already cleared" message from
2094  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2095  * e2fsck was run on this filesystem, and it must have already done the orphan
2096  * inode cleanup for us, so we can safely abort without any further action.
2097  */
2098 static void ext4_orphan_cleanup(struct super_block *sb,
2099 				struct ext4_super_block *es)
2100 {
2101 	unsigned int s_flags = sb->s_flags;
2102 	int nr_orphans = 0, nr_truncates = 0;
2103 #ifdef CONFIG_QUOTA
2104 	int i;
2105 #endif
2106 	if (!es->s_last_orphan) {
2107 		jbd_debug(4, "no orphan inodes to clean up\n");
2108 		return;
2109 	}
2110 
2111 	if (bdev_read_only(sb->s_bdev)) {
2112 		ext4_msg(sb, KERN_ERR, "write access "
2113 			"unavailable, skipping orphan cleanup");
2114 		return;
2115 	}
2116 
2117 	/* Check if feature set would not allow a r/w mount */
2118 	if (!ext4_feature_set_ok(sb, 0)) {
2119 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2120 			 "unknown ROCOMPAT features");
2121 		return;
2122 	}
2123 
2124 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2125 		/* don't clear list on RO mount w/ errors */
2126 		if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2127 			jbd_debug(1, "Errors on filesystem, "
2128 				  "clearing orphan list.\n");
2129 			es->s_last_orphan = 0;
2130 		}
2131 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2132 		return;
2133 	}
2134 
2135 	if (s_flags & MS_RDONLY) {
2136 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2137 		sb->s_flags &= ~MS_RDONLY;
2138 	}
2139 #ifdef CONFIG_QUOTA
2140 	/* Needed for iput() to work correctly and not trash data */
2141 	sb->s_flags |= MS_ACTIVE;
2142 	/* Turn on quotas so that they are updated correctly */
2143 	for (i = 0; i < MAXQUOTAS; i++) {
2144 		if (EXT4_SB(sb)->s_qf_names[i]) {
2145 			int ret = ext4_quota_on_mount(sb, i);
2146 			if (ret < 0)
2147 				ext4_msg(sb, KERN_ERR,
2148 					"Cannot turn on journaled "
2149 					"quota: error %d", ret);
2150 		}
2151 	}
2152 #endif
2153 
2154 	while (es->s_last_orphan) {
2155 		struct inode *inode;
2156 
2157 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2158 		if (IS_ERR(inode)) {
2159 			es->s_last_orphan = 0;
2160 			break;
2161 		}
2162 
2163 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2164 		dquot_initialize(inode);
2165 		if (inode->i_nlink) {
2166 			ext4_msg(sb, KERN_DEBUG,
2167 				"%s: truncating inode %lu to %lld bytes",
2168 				__func__, inode->i_ino, inode->i_size);
2169 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2170 				  inode->i_ino, inode->i_size);
2171 			mutex_lock(&inode->i_mutex);
2172 			ext4_truncate(inode);
2173 			mutex_unlock(&inode->i_mutex);
2174 			nr_truncates++;
2175 		} else {
2176 			ext4_msg(sb, KERN_DEBUG,
2177 				"%s: deleting unreferenced inode %lu",
2178 				__func__, inode->i_ino);
2179 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2180 				  inode->i_ino);
2181 			nr_orphans++;
2182 		}
2183 		iput(inode);  /* The delete magic happens here! */
2184 	}
2185 
2186 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2187 
2188 	if (nr_orphans)
2189 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2190 		       PLURAL(nr_orphans));
2191 	if (nr_truncates)
2192 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2193 		       PLURAL(nr_truncates));
2194 #ifdef CONFIG_QUOTA
2195 	/* Turn quotas off */
2196 	for (i = 0; i < MAXQUOTAS; i++) {
2197 		if (sb_dqopt(sb)->files[i])
2198 			dquot_quota_off(sb, i);
2199 	}
2200 #endif
2201 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2202 }
2203 
2204 /*
2205  * Maximal extent format file size.
2206  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2207  * extent format containers, within a sector_t, and within i_blocks
2208  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2209  * so that won't be a limiting factor.
2210  *
2211  * However there is other limiting factor. We do store extents in the form
2212  * of starting block and length, hence the resulting length of the extent
2213  * covering maximum file size must fit into on-disk format containers as
2214  * well. Given that length is always by 1 unit bigger than max unit (because
2215  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2216  *
2217  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2218  */
2219 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2220 {
2221 	loff_t res;
2222 	loff_t upper_limit = MAX_LFS_FILESIZE;
2223 
2224 	/* small i_blocks in vfs inode? */
2225 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2226 		/*
2227 		 * CONFIG_LBDAF is not enabled implies the inode
2228 		 * i_block represent total blocks in 512 bytes
2229 		 * 32 == size of vfs inode i_blocks * 8
2230 		 */
2231 		upper_limit = (1LL << 32) - 1;
2232 
2233 		/* total blocks in file system block size */
2234 		upper_limit >>= (blkbits - 9);
2235 		upper_limit <<= blkbits;
2236 	}
2237 
2238 	/*
2239 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2240 	 * by one fs block, so ee_len can cover the extent of maximum file
2241 	 * size
2242 	 */
2243 	res = (1LL << 32) - 1;
2244 	res <<= blkbits;
2245 
2246 	/* Sanity check against vm- & vfs- imposed limits */
2247 	if (res > upper_limit)
2248 		res = upper_limit;
2249 
2250 	return res;
2251 }
2252 
2253 /*
2254  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2255  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2256  * We need to be 1 filesystem block less than the 2^48 sector limit.
2257  */
2258 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2259 {
2260 	loff_t res = EXT4_NDIR_BLOCKS;
2261 	int meta_blocks;
2262 	loff_t upper_limit;
2263 	/* This is calculated to be the largest file size for a dense, block
2264 	 * mapped file such that the file's total number of 512-byte sectors,
2265 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2266 	 *
2267 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2268 	 * number of 512-byte sectors of the file.
2269 	 */
2270 
2271 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2272 		/*
2273 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2274 		 * the inode i_block field represents total file blocks in
2275 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2276 		 */
2277 		upper_limit = (1LL << 32) - 1;
2278 
2279 		/* total blocks in file system block size */
2280 		upper_limit >>= (bits - 9);
2281 
2282 	} else {
2283 		/*
2284 		 * We use 48 bit ext4_inode i_blocks
2285 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2286 		 * represent total number of blocks in
2287 		 * file system block size
2288 		 */
2289 		upper_limit = (1LL << 48) - 1;
2290 
2291 	}
2292 
2293 	/* indirect blocks */
2294 	meta_blocks = 1;
2295 	/* double indirect blocks */
2296 	meta_blocks += 1 + (1LL << (bits-2));
2297 	/* tripple indirect blocks */
2298 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2299 
2300 	upper_limit -= meta_blocks;
2301 	upper_limit <<= bits;
2302 
2303 	res += 1LL << (bits-2);
2304 	res += 1LL << (2*(bits-2));
2305 	res += 1LL << (3*(bits-2));
2306 	res <<= bits;
2307 	if (res > upper_limit)
2308 		res = upper_limit;
2309 
2310 	if (res > MAX_LFS_FILESIZE)
2311 		res = MAX_LFS_FILESIZE;
2312 
2313 	return res;
2314 }
2315 
2316 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2317 				   ext4_fsblk_t logical_sb_block, int nr)
2318 {
2319 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2320 	ext4_group_t bg, first_meta_bg;
2321 	int has_super = 0;
2322 
2323 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2324 
2325 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2326 	    nr < first_meta_bg)
2327 		return logical_sb_block + nr + 1;
2328 	bg = sbi->s_desc_per_block * nr;
2329 	if (ext4_bg_has_super(sb, bg))
2330 		has_super = 1;
2331 
2332 	return (has_super + ext4_group_first_block_no(sb, bg));
2333 }
2334 
2335 /**
2336  * ext4_get_stripe_size: Get the stripe size.
2337  * @sbi: In memory super block info
2338  *
2339  * If we have specified it via mount option, then
2340  * use the mount option value. If the value specified at mount time is
2341  * greater than the blocks per group use the super block value.
2342  * If the super block value is greater than blocks per group return 0.
2343  * Allocator needs it be less than blocks per group.
2344  *
2345  */
2346 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2347 {
2348 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2349 	unsigned long stripe_width =
2350 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2351 	int ret;
2352 
2353 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2354 		ret = sbi->s_stripe;
2355 	else if (stripe_width <= sbi->s_blocks_per_group)
2356 		ret = stripe_width;
2357 	else if (stride <= sbi->s_blocks_per_group)
2358 		ret = stride;
2359 	else
2360 		ret = 0;
2361 
2362 	/*
2363 	 * If the stripe width is 1, this makes no sense and
2364 	 * we set it to 0 to turn off stripe handling code.
2365 	 */
2366 	if (ret <= 1)
2367 		ret = 0;
2368 
2369 	return ret;
2370 }
2371 
2372 /* sysfs supprt */
2373 
2374 struct ext4_attr {
2375 	struct attribute attr;
2376 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2377 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2378 			 const char *, size_t);
2379 	int offset;
2380 };
2381 
2382 static int parse_strtoul(const char *buf,
2383 		unsigned long max, unsigned long *value)
2384 {
2385 	char *endp;
2386 
2387 	*value = simple_strtoul(skip_spaces(buf), &endp, 0);
2388 	endp = skip_spaces(endp);
2389 	if (*endp || *value > max)
2390 		return -EINVAL;
2391 
2392 	return 0;
2393 }
2394 
2395 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2396 					      struct ext4_sb_info *sbi,
2397 					      char *buf)
2398 {
2399 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2400 		(s64) EXT4_C2B(sbi,
2401 			percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2402 }
2403 
2404 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2405 					 struct ext4_sb_info *sbi, char *buf)
2406 {
2407 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2408 
2409 	if (!sb->s_bdev->bd_part)
2410 		return snprintf(buf, PAGE_SIZE, "0\n");
2411 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2412 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2413 			 sbi->s_sectors_written_start) >> 1);
2414 }
2415 
2416 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2417 					  struct ext4_sb_info *sbi, char *buf)
2418 {
2419 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2420 
2421 	if (!sb->s_bdev->bd_part)
2422 		return snprintf(buf, PAGE_SIZE, "0\n");
2423 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2424 			(unsigned long long)(sbi->s_kbytes_written +
2425 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2426 			  EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2427 }
2428 
2429 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2430 					  struct ext4_sb_info *sbi,
2431 					  const char *buf, size_t count)
2432 {
2433 	unsigned long t;
2434 
2435 	if (parse_strtoul(buf, 0x40000000, &t))
2436 		return -EINVAL;
2437 
2438 	if (t && !is_power_of_2(t))
2439 		return -EINVAL;
2440 
2441 	sbi->s_inode_readahead_blks = t;
2442 	return count;
2443 }
2444 
2445 static ssize_t sbi_ui_show(struct ext4_attr *a,
2446 			   struct ext4_sb_info *sbi, char *buf)
2447 {
2448 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2449 
2450 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2451 }
2452 
2453 static ssize_t sbi_ui_store(struct ext4_attr *a,
2454 			    struct ext4_sb_info *sbi,
2455 			    const char *buf, size_t count)
2456 {
2457 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2458 	unsigned long t;
2459 
2460 	if (parse_strtoul(buf, 0xffffffff, &t))
2461 		return -EINVAL;
2462 	*ui = t;
2463 	return count;
2464 }
2465 
2466 static ssize_t trigger_test_error(struct ext4_attr *a,
2467 				  struct ext4_sb_info *sbi,
2468 				  const char *buf, size_t count)
2469 {
2470 	int len = count;
2471 
2472 	if (!capable(CAP_SYS_ADMIN))
2473 		return -EPERM;
2474 
2475 	if (len && buf[len-1] == '\n')
2476 		len--;
2477 
2478 	if (len)
2479 		ext4_error(sbi->s_sb, "%.*s", len, buf);
2480 	return count;
2481 }
2482 
2483 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2484 static struct ext4_attr ext4_attr_##_name = {			\
2485 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2486 	.show	= _show,					\
2487 	.store	= _store,					\
2488 	.offset = offsetof(struct ext4_sb_info, _elname),	\
2489 }
2490 #define EXT4_ATTR(name, mode, show, store) \
2491 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2492 
2493 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2494 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2495 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2496 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2497 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2498 #define ATTR_LIST(name) &ext4_attr_##name.attr
2499 
2500 EXT4_RO_ATTR(delayed_allocation_blocks);
2501 EXT4_RO_ATTR(session_write_kbytes);
2502 EXT4_RO_ATTR(lifetime_write_kbytes);
2503 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2504 		 inode_readahead_blks_store, s_inode_readahead_blks);
2505 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2506 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2507 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2508 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2509 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2510 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2511 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2512 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2513 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2514 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2515 
2516 static struct attribute *ext4_attrs[] = {
2517 	ATTR_LIST(delayed_allocation_blocks),
2518 	ATTR_LIST(session_write_kbytes),
2519 	ATTR_LIST(lifetime_write_kbytes),
2520 	ATTR_LIST(inode_readahead_blks),
2521 	ATTR_LIST(inode_goal),
2522 	ATTR_LIST(mb_stats),
2523 	ATTR_LIST(mb_max_to_scan),
2524 	ATTR_LIST(mb_min_to_scan),
2525 	ATTR_LIST(mb_order2_req),
2526 	ATTR_LIST(mb_stream_req),
2527 	ATTR_LIST(mb_group_prealloc),
2528 	ATTR_LIST(max_writeback_mb_bump),
2529 	ATTR_LIST(extent_max_zeroout_kb),
2530 	ATTR_LIST(trigger_fs_error),
2531 	NULL,
2532 };
2533 
2534 /* Features this copy of ext4 supports */
2535 EXT4_INFO_ATTR(lazy_itable_init);
2536 EXT4_INFO_ATTR(batched_discard);
2537 EXT4_INFO_ATTR(meta_bg_resize);
2538 
2539 static struct attribute *ext4_feat_attrs[] = {
2540 	ATTR_LIST(lazy_itable_init),
2541 	ATTR_LIST(batched_discard),
2542 	ATTR_LIST(meta_bg_resize),
2543 	NULL,
2544 };
2545 
2546 static ssize_t ext4_attr_show(struct kobject *kobj,
2547 			      struct attribute *attr, char *buf)
2548 {
2549 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2550 						s_kobj);
2551 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2552 
2553 	return a->show ? a->show(a, sbi, buf) : 0;
2554 }
2555 
2556 static ssize_t ext4_attr_store(struct kobject *kobj,
2557 			       struct attribute *attr,
2558 			       const char *buf, size_t len)
2559 {
2560 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2561 						s_kobj);
2562 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2563 
2564 	return a->store ? a->store(a, sbi, buf, len) : 0;
2565 }
2566 
2567 static void ext4_sb_release(struct kobject *kobj)
2568 {
2569 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2570 						s_kobj);
2571 	complete(&sbi->s_kobj_unregister);
2572 }
2573 
2574 static const struct sysfs_ops ext4_attr_ops = {
2575 	.show	= ext4_attr_show,
2576 	.store	= ext4_attr_store,
2577 };
2578 
2579 static struct kobj_type ext4_ktype = {
2580 	.default_attrs	= ext4_attrs,
2581 	.sysfs_ops	= &ext4_attr_ops,
2582 	.release	= ext4_sb_release,
2583 };
2584 
2585 static void ext4_feat_release(struct kobject *kobj)
2586 {
2587 	complete(&ext4_feat->f_kobj_unregister);
2588 }
2589 
2590 static struct kobj_type ext4_feat_ktype = {
2591 	.default_attrs	= ext4_feat_attrs,
2592 	.sysfs_ops	= &ext4_attr_ops,
2593 	.release	= ext4_feat_release,
2594 };
2595 
2596 /*
2597  * Check whether this filesystem can be mounted based on
2598  * the features present and the RDONLY/RDWR mount requested.
