xref: /openbmc/linux/fs/ext4/super.c (revision 39b6f3aa)
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 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
85 
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
88 	.owner		= THIS_MODULE,
89 	.name		= "ext2",
90 	.mount		= ext4_mount,
91 	.kill_sb	= kill_block_super,
92 	.fs_flags	= FS_REQUIRES_DEV,
93 };
94 MODULE_ALIAS_FS("ext2");
95 MODULE_ALIAS("ext2");
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #else
98 #define IS_EXT2_SB(sb) (0)
99 #endif
100 
101 
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 	.owner		= THIS_MODULE,
105 	.name		= "ext3",
106 	.mount		= ext4_mount,
107 	.kill_sb	= kill_block_super,
108 	.fs_flags	= FS_REQUIRES_DEV,
109 };
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113 #else
114 #define IS_EXT3_SB(sb) (0)
115 #endif
116 
117 static int ext4_verify_csum_type(struct super_block *sb,
118 				 struct ext4_super_block *es)
119 {
120 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 		return 1;
123 
124 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125 }
126 
127 static __le32 ext4_superblock_csum(struct super_block *sb,
128 				   struct ext4_super_block *es)
129 {
130 	struct ext4_sb_info *sbi = EXT4_SB(sb);
131 	int offset = offsetof(struct ext4_super_block, s_checksum);
132 	__u32 csum;
133 
134 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135 
136 	return cpu_to_le32(csum);
137 }
138 
139 int ext4_superblock_csum_verify(struct super_block *sb,
140 				struct ext4_super_block *es)
141 {
142 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
143 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
144 		return 1;
145 
146 	return es->s_checksum == ext4_superblock_csum(sb, es);
147 }
148 
149 void ext4_superblock_csum_set(struct super_block *sb)
150 {
151 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152 
153 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
154 		EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
155 		return;
156 
157 	es->s_checksum = ext4_superblock_csum(sb, es);
158 }
159 
160 void *ext4_kvmalloc(size_t size, gfp_t flags)
161 {
162 	void *ret;
163 
164 	ret = kmalloc(size, flags);
165 	if (!ret)
166 		ret = __vmalloc(size, flags, PAGE_KERNEL);
167 	return ret;
168 }
169 
170 void *ext4_kvzalloc(size_t size, gfp_t flags)
171 {
172 	void *ret;
173 
174 	ret = kzalloc(size, flags);
175 	if (!ret)
176 		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
177 	return ret;
178 }
179 
180 void ext4_kvfree(void *ptr)
181 {
182 	if (is_vmalloc_addr(ptr))
183 		vfree(ptr);
184 	else
185 		kfree(ptr);
186 
187 }
188 
189 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
190 			       struct ext4_group_desc *bg)
191 {
192 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
193 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
194 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
195 }
196 
197 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
198 			       struct ext4_group_desc *bg)
199 {
200 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
201 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
202 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
203 }
204 
205 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
206 			      struct ext4_group_desc *bg)
207 {
208 	return le32_to_cpu(bg->bg_inode_table_lo) |
209 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
210 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
211 }
212 
213 __u32 ext4_free_group_clusters(struct super_block *sb,
214 			       struct ext4_group_desc *bg)
215 {
216 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
217 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
218 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
219 }
220 
221 __u32 ext4_free_inodes_count(struct super_block *sb,
222 			      struct ext4_group_desc *bg)
223 {
224 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
225 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
226 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
227 }
228 
229 __u32 ext4_used_dirs_count(struct super_block *sb,
230 			      struct ext4_group_desc *bg)
231 {
232 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
233 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
234 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
235 }
236 
237 __u32 ext4_itable_unused_count(struct super_block *sb,
238 			      struct ext4_group_desc *bg)
239 {
240 	return le16_to_cpu(bg->bg_itable_unused_lo) |
241 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
242 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
243 }
244 
245 void ext4_block_bitmap_set(struct super_block *sb,
246 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
247 {
248 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
249 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
250 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
251 }
252 
253 void ext4_inode_bitmap_set(struct super_block *sb,
254 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
255 {
256 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
257 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
258 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
259 }
260 
261 void ext4_inode_table_set(struct super_block *sb,
262 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
263 {
264 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
265 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
267 }
268 
269 void ext4_free_group_clusters_set(struct super_block *sb,
270 				  struct ext4_group_desc *bg, __u32 count)
271 {
272 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
273 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
275 }
276 
277 void ext4_free_inodes_set(struct super_block *sb,
278 			  struct ext4_group_desc *bg, __u32 count)
279 {
280 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
281 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
283 }
284 
285 void ext4_used_dirs_set(struct super_block *sb,
286 			  struct ext4_group_desc *bg, __u32 count)
287 {
288 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
289 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
290 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
291 }
292 
293 void ext4_itable_unused_set(struct super_block *sb,
294 			  struct ext4_group_desc *bg, __u32 count)
295 {
296 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
297 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
298 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
299 }
300 
301 
302 static void __save_error_info(struct super_block *sb, const char *func,
303 			    unsigned int line)
304 {
305 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
306 
307 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
308 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
309 	es->s_last_error_time = cpu_to_le32(get_seconds());
310 	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
311 	es->s_last_error_line = cpu_to_le32(line);
312 	if (!es->s_first_error_time) {
313 		es->s_first_error_time = es->s_last_error_time;
314 		strncpy(es->s_first_error_func, func,
315 			sizeof(es->s_first_error_func));
316 		es->s_first_error_line = cpu_to_le32(line);
317 		es->s_first_error_ino = es->s_last_error_ino;
318 		es->s_first_error_block = es->s_last_error_block;
319 	}
320 	/*
321 	 * Start the daily error reporting function if it hasn't been
322 	 * started already
323 	 */
324 	if (!es->s_error_count)
325 		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
326 	le32_add_cpu(&es->s_error_count, 1);
327 }
328 
329 static void save_error_info(struct super_block *sb, const char *func,
330 			    unsigned int line)
331 {
332 	__save_error_info(sb, func, line);
333 	ext4_commit_super(sb, 1);
334 }
335 
336 /*
337  * The del_gendisk() function uninitializes the disk-specific data
338  * structures, including the bdi structure, without telling anyone
339  * else.  Once this happens, any attempt to call mark_buffer_dirty()
340  * (for example, by ext4_commit_super), will cause a kernel OOPS.
341  * This is a kludge to prevent these oops until we can put in a proper
342  * hook in del_gendisk() to inform the VFS and file system layers.
343  */
344 static int block_device_ejected(struct super_block *sb)
345 {
346 	struct inode *bd_inode = sb->s_bdev->bd_inode;
347 	struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
348 
349 	return bdi->dev == NULL;
350 }
351 
352 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
353 {
354 	struct super_block		*sb = journal->j_private;
355 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
356 	int				error = is_journal_aborted(journal);
357 	struct ext4_journal_cb_entry	*jce;
358 
359 	BUG_ON(txn->t_state == T_FINISHED);
360 	spin_lock(&sbi->s_md_lock);
361 	while (!list_empty(&txn->t_private_list)) {
362 		jce = list_entry(txn->t_private_list.next,
363 				 struct ext4_journal_cb_entry, jce_list);
364 		list_del_init(&jce->jce_list);
365 		spin_unlock(&sbi->s_md_lock);
366 		jce->jce_func(sb, jce, error);
367 		spin_lock(&sbi->s_md_lock);
368 	}
369 	spin_unlock(&sbi->s_md_lock);
370 }
371 
372 /* Deal with the reporting of failure conditions on a filesystem such as
373  * inconsistencies detected or read IO failures.
374  *
375  * On ext2, we can store the error state of the filesystem in the
376  * superblock.  That is not possible on ext4, because we may have other
377  * write ordering constraints on the superblock which prevent us from
378  * writing it out straight away; and given that the journal is about to
379  * be aborted, we can't rely on the current, or future, transactions to
380  * write out the superblock safely.
381  *
382  * We'll just use the jbd2_journal_abort() error code to record an error in
383  * the journal instead.  On recovery, the journal will complain about
384  * that error until we've noted it down and cleared it.
385  */
386 
387 static void ext4_handle_error(struct super_block *sb)
388 {
389 	if (sb->s_flags & MS_RDONLY)
390 		return;
391 
392 	if (!test_opt(sb, ERRORS_CONT)) {
393 		journal_t *journal = EXT4_SB(sb)->s_journal;
394 
395 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
396 		if (journal)
397 			jbd2_journal_abort(journal, -EIO);
398 	}
399 	if (test_opt(sb, ERRORS_RO)) {
400 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
401 		sb->s_flags |= MS_RDONLY;
402 	}
403 	if (test_opt(sb, ERRORS_PANIC))
404 		panic("EXT4-fs (device %s): panic forced after error\n",
405 			sb->s_id);
406 }
407 
408 void __ext4_error(struct super_block *sb, const char *function,
409 		  unsigned int line, const char *fmt, ...)
410 {
411 	struct va_format vaf;
412 	va_list args;
413 
414 	va_start(args, fmt);
415 	vaf.fmt = fmt;
416 	vaf.va = &args;
417 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
418 	       sb->s_id, function, line, current->comm, &vaf);
419 	va_end(args);
420 	save_error_info(sb, function, line);
421 
422 	ext4_handle_error(sb);
423 }
424 
425 void ext4_error_inode(struct inode *inode, const char *function,
426 		      unsigned int line, ext4_fsblk_t block,
427 		      const char *fmt, ...)
428 {
429 	va_list args;
430 	struct va_format vaf;
431 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
432 
433 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
434 	es->s_last_error_block = cpu_to_le64(block);
435 	save_error_info(inode->i_sb, function, line);
436 	va_start(args, fmt);
437 	vaf.fmt = fmt;
438 	vaf.va = &args;
439 	if (block)
440 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
441 		       "inode #%lu: block %llu: comm %s: %pV\n",
442 		       inode->i_sb->s_id, function, line, inode->i_ino,
443 		       block, current->comm, &vaf);
444 	else
445 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
446 		       "inode #%lu: comm %s: %pV\n",
447 		       inode->i_sb->s_id, function, line, inode->i_ino,
448 		       current->comm, &vaf);
449 	va_end(args);
450 
451 	ext4_handle_error(inode->i_sb);
452 }
453 
454 void ext4_error_file(struct file *file, const char *function,
455 		     unsigned int line, ext4_fsblk_t block,
456 		     const char *fmt, ...)
457 {
458 	va_list args;
459 	struct va_format vaf;
460 	struct ext4_super_block *es;
461 	struct inode *inode = file_inode(file);
462 	char pathname[80], *path;
463 
464 	es = EXT4_SB(inode->i_sb)->s_es;
465 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
466 	save_error_info(inode->i_sb, function, line);
467 	path = d_path(&(file->f_path), pathname, sizeof(pathname));
468 	if (IS_ERR(path))
469 		path = "(unknown)";
470 	va_start(args, fmt);
471 	vaf.fmt = fmt;
472 	vaf.va = &args;
473 	if (block)
474 		printk(KERN_CRIT
475 		       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
476 		       "block %llu: comm %s: path %s: %pV\n",
477 		       inode->i_sb->s_id, function, line, inode->i_ino,
478 		       block, current->comm, path, &vaf);
479 	else
480 		printk(KERN_CRIT
481 		       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 		       "comm %s: path %s: %pV\n",
483 		       inode->i_sb->s_id, function, line, inode->i_ino,
484 		       current->comm, path, &vaf);
485 	va_end(args);
486 
487 	ext4_handle_error(inode->i_sb);
488 }
489 
490 const char *ext4_decode_error(struct super_block *sb, int errno,
491 			      char nbuf[16])
492 {
493 	char *errstr = NULL;
494 
495 	switch (errno) {
496 	case -EIO:
497 		errstr = "IO failure";
498 		break;
499 	case -ENOMEM:
500 		errstr = "Out of memory";
501 		break;
502 	case -EROFS:
503 		if (!sb || (EXT4_SB(sb)->s_journal &&
504 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
505 			errstr = "Journal has aborted";
506 		else
507 			errstr = "Readonly filesystem";
508 		break;
509 	default:
510 		/* If the caller passed in an extra buffer for unknown
511 		 * errors, textualise them now.  Else we just return
512 		 * NULL. */
513 		if (nbuf) {
514 			/* Check for truncated error codes... */
515 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
516 				errstr = nbuf;
517 		}
518 		break;
519 	}
520 
521 	return errstr;
522 }
523 
524 /* __ext4_std_error decodes expected errors from journaling functions
525  * automatically and invokes the appropriate error response.  */
526 
527 void __ext4_std_error(struct super_block *sb, const char *function,
528 		      unsigned int line, int errno)
529 {
530 	char nbuf[16];
531 	const char *errstr;
532 
533 	/* Special case: if the error is EROFS, and we're not already
534 	 * inside a transaction, then there's really no point in logging
535 	 * an error. */
536 	if (errno == -EROFS && journal_current_handle() == NULL &&
537 	    (sb->s_flags & MS_RDONLY))
538 		return;
539 
540 	errstr = ext4_decode_error(sb, errno, nbuf);
541 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
542 	       sb->s_id, function, line, errstr);
543 	save_error_info(sb, function, line);
544 
545 	ext4_handle_error(sb);
546 }
547 
548 /*
549  * ext4_abort is a much stronger failure handler than ext4_error.  The
550  * abort function may be used to deal with unrecoverable failures such
551  * as journal IO errors or ENOMEM at a critical moment in log management.
