xref: /linux/fs/ext4/super.c (revision 61d0b5a4b2777dcf5daef245e212b3c1fa8091ca)
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 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
94 #else
95 #define IS_EXT2_SB(sb) (0)
96 #endif
97 
98 
99 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
100 static struct file_system_type ext3_fs_type = {
101 	.owner		= THIS_MODULE,
102 	.name		= "ext3",
103 	.mount		= ext4_mount,
104 	.kill_sb	= kill_block_super,
105 	.fs_flags	= FS_REQUIRES_DEV,
106 };
107 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
108 #else
109 #define IS_EXT3_SB(sb) (0)
110 #endif
111 
112 static int ext4_verify_csum_type(struct super_block *sb,
113 				 struct ext4_super_block *es)
114 {
115 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
116 					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
117 		return 1;
118 
119 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
120 }
121 
122 static __le32 ext4_superblock_csum(struct super_block *sb,
123 				   struct ext4_super_block *es)
124 {
125 	struct ext4_sb_info *sbi = EXT4_SB(sb);
126 	int offset = offsetof(struct ext4_super_block, s_checksum);
127 	__u32 csum;
128 
129 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
130 
131 	return cpu_to_le32(csum);
132 }
133 
134 int ext4_superblock_csum_verify(struct super_block *sb,
135 				struct ext4_super_block *es)
136 {
137 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
138 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
139 		return 1;
140 
141 	return es->s_checksum == ext4_superblock_csum(sb, es);
142 }
143 
144 void ext4_superblock_csum_set(struct super_block *sb)
145 {
146 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
147 
148 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
149 		EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
150 		return;
151 
152 	es->s_checksum = ext4_superblock_csum(sb, es);
153 }
154 
155 void *ext4_kvmalloc(size_t size, gfp_t flags)
156 {
157 	void *ret;
158 
159 	ret = kmalloc(size, flags);
160 	if (!ret)
161 		ret = __vmalloc(size, flags, PAGE_KERNEL);
162 	return ret;
163 }
164 
165 void *ext4_kvzalloc(size_t size, gfp_t flags)
166 {
167 	void *ret;
168 
169 	ret = kzalloc(size, flags);
170 	if (!ret)
171 		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
172 	return ret;
173 }
174 
175 void ext4_kvfree(void *ptr)
176 {
177 	if (is_vmalloc_addr(ptr))
178 		vfree(ptr);
179 	else
180 		kfree(ptr);
181 
182 }
183 
184 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
185 			       struct ext4_group_desc *bg)
186 {
187 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
188 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
189 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
190 }
191 
192 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
193 			       struct ext4_group_desc *bg)
194 {
195 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
196 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
197 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
198 }
199 
200 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
201 			      struct ext4_group_desc *bg)
202 {
203 	return le32_to_cpu(bg->bg_inode_table_lo) |
204 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
205 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
206 }
207 
208 __u32 ext4_free_group_clusters(struct super_block *sb,
209 			       struct ext4_group_desc *bg)
210 {
211 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
212 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
213 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
214 }
215 
216 __u32 ext4_free_inodes_count(struct super_block *sb,
217 			      struct ext4_group_desc *bg)
218 {
219 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
220 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
221 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
222 }
223 
224 __u32 ext4_used_dirs_count(struct super_block *sb,
225 			      struct ext4_group_desc *bg)
226 {
227 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
228 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
229 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
230 }
231 
232 __u32 ext4_itable_unused_count(struct super_block *sb,
233 			      struct ext4_group_desc *bg)
234 {
235 	return le16_to_cpu(bg->bg_itable_unused_lo) |
236 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
237 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
238 }
239 
240 void ext4_block_bitmap_set(struct super_block *sb,
241 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
242 {
243 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
244 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
245 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
246 }
247 
248 void ext4_inode_bitmap_set(struct super_block *sb,
249 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
250 {
251 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
252 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
253 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
254 }
255 
256 void ext4_inode_table_set(struct super_block *sb,
257 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
258 {
259 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
260 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
261 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
262 }
263 
264 void ext4_free_group_clusters_set(struct super_block *sb,
265 				  struct ext4_group_desc *bg, __u32 count)
266 {
267 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
268 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
269 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
270 }
271 
272 void ext4_free_inodes_set(struct super_block *sb,
273 			  struct ext4_group_desc *bg, __u32 count)
274 {
275 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
276 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
277 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
278 }
279 
280 void ext4_used_dirs_set(struct super_block *sb,
281 			  struct ext4_group_desc *bg, __u32 count)
282 {
283 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
284 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
285 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
286 }
287 
288 void ext4_itable_unused_set(struct super_block *sb,
289 			  struct ext4_group_desc *bg, __u32 count)
290 {
291 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
292 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
293 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
294 }
295 
296 
297 static void __save_error_info(struct super_block *sb, const char *func,
298 			    unsigned int line)
299 {
300 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
301 
302 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
303 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
304 	es->s_last_error_time = cpu_to_le32(get_seconds());
305 	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
306 	es->s_last_error_line = cpu_to_le32(line);
307 	if (!es->s_first_error_time) {
308 		es->s_first_error_time = es->s_last_error_time;
309 		strncpy(es->s_first_error_func, func,
310 			sizeof(es->s_first_error_func));
311 		es->s_first_error_line = cpu_to_le32(line);
312 		es->s_first_error_ino = es->s_last_error_ino;
313 		es->s_first_error_block = es->s_last_error_block;
314 	}
315 	/*
316 	 * Start the daily error reporting function if it hasn't been
317 	 * started already
318 	 */
319 	if (!es->s_error_count)
320 		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
321 	le32_add_cpu(&es->s_error_count, 1);
322 }
323 
324 static void save_error_info(struct super_block *sb, const char *func,
325 			    unsigned int line)
326 {
327 	__save_error_info(sb, func, line);
328 	ext4_commit_super(sb, 1);
329 }
330 
331 /*
332  * The del_gendisk() function uninitializes the disk-specific data
333  * structures, including the bdi structure, without telling anyone
334  * else.  Once this happens, any attempt to call mark_buffer_dirty()
335  * (for example, by ext4_commit_super), will cause a kernel OOPS.
336  * This is a kludge to prevent these oops until we can put in a proper
337  * hook in del_gendisk() to inform the VFS and file system layers.
338  */
339 static int block_device_ejected(struct super_block *sb)
340 {
341 	struct inode *bd_inode = sb->s_bdev->bd_inode;
342 	struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
343 
344 	return bdi->dev == NULL;
345 }
346 
347 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
348 {
349 	struct super_block		*sb = journal->j_private;
350 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
351 	int				error = is_journal_aborted(journal);
352 	struct ext4_journal_cb_entry	*jce, *tmp;
353 
354 	spin_lock(&sbi->s_md_lock);
355 	list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
356 		list_del_init(&jce->jce_list);
357 		spin_unlock(&sbi->s_md_lock);
358 		jce->jce_func(sb, jce, error);
359 		spin_lock(&sbi->s_md_lock);
360 	}
361 	spin_unlock(&sbi->s_md_lock);
362 }
363 
364 /* Deal with the reporting of failure conditions on a filesystem such as
365  * inconsistencies detected or read IO failures.
366  *
367  * On ext2, we can store the error state of the filesystem in the
368  * superblock.  That is not possible on ext4, because we may have other
369  * write ordering constraints on the superblock which prevent us from
370  * writing it out straight away; and given that the journal is about to
371  * be aborted, we can't rely on the current, or future, transactions to
372  * write out the superblock safely.
373  *
374  * We'll just use the jbd2_journal_abort() error code to record an error in
375  * the journal instead.  On recovery, the journal will complain about
376  * that error until we've noted it down and cleared it.
377  */
378 
379 static void ext4_handle_error(struct super_block *sb)
380 {
381 	if (sb->s_flags & MS_RDONLY)
382 		return;
383 
384 	if (!test_opt(sb, ERRORS_CONT)) {
385 		journal_t *journal = EXT4_SB(sb)->s_journal;
386 
387 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
388 		if (journal)
389 			jbd2_journal_abort(journal, -EIO);
390 	}
391 	if (test_opt(sb, ERRORS_RO)) {
392 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
393 		sb->s_flags |= MS_RDONLY;
394 	}
395 	if (test_opt(sb, ERRORS_PANIC))
396 		panic("EXT4-fs (device %s): panic forced after error\n",
397 			sb->s_id);
398 }
399 
400 void __ext4_error(struct super_block *sb, const char *function,
401 		  unsigned int line, const char *fmt, ...)
402 {
403 	struct va_format vaf;
404 	va_list args;
405 
406 	va_start(args, fmt);
407 	vaf.fmt = fmt;
408 	vaf.va = &args;
409 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
410 	       sb->s_id, function, line, current->comm, &vaf);
411 	va_end(args);
412 	save_error_info(sb, function, line);
413 
414 	ext4_handle_error(sb);
415 }
416 
417 void ext4_error_inode(struct inode *inode, const char *function,
418 		      unsigned int line, ext4_fsblk_t block,
419 		      const char *fmt, ...)
420 {
421 	va_list args;
422 	struct va_format vaf;
423 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
424 
425 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
426 	es->s_last_error_block = cpu_to_le64(block);
427 	save_error_info(inode->i_sb, function, line);
428 	va_start(args, fmt);
429 	vaf.fmt = fmt;
430 	vaf.va = &args;
431 	if (block)
432 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
433 		       "inode #%lu: block %llu: comm %s: %pV\n",
434 		       inode->i_sb->s_id, function, line, inode->i_ino,
435 		       block, current->comm, &vaf);
436 	else
437 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
438 		       "inode #%lu: comm %s: %pV\n",
439 		       inode->i_sb->s_id, function, line, inode->i_ino,
440 		       current->comm, &vaf);
441 	va_end(args);
442 
443 	ext4_handle_error(inode->i_sb);
444 }
445 
446 void ext4_error_file(struct file *file, const char *function,
447 		     unsigned int line, ext4_fsblk_t block,
448 		     const char *fmt, ...)
449 {
450 	va_list args;
451 	struct va_format vaf;
452 	struct ext4_super_block *es;
453 	struct inode *inode = file_inode(file);
454 	char pathname[80], *path;
455 
456 	es = EXT4_SB(inode->i_sb)->s_es;
457 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
458 	save_error_info(inode->i_sb, function, line);
459 	path = d_path(&(file->f_path), pathname, sizeof(pathname));
460 	if (IS_ERR(path))
461 		path = "(unknown)";
462 	va_start(args, fmt);
463 	vaf.fmt = fmt;
464 	vaf.va = &args;
465 	if (block)
466 		printk(KERN_CRIT
467 		       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
468 		       "block %llu: comm %s: path %s: %pV\n",
469 		       inode->i_sb->s_id, function, line, inode->i_ino,
470 		       block, current->comm, path, &vaf);
471 	else
472 		printk(KERN_CRIT
473 		       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
474 		       "comm %s: path %s: %pV\n",
475 		       inode->i_sb->s_id, function, line, inode->i_ino,
476 		       current->comm, path, &vaf);
477 	va_end(args);
478 
479 	ext4_handle_error(inode->i_sb);
480 }
481 
482 const char *ext4_decode_error(struct super_block *sb, int errno,
483 			      char nbuf[16])
484 {
485 	char *errstr = NULL;
486 
487 	switch (errno) {
488 	case -EIO:
489 		errstr = "IO failure";
490 		break;
491 	case -ENOMEM:
492 		errstr = "Out of memory";
493 		break;
494 	case -EROFS:
495 		if (!sb || (EXT4_SB(sb)->s_journal &&
496 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
497 			errstr = "Journal has aborted";
498 		else
499 			errstr = "Readonly filesystem";
500 		break;
501 	default:
502 		/* If the caller passed in an extra buffer for unknown
503 		 * errors, textualise them now.  Else we just return
504 		 * NULL. */
505 		if (nbuf) {
506 			/* Check for truncated error codes... */
507 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
508 				errstr = nbuf;
509 		}
510 		break;
511 	}
512 
513 	return errstr;
514 }
515 
516 /* __ext4_std_error decodes expected errors from journaling functions
517  * automatically and invokes the appropriate error response.  */
518 
519 void __ext4_std_error(struct super_block *sb, const char *function,
520 		      unsigned int line, int errno)
521 {
522 	char nbuf[16];
523 	const char *errstr;
524 
525 	/* Special case: if the error is EROFS, and we're not already
526 	 * inside a transaction, then there's really no point in logging
527 	 * an error. */
528 	if (errno == -EROFS && journal_current_handle() == NULL &&
529 	    (sb->s_flags & MS_RDONLY))
530 		return;
531 
532 	errstr = ext4_decode_error(sb, errno, nbuf);
533 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
534 	       sb->s_id, function, line, errstr);
535 	save_error_info(sb, function, line);
536 
537 	ext4_handle_error(sb);
538 }
539 
540 /*
541  * ext4_abort is a much stronger failure handler than ext4_error.  The
542  * abort function may be used to deal with unrecoverable failures such
543  * as journal IO errors or ENOMEM at a critical moment in log management.
544  *
545  * We unconditionally force the filesystem into an ABORT|READONLY state,
546  * unless the error response on the fs has been set to panic in which
547  * case we take the easy way out and panic immediately.
548  */
549 
550 void __ext4_abort(struct super_block *sb, const char *function,
551 		unsigned int line, const char *fmt, ...)
552 {
553 	va_list args;
554 
555 	save_error_info(sb, function, line);
556 	va_start(args, fmt);
557 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
558 	       function, line);
559 	vprintk(fmt, args);
560 	printk("\n");
561 	va_end(args);
562 
563 	if ((sb->s_flags & MS_RDONLY) == 0) {
564 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
565 		sb->s_flags |= MS_RDONLY;
566 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
567 		if (EXT4_SB(sb)->s_journal)
568 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
569 		save_error_info(sb, function, line);
570 	}
571 	if (test_opt(sb, ERRORS_PANIC))
572 		panic("EXT4-fs panic from previous error\n");
573 }
574 
575 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
576 {
577 	struct va_format vaf;
578 	va_list args;
579 
580 	va_start(args, fmt);
581 	vaf.fmt = fmt;
582 	vaf.va = &args;
583 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
584 	va_end(args);
585 }
586 
587 void __ext4_warning(struct super_block *sb, const char *function,
588 		    unsigned int line, const char *fmt, ...)
589 {
590 	struct va_format vaf;
591 	va_list args;
592 
593 	va_start(args, fmt);
594 	vaf.fmt = fmt;
595 	vaf.va = &args;
596 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
597 	       sb->s_id, function, line, &vaf);
598 	va_end(args);
599 }
600 
601 void __ext4_grp_locked_error(const char *function, unsigned int line,
602 			     struct super_block *sb, ext4_group_t grp,
603 			     unsigned long ino, ext4_fsblk_t block,
604 			     const char *fmt, ...)
605 __releases(bitlock)
606 __acquires(bitlock)
607 {
608 	struct va_format vaf;
609 	va_list args;
610 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
611 
612 	es->s_last_error_ino = cpu_to_le32(ino);
613 	es->s_last_error_block = cpu_to_le64(block);
614 	__save_error_info(sb, function, line);
615 
616 	va_start(args, fmt);
617 
618 	vaf.fmt = fmt;
619 	vaf.va = &args;
620 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
621 	       sb->s_id, function, line, grp);
622 	if (ino)
623 		printk(KERN_CONT "inode %lu: ", ino);
624 	if (block)
625 		printk(KERN_CONT "block %llu:", (unsigned long long) block);
626 	printk(KERN_CONT "%pV\n", &vaf);
627 	va_end(args);
628 
629 	if (test_opt(sb, ERRORS_CONT)) {
630 		ext4_commit_super(sb, 0);
631 		return;
632 	}
633 
634 	ext4_unlock_group(sb, grp);
635 	ext4_handle_error(sb);
636 	/*
637 	 * We only get here in the ERRORS_RO case; relocking the group
638 	 * may be dangerous, but nothing bad will happen since the
639 	 * filesystem will have already been marked read/only and the
640 	 * journal has been aborted.  We return 1 as a hint to callers
641 	 * who might what to use the return value from
642 	 * ext4_grp_locked_error() to distinguish between the
643 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
644 	 * aggressively from the ext4 function in question, with a
645 	 * more appropriate error code.
646 	 */
647 	ext4_lock_group(sb, grp);
648 	return;
649 }
650 
651 void ext4_update_dynamic_rev(struct super_block *sb)
652 {
653 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
654 
655 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
656 		return;
657 
658 	ext4_warning(sb,
659 		     "updating to rev %d because of new feature flag, "
660 		     "running e2fsck is recommended",
661 		     EXT4_DYNAMIC_REV);
662 
663 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
664 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
665 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
666 	/* leave es->s_feature_*compat flags alone */
667 	/* es->s_uuid will be set by e2fsck if empty */
668 
669 	/*
670 	 * The rest of the superblock fields should be zero, and if not it
671 	 * means they are likely already in use, so leave them alone.  We
672 	 * can leave it up to e2fsck to clean up any inconsistencies there.
673 	 */
674 }
675 
676 /*
677  * Open the external journal device
678  */
679 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
680 {
681 	struct block_device *bdev;
682 	char b[BDEVNAME_SIZE];
683 
684 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
685 	if (IS_ERR(bdev))
686 		goto fail;
687 	return bdev;
688 
689 fail:
690 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
691 			__bdevname(dev, b), PTR_ERR(bdev));
692 	return NULL;
693 }
694 
695 /*
696  * Release the journal device
697  */
698 static int ext4_blkdev_put(struct block_device *bdev)
699 {
700 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
701 }
702 
703 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
704 {
705 	struct block_device *bdev;
706 	int ret = -ENODEV;
707 
708 	bdev = sbi->journal_bdev;
709 	if (bdev) {
710 		ret = ext4_blkdev_put(bdev);
711 		sbi->journal_bdev = NULL;
712 	}
713 	return ret;
714 }
715 
716 static inline struct inode *orphan_list_entry(struct list_head *l)
717 {
718 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
719 }
720 
721 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
722 {
723 	struct list_head *l;
724 
725 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
726 		 le32_to_cpu(sbi->s_es->s_last_orphan));
727 
728 	printk(KERN_ERR "sb_info orphan list:\n");
729 	list_for_each(l, &sbi->s_orphan) {
730 		struct inode *inode = orphan_list_entry(l);
731 		printk(KERN_ERR "  "
732 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
733 		       inode->i_sb->s_id, inode->i_ino, inode,
734 		       inode->i_mode, inode->i_nlink,
735 		       NEXT_ORPHAN(inode));
736 	}
737 }
738 
739 static void ext4_put_super(struct super_block *sb)
740 {
741 	struct ext4_sb_info *sbi = EXT4_SB(sb);
742 	struct ext4_super_block *es = sbi->s_es;
743 	int i, err;
744 
745 	ext4_unregister_li_request(sb);
746 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
747 
748 	flush_workqueue(sbi->dio_unwritten_wq);
749 	destroy_workqueue(sbi->dio_unwritten_wq);
750 
751 	if (sbi->s_journal) {
752 		err = jbd2_journal_destroy(sbi->s_journal);
753 		sbi->s_journal = NULL;
754 		if (err < 0)
755 			ext4_abort(sb, "Couldn't clean up the journal");
756 	}
757 
758 	ext4_es_unregister_shrinker(sb);
759 	del_timer(&sbi->s_err_report);
760 	ext4_release_system_zone(sb);
761 	ext4_mb_release(sb);
762 	ext4_ext_release(sb);
763 	ext4_xattr_put_super(sb);
764 
765 	if (!(sb->s_flags & MS_RDONLY)) {
766 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
767 		es->s_state = cpu_to_le16(sbi->s_mount_state);
768 	}
769 	if (!(sb->s_flags & MS_RDONLY))
770 		ext4_commit_super(sb, 1);
771 
772 	if (sbi->s_proc) {
773 		remove_proc_entry("options", sbi->s_proc);
774 		remove_proc_entry(sb->s_id, ext4_proc_root);
775 	}
776 	kobject_del(&sbi->s_kobj);
777 
778 	for (i = 0; i < sbi->s_gdb_count; i++)
779 		brelse(sbi->s_group_desc[i]);
780 	ext4_kvfree(sbi->s_group_desc);
781 	ext4_kvfree(sbi->s_flex_groups);
782 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
783 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
784 	percpu_counter_destroy(&sbi->s_dirs_counter);
785 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
786 	percpu_counter_destroy(&sbi->s_extent_cache_cnt);
787 	brelse(sbi->s_sbh);
788 #ifdef CONFIG_QUOTA
789 	for (i = 0; i < MAXQUOTAS; i++)
790 		kfree(sbi->s_qf_names[i]);
791 #endif
792 
793 	/* Debugging code just in case the in-memory inode orphan list
794 	 * isn't empty.  The on-disk one can be non-empty if we've
795 	 * detected an error and taken the fs readonly, but the
796 	 * in-memory list had better be clean by this point. */
797 	if (!list_empty(&sbi->s_orphan))
798 		dump_orphan_list(sb, sbi);
799 	J_ASSERT(list_empty(&sbi->s_orphan));
800 
801 	invalidate_bdev(sb->s_bdev);
802 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
803 		/*
804 		 * Invalidate the journal device's buffers.  We don't want them
805 		 * floating about in memory - the physical journal device may
806 		 * hotswapped, and it breaks the `ro-after' testing code.