2599  * Returns 1 if this filesystem can be mounted as requested,
2600  * 0 if it cannot be.
2601  */
2602 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2603 {
2604 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2605 		ext4_msg(sb, KERN_ERR,
2606 			"Couldn't mount because of "
2607 			"unsupported optional features (%x)",
2608 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2609 			~EXT4_FEATURE_INCOMPAT_SUPP));
2610 		return 0;
2611 	}
2612 
2613 	if (readonly)
2614 		return 1;
2615 
2616 	/* Check that feature set is OK for a read-write mount */
2617 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2618 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2619 			 "unsupported optional features (%x)",
2620 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2621 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2622 		return 0;
2623 	}
2624 	/*
2625 	 * Large file size enabled file system can only be mounted
2626 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2627 	 */
2628 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2629 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2630 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2631 				 "cannot be mounted RDWR without "
2632 				 "CONFIG_LBDAF");
2633 			return 0;
2634 		}
2635 	}
2636 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2637 	    !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2638 		ext4_msg(sb, KERN_ERR,
2639 			 "Can't support bigalloc feature without "
2640 			 "extents feature\n");
2641 		return 0;
2642 	}
2643 
2644 #ifndef CONFIG_QUOTA
2645 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2646 	    !readonly) {
2647 		ext4_msg(sb, KERN_ERR,
2648 			 "Filesystem with quota feature cannot be mounted RDWR "
2649 			 "without CONFIG_QUOTA");
2650 		return 0;
2651 	}
2652 #endif  /* CONFIG_QUOTA */
2653 	return 1;
2654 }
2655 
2656 /*
2657  * This function is called once a day if we have errors logged
2658  * on the file system
2659  */
2660 static void print_daily_error_info(unsigned long arg)
2661 {
2662 	struct super_block *sb = (struct super_block *) arg;
2663 	struct ext4_sb_info *sbi;
2664 	struct ext4_super_block *es;
2665 
2666 	sbi = EXT4_SB(sb);
2667 	es = sbi->s_es;
2668 
2669 	if (es->s_error_count)
2670 		ext4_msg(sb, KERN_NOTICE, "error count: %u",
2671 			 le32_to_cpu(es->s_error_count));
2672 	if (es->s_first_error_time) {
2673 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2674 		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2675 		       (int) sizeof(es->s_first_error_func),
2676 		       es->s_first_error_func,
2677 		       le32_to_cpu(es->s_first_error_line));
2678 		if (es->s_first_error_ino)
2679 			printk(": inode %u",
2680 			       le32_to_cpu(es->s_first_error_ino));
2681 		if (es->s_first_error_block)
2682 			printk(": block %llu", (unsigned long long)
2683 			       le64_to_cpu(es->s_first_error_block));
2684 		printk("\n");
2685 	}
2686 	if (es->s_last_error_time) {
2687 		printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2688 		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2689 		       (int) sizeof(es->s_last_error_func),
2690 		       es->s_last_error_func,
2691 		       le32_to_cpu(es->s_last_error_line));
2692 		if (es->s_last_error_ino)
2693 			printk(": inode %u",
2694 			       le32_to_cpu(es->s_last_error_ino));
2695 		if (es->s_last_error_block)
2696 			printk(": block %llu", (unsigned long long)
2697 			       le64_to_cpu(es->s_last_error_block));
2698 		printk("\n");
2699 	}
2700 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2701 }
2702 
2703 /* Find next suitable group and run ext4_init_inode_table */
2704 static int ext4_run_li_request(struct ext4_li_request *elr)
2705 {
2706 	struct ext4_group_desc *gdp = NULL;
2707 	ext4_group_t group, ngroups;
2708 	struct super_block *sb;
2709 	unsigned long timeout = 0;
2710 	int ret = 0;
2711 
2712 	sb = elr->lr_super;
2713 	ngroups = EXT4_SB(sb)->s_groups_count;
2714 
2715 	sb_start_write(sb);
2716 	for (group = elr->lr_next_group; group < ngroups; group++) {
2717 		gdp = ext4_get_group_desc(sb, group, NULL);
2718 		if (!gdp) {
2719 			ret = 1;
2720 			break;
2721 		}
2722 
2723 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2724 			break;
2725 	}
2726 
2727 	if (group >= ngroups)
2728 		ret = 1;
2729 
2730 	if (!ret) {
2731 		timeout = jiffies;
2732 		ret = ext4_init_inode_table(sb, group,
2733 					    elr->lr_timeout ? 0 : 1);
2734 		if (elr->lr_timeout == 0) {
2735 			timeout = (jiffies - timeout) *
2736 				  elr->lr_sbi->s_li_wait_mult;
2737 			elr->lr_timeout = timeout;
2738 		}
2739 		elr->lr_next_sched = jiffies + elr->lr_timeout;
2740 		elr->lr_next_group = group + 1;
2741 	}
2742 	sb_end_write(sb);
2743 
2744 	return ret;
2745 }
2746 
2747 /*
2748  * Remove lr_request from the list_request and free the
2749  * request structure. Should be called with li_list_mtx held
2750  */
2751 static void ext4_remove_li_request(struct ext4_li_request *elr)
2752 {
2753 	struct ext4_sb_info *sbi;
2754 
2755 	if (!elr)
2756 		return;
2757 
2758 	sbi = elr->lr_sbi;
2759 
2760 	list_del(&elr->lr_request);
2761 	sbi->s_li_request = NULL;
2762 	kfree(elr);
2763 }
2764 
2765 static void ext4_unregister_li_request(struct super_block *sb)
2766 {
2767 	mutex_lock(&ext4_li_mtx);
2768 	if (!ext4_li_info) {
2769 		mutex_unlock(&ext4_li_mtx);
2770 		return;
2771 	}
2772 
2773 	mutex_lock(&ext4_li_info->li_list_mtx);
2774 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2775 	mutex_unlock(&ext4_li_info->li_list_mtx);
2776 	mutex_unlock(&ext4_li_mtx);
2777 }
2778 
2779 static struct task_struct *ext4_lazyinit_task;
2780 
2781 /*
2782  * This is the function where ext4lazyinit thread lives. It walks
2783  * through the request list searching for next scheduled filesystem.
2784  * When such a fs is found, run the lazy initialization request
2785  * (ext4_rn_li_request) and keep track of the time spend in this
2786  * function. Based on that time we compute next schedule time of
2787  * the request. When walking through the list is complete, compute
2788  * next waking time and put itself into sleep.
2789  */
2790 static int ext4_lazyinit_thread(void *arg)
2791 {
2792 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2793 	struct list_head *pos, *n;
2794 	struct ext4_li_request *elr;
2795 	unsigned long next_wakeup, cur;
2796 
2797 	BUG_ON(NULL == eli);
2798 
2799 cont_thread:
2800 	while (true) {
2801 		next_wakeup = MAX_JIFFY_OFFSET;
2802 
2803 		mutex_lock(&eli->li_list_mtx);
2804 		if (list_empty(&eli->li_request_list)) {
2805 			mutex_unlock(&eli->li_list_mtx);
2806 			goto exit_thread;
2807 		}
2808 
2809 		list_for_each_safe(pos, n, &eli->li_request_list) {
2810 			elr = list_entry(pos, struct ext4_li_request,
2811 					 lr_request);
2812 
2813 			if (time_after_eq(jiffies, elr->lr_next_sched)) {
2814 				if (ext4_run_li_request(elr) != 0) {
2815 					/* error, remove the lazy_init job */
2816 					ext4_remove_li_request(elr);
2817 					continue;
2818 				}
2819 			}
2820 
2821 			if (time_before(elr->lr_next_sched, next_wakeup))
2822 				next_wakeup = elr->lr_next_sched;
2823 		}
2824 		mutex_unlock(&eli->li_list_mtx);
2825 
2826 		try_to_freeze();
2827 
2828 		cur = jiffies;
2829 		if ((time_after_eq(cur, next_wakeup)) ||
2830 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
2831 			cond_resched();
2832 			continue;
2833 		}
2834 
2835 		schedule_timeout_interruptible(next_wakeup - cur);
2836 
2837 		if (kthread_should_stop()) {
2838 			ext4_clear_request_list();
2839 			goto exit_thread;
2840 		}
2841 	}
2842 
2843 exit_thread:
2844 	/*
2845 	 * It looks like the request list is empty, but we need
2846 	 * to check it under the li_list_mtx lock, to prevent any
2847 	 * additions into it, and of course we should lock ext4_li_mtx
2848 	 * to atomically free the list and ext4_li_info, because at
2849 	 * this point another ext4 filesystem could be registering
2850 	 * new one.
2851 	 */
2852 	mutex_lock(&ext4_li_mtx);
2853 	mutex_lock(&eli->li_list_mtx);
2854 	if (!list_empty(&eli->li_request_list)) {
2855 		mutex_unlock(&eli->li_list_mtx);
2856 		mutex_unlock(&ext4_li_mtx);
2857 		goto cont_thread;
2858 	}
2859 	mutex_unlock(&eli->li_list_mtx);
2860 	kfree(ext4_li_info);
2861 	ext4_li_info = NULL;
2862 	mutex_unlock(&ext4_li_mtx);
2863 
2864 	return 0;
2865 }
2866 
2867 static void ext4_clear_request_list(void)
2868 {
2869 	struct list_head *pos, *n;
2870 	struct ext4_li_request *elr;
2871 
2872 	mutex_lock(&ext4_li_info->li_list_mtx);
2873 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2874 		elr = list_entry(pos, struct ext4_li_request,
2875 				 lr_request);
2876 		ext4_remove_li_request(elr);
2877 	}
2878 	mutex_unlock(&ext4_li_info->li_list_mtx);
2879 }
2880 
2881 static int ext4_run_lazyinit_thread(void)
2882 {
2883 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2884 					 ext4_li_info, "ext4lazyinit");
2885 	if (IS_ERR(ext4_lazyinit_task)) {
2886 		int err = PTR_ERR(ext4_lazyinit_task);
2887 		ext4_clear_request_list();
2888 		kfree(ext4_li_info);
2889 		ext4_li_info = NULL;
2890 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2891 				 "initialization thread\n",
2892 				 err);
2893 		return err;
2894 	}
2895 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2896 	return 0;
2897 }
2898 
2899 /*
2900  * Check whether it make sense to run itable init. thread or not.