552  *
553  * We unconditionally force the filesystem into an ABORT|READONLY state,
554  * unless the error response on the fs has been set to panic in which
555  * case we take the easy way out and panic immediately.
556  */
557 
558 void __ext4_abort(struct super_block *sb, const char *function,
559 		unsigned int line, const char *fmt, ...)
560 {
561 	va_list args;
562 
563 	save_error_info(sb, function, line);
564 	va_start(args, fmt);
565 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
566 	       function, line);
567 	vprintk(fmt, args);
568 	printk("\n");
569 	va_end(args);
570 
571 	if ((sb->s_flags & MS_RDONLY) == 0) {
572 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
573 		sb->s_flags |= MS_RDONLY;
574 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
575 		if (EXT4_SB(sb)->s_journal)
576 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
577 		save_error_info(sb, function, line);
578 	}
579 	if (test_opt(sb, ERRORS_PANIC))
580 		panic("EXT4-fs panic from previous error\n");
581 }
582 
583 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
584 {
585 	struct va_format vaf;
586 	va_list args;
587 
588 	va_start(args, fmt);
589 	vaf.fmt = fmt;
590 	vaf.va = &args;
591 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
592 	va_end(args);
593 }
594 
595 void __ext4_warning(struct super_block *sb, const char *function,
596 		    unsigned int line, const char *fmt, ...)
597 {
598 	struct va_format vaf;
599 	va_list args;
600 
601 	va_start(args, fmt);
602 	vaf.fmt = fmt;
603 	vaf.va = &args;
604 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
605 	       sb->s_id, function, line, &vaf);
606 	va_end(args);
607 }
608 
609 void __ext4_grp_locked_error(const char *function, unsigned int line,
610 			     struct super_block *sb, ext4_group_t grp,
611 			     unsigned long ino, ext4_fsblk_t block,
612 			     const char *fmt, ...)
613 __releases(bitlock)
614 __acquires(bitlock)
615 {
616 	struct va_format vaf;
617 	va_list args;
618 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
619 
620 	es->s_last_error_ino = cpu_to_le32(ino);
621 	es->s_last_error_block = cpu_to_le64(block);
622 	__save_error_info(sb, function, line);
623 
624 	va_start(args, fmt);
625 
626 	vaf.fmt = fmt;
627 	vaf.va = &args;
628 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
629 	       sb->s_id, function, line, grp);
630 	if (ino)
631 		printk(KERN_CONT "inode %lu: ", ino);
632 	if (block)
633 		printk(KERN_CONT "block %llu:", (unsigned long long) block);
634 	printk(KERN_CONT "%pV\n", &vaf);
635 	va_end(args);
636 
637 	if (test_opt(sb, ERRORS_CONT)) {
638 		ext4_commit_super(sb, 0);
639 		return;
640 	}
641 
642 	ext4_unlock_group(sb, grp);
643 	ext4_handle_error(sb);
644 	/*
645 	 * We only get here in the ERRORS_RO case; relocking the group
646 	 * may be dangerous, but nothing bad will happen since the
647 	 * filesystem will have already been marked read/only and the
648 	 * journal has been aborted.  We return 1 as a hint to callers
649 	 * who might what to use the return value from
650 	 * ext4_grp_locked_error() to distinguish between the
651 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
652 	 * aggressively from the ext4 function in question, with a
653 	 * more appropriate error code.
654 	 */
655 	ext4_lock_group(sb, grp);
656 	return;
657 }
658 
659 void ext4_update_dynamic_rev(struct super_block *sb)
660 {
661 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
662 
663 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
664 		return;
665 
666 	ext4_warning(sb,
667 		     "updating to rev %d because of new feature flag, "
668 		     "running e2fsck is recommended",
669 		     EXT4_DYNAMIC_REV);
670 
671 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
672 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
673 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
674 	/* leave es->s_feature_*compat flags alone */
675 	/* es->s_uuid will be set by e2fsck if empty */
676 
677 	/*
678 	 * The rest of the superblock fields should be zero, and if not it
679 	 * means they are likely already in use, so leave them alone.  We
680 	 * can leave it up to e2fsck to clean up any inconsistencies there.
681 	 */
682 }
683 
684 /*
685  * Open the external journal device
686  */
687 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
688 {
689 	struct block_device *bdev;
690 	char b[BDEVNAME_SIZE];
691 
692 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
693 	if (IS_ERR(bdev))
694 		goto fail;
695 	return bdev;
696 
697 fail:
698 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
699 			__bdevname(dev, b), PTR_ERR(bdev));
700 	return NULL;
701 }
702 
703 /*
704  * Release the journal device
705  */
706 static void ext4_blkdev_put(struct block_device *bdev)
707 {
708 	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
709 }
710 
711 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
712 {
713 	struct block_device *bdev;
714 	bdev = sbi->journal_bdev;
715 	if (bdev) {
716 		ext4_blkdev_put(bdev);
717 		sbi->journal_bdev = NULL;
718 	}
719 }
720 
721 static inline struct inode *orphan_list_entry(struct list_head *l)
722 {
723 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
724 }
725 
726 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
727 {
728 	struct list_head *l;
729 
730 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
731 		 le32_to_cpu(sbi->s_es->s_last_orphan));
732 
733 	printk(KERN_ERR "sb_info orphan list:\n");
734 	list_for_each(l, &sbi->s_orphan) {
735 		struct inode *inode = orphan_list_entry(l);
736 		printk(KERN_ERR "  "
737 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
738 		       inode->i_sb->s_id, inode->i_ino, inode,
739 		       inode->i_mode, inode->i_nlink,
740 		       NEXT_ORPHAN(inode));
741 	}
742 }
743 
744 static void ext4_put_super(struct super_block *sb)
745 {
746 	struct ext4_sb_info *sbi = EXT4_SB(sb);
747 	struct ext4_super_block *es = sbi->s_es;
748 	int i, err;
749 
750 	ext4_unregister_li_request(sb);
751 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
752 
753 	flush_workqueue(sbi->dio_unwritten_wq);
754 	destroy_workqueue(sbi->dio_unwritten_wq);
755 
756 	if (sbi->s_journal) {
757 		err = jbd2_journal_destroy(sbi->s_journal);
758 		sbi->s_journal = NULL;
759 		if (err < 0)
760 			ext4_abort(sb, "Couldn't clean up the journal");
761 	}
762 
763 	ext4_es_unregister_shrinker(sb);
764 	del_timer(&sbi->s_err_report);
765 	ext4_release_system_zone(sb);
766 	ext4_mb_release(sb);
767 	ext4_ext_release(sb);
768 	ext4_xattr_put_super(sb);
769 
770 	if (!(sb->s_flags & MS_RDONLY)) {
771 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
772 		es->s_state = cpu_to_le16(sbi->s_mount_state);
773 	}
774 	if (!(sb->s_flags & MS_RDONLY))
775 		ext4_commit_super(sb, 1);
776 
777 	if (sbi->s_proc) {
778 		remove_proc_entry("options", sbi->s_proc);
779 		remove_proc_entry(sb->s_id, ext4_proc_root);
780 	}
781 	kobject_del(&sbi->s_kobj);
782 
783 	for (i = 0; i < sbi->s_gdb_count; i++)
784 		brelse(sbi->s_group_desc[i]);
785 	ext4_kvfree(sbi->s_group_desc);
786 	ext4_kvfree(sbi->s_flex_groups);
787 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
788 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
789 	percpu_counter_destroy(&sbi->s_dirs_counter);
790 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
791 	percpu_counter_destroy(&sbi->s_extent_cache_cnt);
792 	brelse(sbi->s_sbh);
793 #ifdef CONFIG_QUOTA
794 	for (i = 0; i < MAXQUOTAS; i++)
795 		kfree(sbi->s_qf_names[i]);
796 #endif
797 
798 	/* Debugging code just in case the in-memory inode orphan list
799 	 * isn't empty.  The on-disk one can be non-empty if we've
800 	 * detected an error and taken the fs readonly, but the
801 	 * in-memory list had better be clean by this point. */
802 	if (!list_empty(&sbi->s_orphan))
803 		dump_orphan_list(sb, sbi);
804 	J_ASSERT(list_empty(&sbi->s_orphan));
805 
806 	invalidate_bdev(sb->s_bdev);
807 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
808 		/*
809 		 * Invalidate the journal device's buffers.  We don't want them
810 		 * floating about in memory - the physical journal device may
811 		 * hotswapped, and it breaks the `ro-after' testing code.
812 		 */
813 		sync_blockdev(sbi->journal_bdev);
814 		invalidate_bdev(sbi->journal_bdev);
815 		ext4_blkdev_remove(sbi);
816 	}
817 	if (sbi->s_mmp_tsk)
818 		kthread_stop(sbi->s_mmp_tsk);
819 	sb->s_fs_info = NULL;
820 	/*
821 	 * Now that we are completely done shutting down the
822 	 * superblock, we need to actually destroy the kobject.
823 	 */
824 	kobject_put(&sbi->s_kobj);
825 	wait_for_completion(&sbi->s_kobj_unregister);
826 	if (sbi->s_chksum_driver)
827 		crypto_free_shash(sbi->s_chksum_driver);
828 	kfree(sbi->s_blockgroup_lock);
829 	kfree(sbi);
830 }
831 
832 static struct kmem_cache *ext4_inode_cachep;
833 
834 /*
835  * Called inside transaction, so use GFP_NOFS
836  */
837 static struct inode *ext4_alloc_inode(struct super_block *sb)
838 {
839 	struct ext4_inode_info *ei;
840 
841 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
842 	if (!ei)
843 		return NULL;
844 
845 	ei->vfs_inode.i_version = 1;
846 	INIT_LIST_HEAD(&ei->i_prealloc_list);
847 	spin_lock_init(&ei->i_prealloc_lock);
848 	ext4_es_init_tree(&ei->i_es_tree);
849 	rwlock_init(&ei->i_es_lock);
850 	INIT_LIST_HEAD(&ei->i_es_lru);
851 	ei->i_es_lru_nr = 0;
852 	ei->i_reserved_data_blocks = 0;
853 	ei->i_reserved_meta_blocks = 0;
854 	ei->i_allocated_meta_blocks = 0;
855 	ei->i_da_metadata_calc_len = 0;
856 	ei->i_da_metadata_calc_last_lblock = 0;
857 	spin_lock_init(&(ei->i_block_reservation_lock));
858 #ifdef CONFIG_QUOTA
859 	ei->i_reserved_quota = 0;
860 #endif
861 	ei->jinode = NULL;
862 	INIT_LIST_HEAD(&ei->i_completed_io_list);
863 	spin_lock_init(&ei->i_completed_io_lock);
864 	ei->i_sync_tid = 0;
865 	ei->i_datasync_tid = 0;
866 	atomic_set(&ei->i_ioend_count, 0);
867 	atomic_set(&ei->i_unwritten, 0);
868 	INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
869 
870 	return &ei->vfs_inode;
871 }
872 
873 static int ext4_drop_inode(struct inode *inode)
874 {
875 	int drop = generic_drop_inode(inode);
876 
877 	trace_ext4_drop_inode(inode, drop);
878 	return drop;
879 }
880 
881 static void ext4_i_callback(struct rcu_head *head)
882 {
883 	struct inode *inode = container_of(head, struct inode, i_rcu);
884 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
885 }
886 
887 static void ext4_destroy_inode(struct inode *inode)
888 {
889 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
890 		ext4_msg(inode->i_sb, KERN_ERR,
891 			 "Inode %lu (%p): orphan list check failed!",
892 			 inode->i_ino, EXT4_I(inode));
893 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
894 				EXT4_I(inode), sizeof(struct ext4_inode_info),
895 				true);
896 		dump_stack();
897 	}
898 	call_rcu(&inode->i_rcu, ext4_i_callback);
899 }
900 
901 static void init_once(void *foo)
902 {
903 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
904 
905 	INIT_LIST_HEAD(&ei->i_orphan);
906 	init_rwsem(&ei->xattr_sem);
907 	init_rwsem(&ei->i_data_sem);
908 	inode_init_once(&ei->vfs_inode);
909 }
910 
911 static int init_inodecache(void)
912 {
913 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
914 					     sizeof(struct ext4_inode_info),
915 					     0, (SLAB_RECLAIM_ACCOUNT|
916 						SLAB_MEM_SPREAD),
917 					     init_once);
918 	if (ext4_inode_cachep == NULL)
919 		return -ENOMEM;
920 	return 0;
921 }
922 
923 static void destroy_inodecache(void)
924 {
925 	/*
926 	 * Make sure all delayed rcu free inodes are flushed before we
927 	 * destroy cache.
928 	 */
929 	rcu_barrier();
930 	kmem_cache_destroy(ext4_inode_cachep);
931 }
932 
933 void ext4_clear_inode(struct inode *inode)
934 {
935 	invalidate_inode_buffers(inode);
936 	clear_inode(inode);
937 	dquot_drop(inode);
938 	ext4_discard_preallocations(inode);
939 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
940 	ext4_es_lru_del(inode);
941 	if (EXT4_I(inode)->jinode) {
942 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
943 					       EXT4_I(inode)->jinode);
944 		jbd2_free_inode(EXT4_I(inode)->jinode);
945 		EXT4_I(inode)->jinode = NULL;
946 	}
947 }
948 
949 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
950 					u64 ino, u32 generation)
951 {
952 	struct inode *inode;
953 
954 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
955 		return ERR_PTR(-ESTALE);
956 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
957 		return ERR_PTR(-ESTALE);
958 
959 	/* iget isn't really right if the inode is currently unallocated!!