807 		 */
808 		sync_blockdev(sbi->journal_bdev);
809 		invalidate_bdev(sbi->journal_bdev);
810 		ext4_blkdev_remove(sbi);
811 	}
812 	if (sbi->s_mmp_tsk)
813 		kthread_stop(sbi->s_mmp_tsk);
814 	sb->s_fs_info = NULL;
815 	/*
816 	 * Now that we are completely done shutting down the
817 	 * superblock, we need to actually destroy the kobject.
818 	 */
819 	kobject_put(&sbi->s_kobj);
820 	wait_for_completion(&sbi->s_kobj_unregister);
821 	if (sbi->s_chksum_driver)
822 		crypto_free_shash(sbi->s_chksum_driver);
823 	kfree(sbi->s_blockgroup_lock);
824 	kfree(sbi);
825 }
826 
827 static struct kmem_cache *ext4_inode_cachep;
828 
829 /*
830  * Called inside transaction, so use GFP_NOFS
831  */
832 static struct inode *ext4_alloc_inode(struct super_block *sb)
833 {
834 	struct ext4_inode_info *ei;
835 
836 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
837 	if (!ei)
838 		return NULL;
839 
840 	ei->vfs_inode.i_version = 1;
841 	INIT_LIST_HEAD(&ei->i_prealloc_list);
842 	spin_lock_init(&ei->i_prealloc_lock);
843 	ext4_es_init_tree(&ei->i_es_tree);
844 	rwlock_init(&ei->i_es_lock);
845 	INIT_LIST_HEAD(&ei->i_es_lru);
846 	ei->i_es_lru_nr = 0;
847 	ei->i_reserved_data_blocks = 0;
848 	ei->i_reserved_meta_blocks = 0;
849 	ei->i_allocated_meta_blocks = 0;
850 	ei->i_da_metadata_calc_len = 0;
851 	ei->i_da_metadata_calc_last_lblock = 0;
852 	spin_lock_init(&(ei->i_block_reservation_lock));
853 #ifdef CONFIG_QUOTA
854 	ei->i_reserved_quota = 0;
855 #endif
856 	ei->jinode = NULL;
857 	INIT_LIST_HEAD(&ei->i_completed_io_list);
858 	spin_lock_init(&ei->i_completed_io_lock);
859 	ei->i_sync_tid = 0;
860 	ei->i_datasync_tid = 0;
861 	atomic_set(&ei->i_ioend_count, 0);
862 	atomic_set(&ei->i_unwritten, 0);
863 	INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
864 
865 	return &ei->vfs_inode;
866 }
867 
868 static int ext4_drop_inode(struct inode *inode)
869 {
870 	int drop = generic_drop_inode(inode);
871 
872 	trace_ext4_drop_inode(inode, drop);
873 	return drop;
874 }
875 
876 static void ext4_i_callback(struct rcu_head *head)
877 {
878 	struct inode *inode = container_of(head, struct inode, i_rcu);
879 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
880 }
881 
882 static void ext4_destroy_inode(struct inode *inode)
883 {
884 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
885 		ext4_msg(inode->i_sb, KERN_ERR,
886 			 "Inode %lu (%p): orphan list check failed!",
887 			 inode->i_ino, EXT4_I(inode));
888 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
889 				EXT4_I(inode), sizeof(struct ext4_inode_info),
890 				true);
891 		dump_stack();
892 	}
893 	call_rcu(&inode->i_rcu, ext4_i_callback);
894 }
895 
896 static void init_once(void *foo)
897 {
898 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
899 
900 	INIT_LIST_HEAD(&ei->i_orphan);
901 	init_rwsem(&ei->xattr_sem);
902 	init_rwsem(&ei->i_data_sem);
903 	inode_init_once(&ei->vfs_inode);
904 }
905 
906 static int init_inodecache(void)
907 {
908 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
909 					     sizeof(struct ext4_inode_info),
910 					     0, (SLAB_RECLAIM_ACCOUNT|
911 						SLAB_MEM_SPREAD),
912 					     init_once);
913 	if (ext4_inode_cachep == NULL)
914 		return -ENOMEM;
915 	return 0;
916 }
917 
918 static void destroy_inodecache(void)
919 {
920 	/*
921 	 * Make sure all delayed rcu free inodes are flushed before we
922 	 * destroy cache.
923 	 */
924 	rcu_barrier();
925 	kmem_cache_destroy(ext4_inode_cachep);
926 }
927 
928 void ext4_clear_inode(struct inode *inode)
929 {
930 	invalidate_inode_buffers(inode);
931 	clear_inode(inode);
932 	dquot_drop(inode);
933 	ext4_discard_preallocations(inode);
934 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
935 	ext4_es_lru_del(inode);
936 	if (EXT4_I(inode)->jinode) {
937 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
938 					       EXT4_I(inode)->jinode);
939 		jbd2_free_inode(EXT4_I(inode)->jinode);
940 		EXT4_I(inode)->jinode = NULL;
941 	}
942 }
943 
944 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
945 					u64 ino, u32 generation)
946 {
947 	struct inode *inode;
948 
949 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
950 		return ERR_PTR(-ESTALE);
951 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
952 		return ERR_PTR(-ESTALE);
953 
954 	/* iget isn't really right if the inode is currently unallocated!!
955 	 *
956 	 * ext4_read_inode will return a bad_inode if the inode had been
957 	 * deleted, so we should be safe.
958 	 *
959 	 * Currently we don't know the generation for parent directory, so
960 	 * a generation of 0 means "accept any"
961 	 */
962 	inode = ext4_iget(sb, ino);
963 	if (IS_ERR(inode))
964 		return ERR_CAST(inode);
965 	if (generation && inode->i_generation != generation) {
966 		iput(inode);
967 		return ERR_PTR(-ESTALE);
968 	}
969 
970 	return inode;
971 }
972 
973 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
974 					int fh_len, int fh_type)
975 {
976 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
977 				    ext4_nfs_get_inode);
978 }
979 
980 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
981 					int fh_len, int fh_type)
982 {
983 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
984 				    ext4_nfs_get_inode);
985 }
986 
987 /*
988  * Try to release metadata pages (indirect blocks, directories) which are
989  * mapped via the block device.  Since these pages could have journal heads
990  * which would prevent try_to_free_buffers() from freeing them, we must use
991  * jbd2 layer's try_to_free_buffers() function to release them.
992  */
993 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
994 				 gfp_t wait)
995 {
996 	journal_t *journal = EXT4_SB(sb)->s_journal;
997 
998 	WARN_ON(PageChecked(page));
999 	if (!page_has_buffers(page))
1000 		return 0;
1001 	if (journal)
1002 		return jbd2_journal_try_to_free_buffers(journal, page,
1003 							wait & ~__GFP_WAIT);
1004 	return try_to_free_buffers(page);
1005 }
1006 
1007 #ifdef CONFIG_QUOTA
1008 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1009 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1010 
1011 static int ext4_write_dquot(struct dquot *dquot);
1012 static int ext4_acquire_dquot(struct dquot *dquot);
1013 static int ext4_release_dquot(struct dquot *dquot);
1014 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1015 static int ext4_write_info(struct super_block *sb, int type);
1016 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1017 			 struct path *path);
1018 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1019 				 int format_id);
1020 static int ext4_quota_off(struct super_block *sb, int type);
1021 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1022 static int ext4_quota_on_mount(struct super_block *sb, int type);
1023 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1024 			       size_t len, loff_t off);
1025 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1026 				const char *data, size_t len, loff_t off);
1027 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1028 			     unsigned int flags);
1029 static int ext4_enable_quotas(struct super_block *sb);
1030 
1031 static const struct dquot_operations ext4_quota_operations = {
1032 	.get_reserved_space = ext4_get_reserved_space,
1033 	.write_dquot	= ext4_write_dquot,
1034 	.acquire_dquot	= ext4_acquire_dquot,
1035 	.release_dquot	= ext4_release_dquot,
1036 	.mark_dirty	= ext4_mark_dquot_dirty,
1037 	.write_info	= ext4_write_info,
1038 	.alloc_dquot	= dquot_alloc,
1039 	.destroy_dquot	= dquot_destroy,
1040 };
1041 
1042 static const struct quotactl_ops ext4_qctl_operations = {
1043 	.quota_on	= ext4_quota_on,
1044 	.quota_off	= ext4_quota_off,
1045 	.quota_sync	= dquot_quota_sync,
1046 	.get_info	= dquot_get_dqinfo,
1047 	.set_info	= dquot_set_dqinfo,
1048 	.get_dqblk	= dquot_get_dqblk,
1049 	.set_dqblk	= dquot_set_dqblk
1050 };
1051 
1052 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1053 	.quota_on_meta	= ext4_quota_on_sysfile,
1054 	.quota_off	= ext4_quota_off_sysfile,
1055 	.quota_sync	= dquot_quota_sync,
1056 	.get_info	= dquot_get_dqinfo,
1057 	.set_info	= dquot_set_dqinfo,
1058 	.get_dqblk	= dquot_get_dqblk,
1059 	.set_dqblk	= dquot_set_dqblk
1060 };
1061 #endif
1062 
1063 static const struct super_operations ext4_sops = {
1064 	.alloc_inode	= ext4_alloc_inode,
1065 	.destroy_inode	= ext4_destroy_inode,
1066 	.write_inode	= ext4_write_inode,
1067 	.dirty_inode	= ext4_dirty_inode,
1068 	.drop_inode	= ext4_drop_inode,
1069 	.evict_inode	= ext4_evict_inode,
1070 	.put_super	= ext4_put_super,
1071 	.sync_fs	= ext4_sync_fs,
1072 	.freeze_fs	= ext4_freeze,
1073 	.unfreeze_fs	= ext4_unfreeze,
1074 	.statfs		= ext4_statfs,
1075 	.remount_fs	= ext4_remount,
1076 	.show_options	= ext4_show_options,
1077 #ifdef CONFIG_QUOTA
1078 	.quota_read	= ext4_quota_read,
1079 	.quota_write	= ext4_quota_write,
1080 #endif
1081 	.bdev_try_to_free_page = bdev_try_to_free_page,
1082 };
1083 
1084 static const struct super_operations ext4_nojournal_sops = {
1085 	.alloc_inode	= ext4_alloc_inode,
1086 	.destroy_inode	= ext4_destroy_inode,
1087 	.write_inode	= ext4_write_inode,
1088 	.dirty_inode	= ext4_dirty_inode,
1089 	.drop_inode	= ext4_drop_inode,
1090 	.evict_inode	= ext4_evict_inode,
1091 	.put_super	= ext4_put_super,
1092 	.statfs		= ext4_statfs,
1093 	.remount_fs	= ext4_remount,
1094 	.show_options	= ext4_show_options,
1095 #ifdef CONFIG_QUOTA
1096 	.quota_read	= ext4_quota_read,
1097 	.quota_write	= ext4_quota_write,
1098 #endif
1099 	.bdev_try_to_free_page = bdev_try_to_free_page,
1100 };
1101 
1102 static const struct export_operations ext4_export_ops = {
1103 	.fh_to_dentry = ext4_fh_to_dentry,
1104 	.fh_to_parent = ext4_fh_to_parent,
1105 	.get_parent = ext4_get_parent,
1106 };
1107 
1108 enum {
1109 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1110 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1111 	Opt_nouid32, Opt_debug, Opt_removed,
1112 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1113 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1114 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1115 	Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1116 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1117 	Opt_data_err_abort, Opt_data_err_ignore,
1118 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1119 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1120 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1121 	Opt_usrquota, Opt_grpquota, Opt_i_version,
1122 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1123 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1124 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1125 	Opt_dioread_nolock, Opt_dioread_lock,
1126 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1127 	Opt_max_dir_size_kb,
1128 };
1129 
1130 static const match_table_t tokens = {
1131 	{Opt_bsd_df, "bsddf"},
1132 	{Opt_minix_df, "minixdf"},
1133 	{Opt_grpid, "grpid"},
1134 	{Opt_grpid, "bsdgroups"},
1135 	{Opt_nogrpid, "nogrpid"},
1136 	{Opt_nogrpid, "sysvgroups"},
1137 	{Opt_resgid, "resgid=%u"},
1138 	{Opt_resuid, "resuid=%u"},
1139 	{Opt_sb, "sb=%u"},
1140 	{Opt_err_cont, "errors=continue"},
1141 	{Opt_err_panic, "errors=panic"},
1142 	{Opt_err_ro, "errors=remount-ro"},
1143 	{Opt_nouid32, "nouid32"},
1144 	{Opt_debug, "debug"},
1145 	{Opt_removed, "oldalloc"},
1146 	{Opt_removed, "orlov"},
1147 	{Opt_user_xattr, "user_xattr"},
1148 	{Opt_nouser_xattr, "nouser_xattr"},
1149 	{Opt_acl, "acl"},
1150 	{Opt_noacl, "noacl"},
1151 	{Opt_noload, "norecovery"},
1152 	{Opt_noload, "noload"},
1153 	{Opt_removed, "nobh"},
1154 	{Opt_removed, "bh"},
1155 	{Opt_commit, "commit=%u"},
1156 	{Opt_min_batch_time, "min_batch_time=%u"},
1157 	{Opt_max_batch_time, "max_batch_time=%u"},
1158 	{Opt_journal_dev, "journal_dev=%u"},
1159 	{Opt_journal_checksum, "journal_checksum"},
1160 	{Opt_journal_async_commit, "journal_async_commit"},
1161 	{Opt_abort, "abort"},
1162 	{Opt_data_journal, "data=journal"},
1163 	{Opt_data_ordered, "data=ordered"},
1164 	{Opt_data_writeback, "data=writeback"},
1165 	{Opt_data_err_abort, "data_err=abort"},
1166 	{Opt_data_err_ignore, "data_err=ignore"},
1167 	{Opt_offusrjquota, "usrjquota="},
1168 	{Opt_usrjquota, "usrjquota=%s"},
1169 	{Opt_offgrpjquota, "grpjquota="},
1170 	{Opt_grpjquota, "grpjquota=%s"},
1171 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1172 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1173 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1174 	{Opt_grpquota, "grpquota"},
1175 	{Opt_noquota, "noquota"},
1176 	{Opt_quota, "quota"},
1177 	{Opt_usrquota, "usrquota"},
1178 	{Opt_barrier, "barrier=%u"},
1179 	{Opt_barrier, "barrier"},
1180 	{Opt_nobarrier, "nobarrier"},
1181 	{Opt_i_version, "i_version"},
1182 	{Opt_stripe, "stripe=%u"},
1183 	{Opt_delalloc, "delalloc"},
1184 	{Opt_nodelalloc, "nodelalloc"},
1185 	{Opt_removed, "mblk_io_submit"},
1186 	{Opt_removed, "nomblk_io_submit"},
1187 	{Opt_block_validity, "block_validity"},
1188 	{Opt_noblock_validity, "noblock_validity"},
1189 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1190 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1191 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1192 	{Opt_auto_da_alloc, "auto_da_alloc"},
1193 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1194 	{Opt_dioread_nolock, "dioread_nolock"},
1195 	{Opt_dioread_lock, "dioread_lock"},
1196 	{Opt_discard, "discard"},
1197 	{Opt_nodiscard, "nodiscard"},
1198 	{Opt_init_itable, "init_itable=%u"},
1199 	{Opt_init_itable, "init_itable"},
1200 	{Opt_noinit_itable, "noinit_itable"},
1201 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1202 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1203 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1204 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1205 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1206 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1207 	{Opt_err, NULL},
1208 };
1209 
1210 static ext4_fsblk_t get_sb_block(void **data)
1211 {
1212 	ext4_fsblk_t	sb_block;
1213 	char		*options = (char *) *data;
1214 
1215 	if (!options || strncmp(options, "sb=", 3) != 0)
1216 		return 1;	/* Default location */
1217 
1218 	options += 3;
1219 	/* TODO: use simple_strtoll with >32bit ext4 */
1220 	sb_block = simple_strtoul(options, &options, 0);
1221 	if (*options && *options != ',') {
1222 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1223 		       (char *) *data);
1224 		return 1;
1225 	}
1226 	if (*options == ',')
1227 		options++;
1228 	*data = (void *) options;
1229 
1230 	return sb_block;
1231 }
1232 
1233 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1234 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1235 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1236 
1237 #ifdef CONFIG_QUOTA
1238 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1239 {
1240 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1241 	char *qname;
1242 	int ret = -1;
1243 
1244 	if (sb_any_quota_loaded(sb) &&
1245 		!sbi->s_qf_names[qtype]) {
1246 		ext4_msg(sb, KERN_ERR,
1247 			"Cannot change journaled "
1248 			"quota options when quota turned on");
1249 		return -1;
1250 	}
1251 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1252 		ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1253 			 "when QUOTA feature is enabled");
1254 		return -1;
1255 	}
1256 	qname = match_strdup(args);
1257 	if (!qname) {
1258 		ext4_msg(sb, KERN_ERR,
1259 			"Not enough memory for storing quotafile name");
1260 		return -1;
1261 	}
1262 	if (sbi->s_qf_names[qtype]) {
1263 		if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1264 			ret = 1;
1265 		else
1266 			ext4_msg(sb, KERN_ERR,
1267 				 "%s quota file already specified",
1268 				 QTYPE2NAME(qtype));
1269 		goto errout;
1270 	}
1271 	if (strchr(qname, '/')) {
1272 		ext4_msg(sb, KERN_ERR,
1273 			"quotafile must be on filesystem root");
1274 		goto errout;
1275 	}
1276 	sbi->s_qf_names[qtype] = qname;
1277 	set_opt(sb, QUOTA);
1278 	return 1;
1279 errout:
1280 	kfree(qname);
1281 	return ret;
1282 }
1283 
1284 static int clear_qf_name(struct super_block *sb, int qtype)
1285 {
1286 
1287 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1288 
1289 	if (sb_any_quota_loaded(sb) &&
1290 		sbi->s_qf_names[qtype]) {
1291 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1292 			" when quota turned on");
1293 		return -1;
1294 	}
1295 	kfree(sbi->s_qf_names[qtype]);
1296 	sbi->s_qf_names[qtype] = NULL;
1297 	return 1;
1298 }
1299 #endif
1300 
1301 #define MOPT_SET	0x0001
1302 #define MOPT_CLEAR	0x0002
1303 #define MOPT_NOSUPPORT	0x0004
1304 #define MOPT_EXPLICIT	0x0008
1305 #define MOPT_CLEAR_ERR	0x0010
1306 #define MOPT_GTE0	0x0020
1307 #ifdef CONFIG_QUOTA
1308 #define MOPT_Q		0
1309 #define MOPT_QFMT	0x0040
1310 #else
1311 #define MOPT_Q		MOPT_NOSUPPORT
1312 #define MOPT_QFMT	MOPT_NOSUPPORT
1313 #endif
1314 #define MOPT_DATAJ	0x0080
1315 #define MOPT_NO_EXT2	0x0100
1316 #define MOPT_NO_EXT3	0x0200
1317 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1318 
1319 static const struct mount_opts {
1320 	int	token;
1321 	int	mount_opt;
1322 	int	flags;
1323 } ext4_mount_opts[] = {
1324 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1325 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1326 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1327 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1328 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1329 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1330 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1331 	 MOPT_EXT4_ONLY | MOPT_SET},
1332 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1333 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1334 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1335 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1336 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1337 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1338 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1339 	 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1340 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1341 	 MOPT_EXT4_ONLY | MOPT_SET},
1342 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1343 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1344 	 MOPT_EXT4_ONLY | MOPT_SET},
1345 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1346 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1347 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1348 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1349 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1350 	 MOPT_NO_EXT2 | MOPT_SET},
1351 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1352 	 MOPT_NO_EXT2 | MOPT_CLEAR},
1353 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1354 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1355 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1356 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1357 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1358 	{Opt_commit, 0, MOPT_GTE0},
1359 	{Opt_max_batch_time, 0, MOPT_GTE0},
1360 	{Opt_min_batch_time, 0, MOPT_GTE0},
1361 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1362 	{Opt_init_itable, 0, MOPT_GTE0},
1363 	{Opt_stripe, 0, MOPT_GTE0},
1364 	{Opt_resuid, 0, MOPT_GTE0},
1365 	{Opt_resgid, 0, MOPT_GTE0},
1366 	{Opt_journal_dev, 0, MOPT_GTE0},
1367 	{Opt_journal_ioprio, 0, MOPT_GTE0},
1368 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1369 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1370 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1371 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1372 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1373 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1374 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1375 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1376 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1377 #else
1378 	{Opt_acl, 0, MOPT_NOSUPPORT},
1379 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1380 #endif
1381 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1382 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1383 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1384 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1385 							MOPT_SET | MOPT_Q},
1386 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1387 							MOPT_SET | MOPT_Q},
1388 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1389 		       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1390 	{Opt_usrjquota, 0, MOPT_Q},
1391 	{Opt_grpjquota, 0, MOPT_Q},
1392 	{Opt_offusrjquota, 0, MOPT_Q},
1393 	{Opt_offgrpjquota, 0, MOPT_Q},
1394 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1395 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1396 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1397 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1398 	{Opt_err, 0, 0}
1399 };
1400 
1401 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1402 			    substring_t *args, unsigned long *journal_devnum,
1403 			    unsigned int *journal_ioprio, int is_remount)
1404 {
1405 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1406 	const struct mount_opts *m;
1407 	kuid_t uid;
1408 	kgid_t gid;
1409 	int arg = 0;
1410 
1411 #ifdef CONFIG_QUOTA
1412 	if (token == Opt_usrjquota)
1413 		return set_qf_name(sb, USRQUOTA, &args[0]);
1414 	else if (token == Opt_grpjquota)
1415 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1416 	else if (token == Opt_offusrjquota)
1417 		return clear_qf_name(sb, USRQUOTA);
1418 	else if (token == Opt_offgrpjquota)
1419 		return clear_qf_name(sb, GRPQUOTA);
1420 #endif
1421 	switch (token) {
1422 	case Opt_noacl:
1423 	case Opt_nouser_xattr:
1424 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1425 		break;
1426 	case Opt_sb:
1427 		return 1;	/* handled by get_sb_block() */
1428 	case Opt_removed:
1429 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1430 		return 1;