2901  * If there is at least one uninitialized inode table, return
2902  * corresponding group number, else the loop goes through all
2903  * groups and return total number of groups.
2904  */
2905 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2906 {
2907 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2908 	struct ext4_group_desc *gdp = NULL;
2909 
2910 	for (group = 0; group < ngroups; group++) {
2911 		gdp = ext4_get_group_desc(sb, group, NULL);
2912 		if (!gdp)
2913 			continue;
2914 
2915 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2916 			break;
2917 	}
2918 
2919 	return group;
2920 }
2921 
2922 static int ext4_li_info_new(void)
2923 {
2924 	struct ext4_lazy_init *eli = NULL;
2925 
2926 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2927 	if (!eli)
2928 		return -ENOMEM;
2929 
2930 	INIT_LIST_HEAD(&eli->li_request_list);
2931 	mutex_init(&eli->li_list_mtx);
2932 
2933 	eli->li_state |= EXT4_LAZYINIT_QUIT;
2934 
2935 	ext4_li_info = eli;
2936 
2937 	return 0;
2938 }
2939 
2940 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2941 					    ext4_group_t start)
2942 {
2943 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2944 	struct ext4_li_request *elr;
2945 	unsigned long rnd;
2946 
2947 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2948 	if (!elr)
2949 		return NULL;
2950 
2951 	elr->lr_super = sb;
2952 	elr->lr_sbi = sbi;
2953 	elr->lr_next_group = start;
2954 
2955 	/*
2956 	 * Randomize first schedule time of the request to
2957 	 * spread the inode table initialization requests
2958 	 * better.
2959 	 */
2960 	get_random_bytes(&rnd, sizeof(rnd));
2961 	elr->lr_next_sched = jiffies + (unsigned long)rnd %
2962 			     (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2963 
2964 	return elr;
2965 }
2966 
2967 int ext4_register_li_request(struct super_block *sb,
2968 			     ext4_group_t first_not_zeroed)
2969 {
2970 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2971 	struct ext4_li_request *elr = NULL;
2972 	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2973 	int ret = 0;
2974 
2975 	mutex_lock(&ext4_li_mtx);
2976 	if (sbi->s_li_request != NULL) {
2977 		/*
2978 		 * Reset timeout so it can be computed again, because
2979 		 * s_li_wait_mult might have changed.
2980 		 */
2981 		sbi->s_li_request->lr_timeout = 0;
2982 		goto out;
2983 	}
2984 
2985 	if (first_not_zeroed == ngroups ||
2986 	    (sb->s_flags & MS_RDONLY) ||
2987 	    !test_opt(sb, INIT_INODE_TABLE))
2988 		goto out;
2989 
2990 	elr = ext4_li_request_new(sb, first_not_zeroed);
2991 	if (!elr) {
2992 		ret = -ENOMEM;
2993 		goto out;
2994 	}
2995 
2996 	if (NULL == ext4_li_info) {
2997 		ret = ext4_li_info_new();
2998 		if (ret)
2999 			goto out;
3000 	}
3001 
3002 	mutex_lock(&ext4_li_info->li_list_mtx);
3003 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3004 	mutex_unlock(&ext4_li_info->li_list_mtx);
3005 
3006 	sbi->s_li_request = elr;
3007 	/*
3008 	 * set elr to NULL here since it has been inserted to
3009 	 * the request_list and the removal and free of it is
3010 	 * handled by ext4_clear_request_list from now on.
3011 	 */
3012 	elr = NULL;
3013 
3014 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3015 		ret = ext4_run_lazyinit_thread();
3016 		if (ret)
3017 			goto out;
3018 	}
3019 out:
3020 	mutex_unlock(&ext4_li_mtx);
3021 	if (ret)
3022 		kfree(elr);
3023 	return ret;
3024 }
3025 
3026 /*
3027  * We do not need to lock anything since this is called on
3028  * module unload.
3029  */
3030 static void ext4_destroy_lazyinit_thread(void)
3031 {
3032 	/*
3033 	 * If thread exited earlier
3034 	 * there's nothing to be done.
3035 	 */
3036 	if (!ext4_li_info || !ext4_lazyinit_task)
3037 		return;
3038 
3039 	kthread_stop(ext4_lazyinit_task);
3040 }
3041 
3042 static int set_journal_csum_feature_set(struct super_block *sb)
3043 {
3044 	int ret = 1;
3045 	int compat, incompat;
3046 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3047 
3048 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3049 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3050 		/* journal checksum v2 */
3051 		compat = 0;
3052 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3053 	} else {
3054 		/* journal checksum v1 */
3055 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3056 		incompat = 0;
3057 	}
3058 
3059 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3060 		ret = jbd2_journal_set_features(sbi->s_journal,
3061 				compat, 0,
3062 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3063 				incompat);
3064 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3065 		ret = jbd2_journal_set_features(sbi->s_journal,
3066 				compat, 0,
3067 				incompat);
3068 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3069 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3070 	} else {
3071 		jbd2_journal_clear_features(sbi->s_journal,
3072 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3073 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3074 				JBD2_FEATURE_INCOMPAT_CSUM_V2);
3075 	}
3076 
3077 	return ret;
3078 }
3079 
3080 /*
3081  * Note: calculating the overhead so we can be compatible with
3082  * historical BSD practice is quite difficult in the face of
3083  * clusters/bigalloc.  This is because multiple metadata blocks from
3084  * different block group can end up in the same allocation cluster.
3085  * Calculating the exact overhead in the face of clustered allocation
3086  * requires either O(all block bitmaps) in memory or O(number of block
3087  * groups**2) in time.  We will still calculate the superblock for
3088  * older file systems --- and if we come across with a bigalloc file
3089  * system with zero in s_overhead_clusters the estimate will be close to
3090  * correct especially for very large cluster sizes --- but for newer
3091  * file systems, it's better to calculate this figure once at mkfs
3092  * time, and store it in the superblock.  If the superblock value is
3093  * present (even for non-bigalloc file systems), we will use it.
3094  */
3095 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3096 			  char *buf)
3097 {
3098 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3099 	struct ext4_group_desc	*gdp;
3100 	ext4_fsblk_t		first_block, last_block, b;
3101 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3102 	int			s, j, count = 0;
3103 
3104 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3105 		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3106 			sbi->s_itb_per_group + 2);
3107 
3108 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3109 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3110 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3111 	for (i = 0; i < ngroups; i++) {
3112 		gdp = ext4_get_group_desc(sb, i, NULL);
3113 		b = ext4_block_bitmap(sb, gdp);
3114 		if (b >= first_block && b <= last_block) {
3115 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3116 			count++;
3117 		}
3118 		b = ext4_inode_bitmap(sb, gdp);
3119 		if (b >= first_block && b <= last_block) {
3120 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3121 			count++;
3122 		}
3123 		b = ext4_inode_table(sb, gdp);
3124 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3125 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3126 				int c = EXT4_B2C(sbi, b - first_block);
3127 				ext4_set_bit(c, buf);
3128 				count++;
3129 			}
3130 		if (i != grp)
3131 			continue;
3132 		s = 0;
3133 		if (ext4_bg_has_super(sb, grp)) {
3134 			ext4_set_bit(s++, buf);
3135 			count++;
3136 		}
3137 		for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3138 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3139 			count++;
3140 		}
3141 	}
3142 	if (!count)
3143 		return 0;
3144 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3145 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3146 }
3147 
3148 /*
3149  * Compute the overhead and stash it in sbi->s_overhead
3150  */
3151 int ext4_calculate_overhead(struct super_block *sb)
3152 {
3153 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3154 	struct ext4_super_block *es = sbi->s_es;
3155 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3156 	ext4_fsblk_t overhead = 0;
3157 	char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3158 
3159 	if (!buf)
3160 		return -ENOMEM;
3161 
3162 	/*
3163 	 * Compute the overhead (FS structures).  This is constant
3164 	 * for a given filesystem unless the number of block groups
3165 	 * changes so we cache the previous value until it does.
3166 	 */
3167 
3168 	/*
3169 	 * All of the blocks before first_data_block are overhead
3170 	 */
3171 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3172 
3173 	/*
3174 	 * Add the overhead found in each block group
3175 	 */
3176 	for (i = 0; i < ngroups; i++) {
3177 		int blks;
3178 
3179 		blks = count_overhead(sb, i, buf);
3180 		overhead += blks;
3181 		if (blks)
3182 			memset(buf, 0, PAGE_SIZE);
3183 		cond_resched();
3184 	}
3185 	/* Add the journal blocks as well */
3186 	if (sbi->s_journal)
3187 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3188 
3189 	sbi->s_overhead = overhead;
3190 	smp_wmb();
3191 	free_page((unsigned long) buf);
3192 	return 0;
3193 }
3194 
3195 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3196 {
3197 	char *orig_data = kstrdup(data, GFP_KERNEL);
3198 	struct buffer_head *bh;
3199 	struct ext4_super_block *es = NULL;
3200 	struct ext4_sb_info *sbi;
3201 	ext4_fsblk_t block;
3202 	ext4_fsblk_t sb_block = get_sb_block(&data);
3203 	ext4_fsblk_t logical_sb_block;
3204 	unsigned long offset = 0;
3205 	unsigned long journal_devnum = 0;
3206 	unsigned long def_mount_opts;
3207 	struct inode *root;
3208 	char *cp;
3209 	const char *descr;
3210 	int ret = -ENOMEM;
3211 	int blocksize, clustersize;
3212 	unsigned int db_count;
3213 	unsigned int i;
3214 	int needs_recovery, has_huge_files, has_bigalloc;
3215 	__u64 blocks_count;
3216 	int err = 0;
3217 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3218 	ext4_group_t first_not_zeroed;
3219 
3220 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3221 	if (!sbi)
3222 		goto out_free_orig;
3223 
3224 	sbi->s_blockgroup_lock =
3225 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3226 	if (!sbi->s_blockgroup_lock) {
3227 		kfree(sbi);
3228 		goto out_free_orig;
3229 	}
3230 	sb->s_fs_info = sbi;
3231 	sbi->s_sb = sb;
3232 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3233 	sbi->s_sb_block = sb_block;
3234 	if (sb->s_bdev->bd_part)
3235 		sbi->s_sectors_written_start =
3236 			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3237 
3238 	/* Cleanup superblock name */
3239 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3240 		*cp = '!';
3241 
3242 	/* -EINVAL is default */
3243 	ret = -EINVAL;
3244 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3245 	if (!blocksize) {
3246 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3247 		goto out_fail;
3248 	}
3249 
3250 	/*
3251 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3252 	 * block sizes.  We need to calculate the offset from buffer start.
3253 	 */
3254 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3255 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3256 		offset = do_div(logical_sb_block, blocksize);
3257 	} else {
3258 		logical_sb_block = sb_block;
3259 	}
3260 
3261 	if (!(bh = sb_bread(sb, logical_sb_block))) {
3262 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3263 		goto out_fail;
3264 	}
3265 	/*
3266 	 * Note: s_es must be initialized as soon as possible because
3267 	 *       some ext4 macro-instructions depend on its value
3268 	 */
3269 	es = (struct ext4_super_block *) (bh->b_data + offset);
3270 	sbi->s_es = es;
3271 	sb->s_magic = le16_to_cpu(es->s_magic);
3272 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3273 		goto cantfind_ext4;
3274 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3275 
3276 	/* Warn if metadata_csum and gdt_csum are both set. */
3277 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3278 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3279 	    EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3280 		ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3281 			     "redundant flags; please run fsck.");
3282 
3283 	/* Check for a known checksum algorithm */
3284 	if (!ext4_verify_csum_type(sb, es)) {
3285 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3286 			 "unknown checksum algorithm.");
3287 		silent = 1;
3288 		goto cantfind_ext4;
3289 	}
3290 
3291 	/* Load the checksum driver */
3292 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3293 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3294 		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3295 		if (IS_ERR(sbi->s_chksum_driver)) {
3296 			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3297 			ret = PTR_ERR(sbi->s_chksum_driver);
3298 			sbi->s_chksum_driver = NULL;
3299 			goto failed_mount;
3300 		}
3301 	}
3302 
3303 	/* Check superblock checksum */
3304 	if (!ext4_superblock_csum_verify(sb, es)) {
3305 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3306 			 "invalid superblock checksum.  Run e2fsck?");
3307 		silent = 1;
3308 		goto cantfind_ext4;
3309 	}
3310 
3311 	/* Precompute checksum seed for all metadata */
3312 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3313 			EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3314 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3315 					       sizeof(es->s_uuid));
3316 
3317 	/* Set defaults before we parse the mount options */
3318 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3319 	set_opt(sb, INIT_INODE_TABLE);
3320 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3321 		set_opt(sb, DEBUG);
3322 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3323 		set_opt(sb, GRPID);
3324 	if (def_mount_opts & EXT4_DEFM_UID16)
3325 		set_opt(sb, NO_UID32);
3326 	/* xattr user namespace & acls are now defaulted on */
3327 	set_opt(sb, XATTR_USER);
3328 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3329 	set_opt(sb, POSIX_ACL);
3330 #endif
3331 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3332 		set_opt(sb, JOURNAL_DATA);
3333 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3334 		set_opt(sb, ORDERED_DATA);
3335 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3336 		set_opt(sb, WRITEBACK_DATA);
3337 
3338 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3339 		set_opt(sb, ERRORS_PANIC);
3340 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3341 		set_opt(sb, ERRORS_CONT);
3342 	else
3343 		set_opt(sb, ERRORS_RO);
3344 	if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3345 		set_opt(sb, BLOCK_VALIDITY);
3346 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3347 		set_opt(sb, DISCARD);
3348 
3349 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3350 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3351 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3352 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3353 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3354 
3355 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3356 		set_opt(sb, BARRIER);
3357 
3358 	/*
3359 	 * enable delayed allocation by default
3360 	 * Use -o nodelalloc to turn it off
3361 	 */
3362 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3363 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3364 		set_opt(sb, DELALLOC);
3365 
3366 	/*
3367 	 * set default s_li_wait_mult for lazyinit, for the case there is
3368 	 * no mount option specified.