960 	 *
961 	 * ext4_read_inode will return a bad_inode if the inode had been
962 	 * deleted, so we should be safe.
963 	 *
964 	 * Currently we don't know the generation for parent directory, so
965 	 * a generation of 0 means "accept any"
966 	 */
967 	inode = ext4_iget(sb, ino);
968 	if (IS_ERR(inode))
969 		return ERR_CAST(inode);
970 	if (generation && inode->i_generation != generation) {
971 		iput(inode);
972 		return ERR_PTR(-ESTALE);
973 	}
974 
975 	return inode;
976 }
977 
978 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
979 					int fh_len, int fh_type)
980 {
981 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
982 				    ext4_nfs_get_inode);
983 }
984 
985 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
986 					int fh_len, int fh_type)
987 {
988 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
989 				    ext4_nfs_get_inode);
990 }
991 
992 /*
993  * Try to release metadata pages (indirect blocks, directories) which are
994  * mapped via the block device.  Since these pages could have journal heads
995  * which would prevent try_to_free_buffers() from freeing them, we must use
996  * jbd2 layer's try_to_free_buffers() function to release them.
997  */
998 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
999 				 gfp_t wait)
1000 {
1001 	journal_t *journal = EXT4_SB(sb)->s_journal;
1002 
1003 	WARN_ON(PageChecked(page));
1004 	if (!page_has_buffers(page))
1005 		return 0;
1006 	if (journal)
1007 		return jbd2_journal_try_to_free_buffers(journal, page,
1008 							wait & ~__GFP_WAIT);
1009 	return try_to_free_buffers(page);
1010 }
1011 
1012 #ifdef CONFIG_QUOTA
1013 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1014 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1015 
1016 static int ext4_write_dquot(struct dquot *dquot);
1017 static int ext4_acquire_dquot(struct dquot *dquot);
1018 static int ext4_release_dquot(struct dquot *dquot);
1019 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1020 static int ext4_write_info(struct super_block *sb, int type);
1021 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1022 			 struct path *path);
1023 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1024 				 int format_id);
1025 static int ext4_quota_off(struct super_block *sb, int type);
1026 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1027 static int ext4_quota_on_mount(struct super_block *sb, int type);
1028 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1029 			       size_t len, loff_t off);
1030 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1031 				const char *data, size_t len, loff_t off);
1032 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1033 			     unsigned int flags);
1034 static int ext4_enable_quotas(struct super_block *sb);
1035 
1036 static const struct dquot_operations ext4_quota_operations = {
1037 	.get_reserved_space = ext4_get_reserved_space,
1038 	.write_dquot	= ext4_write_dquot,
1039 	.acquire_dquot	= ext4_acquire_dquot,
1040 	.release_dquot	= ext4_release_dquot,
1041 	.mark_dirty	= ext4_mark_dquot_dirty,
1042 	.write_info	= ext4_write_info,
1043 	.alloc_dquot	= dquot_alloc,
1044 	.destroy_dquot	= dquot_destroy,
1045 };
1046 
1047 static const struct quotactl_ops ext4_qctl_operations = {
1048 	.quota_on	= ext4_quota_on,
1049 	.quota_off	= ext4_quota_off,
1050 	.quota_sync	= dquot_quota_sync,
1051 	.get_info	= dquot_get_dqinfo,
1052 	.set_info	= dquot_set_dqinfo,
1053 	.get_dqblk	= dquot_get_dqblk,
1054 	.set_dqblk	= dquot_set_dqblk
1055 };
1056 
1057 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1058 	.quota_on_meta	= ext4_quota_on_sysfile,
1059 	.quota_off	= ext4_quota_off_sysfile,
1060 	.quota_sync	= dquot_quota_sync,
1061 	.get_info	= dquot_get_dqinfo,
1062 	.set_info	= dquot_set_dqinfo,
1063 	.get_dqblk	= dquot_get_dqblk,
1064 	.set_dqblk	= dquot_set_dqblk
1065 };
1066 #endif
1067 
1068 static const struct super_operations ext4_sops = {
1069 	.alloc_inode	= ext4_alloc_inode,
1070 	.destroy_inode	= ext4_destroy_inode,
1071 	.write_inode	= ext4_write_inode,
1072 	.dirty_inode	= ext4_dirty_inode,
1073 	.drop_inode	= ext4_drop_inode,
1074 	.evict_inode	= ext4_evict_inode,
1075 	.put_super	= ext4_put_super,
1076 	.sync_fs	= ext4_sync_fs,
1077 	.freeze_fs	= ext4_freeze,
1078 	.unfreeze_fs	= ext4_unfreeze,
1079 	.statfs		= ext4_statfs,
1080 	.remount_fs	= ext4_remount,
1081 	.show_options	= ext4_show_options,
1082 #ifdef CONFIG_QUOTA
1083 	.quota_read	= ext4_quota_read,
1084 	.quota_write	= ext4_quota_write,
1085 #endif
1086 	.bdev_try_to_free_page = bdev_try_to_free_page,
1087 };
1088 
1089 static const struct super_operations ext4_nojournal_sops = {
1090 	.alloc_inode	= ext4_alloc_inode,
1091 	.destroy_inode	= ext4_destroy_inode,
1092 	.write_inode	= ext4_write_inode,
1093 	.dirty_inode	= ext4_dirty_inode,
1094 	.drop_inode	= ext4_drop_inode,
1095 	.evict_inode	= ext4_evict_inode,
1096 	.put_super	= ext4_put_super,
1097 	.statfs		= ext4_statfs,
1098 	.remount_fs	= ext4_remount,
1099 	.show_options	= ext4_show_options,
1100 #ifdef CONFIG_QUOTA
1101 	.quota_read	= ext4_quota_read,
1102 	.quota_write	= ext4_quota_write,
1103 #endif
1104 	.bdev_try_to_free_page = bdev_try_to_free_page,
1105 };
1106 
1107 static const struct export_operations ext4_export_ops = {
1108 	.fh_to_dentry = ext4_fh_to_dentry,
1109 	.fh_to_parent = ext4_fh_to_parent,
1110 	.get_parent = ext4_get_parent,
1111 };
1112 
1113 enum {
1114 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1115 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1116 	Opt_nouid32, Opt_debug, Opt_removed,
1117 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1118 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1119 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1120 	Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1121 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1122 	Opt_data_err_abort, Opt_data_err_ignore,
1123 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1124 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1125 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1126 	Opt_usrquota, Opt_grpquota, Opt_i_version,
1127 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1128 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1129 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1130 	Opt_dioread_nolock, Opt_dioread_lock,
1131 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1132 	Opt_max_dir_size_kb,
1133 };
1134 
1135 static const match_table_t tokens = {
1136 	{Opt_bsd_df, "bsddf"},
1137 	{Opt_minix_df, "minixdf"},
1138 	{Opt_grpid, "grpid"},
1139 	{Opt_grpid, "bsdgroups"},
1140 	{Opt_nogrpid, "nogrpid"},
1141 	{Opt_nogrpid, "sysvgroups"},
1142 	{Opt_resgid, "resgid=%u"},
1143 	{Opt_resuid, "resuid=%u"},
1144 	{Opt_sb, "sb=%u"},
1145 	{Opt_err_cont, "errors=continue"},
1146 	{Opt_err_panic, "errors=panic"},
1147 	{Opt_err_ro, "errors=remount-ro"},
1148 	{Opt_nouid32, "nouid32"},
1149 	{Opt_debug, "debug"},
1150 	{Opt_removed, "oldalloc"},
1151 	{Opt_removed, "orlov"},
1152 	{Opt_user_xattr, "user_xattr"},
1153 	{Opt_nouser_xattr, "nouser_xattr"},
1154 	{Opt_acl, "acl"},
1155 	{Opt_noacl, "noacl"},
1156 	{Opt_noload, "norecovery"},
1157 	{Opt_noload, "noload"},
1158 	{Opt_removed, "nobh"},
1159 	{Opt_removed, "bh"},
1160 	{Opt_commit, "commit=%u"},
1161 	{Opt_min_batch_time, "min_batch_time=%u"},
1162 	{Opt_max_batch_time, "max_batch_time=%u"},
1163 	{Opt_journal_dev, "journal_dev=%u"},
1164 	{Opt_journal_checksum, "journal_checksum"},
1165 	{Opt_journal_async_commit, "journal_async_commit"},
1166 	{Opt_abort, "abort"},
1167 	{Opt_data_journal, "data=journal"},
1168 	{Opt_data_ordered, "data=ordered"},
1169 	{Opt_data_writeback, "data=writeback"},
1170 	{Opt_data_err_abort, "data_err=abort"},
1171 	{Opt_data_err_ignore, "data_err=ignore"},
1172 	{Opt_offusrjquota, "usrjquota="},
1173 	{Opt_usrjquota, "usrjquota=%s"},
1174 	{Opt_offgrpjquota, "grpjquota="},
1175 	{Opt_grpjquota, "grpjquota=%s"},
1176 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1177 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1178 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1179 	{Opt_grpquota, "grpquota"},
1180 	{Opt_noquota, "noquota"},
1181 	{Opt_quota, "quota"},
1182 	{Opt_usrquota, "usrquota"},
1183 	{Opt_barrier, "barrier=%u"},
1184 	{Opt_barrier, "barrier"},
1185 	{Opt_nobarrier, "nobarrier"},
1186 	{Opt_i_version, "i_version"},
1187 	{Opt_stripe, "stripe=%u"},
1188 	{Opt_delalloc, "delalloc"},
1189 	{Opt_nodelalloc, "nodelalloc"},
1190 	{Opt_removed, "mblk_io_submit"},
1191 	{Opt_removed, "nomblk_io_submit"},
1192 	{Opt_block_validity, "block_validity"},
1193 	{Opt_noblock_validity, "noblock_validity"},
1194 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1195 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1196 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1197 	{Opt_auto_da_alloc, "auto_da_alloc"},
1198 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1199 	{Opt_dioread_nolock, "dioread_nolock"},
1200 	{Opt_dioread_lock, "dioread_lock"},
1201 	{Opt_discard, "discard"},
1202 	{Opt_nodiscard, "nodiscard"},
1203 	{Opt_init_itable, "init_itable=%u"},
1204 	{Opt_init_itable, "init_itable"},
1205 	{Opt_noinit_itable, "noinit_itable"},
1206 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1207 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1208 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1209 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1210 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1211 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1212 	{Opt_err, NULL},
1213 };
1214 
1215 static ext4_fsblk_t get_sb_block(void **data)
1216 {
1217 	ext4_fsblk_t	sb_block;
1218 	char		*options = (char *) *data;
1219 
1220 	if (!options || strncmp(options, "sb=", 3) != 0)
1221 		return 1;	/* Default location */
1222 
1223 	options += 3;
1224 	/* TODO: use simple_strtoll with >32bit ext4 */
1225 	sb_block = simple_strtoul(options, &options, 0);
1226 	if (*options && *options != ',') {
1227 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1228 		       (char *) *data);
1229 		return 1;
1230 	}
1231 	if (*options == ',')
1232 		options++;
1233 	*data = (void *) options;
1234 
1235 	return sb_block;
1236 }
1237 
1238 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1239 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1240 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1241 
1242 #ifdef CONFIG_QUOTA
1243 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1244 {
1245 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1246 	char *qname;
1247 	int ret = -1;
1248 
1249 	if (sb_any_quota_loaded(sb) &&
1250 		!sbi->s_qf_names[qtype]) {
1251 		ext4_msg(sb, KERN_ERR,
1252 			"Cannot change journaled "
1253 			"quota options when quota turned on");
1254 		return -1;
1255 	}
1256 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1257 		ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1258 			 "when QUOTA feature is enabled");
1259 		return -1;
1260 	}
1261 	qname = match_strdup(args);
1262 	if (!qname) {
1263 		ext4_msg(sb, KERN_ERR,
1264 			"Not enough memory for storing quotafile name");
1265 		return -1;
1266 	}
1267 	if (sbi->s_qf_names[qtype]) {
1268 		if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1269 			ret = 1;
1270 		else
1271 			ext4_msg(sb, KERN_ERR,
1272 				 "%s quota file already specified",
1273 				 QTYPE2NAME(qtype));
1274 		goto errout;
1275 	}
1276 	if (strchr(qname, '/')) {
1277 		ext4_msg(sb, KERN_ERR,