1431 	case Opt_abort:
1432 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1433 		return 1;
1434 	case Opt_i_version:
1435 		sb->s_flags |= MS_I_VERSION;
1436 		return 1;
1437 	}
1438 
1439 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1440 		if (token == m->token)
1441 			break;
1442 
1443 	if (m->token == Opt_err) {
1444 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1445 			 "or missing value", opt);
1446 		return -1;
1447 	}
1448 
1449 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1450 		ext4_msg(sb, KERN_ERR,
1451 			 "Mount option \"%s\" incompatible with ext2", opt);
1452 		return -1;
1453 	}
1454 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1455 		ext4_msg(sb, KERN_ERR,
1456 			 "Mount option \"%s\" incompatible with ext3", opt);
1457 		return -1;
1458 	}
1459 
1460 	if (args->from && match_int(args, &arg))
1461 		return -1;
1462 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1463 		return -1;
1464 	if (m->flags & MOPT_EXPLICIT)
1465 		set_opt2(sb, EXPLICIT_DELALLOC);
1466 	if (m->flags & MOPT_CLEAR_ERR)
1467 		clear_opt(sb, ERRORS_MASK);
1468 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1469 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1470 			 "options when quota turned on");
1471 		return -1;
1472 	}
1473 
1474 	if (m->flags & MOPT_NOSUPPORT) {
1475 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1476 	} else if (token == Opt_commit) {
1477 		if (arg == 0)
1478 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1479 		sbi->s_commit_interval = HZ * arg;
1480 	} else if (token == Opt_max_batch_time) {
1481 		if (arg == 0)
1482 			arg = EXT4_DEF_MAX_BATCH_TIME;
1483 		sbi->s_max_batch_time = arg;
1484 	} else if (token == Opt_min_batch_time) {
1485 		sbi->s_min_batch_time = arg;
1486 	} else if (token == Opt_inode_readahead_blks) {
1487 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1488 			ext4_msg(sb, KERN_ERR,
1489 				 "EXT4-fs: inode_readahead_blks must be "
1490 				 "0 or a power of 2 smaller than 2^31");
1491 			return -1;
1492 		}
1493 		sbi->s_inode_readahead_blks = arg;
1494 	} else if (token == Opt_init_itable) {
1495 		set_opt(sb, INIT_INODE_TABLE);
1496 		if (!args->from)
1497 			arg = EXT4_DEF_LI_WAIT_MULT;
1498 		sbi->s_li_wait_mult = arg;
1499 	} else if (token == Opt_max_dir_size_kb) {
1500 		sbi->s_max_dir_size_kb = arg;
1501 	} else if (token == Opt_stripe) {
1502 		sbi->s_stripe = arg;
1503 	} else if (token == Opt_resuid) {
1504 		uid = make_kuid(current_user_ns(), arg);
1505 		if (!uid_valid(uid)) {
1506 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1507 			return -1;
1508 		}
1509 		sbi->s_resuid = uid;
1510 	} else if (token == Opt_resgid) {
1511 		gid = make_kgid(current_user_ns(), arg);
1512 		if (!gid_valid(gid)) {
1513 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1514 			return -1;
1515 		}
1516 		sbi->s_resgid = gid;
1517 	} else if (token == Opt_journal_dev) {
1518 		if (is_remount) {
1519 			ext4_msg(sb, KERN_ERR,
1520 				 "Cannot specify journal on remount");
1521 			return -1;
1522 		}
1523 		*journal_devnum = arg;
1524 	} else if (token == Opt_journal_ioprio) {
1525 		if (arg > 7) {
1526 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1527 				 " (must be 0-7)");
1528 			return -1;
1529 		}
1530 		*journal_ioprio =
1531 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1532 	} else if (m->flags & MOPT_DATAJ) {
1533 		if (is_remount) {
1534 			if (!sbi->s_journal)
1535 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1536 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1537 				ext4_msg(sb, KERN_ERR,
1538 					 "Cannot change data mode on remount");
1539 				return -1;
1540 			}
1541 		} else {
1542 			clear_opt(sb, DATA_FLAGS);
1543 			sbi->s_mount_opt |= m->mount_opt;
1544 		}
1545 #ifdef CONFIG_QUOTA
1546 	} else if (m->flags & MOPT_QFMT) {
1547 		if (sb_any_quota_loaded(sb) &&
1548 		    sbi->s_jquota_fmt != m->mount_opt) {
1549 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1550 				 "quota options when quota turned on");
1551 			return -1;
1552 		}
1553 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1554 					       EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1555 			ext4_msg(sb, KERN_ERR,
1556 				 "Cannot set journaled quota options "
1557 				 "when QUOTA feature is enabled");
1558 			return -1;
1559 		}
1560 		sbi->s_jquota_fmt = m->mount_opt;
1561 #endif
1562 	} else {
1563 		if (!args->from)
1564 			arg = 1;
1565 		if (m->flags & MOPT_CLEAR)
1566 			arg = !arg;
1567 		else if (unlikely(!(m->flags & MOPT_SET))) {
1568 			ext4_msg(sb, KERN_WARNING,
1569 				 "buggy handling of option %s", opt);
1570 			WARN_ON(1);
1571 			return -1;
1572 		}
1573 		if (arg != 0)
1574 			sbi->s_mount_opt |= m->mount_opt;
1575 		else
1576 			sbi->s_mount_opt &= ~m->mount_opt;
1577 	}
1578 	return 1;
1579 }
1580 
1581 static int parse_options(char *options, struct super_block *sb,
1582 			 unsigned long *journal_devnum,
1583 			 unsigned int *journal_ioprio,
1584 			 int is_remount)
1585 {
1586 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1587 	char *p;
1588 	substring_t args[MAX_OPT_ARGS];
1589 	int token;
1590 
1591 	if (!options)
1592 		return 1;
1593 
1594 	while ((p = strsep(&options, ",")) != NULL) {
1595 		if (!*p)
1596 			continue;
1597 		/*
1598 		 * Initialize args struct so we know whether arg was
1599 		 * found; some options take optional arguments.
1600 		 */
1601 		args[0].to = args[0].from = NULL;
1602 		token = match_token(p, tokens, args);
1603 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1604 				     journal_ioprio, is_remount) < 0)
1605 			return 0;
1606 	}
1607 #ifdef CONFIG_QUOTA
1608 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1609 	    (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1610 		ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1611 			 "feature is enabled");
1612 		return 0;
1613 	}
1614 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1615 		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1616 			clear_opt(sb, USRQUOTA);
1617 
1618 		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1619 			clear_opt(sb, GRPQUOTA);
1620 
1621 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1622 			ext4_msg(sb, KERN_ERR, "old and new quota "
1623 					"format mixing");
1624 			return 0;
1625 		}
1626 
1627 		if (!sbi->s_jquota_fmt) {
1628 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1629 					"not specified");
1630 			return 0;
1631 		}
1632 	} else {
1633 		if (sbi->s_jquota_fmt) {
1634 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1635 					"specified with no journaling "
1636 					"enabled");
1637 			return 0;
1638 		}
1639 	}
1640 #endif
1641 	if (test_opt(sb, DIOREAD_NOLOCK)) {
1642 		int blocksize =
1643 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1644 
1645 		if (blocksize < PAGE_CACHE_SIZE) {
1646 			ext4_msg(sb, KERN_ERR, "can't mount with "
1647 				 "dioread_nolock if block size != PAGE_SIZE");
1648 			return 0;
1649 		}
1650 	}
1651 	return 1;
1652 }
1653 
1654 static inline void ext4_show_quota_options(struct seq_file *seq,
1655 					   struct super_block *sb)
1656 {
1657 #if defined(CONFIG_QUOTA)
1658 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1659 
1660 	if (sbi->s_jquota_fmt) {
1661 		char *fmtname = "";
1662 
1663 		switch (sbi->s_jquota_fmt) {
1664 		case QFMT_VFS_OLD:
1665 			fmtname = "vfsold";
1666 			break;
1667 		case QFMT_VFS_V0:
1668 			fmtname = "vfsv0";
1669 			break;
1670 		case QFMT_VFS_V1:
1671 			fmtname = "vfsv1";
1672 			break;
1673 		}
1674 		seq_printf(seq, ",jqfmt=%s", fmtname);
1675 	}
1676 
1677 	if (sbi->s_qf_names[USRQUOTA])
1678 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1679 
1680 	if (sbi->s_qf_names[GRPQUOTA])
1681 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1682 
1683 	if (test_opt(sb, USRQUOTA))
1684 		seq_puts(seq, ",usrquota");
1685 
1686 	if (test_opt(sb, GRPQUOTA))
1687 		seq_puts(seq, ",grpquota");
1688 #endif
1689 }
1690 
1691 static const char *token2str(int token)
1692 {
1693 	const struct match_token *t;
1694 
1695 	for (t = tokens; t->token != Opt_err; t++)
1696 		if (t->token == token && !strchr(t->pattern, '='))
1697 			break;
1698 	return t->pattern;
1699 }
1700 
1701 /*
1702  * Show an option if
1703  *  - it's set to a non-default value OR
1704  *  - if the per-sb default is different from the global default
1705  */
1706 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1707 			      int nodefs)
1708 {
1709 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1710 	struct ext4_super_block *es = sbi->s_es;
1711 	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1712 	const struct mount_opts *m;
1713 	char sep = nodefs ? '\n' : ',';
1714 
1715 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1716 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1717 
1718 	if (sbi->s_sb_block != 1)
1719 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1720 
1721 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1722 		int want_set = m->flags & MOPT_SET;
1723 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1724 		    (m->flags & MOPT_CLEAR_ERR))
1725 			continue;
1726 		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1727 			continue; /* skip if same as the default */
1728 		if ((want_set &&
1729 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1730 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1731 			continue; /* select Opt_noFoo vs Opt_Foo */
1732 		SEQ_OPTS_PRINT("%s", token2str(m->token));
1733 	}
1734 
1735 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1736 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1737 		SEQ_OPTS_PRINT("resuid=%u",
1738 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
1739 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1740 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1741 		SEQ_OPTS_PRINT("resgid=%u",
1742 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
1743 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1744 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1745 		SEQ_OPTS_PUTS("errors=remount-ro");
1746 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1747 		SEQ_OPTS_PUTS("errors=continue");
1748 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1749 		SEQ_OPTS_PUTS("errors=panic");
1750 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1751 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1752 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1753 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1754 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1755 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1756 	if (sb->s_flags & MS_I_VERSION)
1757 		SEQ_OPTS_PUTS("i_version");
1758 	if (nodefs || sbi->s_stripe)
1759 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1760 	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1761 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1762 			SEQ_OPTS_PUTS("data=journal");
1763 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1764 			SEQ_OPTS_PUTS("data=ordered");
1765 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1766 			SEQ_OPTS_PUTS("data=writeback");
1767 	}
1768 	if (nodefs ||
1769 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1770 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1771 			       sbi->s_inode_readahead_blks);
1772 
1773 	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1774 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1775 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1776 	if (nodefs || sbi->s_max_dir_size_kb)
1777 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1778 
1779 	ext4_show_quota_options(seq, sb);
1780 	return 0;
1781 }
1782 
1783 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1784 {
1785 	return _ext4_show_options(seq, root->d_sb, 0);
1786 }
1787 
1788 static int options_seq_show(struct seq_file *seq, void *offset)
1789 {
1790 	struct super_block *sb = seq->private;
1791 	int rc;
1792 
1793 	seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1794 	rc = _ext4_show_options(seq, sb, 1);
1795 	seq_puts(seq, "\n");
1796 	return rc;
1797 }
1798 
1799 static int options_open_fs(struct inode *inode, struct file *file)
1800 {
1801 	return single_open(file, options_seq_show, PDE(inode)->data);
1802 }
1803 
1804 static const struct file_operations ext4_seq_options_fops = {
1805 	.owner = THIS_MODULE,
1806 	.open = options_open_fs,
1807 	.read = seq_read,
1808 	.llseek = seq_lseek,
1809 	.release = single_release,
1810 };
1811 
1812 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1813 			    int read_only)
1814 {
1815 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1816 	int res = 0;
1817 
1818 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1819 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1820 			 "forcing read-only mode");
1821 		res = MS_RDONLY;
1822 	}
1823 	if (read_only)
1824 		goto done;
1825 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1826 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1827 			 "running e2fsck is recommended");
1828 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1829 		ext4_msg(sb, KERN_WARNING,
1830 			 "warning: mounting fs with errors, "
1831 			 "running e2fsck is recommended");
1832 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1833 		 le16_to_cpu(es->s_mnt_count) >=
1834 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1835 		ext4_msg(sb, KERN_WARNING,
1836 			 "warning: maximal mount count reached, "
1837 			 "running e2fsck is recommended");
1838 	else if (le32_to_cpu(es->s_checkinterval) &&
1839 		(le32_to_cpu(es->s_lastcheck) +
1840 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1841 		ext4_msg(sb, KERN_WARNING,
1842 			 "warning: checktime reached, "
1843 			 "running e2fsck is recommended");
1844 	if (!sbi->s_journal)
1845 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1846 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1847 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1848 	le16_add_cpu(&es->s_mnt_count, 1);
1849 	es->s_mtime = cpu_to_le32(get_seconds());
1850 	ext4_update_dynamic_rev(sb);
1851 	if (sbi->s_journal)
1852 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1853 
1854 	ext4_commit_super(sb, 1);
1855 done:
1856 	if (test_opt(sb, DEBUG))
1857 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1858 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1859 			sb->s_blocksize,
1860 			sbi->s_groups_count,
1861 			EXT4_BLOCKS_PER_GROUP(sb),
1862 			EXT4_INODES_PER_GROUP(sb),
1863 			sbi->s_mount_opt, sbi->s_mount_opt2);
1864 
1865 	cleancache_init_fs(sb);
1866 	return res;
1867 }
1868 
1869 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1870 {
1871 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1872 	struct flex_groups *new_groups;
1873 	int size;
1874 
1875 	if (!sbi->s_log_groups_per_flex)
1876 		return 0;
1877 
1878 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
1879 	if (size <= sbi->s_flex_groups_allocated)
1880 		return 0;
1881 
1882 	size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1883 	new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1884 	if (!new_groups) {
1885 		ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1886 			 size / (int) sizeof(struct flex_groups));
1887 		return -ENOMEM;
1888 	}
1889 
1890 	if (sbi->s_flex_groups) {
1891 		memcpy(new_groups, sbi->s_flex_groups,
1892 		       (sbi->s_flex_groups_allocated *
1893 			sizeof(struct flex_groups)));
1894 		ext4_kvfree(sbi->s_flex_groups);
1895 	}
1896 	sbi->s_flex_groups = new_groups;
1897 	sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1898 	return 0;
1899 }
1900 
1901 static int ext4_fill_flex_info(struct super_block *sb)
1902 {
1903 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1904 	struct ext4_group_desc *gdp = NULL;
1905 	ext4_group_t flex_group;
1906 	unsigned int groups_per_flex = 0;
1907 	int i, err;
1908 
1909 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1910 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1911 		sbi->s_log_groups_per_flex = 0;
1912 		return 1;
1913 	}
1914 	groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1915 
1916 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1917 	if (err)
1918 		goto failed;
1919 
1920 	for (i = 0; i < sbi->s_groups_count; i++) {
1921 		gdp = ext4_get_group_desc(sb, i, NULL);
1922 
1923 		flex_group = ext4_flex_group(sbi, i);
1924 		atomic_add(ext4_free_inodes_count(sb, gdp),
1925 			   &sbi->s_flex_groups[flex_group].free_inodes);
1926 		atomic_add(ext4_free_group_clusters(sb, gdp),
1927 			   &sbi->s_flex_groups[flex_group].free_clusters);
1928 		atomic_add(ext4_used_dirs_count(sb, gdp),
1929 			   &sbi->s_flex_groups[flex_group].used_dirs);
1930 	}
1931 
1932 	return 1;
1933 failed:
1934 	return 0;
1935 }
1936 
1937 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1938 				   struct ext4_group_desc *gdp)
1939 {
1940 	int offset;
1941 	__u16 crc = 0;
1942 	__le32 le_group = cpu_to_le32(block_group);
1943 
1944 	if ((sbi->s_es->s_feature_ro_compat &
1945 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1946 		/* Use new metadata_csum algorithm */
1947 		__u16 old_csum;
1948 		__u32 csum32;
1949 
1950 		old_csum = gdp->bg_checksum;
1951 		gdp->bg_checksum = 0;
1952 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1953 				     sizeof(le_group));
1954 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1955 				     sbi->s_desc_size);
1956 		gdp->bg_checksum = old_csum;
1957 
1958 		crc = csum32 & 0xFFFF;
1959 		goto out;
1960 	}
1961 
1962 	/* old crc16 code */
1963 	offset = offsetof(struct ext4_group_desc, bg_checksum);
1964 
1965 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1966 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1967 	crc = crc16(crc, (__u8 *)gdp, offset);
1968 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
1969 	/* for checksum of struct ext4_group_desc do the rest...*/
1970 	if ((sbi->s_es->s_feature_incompat &
1971 	     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1972 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1973 		crc = crc16(crc, (__u8 *)gdp + offset,
1974 			    le16_to_cpu(sbi->s_es->s_desc_size) -
1975 				offset);
1976 
1977 out:
1978 	return cpu_to_le16(crc);
1979 }
1980 
1981 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1982 				struct ext4_group_desc *gdp)
1983 {
1984 	if (ext4_has_group_desc_csum(sb) &&
1985 	    (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1986 						      block_group, gdp)))
1987 		return 0;
1988 
1989 	return 1;
1990 }
1991 
1992 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1993 			      struct ext4_group_desc *gdp)
1994 {
1995 	if (!ext4_has_group_desc_csum(sb))
1996 		return;
1997 	gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
1998 }
1999 
2000 /* Called at mount-time, super-block is locked */
2001 static int ext4_check_descriptors(struct super_block *sb,
2002 				  ext4_group_t *first_not_zeroed)
2003 {
2004 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2005 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2006 	ext4_fsblk_t last_block;
2007 	ext4_fsblk_t block_bitmap;
2008 	ext4_fsblk_t inode_bitmap;
2009 	ext4_fsblk_t inode_table;
2010 	int flexbg_flag = 0;
2011 	ext4_group_t i, grp = sbi->s_groups_count;
2012 
2013 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2014 		flexbg_flag = 1;
2015 
2016 	ext4_debug("Checking group descriptors");
2017 
2018 	for (i = 0; i < sbi->s_groups_count; i++) {
2019 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2020 
2021 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2022 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2023 		else
2024 			last_block = first_block +
2025 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2026 
2027 		if ((grp == sbi->s_groups_count) &&
2028 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2029 			grp = i;
2030 
2031 		block_bitmap = ext4_block_bitmap(sb, gdp);
2032 		if (block_bitmap < first_block || block_bitmap > last_block) {
2033 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2034 			       "Block bitmap for group %u not in group "
2035 			       "(block %llu)!", i, block_bitmap);
2036 			return 0;
2037 		}
2038 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2039 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2040 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2041 			       "Inode bitmap for group %u not in group "
2042 			       "(block %llu)!", i, inode_bitmap);
2043 			return 0;
2044 		}
2045 		inode_table = ext4_inode_table(sb, gdp);
2046 		if (inode_table < first_block ||
2047 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2048 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2049 			       "Inode table for group %u not in group "
2050 			       "(block %llu)!", i, inode_table);
2051 			return 0;
2052 		}
2053 		ext4_lock_group(sb, i);
2054 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2055 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2056 				 "Checksum for group %u failed (%u!=%u)",
2057 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2058 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2059 			if (!(sb->s_flags & MS_RDONLY)) {
2060 				ext4_unlock_group(sb, i);
2061 				return 0;
2062 			}
2063 		}
2064 		ext4_unlock_group(sb, i);
2065 		if (!flexbg_flag)
2066 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2067 	}
2068 	if (NULL != first_not_zeroed)
2069 		*first_not_zeroed = grp;
2070 
2071 	ext4_free_blocks_count_set(sbi->s_es,
2072 				   EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2073 	sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2074 	return 1;
2075 }
2076 
2077 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2078  * the superblock) which were deleted from all directories, but held open by
2079  * a process at the time of a crash.  We walk the list and try to delete these
2080  * inodes at recovery time (only with a read-write filesystem).