3369 	 */
3370 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3371 
3372 	if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3373 			   &journal_devnum, &journal_ioprio, 0)) {
3374 		ext4_msg(sb, KERN_WARNING,
3375 			 "failed to parse options in superblock: %s",
3376 			 sbi->s_es->s_mount_opts);
3377 	}
3378 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3379 	if (!parse_options((char *) data, sb, &journal_devnum,
3380 			   &journal_ioprio, 0))
3381 		goto failed_mount;
3382 
3383 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3384 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3385 			    "with data=journal disables delayed "
3386 			    "allocation and O_DIRECT support!\n");
3387 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3388 			ext4_msg(sb, KERN_ERR, "can't mount with "
3389 				 "both data=journal and delalloc");
3390 			goto failed_mount;
3391 		}
3392 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3393 			ext4_msg(sb, KERN_ERR, "can't mount with "
3394 				 "both data=journal and delalloc");
3395 			goto failed_mount;
3396 		}
3397 		if (test_opt(sb, DELALLOC))
3398 			clear_opt(sb, DELALLOC);
3399 	}
3400 
3401 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3402 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3403 
3404 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3405 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3406 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3407 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3408 		ext4_msg(sb, KERN_WARNING,
3409 		       "feature flags set on rev 0 fs, "
3410 		       "running e2fsck is recommended");
3411 
3412 	if (IS_EXT2_SB(sb)) {
3413 		if (ext2_feature_set_ok(sb))
3414 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3415 				 "using the ext4 subsystem");
3416 		else {
3417 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3418 				 "to feature incompatibilities");
3419 			goto failed_mount;
3420 		}
3421 	}
3422 
3423 	if (IS_EXT3_SB(sb)) {
3424 		if (ext3_feature_set_ok(sb))
3425 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3426 				 "using the ext4 subsystem");
3427 		else {
3428 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3429 				 "to feature incompatibilities");
3430 			goto failed_mount;
3431 		}
3432 	}
3433 
3434 	/*
3435 	 * Check feature flags regardless of the revision level, since we
3436 	 * previously didn't change the revision level when setting the flags,
3437 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3438 	 */
3439 	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3440 		goto failed_mount;
3441 
3442 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3443 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3444 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3445 		ext4_msg(sb, KERN_ERR,
3446 		       "Unsupported filesystem blocksize %d", blocksize);
3447 		goto failed_mount;
3448 	}
3449 
3450 	if (sb->s_blocksize != blocksize) {
3451 		/* Validate the filesystem blocksize */
3452 		if (!sb_set_blocksize(sb, blocksize)) {
3453 			ext4_msg(sb, KERN_ERR, "bad block size %d",
3454 					blocksize);
3455 			goto failed_mount;
3456 		}
3457 
3458 		brelse(bh);
3459 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3460 		offset = do_div(logical_sb_block, blocksize);
3461 		bh = sb_bread(sb, logical_sb_block);
3462 		if (!bh) {
3463 			ext4_msg(sb, KERN_ERR,
3464 			       "Can't read superblock on 2nd try");
3465 			goto failed_mount;
3466 		}
3467 		es = (struct ext4_super_block *)(bh->b_data + offset);
3468 		sbi->s_es = es;
3469 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3470 			ext4_msg(sb, KERN_ERR,
3471 			       "Magic mismatch, very weird!");
3472 			goto failed_mount;
3473 		}
3474 	}
3475 
3476 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3477 				EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3478 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3479 						      has_huge_files);
3480 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3481 
3482 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3483 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3484 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3485 	} else {
3486 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3487 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3488 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3489 		    (!is_power_of_2(sbi->s_inode_size)) ||
3490 		    (sbi->s_inode_size > blocksize)) {
3491 			ext4_msg(sb, KERN_ERR,
3492 			       "unsupported inode size: %d",
3493 			       sbi->s_inode_size);
3494 			goto failed_mount;
3495 		}
3496 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3497 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3498 	}
3499 
3500 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3501 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3502 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3503 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3504 		    !is_power_of_2(sbi->s_desc_size)) {
3505 			ext4_msg(sb, KERN_ERR,
3506 			       "unsupported descriptor size %lu",
3507 			       sbi->s_desc_size);
3508 			goto failed_mount;
3509 		}
3510 	} else
3511 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3512 
3513 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3514 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3515 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3516 		goto cantfind_ext4;
3517 
3518 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3519 	if (sbi->s_inodes_per_block == 0)
3520 		goto cantfind_ext4;
3521 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3522 					sbi->s_inodes_per_block;
3523 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3524 	sbi->s_sbh = bh;
3525 	sbi->s_mount_state = le16_to_cpu(es->s_state);
3526 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3527 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3528 
3529 	for (i = 0; i < 4; i++)
3530 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3531 	sbi->s_def_hash_version = es->s_def_hash_version;
3532 	i = le32_to_cpu(es->s_flags);
3533 	if (i & EXT2_FLAGS_UNSIGNED_HASH)
3534 		sbi->s_hash_unsigned = 3;
3535 	else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3536 #ifdef __CHAR_UNSIGNED__
3537 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3538 		sbi->s_hash_unsigned = 3;
3539 #else
3540 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3541 #endif
3542 	}
3543 
3544 	/* Handle clustersize */
3545 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3546 	has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3547 				EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3548 	if (has_bigalloc) {
3549 		if (clustersize < blocksize) {
3550 			ext4_msg(sb, KERN_ERR,
3551 				 "cluster size (%d) smaller than "
3552 				 "block size (%d)", clustersize, blocksize);
3553 			goto failed_mount;
3554 		}
3555 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3556 			le32_to_cpu(es->s_log_block_size);
3557 		sbi->s_clusters_per_group =
3558 			le32_to_cpu(es->s_clusters_per_group);
3559 		if (sbi->s_clusters_per_group > blocksize * 8) {
3560 			ext4_msg(sb, KERN_ERR,
3561 				 "#clusters per group too big: %lu",
3562 				 sbi->s_clusters_per_group);
3563 			goto failed_mount;
3564 		}
3565 		if (sbi->s_blocks_per_group !=
3566 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3567 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3568 				 "clusters per group (%lu) inconsistent",
3569 				 sbi->s_blocks_per_group,
3570 				 sbi->s_clusters_per_group);
3571 			goto failed_mount;
3572 		}
3573 	} else {
3574 		if (clustersize != blocksize) {
3575 			ext4_warning(sb, "fragment/cluster size (%d) != "
3576 				     "block size (%d)", clustersize,
3577 				     blocksize);
3578 			clustersize = blocksize;
3579 		}
3580 		if (sbi->s_blocks_per_group > blocksize * 8) {
3581 			ext4_msg(sb, KERN_ERR,
3582 				 "#blocks per group too big: %lu",
3583 				 sbi->s_blocks_per_group);
3584 			goto failed_mount;
3585 		}
3586 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3587 		sbi->s_cluster_bits = 0;
3588 	}
3589 	sbi->s_cluster_ratio = clustersize / blocksize;
3590 
3591 	if (sbi->s_inodes_per_group > blocksize * 8) {
3592 		ext4_msg(sb, KERN_ERR,
3593 		       "#inodes per group too big: %lu",
3594 		       sbi->s_inodes_per_group);
3595 		goto failed_mount;
3596 	}
3597 
3598 	/*
3599 	 * Test whether we have more sectors than will fit in sector_t,
3600 	 * and whether the max offset is addressable by the page cache.
3601 	 */
3602 	err = generic_check_addressable(sb->s_blocksize_bits,
3603 					ext4_blocks_count(es));
3604 	if (err) {
3605 		ext4_msg(sb, KERN_ERR, "filesystem"
3606 			 " too large to mount safely on this system");
3607 		if (sizeof(sector_t) < 8)
3608 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3609 		goto failed_mount;
3610 	}
3611 
3612 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3613 		goto cantfind_ext4;
3614 
3615 	/* check blocks count against device size */
3616 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3617 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3618 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3619 		       "exceeds size of device (%llu blocks)",
3620 		       ext4_blocks_count(es), blocks_count);
3621 		goto failed_mount;
3622 	}
3623 
3624 	/*
3625 	 * It makes no sense for the first data block to be beyond the end
3626 	 * of the filesystem.
3627 	 */
3628 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3629 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3630 			 "block %u is beyond end of filesystem (%llu)",
3631 			 le32_to_cpu(es->s_first_data_block),
3632 			 ext4_blocks_count(es));
3633 		goto failed_mount;
3634 	}
3635 	blocks_count = (ext4_blocks_count(es) -
3636 			le32_to_cpu(es->s_first_data_block) +
3637 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3638 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3639 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3640 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3641 		       "(block count %llu, first data block %u, "
3642 		       "blocks per group %lu)", sbi->s_groups_count,
3643 		       ext4_blocks_count(es),
3644 		       le32_to_cpu(es->s_first_data_block),
3645 		       EXT4_BLOCKS_PER_GROUP(sb));
3646 		goto failed_mount;
3647 	}
3648 	sbi->s_groups_count = blocks_count;
3649 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3650 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3651 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3652 		   EXT4_DESC_PER_BLOCK(sb);
3653 	sbi->s_group_desc = ext4_kvmalloc(db_count *
3654 					  sizeof(struct buffer_head *),
3655 					  GFP_KERNEL);
3656 	if (sbi->s_group_desc == NULL) {
3657 		ext4_msg(sb, KERN_ERR, "not enough memory");
3658 		ret = -ENOMEM;
3659 		goto failed_mount;
3660 	}
3661 
3662 	if (ext4_proc_root)
3663 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3664 
3665 	if (sbi->s_proc)
3666 		proc_create_data("options", S_IRUGO, sbi->s_proc,
3667 				 &ext4_seq_options_fops, sb);
3668 
3669 	bgl_lock_init(sbi->s_blockgroup_lock);
3670 
3671 	for (i = 0; i < db_count; i++) {
3672 		block = descriptor_loc(sb, logical_sb_block, i);
3673 		sbi->s_group_desc[i] = sb_bread(sb, block);
3674 		if (!sbi->s_group_desc[i]) {
3675 			ext4_msg(sb, KERN_ERR,
3676 			       "can't read group descriptor %d", i);
3677 			db_count = i;
3678 			goto failed_mount2;
3679 		}
3680 	}
3681 	if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3682 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3683 		goto failed_mount2;
3684 	}
3685 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3686 		if (!ext4_fill_flex_info(sb)) {
3687 			ext4_msg(sb, KERN_ERR,
3688 			       "unable to initialize "
3689 			       "flex_bg meta info!");
3690 			goto failed_mount2;
3691 		}
3692 
3693 	sbi->s_gdb_count = db_count;
3694 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3695 	spin_lock_init(&sbi->s_next_gen_lock);
3696 
3697 	init_timer(&sbi->s_err_report);
3698 	sbi->s_err_report.function = print_daily_error_info;
3699 	sbi->s_err_report.data = (unsigned long) sb;
3700 
3701 	err = percpu_counter_init(&sbi->s_freeclusters_counter,
3702 			ext4_count_free_clusters(sb));
3703 	if (!err) {
3704 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
3705 				ext4_count_free_inodes(sb));
3706 	}
3707 	if (!err) {
3708 		err = percpu_counter_init(&sbi->s_dirs_counter,
3709 				ext4_count_dirs(sb));
3710 	}
3711 	if (!err) {
3712 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3713 	}
3714 	if (!err) {
3715 		err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3716 	}
3717 	if (err) {
3718 		ext4_msg(sb, KERN_ERR, "insufficient memory");
3719 		goto failed_mount3;
3720 	}
3721 
3722 	sbi->s_stripe = ext4_get_stripe_size(sbi);
3723 	sbi->s_max_writeback_mb_bump = 128;
3724 	sbi->s_extent_max_zeroout_kb = 32;
3725 
3726 	/* Register extent status tree shrinker */
3727 	ext4_es_register_shrinker(sb);
3728 
3729 	/*
3730 	 * set up enough so that it can read an inode
3731 	 */
3732 	if (!test_opt(sb, NOLOAD) &&
3733 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3734 		sb->s_op = &ext4_sops;
3735 	else
3736 		sb->s_op = &ext4_nojournal_sops;
3737 	sb->s_export_op = &ext4_export_ops;
3738 	sb->s_xattr = ext4_xattr_handlers;
3739 #ifdef CONFIG_QUOTA
3740 	sb->dq_op = &ext4_quota_operations;
3741 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3742 		sb->s_qcop = &ext4_qctl_sysfile_operations;
3743 	else
3744 		sb->s_qcop = &ext4_qctl_operations;
3745 #endif
3746 	memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3747 
3748 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3749 	mutex_init(&sbi->s_orphan_lock);
3750 
3751 	sb->s_root = NULL;
3752 
3753 	needs_recovery = (es->s_last_orphan != 0 ||
3754 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
3755 				    EXT4_FEATURE_INCOMPAT_RECOVER));
3756 
3757 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3758 	    !(sb->s_flags & MS_RDONLY))
3759 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3760 			goto failed_mount3;
3761 
3762 	/*
3763 	 * The first inode we look at is the journal inode.  Don't try
3764 	 * root first: it may be modified in the journal!