1278 			"quotafile must be on filesystem root");
1279 		goto errout;
1280 	}
1281 	sbi->s_qf_names[qtype] = qname;
1282 	set_opt(sb, QUOTA);
1283 	return 1;
1284 errout:
1285 	kfree(qname);
1286 	return ret;
1287 }
1288 
1289 static int clear_qf_name(struct super_block *sb, int qtype)
1290 {
1291 
1292 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1293 
1294 	if (sb_any_quota_loaded(sb) &&
1295 		sbi->s_qf_names[qtype]) {
1296 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1297 			" when quota turned on");
1298 		return -1;
1299 	}
1300 	kfree(sbi->s_qf_names[qtype]);
1301 	sbi->s_qf_names[qtype] = NULL;
1302 	return 1;
1303 }
1304 #endif
1305 
1306 #define MOPT_SET	0x0001
1307 #define MOPT_CLEAR	0x0002
1308 #define MOPT_NOSUPPORT	0x0004
1309 #define MOPT_EXPLICIT	0x0008
1310 #define MOPT_CLEAR_ERR	0x0010
1311 #define MOPT_GTE0	0x0020
1312 #ifdef CONFIG_QUOTA
1313 #define MOPT_Q		0
1314 #define MOPT_QFMT	0x0040
1315 #else
1316 #define MOPT_Q		MOPT_NOSUPPORT
1317 #define MOPT_QFMT	MOPT_NOSUPPORT
1318 #endif
1319 #define MOPT_DATAJ	0x0080
1320 #define MOPT_NO_EXT2	0x0100
1321 #define MOPT_NO_EXT3	0x0200
1322 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1323 
1324 static const struct mount_opts {
1325 	int	token;
1326 	int	mount_opt;
1327 	int	flags;
1328 } ext4_mount_opts[] = {
1329 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1330 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1331 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1332 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1333 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1334 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1335 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1336 	 MOPT_EXT4_ONLY | MOPT_SET},
1337 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1338 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1339 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1340 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1341 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1342 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1343 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1344 	 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1345 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1346 	 MOPT_EXT4_ONLY | MOPT_SET},
1347 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1348 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1349 	 MOPT_EXT4_ONLY | MOPT_SET},
1350 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1351 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1352 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1353 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1354 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1355 	 MOPT_NO_EXT2 | MOPT_SET},
1356 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1357 	 MOPT_NO_EXT2 | MOPT_CLEAR},
1358 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1359 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1360 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1361 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1362 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1363 	{Opt_commit, 0, MOPT_GTE0},
1364 	{Opt_max_batch_time, 0, MOPT_GTE0},
1365 	{Opt_min_batch_time, 0, MOPT_GTE0},
1366 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1367 	{Opt_init_itable, 0, MOPT_GTE0},
1368 	{Opt_stripe, 0, MOPT_GTE0},
1369 	{Opt_resuid, 0, MOPT_GTE0},
1370 	{Opt_resgid, 0, MOPT_GTE0},
1371 	{Opt_journal_dev, 0, MOPT_GTE0},
1372 	{Opt_journal_ioprio, 0, MOPT_GTE0},
1373 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1374 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1375 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1376 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1377 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1378 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1379 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1380 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1381 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1382 #else
1383 	{Opt_acl, 0, MOPT_NOSUPPORT},
1384 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1385 #endif
1386 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1387 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1388 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1389 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1390 							MOPT_SET | MOPT_Q},
1391 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1392 							MOPT_SET | MOPT_Q},
1393 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1394 		       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1395 	{Opt_usrjquota, 0, MOPT_Q},
1396 	{Opt_grpjquota, 0, MOPT_Q},
1397 	{Opt_offusrjquota, 0, MOPT_Q},
1398 	{Opt_offgrpjquota, 0, MOPT_Q},
1399 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1400 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1401 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1402 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1403 	{Opt_err, 0, 0}
1404 };
1405 
1406 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1407 			    substring_t *args, unsigned long *journal_devnum,
1408 			    unsigned int *journal_ioprio, int is_remount)
1409 {
1410 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1411 	const struct mount_opts *m;
1412 	kuid_t uid;
1413 	kgid_t gid;
1414 	int arg = 0;
1415 
1416 #ifdef CONFIG_QUOTA
1417 	if (token == Opt_usrjquota)
1418 		return set_qf_name(sb, USRQUOTA, &args[0]);
1419 	else if (token == Opt_grpjquota)
1420 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1421 	else if (token == Opt_offusrjquota)
1422 		return clear_qf_name(sb, USRQUOTA);
1423 	else if (token == Opt_offgrpjquota)
1424 		return clear_qf_name(sb, GRPQUOTA);
1425 #endif
1426 	switch (token) {
1427 	case Opt_noacl:
1428 	case Opt_nouser_xattr:
1429 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1430 		break;
1431 	case Opt_sb:
1432 		return 1;	/* handled by get_sb_block() */
1433 	case Opt_removed:
1434 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1435 		return 1;
1436 	case Opt_abort:
1437 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1438 		return 1;
1439 	case Opt_i_version:
1440 		sb->s_flags |= MS_I_VERSION;
1441 		return 1;
1442 	}
1443 
1444 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1445 		if (token == m->token)
1446 			break;
1447 
1448 	if (m->token == Opt_err) {
1449 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1450 			 "or missing value", opt);
1451 		return -1;
1452 	}
1453 
1454 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1455 		ext4_msg(sb, KERN_ERR,
1456 			 "Mount option \"%s\" incompatible with ext2", opt);
1457 		return -1;
1458 	}
1459 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1460 		ext4_msg(sb, KERN_ERR,
1461 			 "Mount option \"%s\" incompatible with ext3", opt);
1462 		return -1;
1463 	}
1464 
1465 	if (args->from && match_int(args, &arg))
1466 		return -1;
1467 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1468 		return -1;
1469 	if (m->flags & MOPT_EXPLICIT)
1470 		set_opt2(sb, EXPLICIT_DELALLOC);
1471 	if (m->flags & MOPT_CLEAR_ERR)
1472 		clear_opt(sb, ERRORS_MASK);
1473 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1474 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1475 			 "options when quota turned on");
1476 		return -1;
1477 	}
1478 
1479 	if (m->flags & MOPT_NOSUPPORT) {
1480 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1481 	} else if (token == Opt_commit) {
1482 		if (arg == 0)
1483 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1484 		sbi->s_commit_interval = HZ * arg;
1485 	} else if (token == Opt_max_batch_time) {
1486 		if (arg == 0)
1487 			arg = EXT4_DEF_MAX_BATCH_TIME;
1488 		sbi->s_max_batch_time = arg;
1489 	} else if (token == Opt_min_batch_time) {
1490 		sbi->s_min_batch_time = arg;
1491 	} else if (token == Opt_inode_readahead_blks) {
1492 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1493 			ext4_msg(sb, KERN_ERR,
1494 				 "EXT4-fs: inode_readahead_blks must be "
1495 				 "0 or a power of 2 smaller than 2^31");
1496 			return -1;
1497 		}
1498 		sbi->s_inode_readahead_blks = arg;
1499 	} else if (token == Opt_init_itable) {
1500 		set_opt(sb, INIT_INODE_TABLE);
1501 		if (!args->from)
1502 			arg = EXT4_DEF_LI_WAIT_MULT;
1503 		sbi->s_li_wait_mult = arg;
1504 	} else if (token == Opt_max_dir_size_kb) {
1505 		sbi->s_max_dir_size_kb = arg;
1506 	} else if (token == Opt_stripe) {
1507 		sbi->s_stripe = arg;
1508 	} else if (token == Opt_resuid) {
1509 		uid = make_kuid(current_user_ns(), arg);
1510 		if (!uid_valid(uid)) {
1511 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1512 			return -1;
1513 		}
1514 		sbi->s_resuid = uid;
1515 	} else if (token == Opt_resgid) {
1516 		gid = make_kgid(current_user_ns(), arg);
1517 		if (!gid_valid(gid)) {
1518 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1519 			return -1;
1520 		}
1521 		sbi->s_resgid = gid;
1522 	} else if (token == Opt_journal_dev) {
1523 		if (is_remount) {
1524 			ext4_msg(sb, KERN_ERR,
1525 				 "Cannot specify journal on remount");
1526 			return -1;
1527 		}
1528 		*journal_devnum = arg;
1529 	} else if (token == Opt_journal_ioprio) {
1530 		if (arg > 7) {
1531 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1532 				 " (must be 0-7)");
1533 			return -1;
1534 		}
1535 		*journal_ioprio =
1536 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1537 	} else if (m->flags & MOPT_DATAJ) {
1538 		if (is_remount) {
1539 			if (!sbi->s_journal)
1540 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1541 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1542 				ext4_msg(sb, KERN_ERR,
1543 					 "Cannot change data mode on remount");
1544 				return -1;
1545 			}
1546 		} else {
1547 			clear_opt(sb, DATA_FLAGS);
1548 			sbi->s_mount_opt |= m->mount_opt;
1549 		}
1550 #ifdef CONFIG_QUOTA
1551 	} else if (m->flags & MOPT_QFMT) {
1552 		if (sb_any_quota_loaded(sb) &&
1553 		    sbi->s_jquota_fmt != m->mount_opt) {
1554 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1555 				 "quota options when quota turned on");
1556 			return -1;
1557 		}
1558 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1559 					       EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1560 			ext4_msg(sb, KERN_ERR,
1561 				 "Cannot set journaled quota options "
1562 				 "when QUOTA feature is enabled");
1563 			return -1;
1564 		}
1565 		sbi->s_jquota_fmt = m->mount_opt;
1566 #endif
1567 	} else {
1568 		if (!args->from)
1569 			arg = 1;
1570 		if (m->flags & MOPT_CLEAR)
1571 			arg = !arg;
1572 		else if (unlikely(!(m->flags & MOPT_SET))) {
1573 			ext4_msg(sb, KERN_WARNING,
1574 				 "buggy handling of option %s", opt);
1575 			WARN_ON(1);
1576 			return -1;
1577 		}
1578 		if (arg != 0)
1579 			sbi->s_mount_opt |= m->mount_opt;
1580 		else
1581 			sbi->s_mount_opt &= ~m->mount_opt;
1582 	}
1583 	return 1;
1584 }
1585 
1586 static int parse_options(char *options, struct super_block *sb,
1587 			 unsigned long *journal_devnum,
1588 			 unsigned int *journal_ioprio,
1589 			 int is_remount)
1590 {
1591 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1592 	char *p;
1593 	substring_t args[MAX_OPT_ARGS];
1594 	int token;
1595 
1596 	if (!options)
1597 		return 1;
1598 
1599 	while ((p = strsep(&options, ",")) != NULL) {
1600 		if (!*p)
1601 			continue;
1602 		/*
1603 		 * Initialize args struct so we know whether arg was
1604 		 * found; some options take optional arguments.