2081  *
2082  * In order to keep the orphan inode chain consistent during traversal (in
2083  * case of crash during recovery), we link each inode into the superblock
2084  * orphan list_head and handle it the same way as an inode deletion during
2085  * normal operation (which journals the operations for us).
2086  *
2087  * We only do an iget() and an iput() on each inode, which is very safe if we
2088  * accidentally point at an in-use or already deleted inode.  The worst that
2089  * can happen in this case is that we get a "bit already cleared" message from
2090  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2091  * e2fsck was run on this filesystem, and it must have already done the orphan
2092  * inode cleanup for us, so we can safely abort without any further action.
2093  */
2094 static void ext4_orphan_cleanup(struct super_block *sb,
2095 				struct ext4_super_block *es)
2096 {
2097 	unsigned int s_flags = sb->s_flags;
2098 	int nr_orphans = 0, nr_truncates = 0;
2099 #ifdef CONFIG_QUOTA
2100 	int i;
2101 #endif
2102 	if (!es->s_last_orphan) {
2103 		jbd_debug(4, "no orphan inodes to clean up\n");
2104 		return;
2105 	}
2106 
2107 	if (bdev_read_only(sb->s_bdev)) {
2108 		ext4_msg(sb, KERN_ERR, "write access "
2109 			"unavailable, skipping orphan cleanup");
2110 		return;
2111 	}
2112 
2113 	/* Check if feature set would not allow a r/w mount */
2114 	if (!ext4_feature_set_ok(sb, 0)) {
2115 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2116 			 "unknown ROCOMPAT features");
2117 		return;
2118 	}
2119 
2120 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2121 		/* don't clear list on RO mount w/ errors */
2122 		if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2123 			jbd_debug(1, "Errors on filesystem, "
2124 				  "clearing orphan list.\n");
2125 			es->s_last_orphan = 0;
2126 		}
2127 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2128 		return;
2129 	}
2130 
2131 	if (s_flags & MS_RDONLY) {
2132 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2133 		sb->s_flags &= ~MS_RDONLY;
2134 	}
2135 #ifdef CONFIG_QUOTA
2136 	/* Needed for iput() to work correctly and not trash data */
2137 	sb->s_flags |= MS_ACTIVE;
2138 	/* Turn on quotas so that they are updated correctly */
2139 	for (i = 0; i < MAXQUOTAS; i++) {
2140 		if (EXT4_SB(sb)->s_qf_names[i]) {
2141 			int ret = ext4_quota_on_mount(sb, i);
2142 			if (ret < 0)
2143 				ext4_msg(sb, KERN_ERR,
2144 					"Cannot turn on journaled "
2145 					"quota: error %d", ret);
2146 		}
2147 	}
2148 #endif
2149 
2150 	while (es->s_last_orphan) {
2151 		struct inode *inode;
2152 
2153 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2154 		if (IS_ERR(inode)) {
2155 			es->s_last_orphan = 0;
2156 			break;
2157 		}
2158 
2159 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2160 		dquot_initialize(inode);
2161 		if (inode->i_nlink) {
2162 			ext4_msg(sb, KERN_DEBUG,
2163 				"%s: truncating inode %lu to %lld bytes",
2164 				__func__, inode->i_ino, inode->i_size);
2165 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2166 				  inode->i_ino, inode->i_size);
2167 			mutex_lock(&inode->i_mutex);
2168 			ext4_truncate(inode);
2169 			mutex_unlock(&inode->i_mutex);
2170 			nr_truncates++;
2171 		} else {
2172 			ext4_msg(sb, KERN_DEBUG,
2173 				"%s: deleting unreferenced inode %lu",
2174 				__func__, inode->i_ino);
2175 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2176 				  inode->i_ino);
2177 			nr_orphans++;
2178 		}
2179 		iput(inode);  /* The delete magic happens here! */
2180 	}
2181 
2182 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2183 
2184 	if (nr_orphans)
2185 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2186 		       PLURAL(nr_orphans));
2187 	if (nr_truncates)
2188 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2189 		       PLURAL(nr_truncates));
2190 #ifdef CONFIG_QUOTA
2191 	/* Turn quotas off */
2192 	for (i = 0; i < MAXQUOTAS; i++) {
2193 		if (sb_dqopt(sb)->files[i])
2194 			dquot_quota_off(sb, i);
2195 	}
2196 #endif
2197 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2198 }
2199 
2200 /*
2201  * Maximal extent format file size.
2202  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2203  * extent format containers, within a sector_t, and within i_blocks
2204  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2205  * so that won't be a limiting factor.
2206  *
2207  * However there is other limiting factor. We do store extents in the form
2208  * of starting block and length, hence the resulting length of the extent
2209  * covering maximum file size must fit into on-disk format containers as
2210  * well. Given that length is always by 1 unit bigger than max unit (because
2211  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2212  *
2213  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2214  */
2215 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2216 {
2217 	loff_t res;
2218 	loff_t upper_limit = MAX_LFS_FILESIZE;
2219 
2220 	/* small i_blocks in vfs inode? */
2221 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2222 		/*
2223 		 * CONFIG_LBDAF is not enabled implies the inode
2224 		 * i_block represent total blocks in 512 bytes
2225 		 * 32 == size of vfs inode i_blocks * 8
2226 		 */
2227 		upper_limit = (1LL << 32) - 1;
2228 
2229 		/* total blocks in file system block size */
2230 		upper_limit >>= (blkbits - 9);
2231 		upper_limit <<= blkbits;
2232 	}
2233 
2234 	/*
2235 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2236 	 * by one fs block, so ee_len can cover the extent of maximum file
2237 	 * size
2238 	 */
2239 	res = (1LL << 32) - 1;
2240 	res <<= blkbits;
2241 
2242 	/* Sanity check against vm- & vfs- imposed limits */
2243 	if (res > upper_limit)
2244 		res = upper_limit;
2245 
2246 	return res;
2247 }
2248 
2249 /*
2250  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2251  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2252  * We need to be 1 filesystem block less than the 2^48 sector limit.
2253  */
2254 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2255 {
2256 	loff_t res = EXT4_NDIR_BLOCKS;
2257 	int meta_blocks;
2258 	loff_t upper_limit;
2259 	/* This is calculated to be the largest file size for a dense, block
2260 	 * mapped file such that the file's total number of 512-byte sectors,
2261 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2262 	 *
2263 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2264 	 * number of 512-byte sectors of the file.
2265 	 */
2266 
2267 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2268 		/*
2269 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2270 		 * the inode i_block field represents total file blocks in
2271 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2272 		 */
2273 		upper_limit = (1LL << 32) - 1;
2274 
2275 		/* total blocks in file system block size */
2276 		upper_limit >>= (bits - 9);
2277 
2278 	} else {
2279 		/*
2280 		 * We use 48 bit ext4_inode i_blocks
2281 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2282 		 * represent total number of blocks in
2283 		 * file system block size
2284 		 */
2285 		upper_limit = (1LL << 48) - 1;
2286 
2287 	}
2288 
2289 	/* indirect blocks */
2290 	meta_blocks = 1;
2291 	/* double indirect blocks */
2292 	meta_blocks += 1 + (1LL << (bits-2));
2293 	/* tripple indirect blocks */
2294 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2295 
2296 	upper_limit -= meta_blocks;
2297 	upper_limit <<= bits;
2298 
2299 	res += 1LL << (bits-2);
2300 	res += 1LL << (2*(bits-2));
2301 	res += 1LL << (3*(bits-2));
2302 	res <<= bits;
2303 	if (res > upper_limit)
2304 		res = upper_limit;
2305 
2306 	if (res > MAX_LFS_FILESIZE)
2307 		res = MAX_LFS_FILESIZE;
2308 
2309 	return res;
2310 }
2311 
2312 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2313 				   ext4_fsblk_t logical_sb_block, int nr)
2314 {
2315 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2316 	ext4_group_t bg, first_meta_bg;
2317 	int has_super = 0;
2318 
2319 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2320 
2321 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2322 	    nr < first_meta_bg)
2323 		return logical_sb_block + nr + 1;
2324 	bg = sbi->s_desc_per_block * nr;
2325 	if (ext4_bg_has_super(sb, bg))
2326 		has_super = 1;
2327 
2328 	return (has_super + ext4_group_first_block_no(sb, bg));
2329 }
2330 
2331 /**
2332  * ext4_get_stripe_size: Get the stripe size.
2333  * @sbi: In memory super block info
2334  *
2335  * If we have specified it via mount option, then
2336  * use the mount option value. If the value specified at mount time is
2337  * greater than the blocks per group use the super block value.
2338  * If the super block value is greater than blocks per group return 0.
2339  * Allocator needs it be less than blocks per group.
2340  *
2341  */
2342 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2343 {
2344 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2345 	unsigned long stripe_width =
2346 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2347 	int ret;
2348 
2349 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2350 		ret = sbi->s_stripe;
2351 	else if (stripe_width <= sbi->s_blocks_per_group)
2352 		ret = stripe_width;
2353 	else if (stride <= sbi->s_blocks_per_group)
2354 		ret = stride;
2355 	else
2356 		ret = 0;
2357 
2358 	/*
2359 	 * If the stripe width is 1, this makes no sense and
2360 	 * we set it to 0 to turn off stripe handling code.
2361 	 */
2362 	if (ret <= 1)
2363 		ret = 0;
2364 
2365 	return ret;
2366 }
2367 
2368 /* sysfs supprt */
2369 
2370 struct ext4_attr {
2371 	struct attribute attr;
2372 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2373 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2374 			 const char *, size_t);
2375 	int offset;
2376 };
2377 
2378 static int parse_strtoul(const char *buf,
2379 		unsigned long max, unsigned long *value)
2380 {
2381 	char *endp;
2382 
2383 	*value = simple_strtoul(skip_spaces(buf), &endp, 0);
2384 	endp = skip_spaces(endp);
2385 	if (*endp || *value > max)
2386 		return -EINVAL;
2387 
2388 	return 0;
2389 }
2390 
2391 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2392 					      struct ext4_sb_info *sbi,
2393 					      char *buf)
2394 {
2395 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2396 		(s64) EXT4_C2B(sbi,
2397 			percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2398 }
2399 
2400 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2401 					 struct ext4_sb_info *sbi, char *buf)
2402 {
2403 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2404 
2405 	if (!sb->s_bdev->bd_part)
2406 		return snprintf(buf, PAGE_SIZE, "0\n");
2407 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2408 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2409 			 sbi->s_sectors_written_start) >> 1);
2410 }
2411 
2412 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2413 					  struct ext4_sb_info *sbi, char *buf)
2414 {
2415 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2416 
2417 	if (!sb->s_bdev->bd_part)
2418 		return snprintf(buf, PAGE_SIZE, "0\n");
2419 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2420 			(unsigned long long)(sbi->s_kbytes_written +
2421 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2422 			  EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2423 }
2424 
2425 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2426 					  struct ext4_sb_info *sbi,
2427 					  const char *buf, size_t count)
2428 {
2429 	unsigned long t;
2430 
2431 	if (parse_strtoul(buf, 0x40000000, &t))
2432 		return -EINVAL;
2433 
2434 	if (t && !is_power_of_2(t))
2435 		return -EINVAL;
2436 
2437 	sbi->s_inode_readahead_blks = t;
2438 	return count;
2439 }
2440 
2441 static ssize_t sbi_ui_show(struct ext4_attr *a,
2442 			   struct ext4_sb_info *sbi, char *buf)
2443 {
2444 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2445 
2446 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2447 }
2448 
2449 static ssize_t sbi_ui_store(struct ext4_attr *a,
2450 			    struct ext4_sb_info *sbi,
2451 			    const char *buf, size_t count)
2452 {
2453 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2454 	unsigned long t;
2455 
2456 	if (parse_strtoul(buf, 0xffffffff, &t))
2457 		return -EINVAL;
2458 	*ui = t;
2459 	return count;
2460 }
2461 
2462 static ssize_t trigger_test_error(struct ext4_attr *a,
2463 				  struct ext4_sb_info *sbi,
2464 				  const char *buf, size_t count)
2465 {
2466 	int len = count;
2467 
2468 	if (!capable(CAP_SYS_ADMIN))
2469 		return -EPERM;
2470 
2471 	if (len && buf[len-1] == '\n')
2472 		len--;
2473 
2474 	if (len)
2475 		ext4_error(sbi->s_sb, "%.*s", len, buf);
2476 	return count;
2477 }
2478 
2479 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2480 static struct ext4_attr ext4_attr_##_name = {			\
2481 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2482 	.show	= _show,					\
2483 	.store	= _store,					\
2484 	.offset = offsetof(struct ext4_sb_info, _elname),	\
2485 }
2486 #define EXT4_ATTR(name, mode, show, store) \
2487 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2488 
2489 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2490 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2491 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2492 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2493 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2494 #define ATTR_LIST(name) &ext4_attr_##name.attr
2495 
2496 EXT4_RO_ATTR(delayed_allocation_blocks);
2497 EXT4_RO_ATTR(session_write_kbytes);
2498 EXT4_RO_ATTR(lifetime_write_kbytes);
2499 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2500 		 inode_readahead_blks_store, s_inode_readahead_blks);
2501 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2502 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2503 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2504 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2505 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2506 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2507 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2508 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2509 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2510 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2511 
2512 static struct attribute *ext4_attrs[] = {
2513 	ATTR_LIST(delayed_allocation_blocks),
2514 	ATTR_LIST(session_write_kbytes),
2515 	ATTR_LIST(lifetime_write_kbytes),
2516 	ATTR_LIST(inode_readahead_blks),
2517 	ATTR_LIST(inode_goal),
2518 	ATTR_LIST(mb_stats),
2519 	ATTR_LIST(mb_max_to_scan),
2520 	ATTR_LIST(mb_min_to_scan),
2521 	ATTR_LIST(mb_order2_req),
2522 	ATTR_LIST(mb_stream_req),
2523 	ATTR_LIST(mb_group_prealloc),
2524 	ATTR_LIST(max_writeback_mb_bump),
2525 	ATTR_LIST(extent_max_zeroout_kb),
2526 	ATTR_LIST(trigger_fs_error),
2527 	NULL,
2528 };
2529 
2530 /* Features this copy of ext4 supports */
2531 EXT4_INFO_ATTR(lazy_itable_init);
2532 EXT4_INFO_ATTR(batched_discard);
2533 EXT4_INFO_ATTR(meta_bg_resize);
2534 
2535 static struct attribute *ext4_feat_attrs[] = {
2536 	ATTR_LIST(lazy_itable_init),
2537 	ATTR_LIST(batched_discard),
2538 	ATTR_LIST(meta_bg_resize),
2539 	NULL,
2540 };
2541 
2542 static ssize_t ext4_attr_show(struct kobject *kobj,
2543 			      struct attribute *attr, char *buf)
2544 {
2545 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2546 						s_kobj);
2547 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2548 
2549 	return a->show ? a->show(a, sbi, buf) : 0;
2550 }
2551 
2552 static ssize_t ext4_attr_store(struct kobject *kobj,
2553 			       struct attribute *attr,
2554 			       const char *buf, size_t len)
2555 {
2556 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2557 						s_kobj);
2558 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2559 
2560 	return a->store ? a->store(a, sbi, buf, len) : 0;
2561 }
2562 
2563 static void ext4_sb_release(struct kobject *kobj)
2564 {
2565 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2566 						s_kobj);
2567 	complete(&sbi->s_kobj_unregister);
2568 }
2569 
2570 static const struct sysfs_ops ext4_attr_ops = {
2571 	.show	= ext4_attr_show,
2572 	.store	= ext4_attr_store,
2573 };
2574 
2575 static struct kobj_type ext4_ktype = {
2576 	.default_attrs	= ext4_attrs,
2577 	.sysfs_ops	= &ext4_attr_ops,
2578 	.release	= ext4_sb_release,
2579 };
2580 
2581 static void ext4_feat_release(struct kobject *kobj)
2582 {
2583 	complete(&ext4_feat->f_kobj_unregister);
2584 }
2585 
2586 static struct kobj_type ext4_feat_ktype = {
2587 	.default_attrs	= ext4_feat_attrs,
2588 	.sysfs_ops	= &ext4_attr_ops,
2589 	.release	= ext4_feat_release,
2590 };
2591 
2592 /*
2593  * Check whether this filesystem can be mounted based on
2594  * the features present and the RDONLY/RDWR mount requested.
2595  * Returns 1 if this filesystem can be mounted as requested,
2596  * 0 if it cannot be.
2597  */
2598 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2599 {
2600 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2601 		ext4_msg(sb, KERN_ERR,
2602 			"Couldn't mount because of "
2603 			"unsupported optional features (%x)",
2604 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2605 			~EXT4_FEATURE_INCOMPAT_SUPP));
2606 		return 0;
2607 	}
2608 
2609 	if (readonly)
2610 		return 1;
2611 
2612 	/* Check that feature set is OK for a read-write mount */
2613 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2614 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2615 			 "unsupported optional features (%x)",
2616 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2617 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2618 		return 0;
2619 	}
2620 	/*
2621 	 * Large file size enabled file system can only be mounted
2622 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2623 	 */
2624 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2625 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2626 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2627 				 "cannot be mounted RDWR without "
2628 				 "CONFIG_LBDAF");
2629 			return 0;
2630 		}
2631 	}
2632 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2633 	    !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2634 		ext4_msg(sb, KERN_ERR,
2635 			 "Can't support bigalloc feature without "
2636 			 "extents feature\n");
2637 		return 0;
2638 	}
2639 
2640 #ifndef CONFIG_QUOTA
2641 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2642 	    !readonly) {
2643 		ext4_msg(sb, KERN_ERR,
2644 			 "Filesystem with quota feature cannot be mounted RDWR "
2645 			 "without CONFIG_QUOTA");
2646 		return 0;
2647 	}
2648 #endif  /* CONFIG_QUOTA */
2649 	return 1;
2650 }
2651 
2652 /*
2653  * This function is called once a day if we have errors logged
2654  * on the file system
2655  */
2656 static void print_daily_error_info(unsigned long arg)
2657 {
2658 	struct super_block *sb = (struct super_block *) arg;
2659 	struct ext4_sb_info *sbi;
2660 	struct ext4_super_block *es;
2661 
2662 	sbi = EXT4_SB(sb);
2663 	es = sbi->s_es;
2664 
2665 	if (es->s_error_count)
2666 		ext4_msg(sb, KERN_NOTICE, "error count: %u",
2667 			 le32_to_cpu(es->s_error_count));
2668 	if (es->s_first_error_time) {
2669 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2670 		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2671 		       (int) sizeof(es->s_first_error_func),
2672 		       es->s_first_error_func,
2673 		       le32_to_cpu(es->s_first_error_line));
2674 		if (es->s_first_error_ino)
2675 			printk(": inode %u",
2676 			       le32_to_cpu(es->s_first_error_ino));
2677 		if (es->s_first_error_block)
2678 			printk(": block %llu", (unsigned long long)
2679 			       le64_to_cpu(es->s_first_error_block));
2680 		printk("\n");
2681 	}
2682 	if (es->s_last_error_time) {
2683 		printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2684 		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2685 		       (int) sizeof(es->s_last_error_func),
2686 		       es->s_last_error_func,
2687 		       le32_to_cpu(es->s_last_error_line));
2688 		if (es->s_last_error_ino)
2689 			printk(": inode %u",
2690 			       le32_to_cpu(es->s_last_error_ino));
2691 		if (es->s_last_error_block)
2692 			printk(": block %llu", (unsigned long long)
2693 			       le64_to_cpu(es->s_last_error_block));
2694 		printk("\n");
2695 	}
2696 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2697 }
2698 
2699 /* Find next suitable group and run ext4_init_inode_table */
2700 static int ext4_run_li_request(struct ext4_li_request *elr)
2701 {
2702 	struct ext4_group_desc *gdp = NULL;
2703 	ext4_group_t group, ngroups;
2704 	struct super_block *sb;
2705 	unsigned long timeout = 0;
2706 	int ret = 0;
2707 
2708 	sb = elr->lr_super;
2709 	ngroups = EXT4_SB(sb)->s_groups_count;
2710 
2711 	sb_start_write(sb);
2712 	for (group = elr->lr_next_group; group < ngroups; group++) {
2713 		gdp = ext4_get_group_desc(sb, group, NULL);
2714 		if (!gdp) {
2715 			ret = 1;
2716 			break;
2717 		}
2718 
2719 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2720 			break;
2721 	}
2722 
2723 	if (group >= ngroups)
2724 		ret = 1;
2725 
2726 	if (!ret) {
2727 		timeout = jiffies;
2728 		ret = ext4_init_inode_table(sb, group,
2729 					    elr->lr_timeout ? 0 : 1);
2730 		if (elr->lr_timeout == 0) {
2731 			timeout = (jiffies - timeout) *
2732 				  elr->lr_sbi->s_li_wait_mult;
2733 			elr->lr_timeout = timeout;
2734 		}
2735 		elr->lr_next_sched = jiffies + elr->lr_timeout;
2736 		elr->lr_next_group = group + 1;
2737 	}
2738 	sb_end_write(sb);
2739 
2740 	return ret;
2741 }
2742 
2743 /*
2744  * Remove lr_request from the list_request and free the
2745  * request structure. Should be called with li_list_mtx held
2746  */
2747 static void ext4_remove_li_request(struct ext4_li_request *elr)
2748 {
2749 	struct ext4_sb_info *sbi;
2750 
2751 	if (!elr)
2752 		return;
2753 
2754 	sbi = elr->lr_sbi;
2755 
2756 	list_del(&elr->lr_request);
2757 	sbi->s_li_request = NULL;
2758 	kfree(elr);
2759 }
2760 
2761 static void ext4_unregister_li_request(struct super_block *sb)
2762 {
2763 	mutex_lock(&ext4_li_mtx);
2764 	if (!ext4_li_info) {
2765 		mutex_unlock(&ext4_li_mtx);
2766 		return;
2767 	}
2768 
2769 	mutex_lock(&ext4_li_info->li_list_mtx);
2770 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2771 	mutex_unlock(&ext4_li_info->li_list_mtx);
2772 	mutex_unlock(&ext4_li_mtx);
2773 }
2774 
2775 static struct task_struct *ext4_lazyinit_task;
2776 
2777 /*
2778  * This is the function where ext4lazyinit thread lives. It walks
2779  * through the request list searching for next scheduled filesystem.