3765 	 */
3766 	if (!test_opt(sb, NOLOAD) &&
3767 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3768 		if (ext4_load_journal(sb, es, journal_devnum))
3769 			goto failed_mount3;
3770 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3771 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3772 		ext4_msg(sb, KERN_ERR, "required journal recovery "
3773 		       "suppressed and not mounted read-only");
3774 		goto failed_mount_wq;
3775 	} else {
3776 		clear_opt(sb, DATA_FLAGS);
3777 		sbi->s_journal = NULL;
3778 		needs_recovery = 0;
3779 		goto no_journal;
3780 	}
3781 
3782 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3783 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3784 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
3785 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3786 		goto failed_mount_wq;
3787 	}
3788 
3789 	if (!set_journal_csum_feature_set(sb)) {
3790 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3791 			 "feature set");
3792 		goto failed_mount_wq;
3793 	}
3794 
3795 	/* We have now updated the journal if required, so we can
3796 	 * validate the data journaling mode. */
3797 	switch (test_opt(sb, DATA_FLAGS)) {
3798 	case 0:
3799 		/* No mode set, assume a default based on the journal
3800 		 * capabilities: ORDERED_DATA if the journal can
3801 		 * cope, else JOURNAL_DATA
3802 		 */
3803 		if (jbd2_journal_check_available_features
3804 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3805 			set_opt(sb, ORDERED_DATA);
3806 		else
3807 			set_opt(sb, JOURNAL_DATA);
3808 		break;
3809 
3810 	case EXT4_MOUNT_ORDERED_DATA:
3811 	case EXT4_MOUNT_WRITEBACK_DATA:
3812 		if (!jbd2_journal_check_available_features
3813 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3814 			ext4_msg(sb, KERN_ERR, "Journal does not support "
3815 			       "requested data journaling mode");
3816 			goto failed_mount_wq;
3817 		}
3818 	default:
3819 		break;
3820 	}
3821 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3822 
3823 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3824 
3825 	/*
3826 	 * The journal may have updated the bg summary counts, so we
3827 	 * need to update the global counters.
3828 	 */
3829 	percpu_counter_set(&sbi->s_freeclusters_counter,
3830 			   ext4_count_free_clusters(sb));
3831 	percpu_counter_set(&sbi->s_freeinodes_counter,
3832 			   ext4_count_free_inodes(sb));
3833 	percpu_counter_set(&sbi->s_dirs_counter,
3834 			   ext4_count_dirs(sb));
3835 	percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3836 
3837 no_journal:
3838 	/*
3839 	 * Get the # of file system overhead blocks from the
3840 	 * superblock if present.
3841 	 */
3842 	if (es->s_overhead_clusters)
3843 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3844 	else {
3845 		err = ext4_calculate_overhead(sb);
3846 		if (err)
3847 			goto failed_mount_wq;
3848 	}
3849 
3850 	/*
3851 	 * The maximum number of concurrent works can be high and
3852 	 * concurrency isn't really necessary.  Limit it to 1.
3853 	 */
3854 	EXT4_SB(sb)->dio_unwritten_wq =
3855 		alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3856 	if (!EXT4_SB(sb)->dio_unwritten_wq) {
3857 		printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3858 		ret = -ENOMEM;
3859 		goto failed_mount_wq;
3860 	}
3861 
3862 	/*
3863 	 * The jbd2_journal_load will have done any necessary log recovery,
3864 	 * so we can safely mount the rest of the filesystem now.
3865 	 */
3866 
3867 	root = ext4_iget(sb, EXT4_ROOT_INO);
3868 	if (IS_ERR(root)) {
3869 		ext4_msg(sb, KERN_ERR, "get root inode failed");
3870 		ret = PTR_ERR(root);
3871 		root = NULL;
3872 		goto failed_mount4;
3873 	}
3874 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3875 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3876 		iput(root);
3877 		goto failed_mount4;
3878 	}
3879 	sb->s_root = d_make_root(root);
3880 	if (!sb->s_root) {
3881 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
3882 		ret = -ENOMEM;
3883 		goto failed_mount4;
3884 	}
3885 
3886 	if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3887 		sb->s_flags |= MS_RDONLY;
3888 
3889 	/* determine the minimum size of new large inodes, if present */
3890 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3891 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3892 						     EXT4_GOOD_OLD_INODE_SIZE;
3893 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3894 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3895 			if (sbi->s_want_extra_isize <
3896 			    le16_to_cpu(es->s_want_extra_isize))
3897 				sbi->s_want_extra_isize =
3898 					le16_to_cpu(es->s_want_extra_isize);
3899 			if (sbi->s_want_extra_isize <
3900 			    le16_to_cpu(es->s_min_extra_isize))
3901 				sbi->s_want_extra_isize =
3902 					le16_to_cpu(es->s_min_extra_isize);
3903 		}
3904 	}
3905 	/* Check if enough inode space is available */
3906 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3907 							sbi->s_inode_size) {
3908 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3909 						       EXT4_GOOD_OLD_INODE_SIZE;
3910 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
3911 			 "available");
3912 	}
3913 
3914 	err = ext4_setup_system_zone(sb);
3915 	if (err) {
3916 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
3917 			 "zone (%d)", err);
3918 		goto failed_mount4a;
3919 	}
3920 
3921 	ext4_ext_init(sb);
3922 	err = ext4_mb_init(sb);
3923 	if (err) {
3924 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3925 			 err);
3926 		goto failed_mount5;
3927 	}
3928 
3929 	err = ext4_register_li_request(sb, first_not_zeroed);
3930 	if (err)
3931 		goto failed_mount6;
3932 
3933 	sbi->s_kobj.kset = ext4_kset;
3934 	init_completion(&sbi->s_kobj_unregister);
3935 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3936 				   "%s", sb->s_id);
3937 	if (err)
3938 		goto failed_mount7;
3939 
3940 #ifdef CONFIG_QUOTA
3941 	/* Enable quota usage during mount. */
3942 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3943 	    !(sb->s_flags & MS_RDONLY)) {
3944 		err = ext4_enable_quotas(sb);
3945 		if (err)
3946 			goto failed_mount8;
3947 	}
3948 #endif  /* CONFIG_QUOTA */
3949 
3950 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3951 	ext4_orphan_cleanup(sb, es);
3952 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3953 	if (needs_recovery) {
3954 		ext4_msg(sb, KERN_INFO, "recovery complete");
3955 		ext4_mark_recovery_complete(sb, es);
3956 	}
3957 	if (EXT4_SB(sb)->s_journal) {
3958 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3959 			descr = " journalled data mode";
3960 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3961 			descr = " ordered data mode";
3962 		else
3963 			descr = " writeback data mode";
3964 	} else
3965 		descr = "out journal";
3966 
3967 	if (test_opt(sb, DISCARD)) {
3968 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
3969 		if (!blk_queue_discard(q))
3970 			ext4_msg(sb, KERN_WARNING,
3971 				 "mounting with \"discard\" option, but "
3972 				 "the device does not support discard");
3973 	}
3974 
3975 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3976 		 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3977 		 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3978 
3979 	if (es->s_error_count)
3980 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3981 
3982 	kfree(orig_data);
3983 	return 0;
3984 
3985 cantfind_ext4:
3986 	if (!silent)
3987 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3988 	goto failed_mount;
3989 
3990 #ifdef CONFIG_QUOTA
3991 failed_mount8:
3992 	kobject_del(&sbi->s_kobj);
3993 #endif
3994 failed_mount7:
3995 	ext4_unregister_li_request(sb);
3996 failed_mount6:
3997 	ext4_mb_release(sb);
3998 failed_mount5:
3999 	ext4_ext_release(sb);
4000 	ext4_release_system_zone(sb);
4001 failed_mount4a:
4002 	dput(sb->s_root);
4003 	sb->s_root = NULL;
4004 failed_mount4:
4005 	ext4_msg(sb, KERN_ERR, "mount failed");
4006 	destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4007 failed_mount_wq:
4008 	if (sbi->s_journal) {
4009 		jbd2_journal_destroy(sbi->s_journal);
4010 		sbi->s_journal = NULL;
4011 	}
4012 failed_mount3:
4013 	del_timer(&sbi->s_err_report);
4014 	if (sbi->s_flex_groups)
4015 		ext4_kvfree(sbi->s_flex_groups);
4016 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
4017 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
4018 	percpu_counter_destroy(&sbi->s_dirs_counter);
4019 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4020 	percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4021 	if (sbi->s_mmp_tsk)
4022 		kthread_stop(sbi->s_mmp_tsk);
4023 failed_mount2:
4024 	for (i = 0; i < db_count; i++)
4025 		brelse(sbi->s_group_desc[i]);
4026 	ext4_kvfree(sbi->s_group_desc);
4027 failed_mount:
4028 	if (sbi->s_chksum_driver)
4029 		crypto_free_shash(sbi->s_chksum_driver);
4030 	if (sbi->s_proc) {
4031 		remove_proc_entry("options", sbi->s_proc);
4032 		remove_proc_entry(sb->s_id, ext4_proc_root);
4033 	}
4034 #ifdef CONFIG_QUOTA
4035 	for (i = 0; i < MAXQUOTAS; i++)
4036 		kfree(sbi->s_qf_names[i]);
4037 #endif
4038 	ext4_blkdev_remove(sbi);
4039 	brelse(bh);
4040 out_fail:
4041 	sb->s_fs_info = NULL;
4042 	kfree(sbi->s_blockgroup_lock);
4043 	kfree(sbi);
4044 out_free_orig:
4045 	kfree(orig_data);
4046 	return err ? err : ret;
4047 }
4048 
4049 /*
4050  * Setup any per-fs journal parameters now.  We'll do this both on
4051  * initial mount, once the journal has been initialised but before we've
4052  * done any recovery; and again on any subsequent remount.