1605 		 */
1606 		args[0].to = args[0].from = NULL;
1607 		token = match_token(p, tokens, args);
1608 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1609 				     journal_ioprio, is_remount) < 0)
1610 			return 0;
1611 	}
1612 #ifdef CONFIG_QUOTA
1613 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1614 	    (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1615 		ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1616 			 "feature is enabled");
1617 		return 0;
1618 	}
1619 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1620 		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1621 			clear_opt(sb, USRQUOTA);
1622 
1623 		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1624 			clear_opt(sb, GRPQUOTA);
1625 
1626 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1627 			ext4_msg(sb, KERN_ERR, "old and new quota "
1628 					"format mixing");
1629 			return 0;
1630 		}
1631 
1632 		if (!sbi->s_jquota_fmt) {
1633 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1634 					"not specified");
1635 			return 0;
1636 		}
1637 	} else {
1638 		if (sbi->s_jquota_fmt) {
1639 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1640 					"specified with no journaling "
1641 					"enabled");
1642 			return 0;
1643 		}
1644 	}
1645 #endif
1646 	if (test_opt(sb, DIOREAD_NOLOCK)) {
1647 		int blocksize =
1648 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1649 
1650 		if (blocksize < PAGE_CACHE_SIZE) {
1651 			ext4_msg(sb, KERN_ERR, "can't mount with "
1652 				 "dioread_nolock if block size != PAGE_SIZE");
1653 			return 0;
1654 		}
1655 	}
1656 	return 1;
1657 }
1658 
1659 static inline void ext4_show_quota_options(struct seq_file *seq,
1660 					   struct super_block *sb)
1661 {
1662 #if defined(CONFIG_QUOTA)
1663 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1664 
1665 	if (sbi->s_jquota_fmt) {
1666 		char *fmtname = "";
1667 
1668 		switch (sbi->s_jquota_fmt) {
1669 		case QFMT_VFS_OLD:
1670 			fmtname = "vfsold";
1671 			break;
1672 		case QFMT_VFS_V0:
1673 			fmtname = "vfsv0";
1674 			break;
1675 		case QFMT_VFS_V1:
1676 			fmtname = "vfsv1";
1677 			break;
1678 		}
1679 		seq_printf(seq, ",jqfmt=%s", fmtname);
1680 	}
1681 
1682 	if (sbi->s_qf_names[USRQUOTA])
1683 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1684 
1685 	if (sbi->s_qf_names[GRPQUOTA])
1686 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1687 
1688 	if (test_opt(sb, USRQUOTA))
1689 		seq_puts(seq, ",usrquota");
1690 
1691 	if (test_opt(sb, GRPQUOTA))
1692 		seq_puts(seq, ",grpquota");
1693 #endif
1694 }
1695 
1696 static const char *token2str(int token)
1697 {
1698 	const struct match_token *t;
1699 
1700 	for (t = tokens; t->token != Opt_err; t++)
1701 		if (t->token == token && !strchr(t->pattern, '='))
1702 			break;
1703 	return t->pattern;
1704 }
1705 
1706 /*
1707  * Show an option if
1708  *  - it's set to a non-default value OR
1709  *  - if the per-sb default is different from the global default
1710  */
1711 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1712 			      int nodefs)
1713 {
1714 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1715 	struct ext4_super_block *es = sbi->s_es;
1716 	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1717 	const struct mount_opts *m;
1718 	char sep = nodefs ? '\n' : ',';
1719 
1720 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1721 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1722 
1723 	if (sbi->s_sb_block != 1)
1724 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1725 
1726 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1727 		int want_set = m->flags & MOPT_SET;
1728 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1729 		    (m->flags & MOPT_CLEAR_ERR))
1730 			continue;
1731 		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1732 			continue; /* skip if same as the default */
1733 		if ((want_set &&
1734 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1735 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1736 			continue; /* select Opt_noFoo vs Opt_Foo */
1737 		SEQ_OPTS_PRINT("%s", token2str(m->token));
1738 	}
1739 
1740 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1741 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1742 		SEQ_OPTS_PRINT("resuid=%u",
1743 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
1744 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1745 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1746 		SEQ_OPTS_PRINT("resgid=%u",
1747 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
1748 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1749 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1750 		SEQ_OPTS_PUTS("errors=remount-ro");
1751 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1752 		SEQ_OPTS_PUTS("errors=continue");
1753 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1754 		SEQ_OPTS_PUTS("errors=panic");
1755 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1756 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1757 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1758 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1759 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1760 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1761 	if (sb->s_flags & MS_I_VERSION)
1762 		SEQ_OPTS_PUTS("i_version");
1763 	if (nodefs || sbi->s_stripe)
1764 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1765 	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1766 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1767 			SEQ_OPTS_PUTS("data=journal");
1768 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1769 			SEQ_OPTS_PUTS("data=ordered");
1770 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1771 			SEQ_OPTS_PUTS("data=writeback");
1772 	}
1773 	if (nodefs ||
1774 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1775 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1776 			       sbi->s_inode_readahead_blks);
1777 
1778 	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1779 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1780 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1781 	if (nodefs || sbi->s_max_dir_size_kb)
1782 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1783 
1784 	ext4_show_quota_options(seq, sb);
1785 	return 0;
1786 }
1787 
1788 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1789 {
1790 	return _ext4_show_options(seq, root->d_sb, 0);
1791 }
1792 
1793 static int options_seq_show(struct seq_file *seq, void *offset)
1794 {
1795 	struct super_block *sb = seq->private;
1796 	int rc;
1797 
1798 	seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1799 	rc = _ext4_show_options(seq, sb, 1);
1800 	seq_puts(seq, "\n");
1801 	return rc;
1802 }
1803 
1804 static int options_open_fs(struct inode *inode, struct file *file)
1805 {
1806 	return single_open(file, options_seq_show, PDE_DATA(inode));
1807 }
1808 
1809 static const struct file_operations ext4_seq_options_fops = {
1810 	.owner = THIS_MODULE,
1811 	.open = options_open_fs,
1812 	.read = seq_read,
1813 	.llseek = seq_lseek,
1814 	.release = single_release,
1815 };
1816 
1817 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1818 			    int read_only)
1819 {
1820 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1821 	int res = 0;
1822 
1823 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1824 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1825 			 "forcing read-only mode");
1826 		res = MS_RDONLY;
1827 	}
1828 	if (read_only)
1829 		goto done;
1830 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1831 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1832 			 "running e2fsck is recommended");
1833 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1834 		ext4_msg(sb, KERN_WARNING,
1835 			 "warning: mounting fs with errors, "
1836 			 "running e2fsck is recommended");
1837 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1838 		 le16_to_cpu(es->s_mnt_count) >=
1839 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1840 		ext4_msg(sb, KERN_WARNING,
1841 			 "warning: maximal mount count reached, "
1842 			 "running e2fsck is recommended");
1843 	else if (le32_to_cpu(es->s_checkinterval) &&
1844 		(le32_to_cpu(es->s_lastcheck) +
1845 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1846 		ext4_msg(sb, KERN_WARNING,
1847 			 "warning: checktime reached, "
1848 			 "running e2fsck is recommended");
1849 	if (!sbi->s_journal)
1850 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1851 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1852 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1853 	le16_add_cpu(&es->s_mnt_count, 1);
1854 	es->s_mtime = cpu_to_le32(get_seconds());
1855 	ext4_update_dynamic_rev(sb);
1856 	if (sbi->s_journal)
1857 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1858 
1859 	ext4_commit_super(sb, 1);
1860 done:
1861 	if (test_opt(sb, DEBUG))
1862 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1863 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1864 			sb->s_blocksize,
1865 			sbi->s_groups_count,
1866 			EXT4_BLOCKS_PER_GROUP(sb),
1867 			EXT4_INODES_PER_GROUP(sb),
1868 			sbi->s_mount_opt, sbi->s_mount_opt2);
1869 
1870 	cleancache_init_fs(sb);
1871 	return res;
1872 }
1873 
1874 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1875 {
1876 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1877 	struct flex_groups *new_groups;
1878 	int size;
1879 
1880 	if (!sbi->s_log_groups_per_flex)
1881 		return 0;
1882 
1883 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
1884 	if (size <= sbi->s_flex_groups_allocated)
1885 		return 0;
1886 
1887 	size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1888 	new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1889 	if (!new_groups) {
1890 		ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1891 			 size / (int) sizeof(struct flex_groups));
1892 		return -ENOMEM;
1893 	}
1894 
1895 	if (sbi->s_flex_groups) {
1896 		memcpy(new_groups, sbi->s_flex_groups,
1897 		       (sbi->s_flex_groups_allocated *
1898 			sizeof(struct flex_groups)));
1899 		ext4_kvfree(sbi->s_flex_groups);
1900 	}
1901 	sbi->s_flex_groups = new_groups;
1902 	sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1903 	return 0;
1904 }
1905 
1906 static int ext4_fill_flex_info(struct super_block *sb)
1907 {
1908 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1909 	struct ext4_group_desc *gdp = NULL;
1910 	ext4_group_t flex_group;
1911 	unsigned int groups_per_flex = 0;
1912 	int i, err;
1913 
1914 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1915 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1916 		sbi->s_log_groups_per_flex = 0;
1917 		return 1;
1918 	}
1919 	groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1920 
1921 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1922 	if (err)
1923 		goto failed;
1924 
1925 	for (i = 0; i < sbi->s_groups_count; i++) {
1926 		gdp = ext4_get_group_desc(sb, i, NULL);
1927 
1928 		flex_group = ext4_flex_group(sbi, i);
1929 		atomic_add(ext4_free_inodes_count(sb, gdp),
1930 			   &sbi->s_flex_groups[flex_group].free_inodes);
1931 		atomic64_add(ext4_free_group_clusters(sb, gdp),
1932 			     &sbi->s_flex_groups[flex_group].free_clusters);
1933 		atomic_add(ext4_used_dirs_count(sb, gdp),
1934 			   &sbi->s_flex_groups[flex_group].used_dirs);
1935 	}
1936 
1937 	return 1;
1938 failed:
1939 	return 0;
1940 }
1941 
1942 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1943 				   struct ext4_group_desc *gdp)
1944 {
1945 	int offset;
1946 	__u16 crc = 0;
1947 	__le32 le_group = cpu_to_le32(block_group);
1948 
1949 	if ((sbi->s_es->s_feature_ro_compat &
1950 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1951 		/* Use new metadata_csum algorithm */
1952 		__le16 save_csum;
1953 		__u32 csum32;
1954 
1955 		save_csum = gdp->bg_checksum;
1956 		gdp->bg_checksum = 0;
1957 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1958 				     sizeof(le_group));
1959 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1960 				     sbi->s_desc_size);
1961 		gdp->bg_checksum = save_csum;
1962 
1963 		crc = csum32 & 0xFFFF;
1964 		goto out;
1965 	}
1966 
1967 	/* old crc16 code */
1968 	offset = offsetof(struct ext4_group_desc, bg_checksum);
1969 
1970 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1971 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1972 	crc = crc16(crc, (__u8 *)gdp, offset);
1973 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
1974 	/* for checksum of struct ext4_group_desc do the rest...*/
1975 	if ((sbi->s_es->s_feature_incompat &
1976 	     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1977 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
1978 		crc = crc16(crc, (__u8 *)gdp + offset,
1979 			    le16_to_cpu(sbi->s_es->s_desc_size) -
1980 				offset);
1981 
1982 out:
1983 	return cpu_to_le16(crc);
1984 }
1985 
1986 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1987 				struct ext4_group_desc *gdp)
1988 {
1989 	if (ext4_has_group_desc_csum(sb) &&
1990 	    (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1991 						      block_group, gdp)))
1992 		return 0;
1993 
1994 	return 1;
1995 }
1996 
1997 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1998 			      struct ext4_group_desc *gdp)
1999 {
2000 	if (!ext4_has_group_desc_csum(sb))
2001 		return;
2002 	gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2003 }
2004 
2005 /* Called at mount-time, super-block is locked */
2006 static int ext4_check_descriptors(struct super_block *sb,
2007 				  ext4_group_t *first_not_zeroed)
2008 {
2009 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2010 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2011 	ext4_fsblk_t last_block;
2012 	ext4_fsblk_t block_bitmap;
2013 	ext4_fsblk_t inode_bitmap;
2014 	ext4_fsblk_t inode_table;
2015 	int flexbg_flag = 0;
2016 	ext4_group_t i, grp = sbi->s_groups_count;
2017 
2018 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2019 		flexbg_flag = 1;
2020 
2021 	ext4_debug("Checking group descriptors");
2022 
2023 	for (i = 0; i < sbi->s_groups_count; i++) {
2024 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2025 
2026 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2027 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2028 		else
2029 			last_block = first_block +
2030 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2031 
2032 		if ((grp == sbi->s_groups_count) &&
2033 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2034 			grp = i;
2035 
2036 		block_bitmap = ext4_block_bitmap(sb, gdp);
2037 		if (block_bitmap < first_block || block_bitmap > last_block) {
2038 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2039 			       "Block bitmap for group %u not in group "
2040 			       "(block %llu)!", i, block_bitmap);
2041 			return 0;
2042 		}
2043 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2044 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2045 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2046 			       "Inode bitmap for group %u not in group "
2047 			       "(block %llu)!", i, inode_bitmap);
2048 			return 0;
2049 		}
2050 		inode_table = ext4_inode_table(sb, gdp);
2051 		if (inode_table < first_block ||
2052 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2053 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2054 			       "Inode table for group %u not in group "
2055 			       "(block %llu)!", i, inode_table);
2056 			return 0;
2057 		}
2058 		ext4_lock_group(sb, i);
2059 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2060 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2061 				 "Checksum for group %u failed (%u!=%u)",
2062 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2063 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2064 			if (!(sb->s_flags & MS_RDONLY)) {
2065 				ext4_unlock_group(sb, i);
2066 				return 0;
2067 			}
2068 		}
2069 		ext4_unlock_group(sb, i);
2070 		if (!flexbg_flag)
2071 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2072 	}
2073 	if (NULL != first_not_zeroed)
2074 		*first_not_zeroed = grp;
2075 
2076 	ext4_free_blocks_count_set(sbi->s_es,
2077 				   EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2078 	sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2079 	return 1;
2080 }
2081 
2082 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2083  * the superblock) which were deleted from all directories, but held open by
2084  * a process at the time of a crash.  We walk the list and try to delete these
2085  * inodes at recovery time (only with a read-write filesystem).
2086  *
2087  * In order to keep the orphan inode chain consistent during traversal (in
2088  * case of crash during recovery), we link each inode into the superblock
2089  * orphan list_head and handle it the same way as an inode deletion during
2090  * normal operation (which journals the operations for us).
2091  *
2092  * We only do an iget() and an iput() on each inode, which is very safe if we
2093  * accidentally point at an in-use or already deleted inode.  The worst that
2094  * can happen in this case is that we get a "bit already cleared" message from
2095  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2096  * e2fsck was run on this filesystem, and it must have already done the orphan
2097  * inode cleanup for us, so we can safely abort without any further action.
2098  */
2099 static void ext4_orphan_cleanup(struct super_block *sb,
2100 				struct ext4_super_block *es)
2101 {
2102 	unsigned int s_flags = sb->s_flags;
2103 	int nr_orphans = 0, nr_truncates = 0;
2104 #ifdef CONFIG_QUOTA
2105 	int i;
2106 #endif
2107 	if (!es->s_last_orphan) {
2108 		jbd_debug(4, "no orphan inodes to clean up\n");
2109 		return;
2110 	}
2111 
2112 	if (bdev_read_only(sb->s_bdev)) {
2113 		ext4_msg(sb, KERN_ERR, "write access "
2114 			"unavailable, skipping orphan cleanup");
2115 		return;
2116 	}
2117 
2118 	/* Check if feature set would not allow a r/w mount */
2119 	if (!ext4_feature_set_ok(sb, 0)) {
2120 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2121 			 "unknown ROCOMPAT features");
2122 		return;
2123 	}
2124 
2125 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2126 		/* don't clear list on RO mount w/ errors */
2127 		if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2128 			jbd_debug(1, "Errors on filesystem, "
2129 				  "clearing orphan list.\n");
2130 			es->s_last_orphan = 0;
2131 		}
2132 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2133 		return;
2134 	}
2135 
2136 	if (s_flags & MS_RDONLY) {
2137 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2138 		sb->s_flags &= ~MS_RDONLY;
2139 	}
2140 #ifdef CONFIG_QUOTA
2141 	/* Needed for iput() to work correctly and not trash data */
2142 	sb->s_flags |= MS_ACTIVE;
2143 	/* Turn on quotas so that they are updated correctly */
2144 	for (i = 0; i < MAXQUOTAS; i++) {
2145 		if (EXT4_SB(sb)->s_qf_names[i]) {
2146 			int ret = ext4_quota_on_mount(sb, i);
2147 			if (ret < 0)
2148 				ext4_msg(sb, KERN_ERR,
2149 					"Cannot turn on journaled "
2150 					"quota: error %d", ret);
2151 		}
2152 	}
2153 #endif
2154 
2155 	while (es->s_last_orphan) {
2156 		struct inode *inode;
2157 
2158 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2159 		if (IS_ERR(inode)) {
2160 			es->s_last_orphan = 0;
2161 			break;
2162 		}
2163 
2164 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2165 		dquot_initialize(inode);
2166 		if (inode->i_nlink) {
2167 			ext4_msg(sb, KERN_DEBUG,
2168 				"%s: truncating inode %lu to %lld bytes",
2169 				__func__, inode->i_ino, inode->i_size);
2170 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2171 				  inode->i_ino, inode->i_size);
2172 			mutex_lock(&inode->i_mutex);
2173 			ext4_truncate(inode);
2174 			mutex_unlock(&inode->i_mutex);
2175 			nr_truncates++;
2176 		} else {
2177 			ext4_msg(sb, KERN_DEBUG,
2178 				"%s: deleting unreferenced inode %lu",
2179 				__func__, inode->i_ino);
2180 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2181 				  inode->i_ino);
2182 			nr_orphans++;
2183 		}
2184 		iput(inode);  /* The delete magic happens here! */
2185 	}
2186 
2187 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2188 
2189 	if (nr_orphans)
2190 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2191 		       PLURAL(nr_orphans));
2192 	if (nr_truncates)
2193 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2194 		       PLURAL(nr_truncates));
2195 #ifdef CONFIG_QUOTA
2196 	/* Turn quotas off */
2197 	for (i = 0; i < MAXQUOTAS; i++) {
2198 		if (sb_dqopt(sb)->files[i])
2199 			dquot_quota_off(sb, i);
2200 	}
2201 #endif
2202 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2203 }
2204 
2205 /*
2206  * Maximal extent format file size.