2780  * When such a fs is found, run the lazy initialization request
2781  * (ext4_rn_li_request) and keep track of the time spend in this
2782  * function. Based on that time we compute next schedule time of
2783  * the request. When walking through the list is complete, compute
2784  * next waking time and put itself into sleep.
2785  */
2786 static int ext4_lazyinit_thread(void *arg)
2787 {
2788 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2789 	struct list_head *pos, *n;
2790 	struct ext4_li_request *elr;
2791 	unsigned long next_wakeup, cur;
2792 
2793 	BUG_ON(NULL == eli);
2794 
2795 cont_thread:
2796 	while (true) {
2797 		next_wakeup = MAX_JIFFY_OFFSET;
2798 
2799 		mutex_lock(&eli->li_list_mtx);
2800 		if (list_empty(&eli->li_request_list)) {
2801 			mutex_unlock(&eli->li_list_mtx);
2802 			goto exit_thread;
2803 		}
2804 
2805 		list_for_each_safe(pos, n, &eli->li_request_list) {
2806 			elr = list_entry(pos, struct ext4_li_request,
2807 					 lr_request);
2808 
2809 			if (time_after_eq(jiffies, elr->lr_next_sched)) {
2810 				if (ext4_run_li_request(elr) != 0) {
2811 					/* error, remove the lazy_init job */
2812 					ext4_remove_li_request(elr);
2813 					continue;
2814 				}
2815 			}
2816 
2817 			if (time_before(elr->lr_next_sched, next_wakeup))
2818 				next_wakeup = elr->lr_next_sched;
2819 		}
2820 		mutex_unlock(&eli->li_list_mtx);
2821 
2822 		try_to_freeze();
2823 
2824 		cur = jiffies;
2825 		if ((time_after_eq(cur, next_wakeup)) ||
2826 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
2827 			cond_resched();
2828 			continue;
2829 		}
2830 
2831 		schedule_timeout_interruptible(next_wakeup - cur);
2832 
2833 		if (kthread_should_stop()) {
2834 			ext4_clear_request_list();
2835 			goto exit_thread;
2836 		}
2837 	}
2838 
2839 exit_thread:
2840 	/*
2841 	 * It looks like the request list is empty, but we need
2842 	 * to check it under the li_list_mtx lock, to prevent any
2843 	 * additions into it, and of course we should lock ext4_li_mtx
2844 	 * to atomically free the list and ext4_li_info, because at
2845 	 * this point another ext4 filesystem could be registering
2846 	 * new one.
2847 	 */
2848 	mutex_lock(&ext4_li_mtx);
2849 	mutex_lock(&eli->li_list_mtx);
2850 	if (!list_empty(&eli->li_request_list)) {
2851 		mutex_unlock(&eli->li_list_mtx);
2852 		mutex_unlock(&ext4_li_mtx);
2853 		goto cont_thread;
2854 	}
2855 	mutex_unlock(&eli->li_list_mtx);
2856 	kfree(ext4_li_info);
2857 	ext4_li_info = NULL;
2858 	mutex_unlock(&ext4_li_mtx);
2859 
2860 	return 0;
2861 }
2862 
2863 static void ext4_clear_request_list(void)
2864 {
2865 	struct list_head *pos, *n;
2866 	struct ext4_li_request *elr;
2867 
2868 	mutex_lock(&ext4_li_info->li_list_mtx);
2869 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2870 		elr = list_entry(pos, struct ext4_li_request,
2871 				 lr_request);
2872 		ext4_remove_li_request(elr);
2873 	}
2874 	mutex_unlock(&ext4_li_info->li_list_mtx);
2875 }
2876 
2877 static int ext4_run_lazyinit_thread(void)
2878 {
2879 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2880 					 ext4_li_info, "ext4lazyinit");
2881 	if (IS_ERR(ext4_lazyinit_task)) {
2882 		int err = PTR_ERR(ext4_lazyinit_task);
2883 		ext4_clear_request_list();
2884 		kfree(ext4_li_info);
2885 		ext4_li_info = NULL;
2886 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2887 				 "initialization thread\n",
2888 				 err);
2889 		return err;
2890 	}
2891 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2892 	return 0;
2893 }
2894 
2895 /*
2896  * Check whether it make sense to run itable init. thread or not.
2897  * If there is at least one uninitialized inode table, return
2898  * corresponding group number, else the loop goes through all
2899  * groups and return total number of groups.
2900  */
2901 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2902 {
2903 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2904 	struct ext4_group_desc *gdp = NULL;
2905 
2906 	for (group = 0; group < ngroups; group++) {
2907 		gdp = ext4_get_group_desc(sb, group, NULL);
2908 		if (!gdp)
2909 			continue;
2910 
2911 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2912 			break;
2913 	}
2914 
2915 	return group;
2916 }
2917 
2918 static int ext4_li_info_new(void)
2919 {
2920 	struct ext4_lazy_init *eli = NULL;
2921 
2922 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2923 	if (!eli)
2924 		return -ENOMEM;
2925 
2926 	INIT_LIST_HEAD(&eli->li_request_list);
2927 	mutex_init(&eli->li_list_mtx);
2928 
2929 	eli->li_state |= EXT4_LAZYINIT_QUIT;
2930 
2931 	ext4_li_info = eli;
2932 
2933 	return 0;
2934 }
2935 
2936 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2937 					    ext4_group_t start)
2938 {
2939 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2940 	struct ext4_li_request *elr;
2941 	unsigned long rnd;
2942 
2943 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2944 	if (!elr)
2945 		return NULL;
2946 
2947 	elr->lr_super = sb;
2948 	elr->lr_sbi = sbi;
2949 	elr->lr_next_group = start;
2950 
2951 	/*
2952 	 * Randomize first schedule time of the request to
2953 	 * spread the inode table initialization requests
2954 	 * better.
2955 	 */
2956 	get_random_bytes(&rnd, sizeof(rnd));
2957 	elr->lr_next_sched = jiffies + (unsigned long)rnd %
2958 			     (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2959 
2960 	return elr;
2961 }
2962 
2963 int ext4_register_li_request(struct super_block *sb,
2964 			     ext4_group_t first_not_zeroed)
2965 {
2966 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2967 	struct ext4_li_request *elr = NULL;
2968 	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2969 	int ret = 0;
2970 
2971 	mutex_lock(&ext4_li_mtx);
2972 	if (sbi->s_li_request != NULL) {
2973 		/*
2974 		 * Reset timeout so it can be computed again, because
2975 		 * s_li_wait_mult might have changed.
2976 		 */
2977 		sbi->s_li_request->lr_timeout = 0;
2978 		goto out;
2979 	}
2980 
2981 	if (first_not_zeroed == ngroups ||
2982 	    (sb->s_flags & MS_RDONLY) ||
2983 	    !test_opt(sb, INIT_INODE_TABLE))
2984 		goto out;
2985 
2986 	elr = ext4_li_request_new(sb, first_not_zeroed);
2987 	if (!elr) {
2988 		ret = -ENOMEM;
2989 		goto out;
2990 	}
2991 
2992 	if (NULL == ext4_li_info) {
2993 		ret = ext4_li_info_new();
2994 		if (ret)
2995 			goto out;
2996 	}
2997 
2998 	mutex_lock(&ext4_li_info->li_list_mtx);
2999 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3000 	mutex_unlock(&ext4_li_info->li_list_mtx);
3001 
3002 	sbi->s_li_request = elr;
3003 	/*
3004 	 * set elr to NULL here since it has been inserted to
3005 	 * the request_list and the removal and free of it is
3006 	 * handled by ext4_clear_request_list from now on.
3007 	 */
3008 	elr = NULL;
3009 
3010 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3011 		ret = ext4_run_lazyinit_thread();
3012 		if (ret)
3013 			goto out;
3014 	}
3015 out:
3016 	mutex_unlock(&ext4_li_mtx);
3017 	if (ret)
3018 		kfree(elr);
3019 	return ret;
3020 }
3021 
3022 /*
3023  * We do not need to lock anything since this is called on
3024  * module unload.
3025  */
3026 static void ext4_destroy_lazyinit_thread(void)
3027 {
3028 	/*
3029 	 * If thread exited earlier
3030 	 * there's nothing to be done.
3031 	 */
3032 	if (!ext4_li_info || !ext4_lazyinit_task)
3033 		return;
3034 
3035 	kthread_stop(ext4_lazyinit_task);
3036 }
3037 
3038 static int set_journal_csum_feature_set(struct super_block *sb)
3039 {
3040 	int ret = 1;
3041 	int compat, incompat;
3042 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3043 
3044 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3045 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3046 		/* journal checksum v2 */
3047 		compat = 0;
3048 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3049 	} else {
3050 		/* journal checksum v1 */
3051 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3052 		incompat = 0;
3053 	}
3054 
3055 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3056 		ret = jbd2_journal_set_features(sbi->s_journal,
3057 				compat, 0,
3058 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3059 				incompat);
3060 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3061 		ret = jbd2_journal_set_features(sbi->s_journal,
3062 				compat, 0,
3063 				incompat);
3064 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3065 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3066 	} else {
3067 		jbd2_journal_clear_features(sbi->s_journal,
3068 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3069 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3070 				JBD2_FEATURE_INCOMPAT_CSUM_V2);
3071 	}
3072 
3073 	return ret;
3074 }
3075 
3076 /*
3077  * Note: calculating the overhead so we can be compatible with
3078  * historical BSD practice is quite difficult in the face of
3079  * clusters/bigalloc.  This is because multiple metadata blocks from
3080  * different block group can end up in the same allocation cluster.
3081  * Calculating the exact overhead in the face of clustered allocation
3082  * requires either O(all block bitmaps) in memory or O(number of block
3083  * groups**2) in time.  We will still calculate the superblock for
3084  * older file systems --- and if we come across with a bigalloc file
3085  * system with zero in s_overhead_clusters the estimate will be close to
3086  * correct especially for very large cluster sizes --- but for newer
3087  * file systems, it's better to calculate this figure once at mkfs
3088  * time, and store it in the superblock.  If the superblock value is
3089  * present (even for non-bigalloc file systems), we will use it.
3090  */
3091 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3092 			  char *buf)
3093 {
3094 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3095 	struct ext4_group_desc	*gdp;
3096 	ext4_fsblk_t		first_block, last_block, b;
3097 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3098 	int			s, j, count = 0;
3099 
3100 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3101 		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3102 			sbi->s_itb_per_group + 2);
3103 
3104 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3105 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3106 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3107 	for (i = 0; i < ngroups; i++) {
3108 		gdp = ext4_get_group_desc(sb, i, NULL);
3109 		b = ext4_block_bitmap(sb, gdp);
3110 		if (b >= first_block && b <= last_block) {
3111 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3112 			count++;
3113 		}
3114 		b = ext4_inode_bitmap(sb, gdp);
3115 		if (b >= first_block && b <= last_block) {
3116 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3117 			count++;
3118 		}
3119 		b = ext4_inode_table(sb, gdp);
3120 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3121 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3122 				int c = EXT4_B2C(sbi, b - first_block);
3123 				ext4_set_bit(c, buf);
3124 				count++;
3125 			}
3126 		if (i != grp)
3127 			continue;
3128 		s = 0;
3129 		if (ext4_bg_has_super(sb, grp)) {
3130 			ext4_set_bit(s++, buf);
3131 			count++;
3132 		}
3133 		for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3134 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3135 			count++;
3136 		}
3137 	}
3138 	if (!count)
3139 		return 0;
3140 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3141 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3142 }
3143 
3144 /*
3145  * Compute the overhead and stash it in sbi->s_overhead
3146  */
3147 int ext4_calculate_overhead(struct super_block *sb)
3148 {
3149 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3150 	struct ext4_super_block *es = sbi->s_es;
3151 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3152 	ext4_fsblk_t overhead = 0;
3153 	char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3154 
3155 	if (!buf)
3156 		return -ENOMEM;
3157 
3158 	/*
3159 	 * Compute the overhead (FS structures).  This is constant
3160 	 * for a given filesystem unless the number of block groups
3161 	 * changes so we cache the previous value until it does.
3162 	 */
3163 
3164 	/*
3165 	 * All of the blocks before first_data_block are overhead
3166 	 */
3167 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3168 
3169 	/*
3170 	 * Add the overhead found in each block group
3171 	 */
3172 	for (i = 0; i < ngroups; i++) {
3173 		int blks;
3174 
3175 		blks = count_overhead(sb, i, buf);
3176 		overhead += blks;
3177 		if (blks)
3178 			memset(buf, 0, PAGE_SIZE);
3179 		cond_resched();
3180 	}
3181 	/* Add the journal blocks as well */
3182 	if (sbi->s_journal)
3183 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3184 
3185 	sbi->s_overhead = overhead;
3186 	smp_wmb();
3187 	free_page((unsigned long) buf);
3188 	return 0;
3189 }
3190 
3191 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3192 {
3193 	char *orig_data = kstrdup(data, GFP_KERNEL);
3194 	struct buffer_head *bh;
3195 	struct ext4_super_block *es = NULL;
3196 	struct ext4_sb_info *sbi;
3197 	ext4_fsblk_t block;
3198 	ext4_fsblk_t sb_block = get_sb_block(&data);
3199 	ext4_fsblk_t logical_sb_block;
3200 	unsigned long offset = 0;
3201 	unsigned long journal_devnum = 0;
3202 	unsigned long def_mount_opts;
3203 	struct inode *root;
3204 	char *cp;
3205 	const char *descr;
3206 	int ret = -ENOMEM;
3207 	int blocksize, clustersize;
3208 	unsigned int db_count;
3209 	unsigned int i;
3210 	int needs_recovery, has_huge_files, has_bigalloc;
3211 	__u64 blocks_count;
3212 	int err = 0;
3213 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3214 	ext4_group_t first_not_zeroed;
3215 
3216 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3217 	if (!sbi)
3218 		goto out_free_orig;
3219 
3220 	sbi->s_blockgroup_lock =
3221 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3222 	if (!sbi->s_blockgroup_lock) {
3223 		kfree(sbi);
3224 		goto out_free_orig;
3225 	}
3226 	sb->s_fs_info = sbi;
3227 	sbi->s_sb = sb;
3228 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3229 	sbi->s_sb_block = sb_block;
3230 	if (sb->s_bdev->bd_part)
3231 		sbi->s_sectors_written_start =
3232 			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3233 
3234 	/* Cleanup superblock name */
3235 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3236 		*cp = '!';
3237 
3238 	/* -EINVAL is default */
3239 	ret = -EINVAL;
3240 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3241 	if (!blocksize) {
3242 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3243 		goto out_fail;
3244 	}
3245 
3246 	/*
3247 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3248 	 * block sizes.  We need to calculate the offset from buffer start.
3249 	 */
3250 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3251 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3252 		offset = do_div(logical_sb_block, blocksize);
3253 	} else {
3254 		logical_sb_block = sb_block;
3255 	}
3256 
3257 	if (!(bh = sb_bread(sb, logical_sb_block))) {
3258 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3259 		goto out_fail;
3260 	}
3261 	/*
3262 	 * Note: s_es must be initialized as soon as possible because
3263 	 *       some ext4 macro-instructions depend on its value
3264 	 */
3265 	es = (struct ext4_super_block *) (bh->b_data + offset);
3266 	sbi->s_es = es;
3267 	sb->s_magic = le16_to_cpu(es->s_magic);
3268 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3269 		goto cantfind_ext4;
3270 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3271 
3272 	/* Warn if metadata_csum and gdt_csum are both set. */
3273 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3274 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3275 	    EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3276 		ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3277 			     "redundant flags; please run fsck.");
3278 
3279 	/* Check for a known checksum algorithm */
3280 	if (!ext4_verify_csum_type(sb, es)) {
3281 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3282 			 "unknown checksum algorithm.");
3283 		silent = 1;
3284 		goto cantfind_ext4;
3285 	}
3286 
3287 	/* Load the checksum driver */
3288 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3289 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3290 		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3291 		if (IS_ERR(sbi->s_chksum_driver)) {
3292 			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3293 			ret = PTR_ERR(sbi->s_chksum_driver);
3294 			sbi->s_chksum_driver = NULL;
3295 			goto failed_mount;
3296 		}
3297 	}
3298 
3299 	/* Check superblock checksum */
3300 	if (!ext4_superblock_csum_verify(sb, es)) {
3301 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3302 			 "invalid superblock checksum.  Run e2fsck?");
3303 		silent = 1;
3304 		goto cantfind_ext4;
3305 	}
3306 
3307 	/* Precompute checksum seed for all metadata */
3308 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3309 			EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3310 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3311 					       sizeof(es->s_uuid));
3312 
3313 	/* Set defaults before we parse the mount options */
3314 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3315 	set_opt(sb, INIT_INODE_TABLE);
3316 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3317 		set_opt(sb, DEBUG);
3318 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3319 		set_opt(sb, GRPID);
3320 	if (def_mount_opts & EXT4_DEFM_UID16)
3321 		set_opt(sb, NO_UID32);
3322 	/* xattr user namespace & acls are now defaulted on */
3323 	set_opt(sb, XATTR_USER);
3324 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3325 	set_opt(sb, POSIX_ACL);
3326 #endif
3327 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3328 		set_opt(sb, JOURNAL_DATA);
3329 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3330 		set_opt(sb, ORDERED_DATA);
3331 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3332 		set_opt(sb, WRITEBACK_DATA);
3333 
3334 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3335 		set_opt(sb, ERRORS_PANIC);
3336 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3337 		set_opt(sb, ERRORS_CONT);
3338 	else
3339 		set_opt(sb, ERRORS_RO);
3340 	if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3341 		set_opt(sb, BLOCK_VALIDITY);
3342 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3343 		set_opt(sb, DISCARD);
3344 
3345 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3346 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3347 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3348 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3349 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3350 
3351 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3352 		set_opt(sb, BARRIER);
3353 
3354 	/*
3355 	 * enable delayed allocation by default
3356 	 * Use -o nodelalloc to turn it off
3357 	 */
3358 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3359 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3360 		set_opt(sb, DELALLOC);
3361 
3362 	/*
3363 	 * set default s_li_wait_mult for lazyinit, for the case there is
3364 	 * no mount option specified.