4053  */
4054 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4055 {
4056 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4057 
4058 	journal->j_commit_interval = sbi->s_commit_interval;
4059 	journal->j_min_batch_time = sbi->s_min_batch_time;
4060 	journal->j_max_batch_time = sbi->s_max_batch_time;
4061 
4062 	write_lock(&journal->j_state_lock);
4063 	if (test_opt(sb, BARRIER))
4064 		journal->j_flags |= JBD2_BARRIER;
4065 	else
4066 		journal->j_flags &= ~JBD2_BARRIER;
4067 	if (test_opt(sb, DATA_ERR_ABORT))
4068 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4069 	else
4070 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4071 	write_unlock(&journal->j_state_lock);
4072 }
4073 
4074 static journal_t *ext4_get_journal(struct super_block *sb,
4075 				   unsigned int journal_inum)
4076 {
4077 	struct inode *journal_inode;
4078 	journal_t *journal;
4079 
4080 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4081 
4082 	/* First, test for the existence of a valid inode on disk.  Bad
4083 	 * things happen if we iget() an unused inode, as the subsequent
4084 	 * iput() will try to delete it. */
4085 
4086 	journal_inode = ext4_iget(sb, journal_inum);
4087 	if (IS_ERR(journal_inode)) {
4088 		ext4_msg(sb, KERN_ERR, "no journal found");
4089 		return NULL;
4090 	}
4091 	if (!journal_inode->i_nlink) {
4092 		make_bad_inode(journal_inode);
4093 		iput(journal_inode);
4094 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4095 		return NULL;
4096 	}
4097 
4098 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4099 		  journal_inode, journal_inode->i_size);
4100 	if (!S_ISREG(journal_inode->i_mode)) {
4101 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4102 		iput(journal_inode);
4103 		return NULL;
4104 	}
4105 
4106 	journal = jbd2_journal_init_inode(journal_inode);
4107 	if (!journal) {
4108 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4109 		iput(journal_inode);
4110 		return NULL;
4111 	}
4112 	journal->j_private = sb;
4113 	ext4_init_journal_params(sb, journal);
4114 	return journal;
4115 }
4116 
4117 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4118 				       dev_t j_dev)
4119 {
4120 	struct buffer_head *bh;
4121 	journal_t *journal;
4122 	ext4_fsblk_t start;
4123 	ext4_fsblk_t len;
4124 	int hblock, blocksize;
4125 	ext4_fsblk_t sb_block;
4126 	unsigned long offset;
4127 	struct ext4_super_block *es;
4128 	struct block_device *bdev;
4129 
4130 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4131 
4132 	bdev = ext4_blkdev_get(j_dev, sb);
4133 	if (bdev == NULL)
4134 		return NULL;
4135 
4136 	blocksize = sb->s_blocksize;
4137 	hblock = bdev_logical_block_size(bdev);
4138 	if (blocksize < hblock) {
4139 		ext4_msg(sb, KERN_ERR,
4140 			"blocksize too small for journal device");
4141 		goto out_bdev;
4142 	}
4143 
4144 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4145 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4146 	set_blocksize(bdev, blocksize);
4147 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4148 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4149 		       "external journal");
4150 		goto out_bdev;
4151 	}
4152 
4153 	es = (struct ext4_super_block *) (bh->b_data + offset);
4154 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4155 	    !(le32_to_cpu(es->s_feature_incompat) &
4156 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4157 		ext4_msg(sb, KERN_ERR, "external journal has "
4158 					"bad superblock");
4159 		brelse(bh);
4160 		goto out_bdev;
4161 	}
4162 
4163 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4164 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4165 		brelse(bh);
4166 		goto out_bdev;
4167 	}
4168 
4169 	len = ext4_blocks_count(es);
4170 	start = sb_block + 1;
4171 	brelse(bh);	/* we're done with the superblock */
4172 
4173 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4174 					start, len, blocksize);
4175 	if (!journal) {
4176 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4177 		goto out_bdev;
4178 	}
4179 	journal->j_private = sb;
4180 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
4181 	wait_on_buffer(journal->j_sb_buffer);
4182 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4183 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4184 		goto out_journal;
4185 	}
4186 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4187 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4188 					"user (unsupported) - %d",
4189 			be32_to_cpu(journal->j_superblock->s_nr_users));
4190 		goto out_journal;
4191 	}
4192 	EXT4_SB(sb)->journal_bdev = bdev;
4193 	ext4_init_journal_params(sb, journal);
4194 	return journal;
4195 
4196 out_journal:
4197 	jbd2_journal_destroy(journal);
4198 out_bdev:
4199 	ext4_blkdev_put(bdev);
4200 	return NULL;
4201 }
4202 
4203 static int ext4_load_journal(struct super_block *sb,
4204 			     struct ext4_super_block *es,
4205 			     unsigned long journal_devnum)
4206 {
4207 	journal_t *journal;
4208 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4209 	dev_t journal_dev;
4210 	int err = 0;
4211 	int really_read_only;
4212 
4213 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4214 
4215 	if (journal_devnum &&
4216 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4217 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4218 			"numbers have changed");
4219 		journal_dev = new_decode_dev(journal_devnum);
4220 	} else
4221 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4222 
4223 	really_read_only = bdev_read_only(sb->s_bdev);
4224 
4225 	/*
4226 	 * Are we loading a blank journal or performing recovery after a
4227 	 * crash?  For recovery, we need to check in advance whether we
4228 	 * can get read-write access to the device.
4229 	 */
4230 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4231 		if (sb->s_flags & MS_RDONLY) {
4232 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4233 					"required on readonly filesystem");
4234 			if (really_read_only) {
4235 				ext4_msg(sb, KERN_ERR, "write access "
4236 					"unavailable, cannot proceed");
4237 				return -EROFS;
4238 			}
4239 			ext4_msg(sb, KERN_INFO, "write access will "
4240 			       "be enabled during recovery");
4241 		}
4242 	}
4243 
4244 	if (journal_inum && journal_dev) {
4245 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4246 		       "and inode journals!");
4247 		return -EINVAL;
4248 	}
4249 
4250 	if (journal_inum) {
4251 		if (!(journal = ext4_get_journal(sb, journal_inum)))
4252 			return -EINVAL;
4253 	} else {
4254 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4255 			return -EINVAL;
4256 	}
4257 
4258 	if (!(journal->j_flags & JBD2_BARRIER))
4259 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4260 
4261 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4262 		err = jbd2_journal_wipe(journal, !really_read_only);
4263 	if (!err) {
4264 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4265 		if (save)
4266 			memcpy(save, ((char *) es) +
4267 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4268 		err = jbd2_journal_load(journal);
4269 		if (save)
4270 			memcpy(((char *) es) + EXT4_S_ERR_START,
4271 			       save, EXT4_S_ERR_LEN);
4272 		kfree(save);
4273 	}
4274 
4275 	if (err) {
4276 		ext4_msg(sb, KERN_ERR, "error loading journal");
4277 		jbd2_journal_destroy(journal);
4278 		return err;
4279 	}
4280 
4281 	EXT4_SB(sb)->s_journal = journal;
4282 	ext4_clear_journal_err(sb, es);
4283 
4284 	if (!really_read_only && journal_devnum &&
4285 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4286 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4287 
4288 		/* Make sure we flush the recovery flag to disk. */
4289 		ext4_commit_super(sb, 1);
4290 	}
4291 
4292 	return 0;
4293 }
4294 
4295 static int ext4_commit_super(struct super_block *sb, int sync)
4296 {
4297 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4298 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4299 	int error = 0;
4300 
4301 	if (!sbh || block_device_ejected(sb))
4302 		return error;
4303 	if (buffer_write_io_error(sbh)) {
4304 		/*
4305 		 * Oh, dear.  A previous attempt to write the
4306 		 * superblock failed.  This could happen because the
4307 		 * USB device was yanked out.  Or it could happen to
4308 		 * be a transient write error and maybe the block will
4309 		 * be remapped.  Nothing we can do but to retry the
4310 		 * write and hope for the best.
4311 		 */
4312 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4313 		       "superblock detected");
4314 		clear_buffer_write_io_error(sbh);
4315 		set_buffer_uptodate(sbh);
4316 	}
4317 	/*
4318 	 * If the file system is mounted read-only, don't update the
4319 	 * superblock write time.  This avoids updating the superblock
4320 	 * write time when we are mounting the root file system
4321 	 * read/only but we need to replay the journal; at that point,
4322 	 * for people who are east of GMT and who make their clock
4323 	 * tick in localtime for Windows bug-for-bug compatibility,
4324 	 * the clock is set in the future, and this will cause e2fsck
4325 	 * to complain and force a full file system check.
4326 	 */
4327 	if (!(sb->s_flags & MS_RDONLY))
4328 		es->s_wtime = cpu_to_le32(get_seconds());
4329 	if (sb->s_bdev->bd_part)
4330 		es->s_kbytes_written =
4331 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4332 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4333 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4334 	else
4335 		es->s_kbytes_written =
4336 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4337 	ext4_free_blocks_count_set(es,
4338 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4339 				&EXT4_SB(sb)->s_freeclusters_counter)));
4340 	es->s_free_inodes_count =
4341 		cpu_to_le32(percpu_counter_sum_positive(
4342 				&EXT4_SB(sb)->s_freeinodes_counter));
4343 	BUFFER_TRACE(sbh, "marking dirty");
4344 	ext4_superblock_csum_set(sb);
4345 	mark_buffer_dirty(sbh);
4346 	if (sync) {
4347 		error = sync_dirty_buffer(sbh);
4348 		if (error)
4349 			return error;
4350 
4351 		error = buffer_write_io_error(sbh);
4352 		if (error) {
4353 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4354 			       "superblock");
4355 			clear_buffer_write_io_error(sbh);
4356 			set_buffer_uptodate(sbh);
4357 		}
4358 	}
4359 	return error;
4360 }
4361 
4362 /*
4363  * Have we just finished recovery?  If so, and if we are mounting (or
4364  * remounting) the filesystem readonly, then we will end up with a
4365  * consistent fs on disk.  Record that fact.
4366  */
4367 static void ext4_mark_recovery_complete(struct super_block *sb,
4368 					struct ext4_super_block *es)
4369 {
4370 	journal_t *journal = EXT4_SB(sb)->s_journal;
4371 
4372 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4373 		BUG_ON(journal != NULL);
4374 		return;
4375 	}
4376 	jbd2_journal_lock_updates(journal);
4377 	if (jbd2_journal_flush(journal) < 0)
4378 		goto out;
4379 
4380 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4381 	    sb->s_flags & MS_RDONLY) {
4382 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4383 		ext4_commit_super(sb, 1);
4384 	}
4385 
4386 out:
4387 	jbd2_journal_unlock_updates(journal);
4388 }
4389 
4390 /*
4391  * If we are mounting (or read-write remounting) a filesystem whose journal
4392  * has recorded an error from a previous lifetime, move that error to the
4393  * main filesystem now.
4394  */
4395 static void ext4_clear_journal_err(struct super_block *sb,
4396 				   struct ext4_super_block *es)
4397 {
4398 	journal_t *journal;
4399 	int j_errno;
4400 	const char *errstr;
4401 
4402 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4403 
4404 	journal = EXT4_SB(sb)->s_journal;
4405 
4406 	/*
4407 	 * Now check for any error status which may have been recorded in the
4408 	 * journal by a prior ext4_error() or ext4_abort()
4409 	 */
4410 
4411 	j_errno = jbd2_journal_errno(journal);
4412 	if (j_errno) {
4413 		char nbuf[16];
4414 
4415 		errstr = ext4_decode_error(sb, j_errno, nbuf);
4416 		ext4_warning(sb, "Filesystem error recorded "
4417 			     "from previous mount: %s", errstr);
4418 		ext4_warning(sb, "Marking fs in need of filesystem check.");
4419 
4420 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4421 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4422 		ext4_commit_super(sb, 1);
4423 
4424 		jbd2_journal_clear_err(journal);
4425 		jbd2_journal_update_sb_errno(journal);
4426 	}
4427 }
4428 
4429 /*
4430  * Force the running and committing transactions to commit,
4431  * and wait on the commit.
4432  */
4433 int ext4_force_commit(struct super_block *sb)
4434 {
4435 	journal_t *journal;
4436 
4437 	if (sb->s_flags & MS_RDONLY)
4438 		return 0;
4439 
4440 	journal = EXT4_SB(sb)->s_journal;
4441 	return ext4_journal_force_commit(journal);
4442 }
4443 
4444 static int ext4_sync_fs(struct super_block *sb, int wait)
4445 {
4446 	int ret = 0;
4447 	tid_t target;
4448 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4449 
4450 	trace_ext4_sync_fs(sb, wait);
4451 	flush_workqueue(sbi->dio_unwritten_wq);
4452 	/*
4453 	 * Writeback quota in non-journalled quota case - journalled quota has
4454 	 * no dirty dquots
4455 	 */
4456 	dquot_writeback_dquots(sb, -1);
4457 	if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4458 		if (wait)
4459 			jbd2_log_wait_commit(sbi->s_journal, target);
4460 	}
4461 	return ret;
4462 }
4463 
4464 /*
4465  * LVM calls this function before a (read-only) snapshot is created.  This
4466  * gives us a chance to flush the journal completely and mark the fs clean.
4467  *
4468  * Note that only this function cannot bring a filesystem to be in a clean
4469  * state independently. It relies on upper layer to stop all data & metadata
4470  * modifications.
4471  */
4472 static int ext4_freeze(struct super_block *sb)
4473 {
4474 	int error = 0;
4475 	journal_t *journal;
4476 
4477 	if (sb->s_flags & MS_RDONLY)
4478 		return 0;
4479 
4480 	journal = EXT4_SB(sb)->s_journal;
4481 
4482 	/* Now we set up the journal barrier. */
4483 	jbd2_journal_lock_updates(journal);
4484 
4485 	/*
4486 	 * Don't clear the needs_recovery flag if we failed to flush
4487 	 * the journal.
4488 	 */
4489 	error = jbd2_journal_flush(journal);
4490 	if (error < 0)
4491 		goto out;
4492 
4493 	/* Journal blocked and flushed, clear needs_recovery flag. */
4494 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4495 	error = ext4_commit_super(sb, 1);
4496 out:
4497 	/* we rely on upper layer to stop further updates */
4498 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4499 	return error;
4500 }
4501 
4502 /*
4503  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4504  * flag here, even though the filesystem is not technically dirty yet.