2207  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2208  * extent format containers, within a sector_t, and within i_blocks
2209  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2210  * so that won't be a limiting factor.
2211  *
2212  * However there is other limiting factor. We do store extents in the form
2213  * of starting block and length, hence the resulting length of the extent
2214  * covering maximum file size must fit into on-disk format containers as
2215  * well. Given that length is always by 1 unit bigger than max unit (because
2216  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2217  *
2218  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2219  */
2220 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2221 {
2222 	loff_t res;
2223 	loff_t upper_limit = MAX_LFS_FILESIZE;
2224 
2225 	/* small i_blocks in vfs inode? */
2226 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2227 		/*
2228 		 * CONFIG_LBDAF is not enabled implies the inode
2229 		 * i_block represent total blocks in 512 bytes
2230 		 * 32 == size of vfs inode i_blocks * 8
2231 		 */
2232 		upper_limit = (1LL << 32) - 1;
2233 
2234 		/* total blocks in file system block size */
2235 		upper_limit >>= (blkbits - 9);
2236 		upper_limit <<= blkbits;
2237 	}
2238 
2239 	/*
2240 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2241 	 * by one fs block, so ee_len can cover the extent of maximum file
2242 	 * size
2243 	 */
2244 	res = (1LL << 32) - 1;
2245 	res <<= blkbits;
2246 
2247 	/* Sanity check against vm- & vfs- imposed limits */
2248 	if (res > upper_limit)
2249 		res = upper_limit;
2250 
2251 	return res;
2252 }
2253 
2254 /*
2255  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2256  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2257  * We need to be 1 filesystem block less than the 2^48 sector limit.
2258  */
2259 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2260 {
2261 	loff_t res = EXT4_NDIR_BLOCKS;
2262 	int meta_blocks;
2263 	loff_t upper_limit;
2264 	/* This is calculated to be the largest file size for a dense, block
2265 	 * mapped file such that the file's total number of 512-byte sectors,
2266 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2267 	 *
2268 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2269 	 * number of 512-byte sectors of the file.
2270 	 */
2271 
2272 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2273 		/*
2274 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2275 		 * the inode i_block field represents total file blocks in
2276 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2277 		 */
2278 		upper_limit = (1LL << 32) - 1;
2279 
2280 		/* total blocks in file system block size */
2281 		upper_limit >>= (bits - 9);
2282 
2283 	} else {
2284 		/*
2285 		 * We use 48 bit ext4_inode i_blocks
2286 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2287 		 * represent total number of blocks in
2288 		 * file system block size
2289 		 */
2290 		upper_limit = (1LL << 48) - 1;
2291 
2292 	}
2293 
2294 	/* indirect blocks */
2295 	meta_blocks = 1;
2296 	/* double indirect blocks */
2297 	meta_blocks += 1 + (1LL << (bits-2));
2298 	/* tripple indirect blocks */
2299 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2300 
2301 	upper_limit -= meta_blocks;
2302 	upper_limit <<= bits;
2303 
2304 	res += 1LL << (bits-2);
2305 	res += 1LL << (2*(bits-2));
2306 	res += 1LL << (3*(bits-2));
2307 	res <<= bits;
2308 	if (res > upper_limit)
2309 		res = upper_limit;
2310 
2311 	if (res > MAX_LFS_FILESIZE)
2312 		res = MAX_LFS_FILESIZE;
2313 
2314 	return res;
2315 }
2316 
2317 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2318 				   ext4_fsblk_t logical_sb_block, int nr)
2319 {
2320 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2321 	ext4_group_t bg, first_meta_bg;
2322 	int has_super = 0;
2323 
2324 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2325 
2326 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2327 	    nr < first_meta_bg)
2328 		return logical_sb_block + nr + 1;
2329 	bg = sbi->s_desc_per_block * nr;
2330 	if (ext4_bg_has_super(sb, bg))
2331 		has_super = 1;
2332 
2333 	return (has_super + ext4_group_first_block_no(sb, bg));
2334 }
2335 
2336 /**
2337  * ext4_get_stripe_size: Get the stripe size.
2338  * @sbi: In memory super block info
2339  *
2340  * If we have specified it via mount option, then
2341  * use the mount option value. If the value specified at mount time is
2342  * greater than the blocks per group use the super block value.
2343  * If the super block value is greater than blocks per group return 0.
2344  * Allocator needs it be less than blocks per group.
2345  *
2346  */
2347 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2348 {
2349 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2350 	unsigned long stripe_width =
2351 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2352 	int ret;
2353 
2354 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2355 		ret = sbi->s_stripe;
2356 	else if (stripe_width <= sbi->s_blocks_per_group)
2357 		ret = stripe_width;
2358 	else if (stride <= sbi->s_blocks_per_group)
2359 		ret = stride;
2360 	else
2361 		ret = 0;
2362 
2363 	/*
2364 	 * If the stripe width is 1, this makes no sense and
2365 	 * we set it to 0 to turn off stripe handling code.
2366 	 */
2367 	if (ret <= 1)
2368 		ret = 0;
2369 
2370 	return ret;
2371 }
2372 
2373 /* sysfs supprt */
2374 
2375 struct ext4_attr {
2376 	struct attribute attr;
2377 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2378 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2379 			 const char *, size_t);
2380 	int offset;
2381 };
2382 
2383 static int parse_strtoull(const char *buf,
2384 		unsigned long long max, unsigned long long *value)
2385 {
2386 	int ret;
2387 
2388 	ret = kstrtoull(skip_spaces(buf), 0, value);
2389 	if (!ret && *value > max)
2390 		ret = -EINVAL;
2391 	return ret;
2392 }
2393 
2394 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2395 					      struct ext4_sb_info *sbi,
2396 					      char *buf)
2397 {
2398 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2399 		(s64) EXT4_C2B(sbi,
2400 			percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2401 }
2402 
2403 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2404 					 struct ext4_sb_info *sbi, char *buf)
2405 {
2406 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2407 
2408 	if (!sb->s_bdev->bd_part)
2409 		return snprintf(buf, PAGE_SIZE, "0\n");
2410 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2411 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2412 			 sbi->s_sectors_written_start) >> 1);
2413 }
2414 
2415 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2416 					  struct ext4_sb_info *sbi, char *buf)
2417 {
2418 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2419 
2420 	if (!sb->s_bdev->bd_part)
2421 		return snprintf(buf, PAGE_SIZE, "0\n");
2422 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2423 			(unsigned long long)(sbi->s_kbytes_written +
2424 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2425 			  EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2426 }
2427 
2428 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2429 					  struct ext4_sb_info *sbi,
2430 					  const char *buf, size_t count)
2431 {
2432 	unsigned long t;
2433 	int ret;
2434 
2435 	ret = kstrtoul(skip_spaces(buf), 0, &t);
2436 	if (ret)
2437 		return ret;
2438 
2439 	if (t && (!is_power_of_2(t) || t > 0x40000000))
2440 		return -EINVAL;
2441 
2442 	sbi->s_inode_readahead_blks = t;
2443 	return count;
2444 }
2445 
2446 static ssize_t sbi_ui_show(struct ext4_attr *a,
2447 			   struct ext4_sb_info *sbi, char *buf)
2448 {
2449 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2450 
2451 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2452 }
2453 
2454 static ssize_t sbi_ui_store(struct ext4_attr *a,
2455 			    struct ext4_sb_info *sbi,
2456 			    const char *buf, size_t count)
2457 {
2458 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2459 	unsigned long t;
2460 	int ret;
2461 
2462 	ret = kstrtoul(skip_spaces(buf), 0, &t);
2463 	if (ret)
2464 		return ret;
2465 	*ui = t;
2466 	return count;
2467 }
2468 
2469 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2470 				  struct ext4_sb_info *sbi, char *buf)
2471 {
2472 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2473 		(unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2474 }
2475 
2476 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2477 				   struct ext4_sb_info *sbi,
2478 				   const char *buf, size_t count)
2479 {
2480 	unsigned long long val;
2481 	int ret;
2482 
2483 	if (parse_strtoull(buf, -1ULL, &val))
2484 		return -EINVAL;
2485 	ret = ext4_reserve_clusters(sbi, val);
2486 
2487 	return ret ? ret : count;
2488 }
2489 
2490 static ssize_t trigger_test_error(struct ext4_attr *a,
2491 				  struct ext4_sb_info *sbi,
2492 				  const char *buf, size_t count)
2493 {
2494 	int len = count;
2495 
2496 	if (!capable(CAP_SYS_ADMIN))
2497 		return -EPERM;
2498 
2499 	if (len && buf[len-1] == '\n')
2500 		len--;
2501 
2502 	if (len)
2503 		ext4_error(sbi->s_sb, "%.*s", len, buf);
2504 	return count;
2505 }
2506 
2507 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2508 static struct ext4_attr ext4_attr_##_name = {			\
2509 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2510 	.show	= _show,					\
2511 	.store	= _store,					\
2512 	.offset = offsetof(struct ext4_sb_info, _elname),	\
2513 }
2514 #define EXT4_ATTR(name, mode, show, store) \
2515 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2516 
2517 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2518 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2519 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2520 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2521 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2522 #define ATTR_LIST(name) &ext4_attr_##name.attr
2523 
2524 EXT4_RO_ATTR(delayed_allocation_blocks);
2525 EXT4_RO_ATTR(session_write_kbytes);
2526 EXT4_RO_ATTR(lifetime_write_kbytes);
2527 EXT4_RW_ATTR(reserved_clusters);
2528 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2529 		 inode_readahead_blks_store, s_inode_readahead_blks);
2530 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2531 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2532 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2533 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2534 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2535 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2536 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2537 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2538 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2539 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2540 
2541 static struct attribute *ext4_attrs[] = {
2542 	ATTR_LIST(delayed_allocation_blocks),
2543 	ATTR_LIST(session_write_kbytes),
2544 	ATTR_LIST(lifetime_write_kbytes),
2545 	ATTR_LIST(reserved_clusters),
2546 	ATTR_LIST(inode_readahead_blks),
2547 	ATTR_LIST(inode_goal),
2548 	ATTR_LIST(mb_stats),
2549 	ATTR_LIST(mb_max_to_scan),
2550 	ATTR_LIST(mb_min_to_scan),
2551 	ATTR_LIST(mb_order2_req),
2552 	ATTR_LIST(mb_stream_req),
2553 	ATTR_LIST(mb_group_prealloc),
2554 	ATTR_LIST(max_writeback_mb_bump),
2555 	ATTR_LIST(extent_max_zeroout_kb),
2556 	ATTR_LIST(trigger_fs_error),
2557 	NULL,
2558 };
2559 
2560 /* Features this copy of ext4 supports */
2561 EXT4_INFO_ATTR(lazy_itable_init);
2562 EXT4_INFO_ATTR(batched_discard);
2563 EXT4_INFO_ATTR(meta_bg_resize);
2564 
2565 static struct attribute *ext4_feat_attrs[] = {
2566 	ATTR_LIST(lazy_itable_init),
2567 	ATTR_LIST(batched_discard),
2568 	ATTR_LIST(meta_bg_resize),
2569 	NULL,
2570 };
2571 
2572 static ssize_t ext4_attr_show(struct kobject *kobj,
2573 			      struct attribute *attr, char *buf)
2574 {
2575 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2576 						s_kobj);
2577 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2578 
2579 	return a->show ? a->show(a, sbi, buf) : 0;
2580 }
2581 
2582 static ssize_t ext4_attr_store(struct kobject *kobj,
2583 			       struct attribute *attr,
2584 			       const char *buf, size_t len)
2585 {
2586 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2587 						s_kobj);
2588 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2589 
2590 	return a->store ? a->store(a, sbi, buf, len) : 0;
2591 }
2592 
2593 static void ext4_sb_release(struct kobject *kobj)
2594 {
2595 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2596 						s_kobj);
2597 	complete(&sbi->s_kobj_unregister);
2598 }
2599 
2600 static const struct sysfs_ops ext4_attr_ops = {
2601 	.show	= ext4_attr_show,
2602 	.store	= ext4_attr_store,
2603 };
2604 
2605 static struct kobj_type ext4_ktype = {
2606 	.default_attrs	= ext4_attrs,
2607 	.sysfs_ops	= &ext4_attr_ops,
2608 	.release	= ext4_sb_release,
2609 };
2610 
2611 static void ext4_feat_release(struct kobject *kobj)
2612 {
2613 	complete(&ext4_feat->f_kobj_unregister);
2614 }
2615 
2616 static struct kobj_type ext4_feat_ktype = {
2617 	.default_attrs	= ext4_feat_attrs,
2618 	.sysfs_ops	= &ext4_attr_ops,
2619 	.release	= ext4_feat_release,
2620 };
2621 
2622 /*
2623  * Check whether this filesystem can be mounted based on
2624  * the features present and the RDONLY/RDWR mount requested.