3365 	 */
3366 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3367 
3368 	if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3369 			   &journal_devnum, &journal_ioprio, 0)) {
3370 		ext4_msg(sb, KERN_WARNING,
3371 			 "failed to parse options in superblock: %s",
3372 			 sbi->s_es->s_mount_opts);
3373 	}
3374 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3375 	if (!parse_options((char *) data, sb, &journal_devnum,
3376 			   &journal_ioprio, 0))
3377 		goto failed_mount;
3378 
3379 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3380 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3381 			    "with data=journal disables delayed "
3382 			    "allocation and O_DIRECT support!\n");
3383 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3384 			ext4_msg(sb, KERN_ERR, "can't mount with "
3385 				 "both data=journal and delalloc");
3386 			goto failed_mount;
3387 		}
3388 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3389 			ext4_msg(sb, KERN_ERR, "can't mount with "
3390 				 "both data=journal and delalloc");
3391 			goto failed_mount;
3392 		}
3393 		if (test_opt(sb, DELALLOC))
3394 			clear_opt(sb, DELALLOC);
3395 	}
3396 
3397 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3398 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3399 
3400 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3401 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3402 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3403 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3404 		ext4_msg(sb, KERN_WARNING,
3405 		       "feature flags set on rev 0 fs, "
3406 		       "running e2fsck is recommended");
3407 
3408 	if (IS_EXT2_SB(sb)) {
3409 		if (ext2_feature_set_ok(sb))
3410 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3411 				 "using the ext4 subsystem");
3412 		else {
3413 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3414 				 "to feature incompatibilities");
3415 			goto failed_mount;
3416 		}
3417 	}
3418 
3419 	if (IS_EXT3_SB(sb)) {
3420 		if (ext3_feature_set_ok(sb))
3421 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3422 				 "using the ext4 subsystem");
3423 		else {
3424 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3425 				 "to feature incompatibilities");
3426 			goto failed_mount;
3427 		}
3428 	}
3429 
3430 	/*
3431 	 * Check feature flags regardless of the revision level, since we
3432 	 * previously didn't change the revision level when setting the flags,
3433 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3434 	 */
3435 	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3436 		goto failed_mount;
3437 
3438 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3439 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3440 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3441 		ext4_msg(sb, KERN_ERR,
3442 		       "Unsupported filesystem blocksize %d", blocksize);
3443 		goto failed_mount;
3444 	}
3445 
3446 	if (sb->s_blocksize != blocksize) {
3447 		/* Validate the filesystem blocksize */
3448 		if (!sb_set_blocksize(sb, blocksize)) {
3449 			ext4_msg(sb, KERN_ERR, "bad block size %d",
3450 					blocksize);
3451 			goto failed_mount;
3452 		}
3453 
3454 		brelse(bh);
3455 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3456 		offset = do_div(logical_sb_block, blocksize);
3457 		bh = sb_bread(sb, logical_sb_block);
3458 		if (!bh) {
3459 			ext4_msg(sb, KERN_ERR,
3460 			       "Can't read superblock on 2nd try");
3461 			goto failed_mount;
3462 		}
3463 		es = (struct ext4_super_block *)(bh->b_data + offset);
3464 		sbi->s_es = es;
3465 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3466 			ext4_msg(sb, KERN_ERR,
3467 			       "Magic mismatch, very weird!");
3468 			goto failed_mount;
3469 		}
3470 	}
3471 
3472 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3473 				EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3474 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3475 						      has_huge_files);
3476 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3477 
3478 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3479 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3480 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3481 	} else {
3482 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3483 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3484 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3485 		    (!is_power_of_2(sbi->s_inode_size)) ||
3486 		    (sbi->s_inode_size > blocksize)) {
3487 			ext4_msg(sb, KERN_ERR,
3488 			       "unsupported inode size: %d",
3489 			       sbi->s_inode_size);
3490 			goto failed_mount;
3491 		}
3492 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3493 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3494 	}
3495 
3496 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3497 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3498 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3499 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3500 		    !is_power_of_2(sbi->s_desc_size)) {
3501 			ext4_msg(sb, KERN_ERR,
3502 			       "unsupported descriptor size %lu",
3503 			       sbi->s_desc_size);
3504 			goto failed_mount;
3505 		}
3506 	} else
3507 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3508 
3509 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3510 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3511 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3512 		goto cantfind_ext4;
3513 
3514 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3515 	if (sbi->s_inodes_per_block == 0)
3516 		goto cantfind_ext4;
3517 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3518 					sbi->s_inodes_per_block;
3519 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3520 	sbi->s_sbh = bh;
3521 	sbi->s_mount_state = le16_to_cpu(es->s_state);
3522 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3523 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3524 
3525 	for (i = 0; i < 4; i++)
3526 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3527 	sbi->s_def_hash_version = es->s_def_hash_version;
3528 	i = le32_to_cpu(es->s_flags);
3529 	if (i & EXT2_FLAGS_UNSIGNED_HASH)
3530 		sbi->s_hash_unsigned = 3;
3531 	else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3532 #ifdef __CHAR_UNSIGNED__
3533 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3534 		sbi->s_hash_unsigned = 3;
3535 #else
3536 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3537 #endif
3538 	}
3539 
3540 	/* Handle clustersize */
3541 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3542 	has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3543 				EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3544 	if (has_bigalloc) {
3545 		if (clustersize < blocksize) {
3546 			ext4_msg(sb, KERN_ERR,
3547 				 "cluster size (%d) smaller than "
3548 				 "block size (%d)", clustersize, blocksize);
3549 			goto failed_mount;
3550 		}
3551 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3552 			le32_to_cpu(es->s_log_block_size);
3553 		sbi->s_clusters_per_group =
3554 			le32_to_cpu(es->s_clusters_per_group);
3555 		if (sbi->s_clusters_per_group > blocksize * 8) {
3556 			ext4_msg(sb, KERN_ERR,
3557 				 "#clusters per group too big: %lu",
3558 				 sbi->s_clusters_per_group);
3559 			goto failed_mount;
3560 		}
3561 		if (sbi->s_blocks_per_group !=
3562 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3563 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3564 				 "clusters per group (%lu) inconsistent",
3565 				 sbi->s_blocks_per_group,
3566 				 sbi->s_clusters_per_group);
3567 			goto failed_mount;
3568 		}
3569 	} else {
3570 		if (clustersize != blocksize) {
3571 			ext4_warning(sb, "fragment/cluster size (%d) != "
3572 				     "block size (%d)", clustersize,
3573 				     blocksize);
3574 			clustersize = blocksize;
3575 		}
3576 		if (sbi->s_blocks_per_group > blocksize * 8) {
3577 			ext4_msg(sb, KERN_ERR,
3578 				 "#blocks per group too big: %lu",
3579 				 sbi->s_blocks_per_group);
3580 			goto failed_mount;
3581 		}
3582 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3583 		sbi->s_cluster_bits = 0;
3584 	}
3585 	sbi->s_cluster_ratio = clustersize / blocksize;
3586 
3587 	if (sbi->s_inodes_per_group > blocksize * 8) {
3588 		ext4_msg(sb, KERN_ERR,
3589 		       "#inodes per group too big: %lu",
3590 		       sbi->s_inodes_per_group);
3591 		goto failed_mount;
3592 	}
3593 
3594 	/*
3595 	 * Test whether we have more sectors than will fit in sector_t,
3596 	 * and whether the max offset is addressable by the page cache.
3597 	 */
3598 	err = generic_check_addressable(sb->s_blocksize_bits,
3599 					ext4_blocks_count(es));
3600 	if (err) {
3601 		ext4_msg(sb, KERN_ERR, "filesystem"
3602 			 " too large to mount safely on this system");
3603 		if (sizeof(sector_t) < 8)
3604 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3605 		goto failed_mount;
3606 	}
3607 
3608 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3609 		goto cantfind_ext4;
3610 
3611 	/* check blocks count against device size */
3612 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3613 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3614 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3615 		       "exceeds size of device (%llu blocks)",
3616 		       ext4_blocks_count(es), blocks_count);
3617 		goto failed_mount;
3618 	}
3619 
3620 	/*
3621 	 * It makes no sense for the first data block to be beyond the end
3622 	 * of the filesystem.
3623 	 */
3624 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3625 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3626 			 "block %u is beyond end of filesystem (%llu)",
3627 			 le32_to_cpu(es->s_first_data_block),
3628 			 ext4_blocks_count(es));
3629 		goto failed_mount;
3630 	}
3631 	blocks_count = (ext4_blocks_count(es) -
3632 			le32_to_cpu(es->s_first_data_block) +
3633 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3634 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3635 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3636 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3637 		       "(block count %llu, first data block %u, "
3638 		       "blocks per group %lu)", sbi->s_groups_count,
3639 		       ext4_blocks_count(es),
3640 		       le32_to_cpu(es->s_first_data_block),
3641 		       EXT4_BLOCKS_PER_GROUP(sb));
3642 		goto failed_mount;
3643 	}
3644 	sbi->s_groups_count = blocks_count;
3645 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3646 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3647 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3648 		   EXT4_DESC_PER_BLOCK(sb);
3649 	sbi->s_group_desc = ext4_kvmalloc(db_count *
3650 					  sizeof(struct buffer_head *),
3651 					  GFP_KERNEL);
3652 	if (sbi->s_group_desc == NULL) {
3653 		ext4_msg(sb, KERN_ERR, "not enough memory");
3654 		ret = -ENOMEM;
3655 		goto failed_mount;
3656 	}
3657 
3658 	if (ext4_proc_root)
3659 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3660 
3661 	if (sbi->s_proc)
3662 		proc_create_data("options", S_IRUGO, sbi->s_proc,
3663 				 &ext4_seq_options_fops, sb);
3664 
3665 	bgl_lock_init(sbi->s_blockgroup_lock);
3666 
3667 	for (i = 0; i < db_count; i++) {
3668 		block = descriptor_loc(sb, logical_sb_block, i);
3669 		sbi->s_group_desc[i] = sb_bread(sb, block);
3670 		if (!sbi->s_group_desc[i]) {
3671 			ext4_msg(sb, KERN_ERR,
3672 			       "can't read group descriptor %d", i);
3673 			db_count = i;
3674 			goto failed_mount2;
3675 		}
3676 	}
3677 	if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3678 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3679 		goto failed_mount2;
3680 	}
3681 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3682 		if (!ext4_fill_flex_info(sb)) {
3683 			ext4_msg(sb, KERN_ERR,
3684 			       "unable to initialize "
3685 			       "flex_bg meta info!");
3686 			goto failed_mount2;
3687 		}
3688 
3689 	sbi->s_gdb_count = db_count;
3690 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3691 	spin_lock_init(&sbi->s_next_gen_lock);
3692 
3693 	init_timer(&sbi->s_err_report);
3694 	sbi->s_err_report.function = print_daily_error_info;
3695 	sbi->s_err_report.data = (unsigned long) sb;
3696 
3697 	err = percpu_counter_init(&sbi->s_freeclusters_counter,
3698 			ext4_count_free_clusters(sb));
3699 	if (!err) {
3700 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
3701 				ext4_count_free_inodes(sb));
3702 	}
3703 	if (!err) {
3704 		err = percpu_counter_init(&sbi->s_dirs_counter,
3705 				ext4_count_dirs(sb));
3706 	}
3707 	if (!err) {
3708 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3709 	}
3710 	if (!err) {
3711 		err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3712 	}
3713 	if (err) {
3714 		ext4_msg(sb, KERN_ERR, "insufficient memory");
3715 		goto failed_mount3;
3716 	}
3717 
3718 	sbi->s_stripe = ext4_get_stripe_size(sbi);
3719 	sbi->s_max_writeback_mb_bump = 128;
3720 	sbi->s_extent_max_zeroout_kb = 32;
3721 
3722 	/* Register extent status tree shrinker */
3723 	ext4_es_register_shrinker(sb);
3724 
3725 	/*
3726 	 * set up enough so that it can read an inode
3727 	 */
3728 	if (!test_opt(sb, NOLOAD) &&
3729 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3730 		sb->s_op = &ext4_sops;
3731 	else
3732 		sb->s_op = &ext4_nojournal_sops;
3733 	sb->s_export_op = &ext4_export_ops;
3734 	sb->s_xattr = ext4_xattr_handlers;
3735 #ifdef CONFIG_QUOTA
3736 	sb->dq_op = &ext4_quota_operations;
3737 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3738 		sb->s_qcop = &ext4_qctl_sysfile_operations;
3739 	else
3740 		sb->s_qcop = &ext4_qctl_operations;
3741 #endif
3742 	memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3743 
3744 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3745 	mutex_init(&sbi->s_orphan_lock);
3746 
3747 	sb->s_root = NULL;
3748 
3749 	needs_recovery = (es->s_last_orphan != 0 ||
3750 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
3751 				    EXT4_FEATURE_INCOMPAT_RECOVER));
3752 
3753 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3754 	    !(sb->s_flags & MS_RDONLY))
3755 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3756 			goto failed_mount3;
3757 
3758 	/*
3759 	 * The first inode we look at is the journal inode.  Don't try
3760 	 * root first: it may be modified in the journal!
3761 	 */
3762 	if (!test_opt(sb, NOLOAD) &&
3763 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3764 		if (ext4_load_journal(sb, es, journal_devnum))
3765 			goto failed_mount3;
3766 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3767 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3768 		ext4_msg(sb, KERN_ERR, "required journal recovery "
3769 		       "suppressed and not mounted read-only");
3770 		goto failed_mount_wq;
3771 	} else {
3772 		clear_opt(sb, DATA_FLAGS);
3773 		sbi->s_journal = NULL;
3774 		needs_recovery = 0;
3775 		goto no_journal;
3776 	}
3777 
3778 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3779 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3780 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
3781 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3782 		goto failed_mount_wq;
3783 	}
3784 
3785 	if (!set_journal_csum_feature_set(sb)) {
3786 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3787 			 "feature set");
3788 		goto failed_mount_wq;
3789 	}
3790 
3791 	/* We have now updated the journal if required, so we can
3792 	 * validate the data journaling mode. */
3793 	switch (test_opt(sb, DATA_FLAGS)) {
3794 	case 0:
3795 		/* No mode set, assume a default based on the journal
3796 		 * capabilities: ORDERED_DATA if the journal can
3797 		 * cope, else JOURNAL_DATA
3798 		 */
3799 		if (jbd2_journal_check_available_features
3800 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3801 			set_opt(sb, ORDERED_DATA);
3802 		else
3803 			set_opt(sb, JOURNAL_DATA);
3804 		break;
3805 
3806 	case EXT4_MOUNT_ORDERED_DATA:
3807 	case EXT4_MOUNT_WRITEBACK_DATA:
3808 		if (!jbd2_journal_check_available_features
3809 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3810 			ext4_msg(sb, KERN_ERR, "Journal does not support "
3811 			       "requested data journaling mode");
3812 			goto failed_mount_wq;
3813 		}
3814 	default:
3815 		break;
3816 	}
3817 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3818 
3819 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3820 
3821 	/*
3822 	 * The journal may have updated the bg summary counts, so we
3823 	 * need to update the global counters.
3824 	 */
3825 	percpu_counter_set(&sbi->s_freeclusters_counter,
3826 			   ext4_count_free_clusters(sb));
3827 	percpu_counter_set(&sbi->s_freeinodes_counter,
3828 			   ext4_count_free_inodes(sb));
3829 	percpu_counter_set(&sbi->s_dirs_counter,
3830 			   ext4_count_dirs(sb));
3831 	percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3832 
3833 no_journal:
3834 	/*
3835 	 * Get the # of file system overhead blocks from the
3836 	 * superblock if present.
3837 	 */
3838 	if (es->s_overhead_clusters)
3839 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3840 	else {
3841 		err = ext4_calculate_overhead(sb);
3842 		if (err)
3843 			goto failed_mount_wq;
3844 	}
3845 
3846 	/*
3847 	 * The maximum number of concurrent works can be high and
3848 	 * concurrency isn't really necessary.  Limit it to 1.
3849 	 */
3850 	EXT4_SB(sb)->dio_unwritten_wq =
3851 		alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3852 	if (!EXT4_SB(sb)->dio_unwritten_wq) {
3853 		printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3854 		ret = -ENOMEM;
3855 		goto failed_mount_wq;
3856 	}
3857 
3858 	/*
3859 	 * The jbd2_journal_load will have done any necessary log recovery,
3860 	 * so we can safely mount the rest of the filesystem now.
3861 	 */
3862 
3863 	root = ext4_iget(sb, EXT4_ROOT_INO);
3864 	if (IS_ERR(root)) {
3865 		ext4_msg(sb, KERN_ERR, "get root inode failed");
3866 		ret = PTR_ERR(root);
3867 		root = NULL;
3868 		goto failed_mount4;
3869 	}
3870 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3871 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3872 		iput(root);
3873 		goto failed_mount4;
3874 	}
3875 	sb->s_root = d_make_root(root);
3876 	if (!sb->s_root) {
3877 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
3878 		ret = -ENOMEM;
3879 		goto failed_mount4;
3880 	}
3881 
3882 	if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3883 		sb->s_flags |= MS_RDONLY;
3884 
3885 	/* determine the minimum size of new large inodes, if present */
3886 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3887 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3888 						     EXT4_GOOD_OLD_INODE_SIZE;
3889 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3890 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3891 			if (sbi->s_want_extra_isize <
3892 			    le16_to_cpu(es->s_want_extra_isize))
3893 				sbi->s_want_extra_isize =
3894 					le16_to_cpu(es->s_want_extra_isize);
3895 			if (sbi->s_want_extra_isize <
3896 			    le16_to_cpu(es->s_min_extra_isize))
3897 				sbi->s_want_extra_isize =
3898 					le16_to_cpu(es->s_min_extra_isize);
3899 		}
3900 	}
3901 	/* Check if enough inode space is available */
3902 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3903 							sbi->s_inode_size) {
3904 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3905 						       EXT4_GOOD_OLD_INODE_SIZE;
3906 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
3907 			 "available");
3908 	}
3909 
3910 	err = ext4_setup_system_zone(sb);
3911 	if (err) {
3912 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
3913 			 "zone (%d)", err);
3914 		goto failed_mount4a;
3915 	}
3916 
3917 	ext4_ext_init(sb);
3918 	err = ext4_mb_init(sb);
3919 	if (err) {
3920 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3921 			 err);
3922 		goto failed_mount5;
3923 	}
3924 
3925 	err = ext4_register_li_request(sb, first_not_zeroed);
3926 	if (err)
3927 		goto failed_mount6;
3928 
3929 	sbi->s_kobj.kset = ext4_kset;
3930 	init_completion(&sbi->s_kobj_unregister);
3931 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3932 				   "%s", sb->s_id);
3933 	if (err)
3934 		goto failed_mount7;
3935 
3936 #ifdef CONFIG_QUOTA
3937 	/* Enable quota usage during mount. */
3938 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3939 	    !(sb->s_flags & MS_RDONLY)) {
3940 		err = ext4_enable_quotas(sb);
3941 		if (err)
3942 			goto failed_mount8;
3943 	}
3944 #endif  /* CONFIG_QUOTA */
3945 
3946 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3947 	ext4_orphan_cleanup(sb, es);
3948 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3949 	if (needs_recovery) {
3950 		ext4_msg(sb, KERN_INFO, "recovery complete");
3951 		ext4_mark_recovery_complete(sb, es);
3952 	}
3953 	if (EXT4_SB(sb)->s_journal) {
3954 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3955 			descr = " journalled data mode";
3956 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3957 			descr = " ordered data mode";
3958 		else
3959 			descr = " writeback data mode";
3960 	} else
3961 		descr = "out journal";
3962 
3963 	if (test_opt(sb, DISCARD)) {
3964 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
3965 		if (!blk_queue_discard(q))
3966 			ext4_msg(sb, KERN_WARNING,
3967 				 "mounting with \"discard\" option, but "
3968 				 "the device does not support discard");
3969 	}
3970 
3971 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3972 		 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3973 		 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3974 
3975 	if (es->s_error_count)
3976 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3977 
3978 	kfree(orig_data);
3979 	return 0;
3980 
3981 cantfind_ext4:
3982 	if (!silent)
3983 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3984 	goto failed_mount;
3985 
3986 #ifdef CONFIG_QUOTA
3987 failed_mount8:
3988 	kobject_del(&sbi->s_kobj);
3989 #endif
3990 failed_mount7:
3991 	ext4_unregister_li_request(sb);
3992 failed_mount6:
3993 	ext4_mb_release(sb);
3994 failed_mount5:
3995 	ext4_ext_release(sb);
3996 	ext4_release_system_zone(sb);
3997 failed_mount4a:
3998 	dput(sb->s_root);
3999 	sb->s_root = NULL;
4000 failed_mount4:
4001 	ext4_msg(sb, KERN_ERR, "mount failed");
4002 	destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4003 failed_mount_wq:
4004 	if (sbi->s_journal) {
4005 		jbd2_journal_destroy(sbi->s_journal);
4006 		sbi->s_journal = NULL;
4007 	}
4008 failed_mount3:
4009 	del_timer(&sbi->s_err_report);
4010 	if (sbi->s_flex_groups)
4011 		ext4_kvfree(sbi->s_flex_groups);
4012 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
4013 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
4014 	percpu_counter_destroy(&sbi->s_dirs_counter);
4015 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4016 	percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4017 	if (sbi->s_mmp_tsk)
4018 		kthread_stop(sbi->s_mmp_tsk);
4019 failed_mount2:
4020 	for (i = 0; i < db_count; i++)
4021 		brelse(sbi->s_group_desc[i]);
4022 	ext4_kvfree(sbi->s_group_desc);
4023 failed_mount:
4024 	if (sbi->s_chksum_driver)
4025 		crypto_free_shash(sbi->s_chksum_driver);
4026 	if (sbi->s_proc) {
4027 		remove_proc_entry("options", sbi->s_proc);
4028 		remove_proc_entry(sb->s_id, ext4_proc_root);
4029 	}
4030 #ifdef CONFIG_QUOTA
4031 	for (i = 0; i < MAXQUOTAS; i++)
4032 		kfree(sbi->s_qf_names[i]);
4033 #endif
4034 	ext4_blkdev_remove(sbi);
4035 	brelse(bh);
4036 out_fail:
4037 	sb->s_fs_info = NULL;
4038 	kfree(sbi->s_blockgroup_lock);
4039 	kfree(sbi);
4040 out_free_orig:
4041 	kfree(orig_data);
4042 	return err ? err : ret;
4043 }
4044 
4045 /*
4046  * Setup any per-fs journal parameters now.  We'll do this both on
4047  * initial mount, once the journal has been initialised but before we've
4048  * done any recovery; and again on any subsequent remount.