4505  */
4506 static int ext4_unfreeze(struct super_block *sb)
4507 {
4508 	if (sb->s_flags & MS_RDONLY)
4509 		return 0;
4510 
4511 	/* Reset the needs_recovery flag before the fs is unlocked. */
4512 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4513 	ext4_commit_super(sb, 1);
4514 	return 0;
4515 }
4516 
4517 /*
4518  * Structure to save mount options for ext4_remount's benefit
4519  */
4520 struct ext4_mount_options {
4521 	unsigned long s_mount_opt;
4522 	unsigned long s_mount_opt2;
4523 	kuid_t s_resuid;
4524 	kgid_t s_resgid;
4525 	unsigned long s_commit_interval;
4526 	u32 s_min_batch_time, s_max_batch_time;
4527 #ifdef CONFIG_QUOTA
4528 	int s_jquota_fmt;
4529 	char *s_qf_names[MAXQUOTAS];
4530 #endif
4531 };
4532 
4533 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4534 {
4535 	struct ext4_super_block *es;
4536 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4537 	unsigned long old_sb_flags;
4538 	struct ext4_mount_options old_opts;
4539 	int enable_quota = 0;
4540 	ext4_group_t g;
4541 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4542 	int err = 0;
4543 #ifdef CONFIG_QUOTA
4544 	int i, j;
4545 #endif
4546 	char *orig_data = kstrdup(data, GFP_KERNEL);
4547 
4548 	/* Store the original options */
4549 	old_sb_flags = sb->s_flags;
4550 	old_opts.s_mount_opt = sbi->s_mount_opt;
4551 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4552 	old_opts.s_resuid = sbi->s_resuid;
4553 	old_opts.s_resgid = sbi->s_resgid;
4554 	old_opts.s_commit_interval = sbi->s_commit_interval;
4555 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4556 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4557 #ifdef CONFIG_QUOTA
4558 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4559 	for (i = 0; i < MAXQUOTAS; i++)
4560 		if (sbi->s_qf_names[i]) {
4561 			old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4562 							 GFP_KERNEL);
4563 			if (!old_opts.s_qf_names[i]) {
4564 				for (j = 0; j < i; j++)
4565 					kfree(old_opts.s_qf_names[j]);
4566 				kfree(orig_data);
4567 				return -ENOMEM;
4568 			}
4569 		} else
4570 			old_opts.s_qf_names[i] = NULL;
4571 #endif
4572 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4573 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4574 
4575 	/*
4576 	 * Allow the "check" option to be passed as a remount option.
4577 	 */
4578 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4579 		err = -EINVAL;
4580 		goto restore_opts;
4581 	}
4582 
4583 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4584 		ext4_abort(sb, "Abort forced by user");
4585 
4586 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4587 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4588 
4589 	es = sbi->s_es;
4590 
4591 	if (sbi->s_journal) {
4592 		ext4_init_journal_params(sb, sbi->s_journal);
4593 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4594 	}
4595 
4596 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4597 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4598 			err = -EROFS;
4599 			goto restore_opts;
4600 		}
4601 
4602 		if (*flags & MS_RDONLY) {
4603 			err = dquot_suspend(sb, -1);
4604 			if (err < 0)
4605 				goto restore_opts;
4606 
4607 			/*
4608 			 * First of all, the unconditional stuff we have to do
4609 			 * to disable replay of the journal when we next remount
4610 			 */
4611 			sb->s_flags |= MS_RDONLY;
4612 
4613 			/*
4614 			 * OK, test if we are remounting a valid rw partition
4615 			 * readonly, and if so set the rdonly flag and then
4616 			 * mark the partition as valid again.
4617 			 */
4618 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4619 			    (sbi->s_mount_state & EXT4_VALID_FS))
4620 				es->s_state = cpu_to_le16(sbi->s_mount_state);
4621 
4622 			if (sbi->s_journal)
4623 				ext4_mark_recovery_complete(sb, es);
4624 		} else {
4625 			/* Make sure we can mount this feature set readwrite */
4626 			if (!ext4_feature_set_ok(sb, 0)) {
4627 				err = -EROFS;
4628 				goto restore_opts;
4629 			}
4630 			/*
4631 			 * Make sure the group descriptor checksums
4632 			 * are sane.  If they aren't, refuse to remount r/w.
4633 			 */
4634 			for (g = 0; g < sbi->s_groups_count; g++) {
4635 				struct ext4_group_desc *gdp =
4636 					ext4_get_group_desc(sb, g, NULL);
4637 
4638 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4639 					ext4_msg(sb, KERN_ERR,
4640 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
4641 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4642 					       le16_to_cpu(gdp->bg_checksum));
4643 					err = -EINVAL;
4644 					goto restore_opts;
4645 				}
4646 			}
4647 
4648 			/*
4649 			 * If we have an unprocessed orphan list hanging
4650 			 * around from a previously readonly bdev mount,
4651 			 * require a full umount/remount for now.
4652 			 */
4653 			if (es->s_last_orphan) {
4654 				ext4_msg(sb, KERN_WARNING, "Couldn't "
4655 				       "remount RDWR because of unprocessed "
4656 				       "orphan inode list.  Please "
4657 				       "umount/remount instead");
4658 				err = -EINVAL;
4659 				goto restore_opts;
4660 			}
4661 
4662 			/*
4663 			 * Mounting a RDONLY partition read-write, so reread
4664 			 * and store the current valid flag.  (It may have
4665 			 * been changed by e2fsck since we originally mounted
4666 			 * the partition.)
4667 			 */
4668 			if (sbi->s_journal)
4669 				ext4_clear_journal_err(sb, es);
4670 			sbi->s_mount_state = le16_to_cpu(es->s_state);
4671 			if (!ext4_setup_super(sb, es, 0))
4672 				sb->s_flags &= ~MS_RDONLY;
4673 			if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4674 						     EXT4_FEATURE_INCOMPAT_MMP))
4675 				if (ext4_multi_mount_protect(sb,
4676 						le64_to_cpu(es->s_mmp_block))) {
4677 					err = -EROFS;
4678 					goto restore_opts;
4679 				}
4680 			enable_quota = 1;
4681 		}
4682 	}
4683 
4684 	/*
4685 	 * Reinitialize lazy itable initialization thread based on
4686 	 * current settings
4687 	 */
4688 	if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4689 		ext4_unregister_li_request(sb);
4690 	else {
4691 		ext4_group_t first_not_zeroed;
4692 		first_not_zeroed = ext4_has_uninit_itable(sb);
4693 		ext4_register_li_request(sb, first_not_zeroed);
4694 	}
4695 
4696 	ext4_setup_system_zone(sb);
4697 	if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4698 		ext4_commit_super(sb, 1);
4699 
4700 #ifdef CONFIG_QUOTA
4701 	/* Release old quota file names */
4702 	for (i = 0; i < MAXQUOTAS; i++)
4703 		kfree(old_opts.s_qf_names[i]);
4704 	if (enable_quota) {
4705 		if (sb_any_quota_suspended(sb))
4706 			dquot_resume(sb, -1);
4707 		else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4708 					EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4709 			err = ext4_enable_quotas(sb);
4710 			if (err)
4711 				goto restore_opts;
4712 		}
4713 	}
4714 #endif
4715 
4716 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4717 	kfree(orig_data);
4718 	return 0;
4719 
4720 restore_opts:
4721 	sb->s_flags = old_sb_flags;
4722 	sbi->s_mount_opt = old_opts.s_mount_opt;
4723 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4724 	sbi->s_resuid = old_opts.s_resuid;
4725 	sbi->s_resgid = old_opts.s_resgid;
4726 	sbi->s_commit_interval = old_opts.s_commit_interval;
4727 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
4728 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
4729 #ifdef CONFIG_QUOTA
4730 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4731 	for (i = 0; i < MAXQUOTAS; i++) {
4732 		kfree(sbi->s_qf_names[i]);
4733 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4734 	}
4735 #endif
4736 	kfree(orig_data);
4737 	return err;
4738 }
4739 
4740 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4741 {
4742 	struct super_block *sb = dentry->d_sb;
4743 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4744 	struct ext4_super_block *es = sbi->s_es;
4745 	ext4_fsblk_t overhead = 0;
4746 	u64 fsid;
4747 	s64 bfree;
4748 
4749 	if (!test_opt(sb, MINIX_DF))
4750 		overhead = sbi->s_overhead;
4751 
4752 	buf->f_type = EXT4_SUPER_MAGIC;
4753 	buf->f_bsize = sb->s_blocksize;
4754 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4755 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4756 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4757 	/* prevent underflow in case that few free space is available */
4758 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4759 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4760 	if (buf->f_bfree < ext4_r_blocks_count(es))
4761 		buf->f_bavail = 0;
4762 	buf->f_files = le32_to_cpu(es->s_inodes_count);
4763 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4764 	buf->f_namelen = EXT4_NAME_LEN;
4765 	fsid = le64_to_cpup((void *)es->s_uuid) ^
4766 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4767 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4768 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4769 
4770 	return 0;
4771 }
4772 
4773 /* Helper function for writing quotas on sync - we need to start transaction
4774  * before quota file is locked for write. Otherwise the are possible deadlocks:
4775  * Process 1                         Process 2
4776  * ext4_create()                     quota_sync()
4777  *   jbd2_journal_start()                  write_dquot()
4778  *   dquot_initialize()                         down(dqio_mutex)
4779  *     down(dqio_mutex)                    jbd2_journal_start()
4780  *
4781  */
4782 
4783 #ifdef CONFIG_QUOTA
4784 
4785 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4786 {
4787 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4788 }
4789 
4790 static int ext4_write_dquot(struct dquot *dquot)
4791 {
4792 	int ret, err;
4793 	handle_t *handle;
4794 	struct inode *inode;
4795 
4796 	inode = dquot_to_inode(dquot);
4797 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4798 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4799 	if (IS_ERR(handle))
4800 		return PTR_ERR(handle);
4801 	ret = dquot_commit(dquot);
4802 	err = ext4_journal_stop(handle);
4803 	if (!ret)
4804 		ret = err;
4805 	return ret;
4806 }
4807 
4808 static int ext4_acquire_dquot(struct dquot *dquot)
4809 {
4810 	int ret, err;
4811 	handle_t *handle;
4812 
4813 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4814 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4815 	if (IS_ERR(handle))
4816 		return PTR_ERR(handle);
4817 	ret = dquot_acquire(dquot);
4818 	err = ext4_journal_stop(handle);
4819 	if (!ret)
4820 		ret = err;
4821 	return ret;
4822 }
4823 
4824 static int ext4_release_dquot(struct dquot *dquot)
4825 {
4826 	int ret, err;
4827 	handle_t *handle;
4828 
4829 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4830 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4831 	if (IS_ERR(handle)) {
4832 		/* Release dquot anyway to avoid endless cycle in dqput() */
4833 		dquot_release(dquot);
4834 		return PTR_ERR(handle);
4835 	}
4836 	ret = dquot_release(dquot);
4837 	err = ext4_journal_stop(handle);
4838 	if (!ret)
4839 		ret = err;
4840 	return ret;
4841 }
4842 
4843 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4844 {
4845 	struct super_block *sb = dquot->dq_sb;
4846 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4847 
4848 	/* Are we journaling quotas? */
4849 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4850 	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4851 		dquot_mark_dquot_dirty(dquot);
4852 		return ext4_write_dquot(dquot);
4853 	} else {
4854 		return dquot_mark_dquot_dirty(dquot);
4855 	}
4856 }
4857 
4858 static int ext4_write_info(struct super_block *sb, int type)
4859 {
4860 	int ret, err;
4861 	handle_t *handle;
4862 
4863 	/* Data block + inode block */
4864 	handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
4865 	if (IS_ERR(handle))
4866 		return PTR_ERR(handle);
4867 	ret = dquot_commit_info(sb, type);
4868 	err = ext4_journal_stop(handle);
4869 	if (!ret)
4870 		ret = err;
4871 	return ret;
4872 }
4873 
4874 /*
4875  * Turn on quotas during mount time - we need to find
4876  * the quota file and such...
4877  */
4878 static int ext4_quota_on_mount(struct super_block *sb, int type)
4879 {
4880 	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4881 					EXT4_SB(sb)->s_jquota_fmt, type);
4882 }
4883 
4884 /*
4885  * Standard function to be called on quota_on
4886  */
4887 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4888 			 struct path *path)
4889 {
4890 	int err;
4891 
4892 	if (!test_opt(sb, QUOTA))
4893 		return -EINVAL;
4894 
4895 	/* Quotafile not on the same filesystem? */
4896 	if (path->dentry->d_sb != sb)
4897 		return -EXDEV;
4898 	/* Journaling quota? */
4899 	if (EXT4_SB(sb)->s_qf_names[type]) {
4900 		/* Quotafile not in fs root? */
4901 		if (path->dentry->d_parent != sb->s_root)
4902 			ext4_msg(sb, KERN_WARNING,
4903 				"Quota file not on filesystem root. "
4904 				"Journaled quota will not work");
4905 	}
4906 
4907 	/*
4908 	 * When we journal data on quota file, we have to flush journal to see
4909 	 * all updates to the file when we bypass pagecache...