2625  * Returns 1 if this filesystem can be mounted as requested,
2626  * 0 if it cannot be.
2627  */
2628 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2629 {
2630 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2631 		ext4_msg(sb, KERN_ERR,
2632 			"Couldn't mount because of "
2633 			"unsupported optional features (%x)",
2634 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2635 			~EXT4_FEATURE_INCOMPAT_SUPP));
2636 		return 0;
2637 	}
2638 
2639 	if (readonly)
2640 		return 1;
2641 
2642 	/* Check that feature set is OK for a read-write mount */
2643 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2644 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2645 			 "unsupported optional features (%x)",
2646 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2647 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2648 		return 0;
2649 	}
2650 	/*
2651 	 * Large file size enabled file system can only be mounted
2652 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2653 	 */
2654 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2655 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2656 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2657 				 "cannot be mounted RDWR without "
2658 				 "CONFIG_LBDAF");
2659 			return 0;
2660 		}
2661 	}
2662 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2663 	    !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2664 		ext4_msg(sb, KERN_ERR,
2665 			 "Can't support bigalloc feature without "
2666 			 "extents feature\n");
2667 		return 0;
2668 	}
2669 
2670 #ifndef CONFIG_QUOTA
2671 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2672 	    !readonly) {
2673 		ext4_msg(sb, KERN_ERR,
2674 			 "Filesystem with quota feature cannot be mounted RDWR "
2675 			 "without CONFIG_QUOTA");
2676 		return 0;
2677 	}
2678 #endif  /* CONFIG_QUOTA */
2679 	return 1;
2680 }
2681 
2682 /*
2683  * This function is called once a day if we have errors logged
2684  * on the file system
2685  */
2686 static void print_daily_error_info(unsigned long arg)
2687 {
2688 	struct super_block *sb = (struct super_block *) arg;
2689 	struct ext4_sb_info *sbi;
2690 	struct ext4_super_block *es;
2691 
2692 	sbi = EXT4_SB(sb);
2693 	es = sbi->s_es;
2694 
2695 	if (es->s_error_count)
2696 		ext4_msg(sb, KERN_NOTICE, "error count: %u",
2697 			 le32_to_cpu(es->s_error_count));
2698 	if (es->s_first_error_time) {
2699 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2700 		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2701 		       (int) sizeof(es->s_first_error_func),
2702 		       es->s_first_error_func,
2703 		       le32_to_cpu(es->s_first_error_line));
2704 		if (es->s_first_error_ino)
2705 			printk(": inode %u",
2706 			       le32_to_cpu(es->s_first_error_ino));
2707 		if (es->s_first_error_block)
2708 			printk(": block %llu", (unsigned long long)
2709 			       le64_to_cpu(es->s_first_error_block));
2710 		printk("\n");
2711 	}
2712 	if (es->s_last_error_time) {
2713 		printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2714 		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2715 		       (int) sizeof(es->s_last_error_func),
2716 		       es->s_last_error_func,
2717 		       le32_to_cpu(es->s_last_error_line));
2718 		if (es->s_last_error_ino)
2719 			printk(": inode %u",
2720 			       le32_to_cpu(es->s_last_error_ino));
2721 		if (es->s_last_error_block)
2722 			printk(": block %llu", (unsigned long long)
2723 			       le64_to_cpu(es->s_last_error_block));
2724 		printk("\n");
2725 	}
2726 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2727 }
2728 
2729 /* Find next suitable group and run ext4_init_inode_table */
2730 static int ext4_run_li_request(struct ext4_li_request *elr)
2731 {
2732 	struct ext4_group_desc *gdp = NULL;
2733 	ext4_group_t group, ngroups;
2734 	struct super_block *sb;
2735 	unsigned long timeout = 0;
2736 	int ret = 0;
2737 
2738 	sb = elr->lr_super;
2739 	ngroups = EXT4_SB(sb)->s_groups_count;
2740 
2741 	sb_start_write(sb);
2742 	for (group = elr->lr_next_group; group < ngroups; group++) {
2743 		gdp = ext4_get_group_desc(sb, group, NULL);
2744 		if (!gdp) {
2745 			ret = 1;
2746 			break;
2747 		}
2748 
2749 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2750 			break;
2751 	}
2752 
2753 	if (group >= ngroups)
2754 		ret = 1;
2755 
2756 	if (!ret) {
2757 		timeout = jiffies;
2758 		ret = ext4_init_inode_table(sb, group,
2759 					    elr->lr_timeout ? 0 : 1);
2760 		if (elr->lr_timeout == 0) {
2761 			timeout = (jiffies - timeout) *
2762 				  elr->lr_sbi->s_li_wait_mult;
2763 			elr->lr_timeout = timeout;
2764 		}
2765 		elr->lr_next_sched = jiffies + elr->lr_timeout;
2766 		elr->lr_next_group = group + 1;
2767 	}
2768 	sb_end_write(sb);
2769 
2770 	return ret;
2771 }
2772 
2773 /*
2774  * Remove lr_request from the list_request and free the
2775  * request structure. Should be called with li_list_mtx held
2776  */
2777 static void ext4_remove_li_request(struct ext4_li_request *elr)
2778 {
2779 	struct ext4_sb_info *sbi;
2780 
2781 	if (!elr)
2782 		return;
2783 
2784 	sbi = elr->lr_sbi;
2785 
2786 	list_del(&elr->lr_request);
2787 	sbi->s_li_request = NULL;
2788 	kfree(elr);
2789 }
2790 
2791 static void ext4_unregister_li_request(struct super_block *sb)
2792 {
2793 	mutex_lock(&ext4_li_mtx);
2794 	if (!ext4_li_info) {
2795 		mutex_unlock(&ext4_li_mtx);
2796 		return;
2797 	}
2798 
2799 	mutex_lock(&ext4_li_info->li_list_mtx);
2800 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2801 	mutex_unlock(&ext4_li_info->li_list_mtx);
2802 	mutex_unlock(&ext4_li_mtx);
2803 }
2804 
2805 static struct task_struct *ext4_lazyinit_task;
2806 
2807 /*
2808  * This is the function where ext4lazyinit thread lives. It walks
2809  * through the request list searching for next scheduled filesystem.
2810  * When such a fs is found, run the lazy initialization request
2811  * (ext4_rn_li_request) and keep track of the time spend in this
2812  * function. Based on that time we compute next schedule time of
2813  * the request. When walking through the list is complete, compute
2814  * next waking time and put itself into sleep.
2815  */
2816 static int ext4_lazyinit_thread(void *arg)
2817 {
2818 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2819 	struct list_head *pos, *n;
2820 	struct ext4_li_request *elr;
2821 	unsigned long next_wakeup, cur;
2822 
2823 	BUG_ON(NULL == eli);
2824 
2825 cont_thread:
2826 	while (true) {
2827 		next_wakeup = MAX_JIFFY_OFFSET;
2828 
2829 		mutex_lock(&eli->li_list_mtx);
2830 		if (list_empty(&eli->li_request_list)) {
2831 			mutex_unlock(&eli->li_list_mtx);
2832 			goto exit_thread;
2833 		}
2834 
2835 		list_for_each_safe(pos, n, &eli->li_request_list) {
2836 			elr = list_entry(pos, struct ext4_li_request,
2837 					 lr_request);
2838 
2839 			if (time_after_eq(jiffies, elr->lr_next_sched)) {
2840 				if (ext4_run_li_request(elr) != 0) {
2841 					/* error, remove the lazy_init job */
2842 					ext4_remove_li_request(elr);
2843 					continue;
2844 				}
2845 			}
2846 
2847 			if (time_before(elr->lr_next_sched, next_wakeup))
2848 				next_wakeup = elr->lr_next_sched;
2849 		}
2850 		mutex_unlock(&eli->li_list_mtx);
2851 
2852 		try_to_freeze();
2853 
2854 		cur = jiffies;
2855 		if ((time_after_eq(cur, next_wakeup)) ||
2856 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
2857 			cond_resched();
2858 			continue;
2859 		}
2860 
2861 		schedule_timeout_interruptible(next_wakeup - cur);
2862 
2863 		if (kthread_should_stop()) {
2864 			ext4_clear_request_list();
2865 			goto exit_thread;
2866 		}
2867 	}
2868 
2869 exit_thread:
2870 	/*
2871 	 * It looks like the request list is empty, but we need
2872 	 * to check it under the li_list_mtx lock, to prevent any
2873 	 * additions into it, and of course we should lock ext4_li_mtx
2874 	 * to atomically free the list and ext4_li_info, because at
2875 	 * this point another ext4 filesystem could be registering
2876 	 * new one.
2877 	 */
2878 	mutex_lock(&ext4_li_mtx);
2879 	mutex_lock(&eli->li_list_mtx);
2880 	if (!list_empty(&eli->li_request_list)) {
2881 		mutex_unlock(&eli->li_list_mtx);
2882 		mutex_unlock(&ext4_li_mtx);
2883 		goto cont_thread;
2884 	}
2885 	mutex_unlock(&eli->li_list_mtx);
2886 	kfree(ext4_li_info);
2887 	ext4_li_info = NULL;
2888 	mutex_unlock(&ext4_li_mtx);
2889 
2890 	return 0;
2891 }
2892 
2893 static void ext4_clear_request_list(void)
2894 {
2895 	struct list_head *pos, *n;
2896 	struct ext4_li_request *elr;
2897 
2898 	mutex_lock(&ext4_li_info->li_list_mtx);
2899 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2900 		elr = list_entry(pos, struct ext4_li_request,
2901 				 lr_request);
2902 		ext4_remove_li_request(elr);
2903 	}
2904 	mutex_unlock(&ext4_li_info->li_list_mtx);
2905 }
2906 
2907 static int ext4_run_lazyinit_thread(void)
2908 {
2909 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2910 					 ext4_li_info, "ext4lazyinit");
2911 	if (IS_ERR(ext4_lazyinit_task)) {
2912 		int err = PTR_ERR(ext4_lazyinit_task);
2913 		ext4_clear_request_list();
2914 		kfree(ext4_li_info);
2915 		ext4_li_info = NULL;
2916 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2917 				 "initialization thread\n",
2918 				 err);
2919 		return err;
2920 	}
2921 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2922 	return 0;
2923 }
2924 
2925 /*
2926  * Check whether it make sense to run itable init. thread or not.
2927  * If there is at least one uninitialized inode table, return
2928  * corresponding group number, else the loop goes through all
2929  * groups and return total number of groups.
2930  */
2931 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2932 {
2933 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2934 	struct ext4_group_desc *gdp = NULL;
2935 
2936 	for (group = 0; group < ngroups; group++) {
2937 		gdp = ext4_get_group_desc(sb, group, NULL);
2938 		if (!gdp)
2939 			continue;
2940 
2941 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2942 			break;
2943 	}
2944 
2945 	return group;
2946 }
2947 
2948 static int ext4_li_info_new(void)
2949 {
2950 	struct ext4_lazy_init *eli = NULL;
2951 
2952 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2953 	if (!eli)
2954 		return -ENOMEM;
2955 
2956 	INIT_LIST_HEAD(&eli->li_request_list);
2957 	mutex_init(&eli->li_list_mtx);
2958 
2959 	eli->li_state |= EXT4_LAZYINIT_QUIT;
2960 
2961 	ext4_li_info = eli;
2962 
2963 	return 0;
2964 }
2965 
2966 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2967 					    ext4_group_t start)
2968 {
2969 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2970 	struct ext4_li_request *elr;
2971 	unsigned long rnd;
2972 
2973 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2974 	if (!elr)
2975 		return NULL;
2976 
2977 	elr->lr_super = sb;
2978 	elr->lr_sbi = sbi;
2979 	elr->lr_next_group = start;
2980 
2981 	/*
2982 	 * Randomize first schedule time of the request to
2983 	 * spread the inode table initialization requests
2984 	 * better.
2985 	 */
2986 	get_random_bytes(&rnd, sizeof(rnd));
2987 	elr->lr_next_sched = jiffies + (unsigned long)rnd %
2988 			     (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2989 
2990 	return elr;
2991 }
2992 
2993 int ext4_register_li_request(struct super_block *sb,
2994 			     ext4_group_t first_not_zeroed)
2995 {
2996 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2997 	struct ext4_li_request *elr = NULL;
2998 	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2999 	int ret = 0;
3000 
3001 	mutex_lock(&ext4_li_mtx);
3002 	if (sbi->s_li_request != NULL) {
3003 		/*
3004 		 * Reset timeout so it can be computed again, because
3005 		 * s_li_wait_mult might have changed.