4049  */
4050 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4051 {
4052 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4053 
4054 	journal->j_commit_interval = sbi->s_commit_interval;
4055 	journal->j_min_batch_time = sbi->s_min_batch_time;
4056 	journal->j_max_batch_time = sbi->s_max_batch_time;
4057 
4058 	write_lock(&journal->j_state_lock);
4059 	if (test_opt(sb, BARRIER))
4060 		journal->j_flags |= JBD2_BARRIER;
4061 	else
4062 		journal->j_flags &= ~JBD2_BARRIER;
4063 	if (test_opt(sb, DATA_ERR_ABORT))
4064 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4065 	else
4066 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4067 	write_unlock(&journal->j_state_lock);
4068 }
4069 
4070 static journal_t *ext4_get_journal(struct super_block *sb,
4071 				   unsigned int journal_inum)
4072 {
4073 	struct inode *journal_inode;
4074 	journal_t *journal;
4075 
4076 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4077 
4078 	/* First, test for the existence of a valid inode on disk.  Bad
4079 	 * things happen if we iget() an unused inode, as the subsequent
4080 	 * iput() will try to delete it. */
4081 
4082 	journal_inode = ext4_iget(sb, journal_inum);
4083 	if (IS_ERR(journal_inode)) {
4084 		ext4_msg(sb, KERN_ERR, "no journal found");
4085 		return NULL;
4086 	}
4087 	if (!journal_inode->i_nlink) {
4088 		make_bad_inode(journal_inode);
4089 		iput(journal_inode);
4090 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4091 		return NULL;
4092 	}
4093 
4094 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4095 		  journal_inode, journal_inode->i_size);
4096 	if (!S_ISREG(journal_inode->i_mode)) {
4097 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4098 		iput(journal_inode);
4099 		return NULL;
4100 	}
4101 
4102 	journal = jbd2_journal_init_inode(journal_inode);
4103 	if (!journal) {
4104 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4105 		iput(journal_inode);
4106 		return NULL;
4107 	}
4108 	journal->j_private = sb;
4109 	ext4_init_journal_params(sb, journal);
4110 	return journal;
4111 }
4112 
4113 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4114 				       dev_t j_dev)
4115 {
4116 	struct buffer_head *bh;
4117 	journal_t *journal;
4118 	ext4_fsblk_t start;
4119 	ext4_fsblk_t len;
4120 	int hblock, blocksize;
4121 	ext4_fsblk_t sb_block;
4122 	unsigned long offset;
4123 	struct ext4_super_block *es;
4124 	struct block_device *bdev;
4125 
4126 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4127 
4128 	bdev = ext4_blkdev_get(j_dev, sb);
4129 	if (bdev == NULL)
4130 		return NULL;
4131 
4132 	blocksize = sb->s_blocksize;
4133 	hblock = bdev_logical_block_size(bdev);
4134 	if (blocksize < hblock) {
4135 		ext4_msg(sb, KERN_ERR,
4136 			"blocksize too small for journal device");
4137 		goto out_bdev;
4138 	}
4139 
4140 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4141 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4142 	set_blocksize(bdev, blocksize);
4143 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4144 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4145 		       "external journal");
4146 		goto out_bdev;
4147 	}
4148 
4149 	es = (struct ext4_super_block *) (bh->b_data + offset);
4150 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4151 	    !(le32_to_cpu(es->s_feature_incompat) &
4152 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4153 		ext4_msg(sb, KERN_ERR, "external journal has "
4154 					"bad superblock");
4155 		brelse(bh);
4156 		goto out_bdev;
4157 	}
4158 
4159 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4160 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4161 		brelse(bh);
4162 		goto out_bdev;
4163 	}
4164 
4165 	len = ext4_blocks_count(es);
4166 	start = sb_block + 1;
4167 	brelse(bh);	/* we're done with the superblock */
4168 
4169 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4170 					start, len, blocksize);
4171 	if (!journal) {
4172 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4173 		goto out_bdev;
4174 	}
4175 	journal->j_private = sb;
4176 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
4177 	wait_on_buffer(journal->j_sb_buffer);
4178 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4179 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4180 		goto out_journal;
4181 	}
4182 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4183 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4184 					"user (unsupported) - %d",
4185 			be32_to_cpu(journal->j_superblock->s_nr_users));
4186 		goto out_journal;
4187 	}
4188 	EXT4_SB(sb)->journal_bdev = bdev;
4189 	ext4_init_journal_params(sb, journal);
4190 	return journal;
4191 
4192 out_journal:
4193 	jbd2_journal_destroy(journal);
4194 out_bdev:
4195 	ext4_blkdev_put(bdev);
4196 	return NULL;
4197 }
4198 
4199 static int ext4_load_journal(struct super_block *sb,
4200 			     struct ext4_super_block *es,
4201 			     unsigned long journal_devnum)
4202 {
4203 	journal_t *journal;
4204 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4205 	dev_t journal_dev;
4206 	int err = 0;
4207 	int really_read_only;
4208 
4209 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4210 
4211 	if (journal_devnum &&
4212 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4213 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4214 			"numbers have changed");
4215 		journal_dev = new_decode_dev(journal_devnum);
4216 	} else
4217 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4218 
4219 	really_read_only = bdev_read_only(sb->s_bdev);
4220 
4221 	/*
4222 	 * Are we loading a blank journal or performing recovery after a
4223 	 * crash?  For recovery, we need to check in advance whether we
4224 	 * can get read-write access to the device.
4225 	 */
4226 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4227 		if (sb->s_flags & MS_RDONLY) {
4228 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4229 					"required on readonly filesystem");
4230 			if (really_read_only) {
4231 				ext4_msg(sb, KERN_ERR, "write access "
4232 					"unavailable, cannot proceed");
4233 				return -EROFS;
4234 			}
4235 			ext4_msg(sb, KERN_INFO, "write access will "
4236 			       "be enabled during recovery");
4237 		}
4238 	}
4239 
4240 	if (journal_inum && journal_dev) {
4241 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4242 		       "and inode journals!");
4243 		return -EINVAL;
4244 	}
4245 
4246 	if (journal_inum) {
4247 		if (!(journal = ext4_get_journal(sb, journal_inum)))
4248 			return -EINVAL;
4249 	} else {
4250 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4251 			return -EINVAL;
4252 	}
4253 
4254 	if (!(journal->j_flags & JBD2_BARRIER))
4255 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4256 
4257 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4258 		err = jbd2_journal_wipe(journal, !really_read_only);
4259 	if (!err) {
4260 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4261 		if (save)
4262 			memcpy(save, ((char *) es) +
4263 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4264 		err = jbd2_journal_load(journal);
4265 		if (save)
4266 			memcpy(((char *) es) + EXT4_S_ERR_START,
4267 			       save, EXT4_S_ERR_LEN);
4268 		kfree(save);
4269 	}
4270 
4271 	if (err) {
4272 		ext4_msg(sb, KERN_ERR, "error loading journal");
4273 		jbd2_journal_destroy(journal);
4274 		return err;
4275 	}
4276 
4277 	EXT4_SB(sb)->s_journal = journal;
4278 	ext4_clear_journal_err(sb, es);
4279 
4280 	if (!really_read_only && journal_devnum &&
4281 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4282 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4283 
4284 		/* Make sure we flush the recovery flag to disk. */
4285 		ext4_commit_super(sb, 1);
4286 	}
4287 
4288 	return 0;
4289 }
4290 
4291 static int ext4_commit_super(struct super_block *sb, int sync)
4292 {
4293 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4294 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4295 	int error = 0;
4296 
4297 	if (!sbh || block_device_ejected(sb))
4298 		return error;
4299 	if (buffer_write_io_error(sbh)) {
4300 		/*
4301 		 * Oh, dear.  A previous attempt to write the
4302 		 * superblock failed.  This could happen because the
4303 		 * USB device was yanked out.  Or it could happen to
4304 		 * be a transient write error and maybe the block will
4305 		 * be remapped.  Nothing we can do but to retry the
4306 		 * write and hope for the best.
4307 		 */
4308 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4309 		       "superblock detected");
4310 		clear_buffer_write_io_error(sbh);
4311 		set_buffer_uptodate(sbh);
4312 	}
4313 	/*
4314 	 * If the file system is mounted read-only, don't update the
4315 	 * superblock write time.  This avoids updating the superblock
4316 	 * write time when we are mounting the root file system
4317 	 * read/only but we need to replay the journal; at that point,
4318 	 * for people who are east of GMT and who make their clock
4319 	 * tick in localtime for Windows bug-for-bug compatibility,
4320 	 * the clock is set in the future, and this will cause e2fsck
4321 	 * to complain and force a full file system check.
4322 	 */
4323 	if (!(sb->s_flags & MS_RDONLY))
4324 		es->s_wtime = cpu_to_le32(get_seconds());
4325 	if (sb->s_bdev->bd_part)
4326 		es->s_kbytes_written =
4327 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4328 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4329 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4330 	else
4331 		es->s_kbytes_written =
4332 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4333 	ext4_free_blocks_count_set(es,
4334 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4335 				&EXT4_SB(sb)->s_freeclusters_counter)));
4336 	es->s_free_inodes_count =
4337 		cpu_to_le32(percpu_counter_sum_positive(
4338 				&EXT4_SB(sb)->s_freeinodes_counter));
4339 	BUFFER_TRACE(sbh, "marking dirty");
4340 	ext4_superblock_csum_set(sb);
4341 	mark_buffer_dirty(sbh);
4342 	if (sync) {
4343 		error = sync_dirty_buffer(sbh);
4344 		if (error)
4345 			return error;
4346 
4347 		error = buffer_write_io_error(sbh);
4348 		if (error) {
4349 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4350 			       "superblock");
4351 			clear_buffer_write_io_error(sbh);
4352 			set_buffer_uptodate(sbh);
4353 		}
4354 	}
4355 	return error;
4356 }
4357 
4358 /*
4359  * Have we just finished recovery?  If so, and if we are mounting (or
4360  * remounting) the filesystem readonly, then we will end up with a
4361  * consistent fs on disk.  Record that fact.
4362  */
4363 static void ext4_mark_recovery_complete(struct super_block *sb,
4364 					struct ext4_super_block *es)
4365 {
4366 	journal_t *journal = EXT4_SB(sb)->s_journal;
4367 
4368 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4369 		BUG_ON(journal != NULL);
4370 		return;
4371 	}
4372 	jbd2_journal_lock_updates(journal);
4373 	if (jbd2_journal_flush(journal) < 0)
4374 		goto out;
4375 
4376 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4377 	    sb->s_flags & MS_RDONLY) {
4378 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4379 		ext4_commit_super(sb, 1);
4380 	}
4381 
4382 out:
4383 	jbd2_journal_unlock_updates(journal);
4384 }
4385 
4386 /*
4387  * If we are mounting (or read-write remounting) a filesystem whose journal
4388  * has recorded an error from a previous lifetime, move that error to the
4389  * main filesystem now.
4390  */
4391 static void ext4_clear_journal_err(struct super_block *sb,
4392 				   struct ext4_super_block *es)
4393 {
4394 	journal_t *journal;
4395 	int j_errno;
4396 	const char *errstr;
4397 
4398 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4399 
4400 	journal = EXT4_SB(sb)->s_journal;
4401 
4402 	/*
4403 	 * Now check for any error status which may have been recorded in the
4404 	 * journal by a prior ext4_error() or ext4_abort()
4405 	 */
4406 
4407 	j_errno = jbd2_journal_errno(journal);
4408 	if (j_errno) {
4409 		char nbuf[16];
4410 
4411 		errstr = ext4_decode_error(sb, j_errno, nbuf);
4412 		ext4_warning(sb, "Filesystem error recorded "
4413 			     "from previous mount: %s", errstr);
4414 		ext4_warning(sb, "Marking fs in need of filesystem check.");
4415 
4416 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4417 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4418 		ext4_commit_super(sb, 1);
4419 
4420 		jbd2_journal_clear_err(journal);
4421 		jbd2_journal_update_sb_errno(journal);
4422 	}
4423 }
4424 
4425 /*
4426  * Force the running and committing transactions to commit,
4427  * and wait on the commit.
4428  */
4429 int ext4_force_commit(struct super_block *sb)
4430 {
4431 	journal_t *journal;
4432 
4433 	if (sb->s_flags & MS_RDONLY)
4434 		return 0;
4435 
4436 	journal = EXT4_SB(sb)->s_journal;
4437 	return ext4_journal_force_commit(journal);
4438 }
4439 
4440 static int ext4_sync_fs(struct super_block *sb, int wait)
4441 {
4442 	int ret = 0;
4443 	tid_t target;
4444 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4445 
4446 	trace_ext4_sync_fs(sb, wait);
4447 	flush_workqueue(sbi->dio_unwritten_wq);
4448 	/*
4449 	 * Writeback quota in non-journalled quota case - journalled quota has
4450 	 * no dirty dquots
4451 	 */
4452 	dquot_writeback_dquots(sb, -1);
4453 	if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4454 		if (wait)
4455 			jbd2_log_wait_commit(sbi->s_journal, target);
4456 	}
4457 	return ret;
4458 }
4459 
4460 /*
4461  * LVM calls this function before a (read-only) snapshot is created.  This
4462  * gives us a chance to flush the journal completely and mark the fs clean.
4463  *
4464  * Note that only this function cannot bring a filesystem to be in a clean
4465  * state independently. It relies on upper layer to stop all data & metadata
4466  * modifications.
4467  */
4468 static int ext4_freeze(struct super_block *sb)
4469 {
4470 	int error = 0;
4471 	journal_t *journal;
4472 
4473 	if (sb->s_flags & MS_RDONLY)
4474 		return 0;
4475 
4476 	journal = EXT4_SB(sb)->s_journal;
4477 
4478 	/* Now we set up the journal barrier. */
4479 	jbd2_journal_lock_updates(journal);
4480 
4481 	/*
4482 	 * Don't clear the needs_recovery flag if we failed to flush
4483 	 * the journal.
4484 	 */
4485 	error = jbd2_journal_flush(journal);
4486 	if (error < 0)
4487 		goto out;
4488 
4489 	/* Journal blocked and flushed, clear needs_recovery flag. */
4490 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4491 	error = ext4_commit_super(sb, 1);
4492 out:
4493 	/* we rely on upper layer to stop further updates */
4494 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4495 	return error;
4496 }
4497 
4498 /*
4499  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4500  * flag here, even though the filesystem is not technically dirty yet.
4501  */
4502 static int ext4_unfreeze(struct super_block *sb)
4503 {
4504 	if (sb->s_flags & MS_RDONLY)
4505 		return 0;
4506 
4507 	/* Reset the needs_recovery flag before the fs is unlocked. */
4508 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4509 	ext4_commit_super(sb, 1);
4510 	return 0;
4511 }
4512 
4513 /*
4514  * Structure to save mount options for ext4_remount's benefit
4515  */
4516 struct ext4_mount_options {
4517 	unsigned long s_mount_opt;
4518 	unsigned long s_mount_opt2;
4519 	kuid_t s_resuid;
4520 	kgid_t s_resgid;
4521 	unsigned long s_commit_interval;
4522 	u32 s_min_batch_time, s_max_batch_time;
4523 #ifdef CONFIG_QUOTA
4524 	int s_jquota_fmt;
4525 	char *s_qf_names[MAXQUOTAS];
4526 #endif
4527 };
4528 
4529 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4530 {
4531 	struct ext4_super_block *es;
4532 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4533 	unsigned long old_sb_flags;
4534 	struct ext4_mount_options old_opts;
4535 	int enable_quota = 0;
4536 	ext4_group_t g;
4537 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4538 	int err = 0;
4539 #ifdef CONFIG_QUOTA
4540 	int i, j;
4541 #endif
4542 	char *orig_data = kstrdup(data, GFP_KERNEL);
4543 
4544 	/* Store the original options */
4545 	old_sb_flags = sb->s_flags;
4546 	old_opts.s_mount_opt = sbi->s_mount_opt;
4547 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4548 	old_opts.s_resuid = sbi->s_resuid;
4549 	old_opts.s_resgid = sbi->s_resgid;
4550 	old_opts.s_commit_interval = sbi->s_commit_interval;
4551 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4552 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4553 #ifdef CONFIG_QUOTA
4554 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4555 	for (i = 0; i < MAXQUOTAS; i++)
4556 		if (sbi->s_qf_names[i]) {
4557 			old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4558 							 GFP_KERNEL);
4559 			if (!old_opts.s_qf_names[i]) {
4560 				for (j = 0; j < i; j++)
4561 					kfree(old_opts.s_qf_names[j]);
4562 				kfree(orig_data);
4563 				return -ENOMEM;
4564 			}
4565 		} else
4566 			old_opts.s_qf_names[i] = NULL;
4567 #endif
4568 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4569 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4570 
4571 	/*
4572 	 * Allow the "check" option to be passed as a remount option.
4573 	 */
4574 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4575 		err = -EINVAL;
4576 		goto restore_opts;
4577 	}
4578 
4579 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4580 		ext4_abort(sb, "Abort forced by user");
4581 
4582 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4583 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4584 
4585 	es = sbi->s_es;
4586 
4587 	if (sbi->s_journal) {
4588 		ext4_init_journal_params(sb, sbi->s_journal);
4589 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4590 	}
4591 
4592 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4593 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4594 			err = -EROFS;
4595 			goto restore_opts;
4596 		}
4597 
4598 		if (*flags & MS_RDONLY) {
4599 			err = dquot_suspend(sb, -1);
4600 			if (err < 0)
4601 				goto restore_opts;
4602 
4603 			/*
4604 			 * First of all, the unconditional stuff we have to do
4605 			 * to disable replay of the journal when we next remount
4606 			 */
4607 			sb->s_flags |= MS_RDONLY;
4608 
4609 			/*
4610 			 * OK, test if we are remounting a valid rw partition
4611 			 * readonly, and if so set the rdonly flag and then
4612 			 * mark the partition as valid again.
4613 			 */
4614 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4615 			    (sbi->s_mount_state & EXT4_VALID_FS))
4616 				es->s_state = cpu_to_le16(sbi->s_mount_state);
4617 
4618 			if (sbi->s_journal)
4619 				ext4_mark_recovery_complete(sb, es);
4620 		} else {
4621 			/* Make sure we can mount this feature set readwrite */
4622 			if (!ext4_feature_set_ok(sb, 0)) {
4623 				err = -EROFS;
4624 				goto restore_opts;
4625 			}
4626 			/*
4627 			 * Make sure the group descriptor checksums
4628 			 * are sane.  If they aren't, refuse to remount r/w.
4629 			 */
4630 			for (g = 0; g < sbi->s_groups_count; g++) {
4631 				struct ext4_group_desc *gdp =
4632 					ext4_get_group_desc(sb, g, NULL);
4633 
4634 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4635 					ext4_msg(sb, KERN_ERR,
4636 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
4637 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4638 					       le16_to_cpu(gdp->bg_checksum));
4639 					err = -EINVAL;
4640 					goto restore_opts;
4641 				}
4642 			}
4643 
4644 			/*
4645 			 * If we have an unprocessed orphan list hanging
4646 			 * around from a previously readonly bdev mount,
4647 			 * require a full umount/remount for now.