4910 	 */
4911 	if (EXT4_SB(sb)->s_journal &&
4912 	    ext4_should_journal_data(path->dentry->d_inode)) {
4913 		/*
4914 		 * We don't need to lock updates but journal_flush() could
4915 		 * otherwise be livelocked...
4916 		 */
4917 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4918 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4919 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4920 		if (err)
4921 			return err;
4922 	}
4923 
4924 	return dquot_quota_on(sb, type, format_id, path);
4925 }
4926 
4927 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4928 			     unsigned int flags)
4929 {
4930 	int err;
4931 	struct inode *qf_inode;
4932 	unsigned long qf_inums[MAXQUOTAS] = {
4933 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4934 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4935 	};
4936 
4937 	BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4938 
4939 	if (!qf_inums[type])
4940 		return -EPERM;
4941 
4942 	qf_inode = ext4_iget(sb, qf_inums[type]);
4943 	if (IS_ERR(qf_inode)) {
4944 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4945 		return PTR_ERR(qf_inode);
4946 	}
4947 
4948 	err = dquot_enable(qf_inode, type, format_id, flags);
4949 	iput(qf_inode);
4950 
4951 	return err;
4952 }
4953 
4954 /* Enable usage tracking for all quota types. */
4955 static int ext4_enable_quotas(struct super_block *sb)
4956 {
4957 	int type, err = 0;
4958 	unsigned long qf_inums[MAXQUOTAS] = {
4959 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4960 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4961 	};
4962 
4963 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4964 	for (type = 0; type < MAXQUOTAS; type++) {
4965 		if (qf_inums[type]) {
4966 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4967 						DQUOT_USAGE_ENABLED);
4968 			if (err) {
4969 				ext4_warning(sb,
4970 					"Failed to enable quota tracking "
4971 					"(type=%d, err=%d). Please run "
4972 					"e2fsck to fix.", type, err);
4973 				return err;
4974 			}
4975 		}
4976 	}
4977 	return 0;
4978 }
4979 
4980 /*
4981  * quota_on function that is used when QUOTA feature is set.
4982  */
4983 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4984 				 int format_id)
4985 {
4986 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4987 		return -EINVAL;
4988 
4989 	/*
4990 	 * USAGE was enabled at mount time. Only need to enable LIMITS now.
4991 	 */
4992 	return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
4993 }
4994 
4995 static int ext4_quota_off(struct super_block *sb, int type)
4996 {
4997 	struct inode *inode = sb_dqopt(sb)->files[type];
4998 	handle_t *handle;
4999 
5000 	/* Force all delayed allocation blocks to be allocated.
5001 	 * Caller already holds s_umount sem */
5002 	if (test_opt(sb, DELALLOC))
5003 		sync_filesystem(sb);
5004 
5005 	if (!inode)
5006 		goto out;
5007 
5008 	/* Update modification times of quota files when userspace can
5009 	 * start looking at them */
5010 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5011 	if (IS_ERR(handle))
5012 		goto out;
5013 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5014 	ext4_mark_inode_dirty(handle, inode);
5015 	ext4_journal_stop(handle);
5016 
5017 out:
5018 	return dquot_quota_off(sb, type);
5019 }
5020 
5021 /*
5022  * quota_off function that is used when QUOTA feature is set.
5023  */
5024 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5025 {
5026 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5027 		return -EINVAL;
5028 
5029 	/* Disable only the limits. */
5030 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5031 }
5032 
5033 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5034  * acquiring the locks... As quota files are never truncated and quota code
5035  * itself serializes the operations (and no one else should touch the files)
5036  * we don't have to be afraid of races */
5037 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5038 			       size_t len, loff_t off)
5039 {
5040 	struct inode *inode = sb_dqopt(sb)->files[type];
5041 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5042 	int err = 0;
5043 	int offset = off & (sb->s_blocksize - 1);
5044 	int tocopy;
5045 	size_t toread;
5046 	struct buffer_head *bh;
5047 	loff_t i_size = i_size_read(inode);
5048 
5049 	if (off > i_size)
5050 		return 0;
5051 	if (off+len > i_size)
5052 		len = i_size-off;
5053 	toread = len;
5054 	while (toread > 0) {
5055 		tocopy = sb->s_blocksize - offset < toread ?
5056 				sb->s_blocksize - offset : toread;
5057 		bh = ext4_bread(NULL, inode, blk, 0, &err);
5058 		if (err)
5059 			return err;
5060 		if (!bh)	/* A hole? */
5061 			memset(data, 0, tocopy);
5062 		else
5063 			memcpy(data, bh->b_data+offset, tocopy);
5064 		brelse(bh);
5065 		offset = 0;
5066 		toread -= tocopy;
5067 		data += tocopy;
5068 		blk++;
5069 	}
5070 	return len;
5071 }
5072 
5073 /* Write to quotafile (we know the transaction is already started and has
5074  * enough credits) */
5075 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5076 				const char *data, size_t len, loff_t off)
5077 {
5078 	struct inode *inode = sb_dqopt(sb)->files[type];
5079 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5080 	int err = 0;
5081 	int offset = off & (sb->s_blocksize - 1);
5082 	struct buffer_head *bh;
5083 	handle_t *handle = journal_current_handle();
5084 
5085 	if (EXT4_SB(sb)->s_journal && !handle) {
5086 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5087 			" cancelled because transaction is not started",
5088 			(unsigned long long)off, (unsigned long long)len);
5089 		return -EIO;
5090 	}
5091 	/*
5092 	 * Since we account only one data block in transaction credits,
5093 	 * then it is impossible to cross a block boundary.
5094 	 */
5095 	if (sb->s_blocksize - offset < len) {
5096 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5097 			" cancelled because not block aligned",
5098 			(unsigned long long)off, (unsigned long long)len);
5099 		return -EIO;
5100 	}
5101 
5102 	bh = ext4_bread(handle, inode, blk, 1, &err);
5103 	if (!bh)
5104 		goto out;
5105 	err = ext4_journal_get_write_access(handle, bh);
5106 	if (err) {
5107 		brelse(bh);
5108 		goto out;
5109 	}
5110 	lock_buffer(bh);
5111 	memcpy(bh->b_data+offset, data, len);
5112 	flush_dcache_page(bh->b_page);
5113 	unlock_buffer(bh);
5114 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5115 	brelse(bh);
5116 out:
5117 	if (err)
5118 		return err;
5119 	if (inode->i_size < off + len) {
5120 		i_size_write(inode, off + len);
5121 		EXT4_I(inode)->i_disksize = inode->i_size;
5122 		ext4_mark_inode_dirty(handle, inode);
5123 	}
5124 	return len;
5125 }
5126 
5127 #endif
5128 
5129 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5130 		       const char *dev_name, void *data)
5131 {
5132 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5133 }
5134 
5135 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5136 static inline void register_as_ext2(void)
5137 {
5138 	int err = register_filesystem(&ext2_fs_type);
5139 	if (err)
5140 		printk(KERN_WARNING
5141 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5142 }
5143 
5144 static inline void unregister_as_ext2(void)
5145 {
5146 	unregister_filesystem(&ext2_fs_type);
5147 }
5148 
5149 static inline int ext2_feature_set_ok(struct super_block *sb)
5150 {
5151 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5152 		return 0;
5153 	if (sb->s_flags & MS_RDONLY)
5154 		return 1;
5155 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5156 		return 0;
5157 	return 1;
5158 }
5159 #else
5160 static inline void register_as_ext2(void) { }
5161 static inline void unregister_as_ext2(void) { }
5162 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5163 #endif
5164 
5165 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5166 static inline void register_as_ext3(void)
5167 {
5168 	int err = register_filesystem(&ext3_fs_type);
5169 	if (err)
5170 		printk(KERN_WARNING
5171 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5172 }
5173 
5174 static inline void unregister_as_ext3(void)
5175 {
5176 	unregister_filesystem(&ext3_fs_type);
5177 }
5178 
5179 static inline int ext3_feature_set_ok(struct super_block *sb)
5180 {
5181 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5182 		return 0;
5183 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5184 		return 0;
5185 	if (sb->s_flags & MS_RDONLY)
5186 		return 1;
5187 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5188 		return 0;
5189 	return 1;
5190 }
5191 #else
5192 static inline void register_as_ext3(void) { }
5193 static inline void unregister_as_ext3(void) { }
5194 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5195 #endif
5196 
5197 static struct file_system_type ext4_fs_type = {
5198 	.owner		= THIS_MODULE,
5199 	.name		= "ext4",
5200 	.mount		= ext4_mount,
5201 	.kill_sb	= kill_block_super,
5202 	.fs_flags	= FS_REQUIRES_DEV,
5203 };
5204 MODULE_ALIAS_FS("ext4");
5205 
5206 static int __init ext4_init_feat_adverts(void)
5207 {
5208 	struct ext4_features *ef;
5209 	int ret = -ENOMEM;
5210 
5211 	ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5212 	if (!ef)
5213 		goto out;
5214 
5215 	ef->f_kobj.kset = ext4_kset;
5216 	init_completion(&ef->f_kobj_unregister);
5217 	ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5218 				   "features");
5219 	if (ret) {
5220 		kfree(ef);
5221 		goto out;
5222 	}
5223 
5224 	ext4_feat = ef;
5225 	ret = 0;
5226 out:
5227 	return ret;
5228 }
5229 
5230 static void ext4_exit_feat_adverts(void)
5231 {
5232 	kobject_put(&ext4_feat->f_kobj);
5233 	wait_for_completion(&ext4_feat->f_kobj_unregister);
5234 	kfree(ext4_feat);
5235 }
5236 
5237 /* Shared across all ext4 file systems */
5238 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5239 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5240 
5241 static int __init ext4_init_fs(void)
5242 {
5243 	int i, err;
5244 
5245 	ext4_li_info = NULL;
5246 	mutex_init(&ext4_li_mtx);
5247 
5248 	/* Build-time check for flags consistency */
5249 	ext4_check_flag_values();
5250 
5251 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5252 		mutex_init(&ext4__aio_mutex[i]);
5253 		init_waitqueue_head(&ext4__ioend_wq[i]);
5254 	}
5255 
5256 	err = ext4_init_es();
5257 	if (err)
5258 		return err;
5259 
5260 	err = ext4_init_pageio();
5261 	if (err)
5262 		goto out7;
5263 
5264 	err = ext4_init_system_zone();
5265 	if (err)
5266 		goto out6;
5267 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5268 	if (!ext4_kset) {
5269 		err = -ENOMEM;
5270 		goto out5;
5271 	}
5272 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5273 
5274 	err = ext4_init_feat_adverts();
5275 	if (err)
5276 		goto out4;
5277 
5278 	err = ext4_init_mballoc();
5279 	if (err)
5280 		goto out3;
5281 
5282 	err = ext4_init_xattr();
5283 	if (err)
5284 		goto out2;
5285 	err = init_inodecache();
5286 	if (err)
5287 		goto out1;
5288 	register_as_ext3();
5289 	register_as_ext2();
5290 	err = register_filesystem(&ext4_fs_type);
5291 	if (err)
5292 		goto out;
5293 
5294 	return 0;
5295 out:
5296 	unregister_as_ext2();
5297 	unregister_as_ext3();
5298 	destroy_inodecache();
5299 out1:
5300 	ext4_exit_xattr();
5301 out2:
5302 	ext4_exit_mballoc();
5303 out3:
5304 	ext4_exit_feat_adverts();
5305 out4:
5306 	if (ext4_proc_root)
5307 		remove_proc_entry("fs/ext4", NULL);
5308 	kset_unregister(ext4_kset);
5309 out5:
5310 	ext4_exit_system_zone();
5311 out6:
5312 	ext4_exit_pageio();
5313 out7:
5314 	ext4_exit_es();
5315 
5316 	return err;
5317 }
5318 
5319 static void __exit ext4_exit_fs(void)
5320 {
5321 	ext4_destroy_lazyinit_thread();
5322 	unregister_as_ext2();
5323 	unregister_as_ext3();
5324 	unregister_filesystem(&ext4_fs_type);
5325 	destroy_inodecache();
5326 	ext4_exit_xattr();
5327 	ext4_exit_mballoc();
5328 	ext4_exit_feat_adverts();
5329 	remove_proc_entry("fs/ext4", NULL);
5330 	kset_unregister(ext4_kset);
5331 	ext4_exit_system_zone();
5332 	ext4_exit_pageio();
5333 }
5334 
5335 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5336 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5337 MODULE_LICENSE("GPL");
5338 module_init(ext4_init_fs)
5339 module_exit(ext4_exit_fs)
5340