4648 			 */
4649 			if (es->s_last_orphan) {
4650 				ext4_msg(sb, KERN_WARNING, "Couldn't "
4651 				       "remount RDWR because of unprocessed "
4652 				       "orphan inode list.  Please "
4653 				       "umount/remount instead");
4654 				err = -EINVAL;
4655 				goto restore_opts;
4656 			}
4657 
4658 			/*
4659 			 * Mounting a RDONLY partition read-write, so reread
4660 			 * and store the current valid flag.  (It may have
4661 			 * been changed by e2fsck since we originally mounted
4662 			 * the partition.)
4663 			 */
4664 			if (sbi->s_journal)
4665 				ext4_clear_journal_err(sb, es);
4666 			sbi->s_mount_state = le16_to_cpu(es->s_state);
4667 			if (!ext4_setup_super(sb, es, 0))
4668 				sb->s_flags &= ~MS_RDONLY;
4669 			if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4670 						     EXT4_FEATURE_INCOMPAT_MMP))
4671 				if (ext4_multi_mount_protect(sb,
4672 						le64_to_cpu(es->s_mmp_block))) {
4673 					err = -EROFS;
4674 					goto restore_opts;
4675 				}
4676 			enable_quota = 1;
4677 		}
4678 	}
4679 
4680 	/*
4681 	 * Reinitialize lazy itable initialization thread based on
4682 	 * current settings
4683 	 */
4684 	if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4685 		ext4_unregister_li_request(sb);
4686 	else {
4687 		ext4_group_t first_not_zeroed;
4688 		first_not_zeroed = ext4_has_uninit_itable(sb);
4689 		ext4_register_li_request(sb, first_not_zeroed);
4690 	}
4691 
4692 	ext4_setup_system_zone(sb);
4693 	if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4694 		ext4_commit_super(sb, 1);
4695 
4696 #ifdef CONFIG_QUOTA
4697 	/* Release old quota file names */
4698 	for (i = 0; i < MAXQUOTAS; i++)
4699 		kfree(old_opts.s_qf_names[i]);
4700 	if (enable_quota) {
4701 		if (sb_any_quota_suspended(sb))
4702 			dquot_resume(sb, -1);
4703 		else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4704 					EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4705 			err = ext4_enable_quotas(sb);
4706 			if (err)
4707 				goto restore_opts;
4708 		}
4709 	}
4710 #endif
4711 
4712 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4713 	kfree(orig_data);
4714 	return 0;
4715 
4716 restore_opts:
4717 	sb->s_flags = old_sb_flags;
4718 	sbi->s_mount_opt = old_opts.s_mount_opt;
4719 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4720 	sbi->s_resuid = old_opts.s_resuid;
4721 	sbi->s_resgid = old_opts.s_resgid;
4722 	sbi->s_commit_interval = old_opts.s_commit_interval;
4723 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
4724 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
4725 #ifdef CONFIG_QUOTA
4726 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4727 	for (i = 0; i < MAXQUOTAS; i++) {
4728 		kfree(sbi->s_qf_names[i]);
4729 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4730 	}
4731 #endif
4732 	kfree(orig_data);
4733 	return err;
4734 }
4735 
4736 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4737 {
4738 	struct super_block *sb = dentry->d_sb;
4739 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4740 	struct ext4_super_block *es = sbi->s_es;
4741 	ext4_fsblk_t overhead = 0;
4742 	u64 fsid;
4743 	s64 bfree;
4744 
4745 	if (!test_opt(sb, MINIX_DF))
4746 		overhead = sbi->s_overhead;
4747 
4748 	buf->f_type = EXT4_SUPER_MAGIC;
4749 	buf->f_bsize = sb->s_blocksize;
4750 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4751 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4752 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4753 	/* prevent underflow in case that few free space is available */
4754 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4755 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4756 	if (buf->f_bfree < ext4_r_blocks_count(es))
4757 		buf->f_bavail = 0;
4758 	buf->f_files = le32_to_cpu(es->s_inodes_count);
4759 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4760 	buf->f_namelen = EXT4_NAME_LEN;
4761 	fsid = le64_to_cpup((void *)es->s_uuid) ^
4762 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4763 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4764 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4765 
4766 	return 0;
4767 }
4768 
4769 /* Helper function for writing quotas on sync - we need to start transaction
4770  * before quota file is locked for write. Otherwise the are possible deadlocks:
4771  * Process 1                         Process 2
4772  * ext4_create()                     quota_sync()
4773  *   jbd2_journal_start()                  write_dquot()
4774  *   dquot_initialize()                         down(dqio_mutex)
4775  *     down(dqio_mutex)                    jbd2_journal_start()
4776  *
4777  */
4778 
4779 #ifdef CONFIG_QUOTA
4780 
4781 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4782 {
4783 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4784 }
4785 
4786 static int ext4_write_dquot(struct dquot *dquot)
4787 {
4788 	int ret, err;
4789 	handle_t *handle;
4790 	struct inode *inode;
4791 
4792 	inode = dquot_to_inode(dquot);
4793 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4794 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4795 	if (IS_ERR(handle))
4796 		return PTR_ERR(handle);
4797 	ret = dquot_commit(dquot);
4798 	err = ext4_journal_stop(handle);
4799 	if (!ret)
4800 		ret = err;
4801 	return ret;
4802 }
4803 
4804 static int ext4_acquire_dquot(struct dquot *dquot)
4805 {
4806 	int ret, err;
4807 	handle_t *handle;
4808 
4809 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4810 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4811 	if (IS_ERR(handle))
4812 		return PTR_ERR(handle);
4813 	ret = dquot_acquire(dquot);
4814 	err = ext4_journal_stop(handle);
4815 	if (!ret)
4816 		ret = err;
4817 	return ret;
4818 }
4819 
4820 static int ext4_release_dquot(struct dquot *dquot)
4821 {
4822 	int ret, err;
4823 	handle_t *handle;
4824 
4825 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4826 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4827 	if (IS_ERR(handle)) {
4828 		/* Release dquot anyway to avoid endless cycle in dqput() */
4829 		dquot_release(dquot);
4830 		return PTR_ERR(handle);
4831 	}
4832 	ret = dquot_release(dquot);
4833 	err = ext4_journal_stop(handle);
4834 	if (!ret)
4835 		ret = err;
4836 	return ret;
4837 }
4838 
4839 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4840 {
4841 	struct super_block *sb = dquot->dq_sb;
4842 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4843 
4844 	/* Are we journaling quotas? */
4845 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4846 	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4847 		dquot_mark_dquot_dirty(dquot);
4848 		return ext4_write_dquot(dquot);
4849 	} else {
4850 		return dquot_mark_dquot_dirty(dquot);
4851 	}
4852 }
4853 
4854 static int ext4_write_info(struct super_block *sb, int type)
4855 {
4856 	int ret, err;
4857 	handle_t *handle;
4858 
4859 	/* Data block + inode block */
4860 	handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
4861 	if (IS_ERR(handle))
4862 		return PTR_ERR(handle);
4863 	ret = dquot_commit_info(sb, type);
4864 	err = ext4_journal_stop(handle);
4865 	if (!ret)
4866 		ret = err;
4867 	return ret;
4868 }
4869 
4870 /*
4871  * Turn on quotas during mount time - we need to find
4872  * the quota file and such...
4873  */
4874 static int ext4_quota_on_mount(struct super_block *sb, int type)
4875 {
4876 	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4877 					EXT4_SB(sb)->s_jquota_fmt, type);
4878 }
4879 
4880 /*
4881  * Standard function to be called on quota_on
4882  */
4883 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4884 			 struct path *path)
4885 {
4886 	int err;
4887 
4888 	if (!test_opt(sb, QUOTA))
4889 		return -EINVAL;
4890 
4891 	/* Quotafile not on the same filesystem? */
4892 	if (path->dentry->d_sb != sb)
4893 		return -EXDEV;
4894 	/* Journaling quota? */
4895 	if (EXT4_SB(sb)->s_qf_names[type]) {
4896 		/* Quotafile not in fs root? */
4897 		if (path->dentry->d_parent != sb->s_root)
4898 			ext4_msg(sb, KERN_WARNING,
4899 				"Quota file not on filesystem root. "
4900 				"Journaled quota will not work");
4901 	}
4902 
4903 	/*
4904 	 * When we journal data on quota file, we have to flush journal to see
4905 	 * all updates to the file when we bypass pagecache...
4906 	 */
4907 	if (EXT4_SB(sb)->s_journal &&
4908 	    ext4_should_journal_data(path->dentry->d_inode)) {
4909 		/*
4910 		 * We don't need to lock updates but journal_flush() could
4911 		 * otherwise be livelocked...
4912 		 */
4913 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4914 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4915 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4916 		if (err)
4917 			return err;
4918 	}
4919 
4920 	return dquot_quota_on(sb, type, format_id, path);
4921 }
4922 
4923 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4924 			     unsigned int flags)
4925 {
4926 	int err;
4927 	struct inode *qf_inode;
4928 	unsigned long qf_inums[MAXQUOTAS] = {
4929 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4930 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4931 	};
4932 
4933 	BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4934 
4935 	if (!qf_inums[type])
4936 		return -EPERM;
4937 
4938 	qf_inode = ext4_iget(sb, qf_inums[type]);
4939 	if (IS_ERR(qf_inode)) {
4940 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4941 		return PTR_ERR(qf_inode);
4942 	}
4943 
4944 	err = dquot_enable(qf_inode, type, format_id, flags);
4945 	iput(qf_inode);
4946 
4947 	return err;
4948 }
4949 
4950 /* Enable usage tracking for all quota types. */
4951 static int ext4_enable_quotas(struct super_block *sb)
4952 {
4953 	int type, err = 0;
4954 	unsigned long qf_inums[MAXQUOTAS] = {
4955 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4956 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4957 	};
4958 
4959 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4960 	for (type = 0; type < MAXQUOTAS; type++) {
4961 		if (qf_inums[type]) {
4962 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4963 						DQUOT_USAGE_ENABLED);
4964 			if (err) {
4965 				ext4_warning(sb,
4966 					"Failed to enable quota tracking "
4967 					"(type=%d, err=%d). Please run "
4968 					"e2fsck to fix.", type, err);
4969 				return err;
4970 			}
4971 		}
4972 	}
4973 	return 0;
4974 }
4975 
4976 /*
4977  * quota_on function that is used when QUOTA feature is set.
4978  */
4979 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4980 				 int format_id)
4981 {
4982 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4983 		return -EINVAL;
4984 
4985 	/*
4986 	 * USAGE was enabled at mount time. Only need to enable LIMITS now.
4987 	 */
4988 	return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
4989 }
4990 
4991 static int ext4_quota_off(struct super_block *sb, int type)
4992 {
4993 	struct inode *inode = sb_dqopt(sb)->files[type];
4994 	handle_t *handle;
4995 
4996 	/* Force all delayed allocation blocks to be allocated.
4997 	 * Caller already holds s_umount sem */
4998 	if (test_opt(sb, DELALLOC))
4999 		sync_filesystem(sb);
5000 
5001 	if (!inode)
5002 		goto out;
5003 
5004 	/* Update modification times of quota files when userspace can
5005 	 * start looking at them */
5006 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5007 	if (IS_ERR(handle))
5008 		goto out;
5009 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5010 	ext4_mark_inode_dirty(handle, inode);
5011 	ext4_journal_stop(handle);
5012 
5013 out:
5014 	return dquot_quota_off(sb, type);
5015 }
5016 
5017 /*
5018  * quota_off function that is used when QUOTA feature is set.
5019  */
5020 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5021 {
5022 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5023 		return -EINVAL;
5024 
5025 	/* Disable only the limits. */
5026 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5027 }
5028 
5029 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5030  * acquiring the locks... As quota files are never truncated and quota code
5031  * itself serializes the operations (and no one else should touch the files)
5032  * we don't have to be afraid of races */
5033 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5034 			       size_t len, loff_t off)
5035 {
5036 	struct inode *inode = sb_dqopt(sb)->files[type];
5037 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5038 	int err = 0;
5039 	int offset = off & (sb->s_blocksize - 1);
5040 	int tocopy;
5041 	size_t toread;
5042 	struct buffer_head *bh;
5043 	loff_t i_size = i_size_read(inode);
5044 
5045 	if (off > i_size)
5046 		return 0;
5047 	if (off+len > i_size)
5048 		len = i_size-off;
5049 	toread = len;
5050 	while (toread > 0) {
5051 		tocopy = sb->s_blocksize - offset < toread ?
5052 				sb->s_blocksize - offset : toread;
5053 		bh = ext4_bread(NULL, inode, blk, 0, &err);
5054 		if (err)
5055 			return err;
5056 		if (!bh)	/* A hole? */
5057 			memset(data, 0, tocopy);
5058 		else
5059 			memcpy(data, bh->b_data+offset, tocopy);
5060 		brelse(bh);
5061 		offset = 0;
5062 		toread -= tocopy;
5063 		data += tocopy;
5064 		blk++;
5065 	}
5066 	return len;
5067 }
5068 
5069 /* Write to quotafile (we know the transaction is already started and has
5070  * enough credits) */
5071 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5072 				const char *data, size_t len, loff_t off)
5073 {
5074 	struct inode *inode = sb_dqopt(sb)->files[type];
5075 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5076 	int err = 0;
5077 	int offset = off & (sb->s_blocksize - 1);
5078 	struct buffer_head *bh;
5079 	handle_t *handle = journal_current_handle();
5080 
5081 	if (EXT4_SB(sb)->s_journal && !handle) {
5082 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5083 			" cancelled because transaction is not started",
5084 			(unsigned long long)off, (unsigned long long)len);
5085 		return -EIO;
5086 	}
5087 	/*
5088 	 * Since we account only one data block in transaction credits,
5089 	 * then it is impossible to cross a block boundary.
5090 	 */
5091 	if (sb->s_blocksize - offset < len) {
5092 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5093 			" cancelled because not block aligned",
5094 			(unsigned long long)off, (unsigned long long)len);
5095 		return -EIO;
5096 	}
5097 
5098 	bh = ext4_bread(handle, inode, blk, 1, &err);
5099 	if (!bh)
5100 		goto out;
5101 	err = ext4_journal_get_write_access(handle, bh);
5102 	if (err) {
5103 		brelse(bh);
5104 		goto out;
5105 	}
5106 	lock_buffer(bh);
5107 	memcpy(bh->b_data+offset, data, len);
5108 	flush_dcache_page(bh->b_page);
5109 	unlock_buffer(bh);
5110 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5111 	brelse(bh);
5112 out:
5113 	if (err)
5114 		return err;
5115 	if (inode->i_size < off + len) {
5116 		i_size_write(inode, off + len);
5117 		EXT4_I(inode)->i_disksize = inode->i_size;
5118 		ext4_mark_inode_dirty(handle, inode);
5119 	}
5120 	return len;
5121 }
5122 
5123 #endif
5124 
5125 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5126 		       const char *dev_name, void *data)
5127 {
5128 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5129 }
5130 
5131 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5132 static inline void register_as_ext2(void)
5133 {
5134 	int err = register_filesystem(&ext2_fs_type);
5135 	if (err)
5136 		printk(KERN_WARNING
5137 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5138 }
5139 
5140 static inline void unregister_as_ext2(void)
5141 {
5142 	unregister_filesystem(&ext2_fs_type);
5143 }
5144 
5145 static inline int ext2_feature_set_ok(struct super_block *sb)
5146 {
5147 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5148 		return 0;
5149 	if (sb->s_flags & MS_RDONLY)
5150 		return 1;
5151 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5152 		return 0;
5153 	return 1;
5154 }
5155 MODULE_ALIAS("ext2");
5156 #else
5157 static inline void register_as_ext2(void) { }
5158 static inline void unregister_as_ext2(void) { }
5159 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5160 #endif
5161 
5162 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5163 static inline void register_as_ext3(void)
5164 {
5165 	int err = register_filesystem(&ext3_fs_type);
5166 	if (err)
5167 		printk(KERN_WARNING
5168 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5169 }
5170 
5171 static inline void unregister_as_ext3(void)
5172 {
5173 	unregister_filesystem(&ext3_fs_type);
5174 }
5175 
5176 static inline int ext3_feature_set_ok(struct super_block *sb)
5177 {
5178 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5179 		return 0;
5180 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5181 		return 0;
5182 	if (sb->s_flags & MS_RDONLY)
5183 		return 1;
5184 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5185 		return 0;
5186 	return 1;
5187 }
5188 MODULE_ALIAS("ext3");
5189 #else
5190 static inline void register_as_ext3(void) { }
5191 static inline void unregister_as_ext3(void) { }
5192 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5193 #endif
5194 
5195 static struct file_system_type ext4_fs_type = {
5196 	.owner		= THIS_MODULE,
5197 	.name		= "ext4",
5198 	.mount		= ext4_mount,
5199 	.kill_sb	= kill_block_super,
5200 	.fs_flags	= FS_REQUIRES_DEV,
5201 };
5202 
5203 static int __init ext4_init_feat_adverts(void)
5204 {
5205 	struct ext4_features *ef;
5206 	int ret = -ENOMEM;
5207 
5208 	ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5209 	if (!ef)
5210 		goto out;
5211 
5212 	ef->f_kobj.kset = ext4_kset;
5213 	init_completion(&ef->f_kobj_unregister);
5214 	ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5215 				   "features");
5216 	if (ret) {
5217 		kfree(ef);
5218 		goto out;
5219 	}
5220 
5221 	ext4_feat = ef;
5222 	ret = 0;
5223 out:
5224 	return ret;
5225 }
5226 
5227 static void ext4_exit_feat_adverts(void)
5228 {
5229 	kobject_put(&ext4_feat->f_kobj);
5230 	wait_for_completion(&ext4_feat->f_kobj_unregister);
5231 	kfree(ext4_feat);
5232 }
5233 
5234 /* Shared across all ext4 file systems */
5235 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5236 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5237 
5238 static int __init ext4_init_fs(void)
5239 {
5240 	int i, err;
5241 
5242 	ext4_li_info = NULL;
5243 	mutex_init(&ext4_li_mtx);
5244 
5245 	/* Build-time check for flags consistency */
5246 	ext4_check_flag_values();
5247 
5248 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5249 		mutex_init(&ext4__aio_mutex[i]);
5250 		init_waitqueue_head(&ext4__ioend_wq[i]);
5251 	}
5252 
5253 	err = ext4_init_es();
5254 	if (err)
5255 		return err;
5256 
5257 	err = ext4_init_pageio();
5258 	if (err)
5259 		goto out7;
5260 
5261 	err = ext4_init_system_zone();
5262 	if (err)
5263 		goto out6;
5264 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5265 	if (!ext4_kset) {
5266 		err = -ENOMEM;
5267 		goto out5;
5268 	}
5269 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5270 
5271 	err = ext4_init_feat_adverts();
5272 	if (err)
5273 		goto out4;
5274 
5275 	err = ext4_init_mballoc();
5276 	if (err)
5277 		goto out3;
5278 
5279 	err = ext4_init_xattr();
5280 	if (err)
5281 		goto out2;
5282 	err = init_inodecache();
5283 	if (err)
5284 		goto out1;
5285 	register_as_ext3();
5286 	register_as_ext2();
5287 	err = register_filesystem(&ext4_fs_type);
5288 	if (err)
5289 		goto out;
5290 
5291 	return 0;
5292 out:
5293 	unregister_as_ext2();
5294 	unregister_as_ext3();
5295 	destroy_inodecache();
5296 out1:
5297 	ext4_exit_xattr();
5298 out2:
5299 	ext4_exit_mballoc();
5300 out3:
5301 	ext4_exit_feat_adverts();
5302 out4:
5303 	if (ext4_proc_root)
5304 		remove_proc_entry("fs/ext4", NULL);
5305 	kset_unregister(ext4_kset);
5306 out5:
5307 	ext4_exit_system_zone();
5308 out6:
5309 	ext4_exit_pageio();
5310 out7:
5311 	ext4_exit_es();
5312 
5313 	return err;
5314 }
5315 
5316 static void __exit ext4_exit_fs(void)
5317 {
5318 	ext4_destroy_lazyinit_thread();
5319 	unregister_as_ext2();
5320 	unregister_as_ext3();
5321 	unregister_filesystem(&ext4_fs_type);
5322 	destroy_inodecache();
5323 	ext4_exit_xattr();
5324 	ext4_exit_mballoc();
5325 	ext4_exit_feat_adverts();
5326 	remove_proc_entry("fs/ext4", NULL);
5327 	kset_unregister(ext4_kset);
5328 	ext4_exit_system_zone();
5329 	ext4_exit_pageio();
5330 }
5331 
5332 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5333 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5334 MODULE_LICENSE("GPL");
5335 module_init(ext4_init_fs)
5336 module_exit(ext4_exit_fs)
5337