3006 		 */
3007 		sbi->s_li_request->lr_timeout = 0;
3008 		goto out;
3009 	}
3010 
3011 	if (first_not_zeroed == ngroups ||
3012 	    (sb->s_flags & MS_RDONLY) ||
3013 	    !test_opt(sb, INIT_INODE_TABLE))
3014 		goto out;
3015 
3016 	elr = ext4_li_request_new(sb, first_not_zeroed);
3017 	if (!elr) {
3018 		ret = -ENOMEM;
3019 		goto out;
3020 	}
3021 
3022 	if (NULL == ext4_li_info) {
3023 		ret = ext4_li_info_new();
3024 		if (ret)
3025 			goto out;
3026 	}
3027 
3028 	mutex_lock(&ext4_li_info->li_list_mtx);
3029 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3030 	mutex_unlock(&ext4_li_info->li_list_mtx);
3031 
3032 	sbi->s_li_request = elr;
3033 	/*
3034 	 * set elr to NULL here since it has been inserted to
3035 	 * the request_list and the removal and free of it is
3036 	 * handled by ext4_clear_request_list from now on.
3037 	 */
3038 	elr = NULL;
3039 
3040 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3041 		ret = ext4_run_lazyinit_thread();
3042 		if (ret)
3043 			goto out;
3044 	}
3045 out:
3046 	mutex_unlock(&ext4_li_mtx);
3047 	if (ret)
3048 		kfree(elr);
3049 	return ret;
3050 }
3051 
3052 /*
3053  * We do not need to lock anything since this is called on
3054  * module unload.
3055  */
3056 static void ext4_destroy_lazyinit_thread(void)
3057 {
3058 	/*
3059 	 * If thread exited earlier
3060 	 * there's nothing to be done.
3061 	 */
3062 	if (!ext4_li_info || !ext4_lazyinit_task)
3063 		return;
3064 
3065 	kthread_stop(ext4_lazyinit_task);
3066 }
3067 
3068 static int set_journal_csum_feature_set(struct super_block *sb)
3069 {
3070 	int ret = 1;
3071 	int compat, incompat;
3072 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3073 
3074 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3075 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3076 		/* journal checksum v2 */
3077 		compat = 0;
3078 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3079 	} else {
3080 		/* journal checksum v1 */
3081 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3082 		incompat = 0;
3083 	}
3084 
3085 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3086 		ret = jbd2_journal_set_features(sbi->s_journal,
3087 				compat, 0,
3088 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3089 				incompat);
3090 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3091 		ret = jbd2_journal_set_features(sbi->s_journal,
3092 				compat, 0,
3093 				incompat);
3094 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3095 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3096 	} else {
3097 		jbd2_journal_clear_features(sbi->s_journal,
3098 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3099 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3100 				JBD2_FEATURE_INCOMPAT_CSUM_V2);
3101 	}
3102 
3103 	return ret;
3104 }
3105 
3106 /*
3107  * Note: calculating the overhead so we can be compatible with
3108  * historical BSD practice is quite difficult in the face of
3109  * clusters/bigalloc.  This is because multiple metadata blocks from
3110  * different block group can end up in the same allocation cluster.
3111  * Calculating the exact overhead in the face of clustered allocation
3112  * requires either O(all block bitmaps) in memory or O(number of block
3113  * groups**2) in time.  We will still calculate the superblock for
3114  * older file systems --- and if we come across with a bigalloc file
3115  * system with zero in s_overhead_clusters the estimate will be close to
3116  * correct especially for very large cluster sizes --- but for newer
3117  * file systems, it's better to calculate this figure once at mkfs
3118  * time, and store it in the superblock.  If the superblock value is
3119  * present (even for non-bigalloc file systems), we will use it.
3120  */
3121 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3122 			  char *buf)
3123 {
3124 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3125 	struct ext4_group_desc	*gdp;
3126 	ext4_fsblk_t		first_block, last_block, b;
3127 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3128 	int			s, j, count = 0;
3129 
3130 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3131 		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3132 			sbi->s_itb_per_group + 2);
3133 
3134 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3135 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3136 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3137 	for (i = 0; i < ngroups; i++) {
3138 		gdp = ext4_get_group_desc(sb, i, NULL);
3139 		b = ext4_block_bitmap(sb, gdp);
3140 		if (b >= first_block && b <= last_block) {
3141 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3142 			count++;
3143 		}
3144 		b = ext4_inode_bitmap(sb, gdp);
3145 		if (b >= first_block && b <= last_block) {
3146 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3147 			count++;
3148 		}
3149 		b = ext4_inode_table(sb, gdp);
3150 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3151 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3152 				int c = EXT4_B2C(sbi, b - first_block);
3153 				ext4_set_bit(c, buf);
3154 				count++;
3155 			}
3156 		if (i != grp)
3157 			continue;
3158 		s = 0;
3159 		if (ext4_bg_has_super(sb, grp)) {
3160 			ext4_set_bit(s++, buf);
3161 			count++;
3162 		}
3163 		for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3164 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3165 			count++;
3166 		}
3167 	}
3168 	if (!count)
3169 		return 0;
3170 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3171 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3172 }
3173 
3174 /*
3175  * Compute the overhead and stash it in sbi->s_overhead
3176  */
3177 int ext4_calculate_overhead(struct super_block *sb)
3178 {
3179 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3180 	struct ext4_super_block *es = sbi->s_es;
3181 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3182 	ext4_fsblk_t overhead = 0;
3183 	char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3184 
3185 	if (!buf)
3186 		return -ENOMEM;
3187 
3188 	/*
3189 	 * Compute the overhead (FS structures).  This is constant
3190 	 * for a given filesystem unless the number of block groups
3191 	 * changes so we cache the previous value until it does.
3192 	 */
3193 
3194 	/*
3195 	 * All of the blocks before first_data_block are overhead
3196 	 */
3197 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3198 
3199 	/*
3200 	 * Add the overhead found in each block group
3201 	 */
3202 	for (i = 0; i < ngroups; i++) {
3203 		int blks;
3204 
3205 		blks = count_overhead(sb, i, buf);
3206 		overhead += blks;
3207 		if (blks)
3208 			memset(buf, 0, PAGE_SIZE);
3209 		cond_resched();
3210 	}
3211 	/* Add the journal blocks as well */
3212 	if (sbi->s_journal)
3213 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3214 
3215 	sbi->s_overhead = overhead;
3216 	smp_wmb();
3217 	free_page((unsigned long) buf);
3218 	return 0;
3219 }
3220 
3221 
3222 static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi)
3223 {
3224 	ext4_fsblk_t resv_clusters;
3225 
3226 	/*
3227 	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3228 	 * This should cover the situations where we can not afford to run
3229 	 * out of space like for example punch hole, or converting
3230 	 * uninitialized extents in delalloc path. In most cases such
3231 	 * allocation would require 1, or 2 blocks, higher numbers are
3232 	 * very rare.
3233 	 */
3234 	resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits;
3235 
3236 	do_div(resv_clusters, 50);
3237 	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3238 
3239 	return resv_clusters;
3240 }
3241 
3242 
3243 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3244 {
3245 	ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3246 				sbi->s_cluster_bits;
3247 
3248 	if (count >= clusters)
3249 		return -EINVAL;
3250 
3251 	atomic64_set(&sbi->s_resv_clusters, count);
3252 	return 0;
3253 }
3254 
3255 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3256 {
3257 	char *orig_data = kstrdup(data, GFP_KERNEL);
3258 	struct buffer_head *bh;
3259 	struct ext4_super_block *es = NULL;
3260 	struct ext4_sb_info *sbi;
3261 	ext4_fsblk_t block;
3262 	ext4_fsblk_t sb_block = get_sb_block(&data);
3263 	ext4_fsblk_t logical_sb_block;
3264 	unsigned long offset = 0;
3265 	unsigned long journal_devnum = 0;
3266 	unsigned long def_mount_opts;
3267 	struct inode *root;
3268 	char *cp;
3269 	const char *descr;
3270 	int ret = -ENOMEM;
3271 	int blocksize, clustersize;
3272 	unsigned int db_count;
3273 	unsigned int i;
3274 	int needs_recovery, has_huge_files, has_bigalloc;
3275 	__u64 blocks_count;
3276 	int err = 0;
3277 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3278 	ext4_group_t first_not_zeroed;
3279 
3280 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3281 	if (!sbi)
3282 		goto out_free_orig;
3283 
3284 	sbi->s_blockgroup_lock =
3285 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3286 	if (!sbi->s_blockgroup_lock) {
3287 		kfree(sbi);
3288 		goto out_free_orig;
3289 	}
3290 	sb->s_fs_info = sbi;
3291 	sbi->s_sb = sb;
3292 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3293 	sbi->s_sb_block = sb_block;
3294 	if (sb->s_bdev->bd_part)
3295 		sbi->s_sectors_written_start =
3296 			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3297 
3298 	/* Cleanup superblock name */
3299 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3300 		*cp = '!';
3301 
3302 	/* -EINVAL is default */
3303 	ret = -EINVAL;
3304 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3305 	if (!blocksize) {
3306 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3307 		goto out_fail;
3308 	}
3309 
3310 	/*
3311 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3312 	 * block sizes.  We need to calculate the offset from buffer start.
3313 	 */
3314 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3315 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3316 		offset = do_div(logical_sb_block, blocksize);
3317 	} else {
3318 		logical_sb_block = sb_block;
3319 	}
3320 
3321 	if (!(bh = sb_bread(sb, logical_sb_block))) {
3322 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3323 		goto out_fail;
3324 	}
3325 	/*
3326 	 * Note: s_es must be initialized as soon as possible because
3327 	 *       some ext4 macro-instructions depend on its value
3328 	 */
3329 	es = (struct ext4_super_block *) (bh->b_data + offset);
3330 	sbi->s_es = es;
3331 	sb->s_magic = le16_to_cpu(es->s_magic);
3332 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3333 		goto cantfind_ext4;
3334 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3335 
3336 	/* Warn if metadata_csum and gdt_csum are both set. */
3337 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3338 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3339 	    EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3340 		ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3341 			     "redundant flags; please run fsck.");
3342 
3343 	/* Check for a known checksum algorithm */
3344 	if (!ext4_verify_csum_type(sb, es)) {
3345 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3346 			 "unknown checksum algorithm.");
3347 		silent = 1;
3348 		goto cantfind_ext4;
3349 	}
3350 
3351 	/* Load the checksum driver */
3352 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3353 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3354 		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3355 		if (IS_ERR(sbi->s_chksum_driver)) {
3356 			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3357 			ret = PTR_ERR(sbi->s_chksum_driver);
3358 			sbi->s_chksum_driver = NULL;
3359 			goto failed_mount;
3360 		}
3361 	}
3362 
3363 	/* Check superblock checksum */
3364 	if (!ext4_superblock_csum_verify(sb, es)) {
3365 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3366 			 "invalid superblock checksum.  Run e2fsck?");
3367 		silent = 1;
3368 		goto cantfind_ext4;
3369 	}
3370 
3371 	/* Precompute checksum seed for all metadata */
3372 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3373 			EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3374 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3375 					       sizeof(es->s_uuid));
3376 
3377 	/* Set defaults before we parse the mount options */
3378 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3379 	set_opt(sb, INIT_INODE_TABLE);
3380 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3381 		set_opt(sb, DEBUG);
3382 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3383 		set_opt(sb, GRPID);
3384 	if (def_mount_opts & EXT4_DEFM_UID16)
3385 		set_opt(sb, NO_UID32);
3386 	/* xattr user namespace & acls are now defaulted on */
3387 	set_opt(sb, XATTR_USER);
3388 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3389 	set_opt(sb, POSIX_ACL);
3390 #endif
3391 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3392 		set_opt(sb, JOURNAL_DATA);
3393 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3394 		set_opt(sb, ORDERED_DATA);
3395 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3396 		set_opt(sb, WRITEBACK_DATA);
3397 
3398 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3399 		set_opt(sb, ERRORS_PANIC);
3400 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3401 		set_opt(sb, ERRORS_CONT);
3402 	else
3403 		set_opt(sb, ERRORS_RO);
3404 	if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3405 		set_opt(sb, BLOCK_VALIDITY);
3406 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3407 		set_opt(sb, DISCARD);
3408 
3409 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3410 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3411 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3412 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3413 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3414 
3415 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3416 		set_opt(sb, BARRIER);
3417 
3418 	/*
3419 	 * enable delayed allocation by default
3420 	 * Use -o nodelalloc to turn it off
3421 	 */
3422 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3423 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3424 		set_opt(sb, DELALLOC);
3425 
3426 	/*
3427 	 * set default s_li_wait_mult for lazyinit, for the case there is
3428 	 * no mount option specified.
3429 	 */
3430 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3431 
3432 	if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3433 			   &journal_devnum, &journal_ioprio, 0)) {
3434 		ext4_msg(sb, KERN_WARNING,
3435 			 "failed to parse options in superblock: %s",
3436 			 sbi->s_es->s_mount_opts);
3437 	}
3438 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3439 	if (!parse_options((char *) data, sb, &journal_devnum,
3440 			   &journal_ioprio, 0))
3441 		goto failed_mount;
3442 
3443 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3444 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3445 			    "with data=journal disables delayed "
3446 			    "allocation and O_DIRECT support!\n");
3447 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3448 			ext4_msg(sb, KERN_ERR, "can't mount with "
3449 				 "both data=journal and delalloc");
3450 			goto failed_mount;
3451 		}
3452 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3453 			ext4_msg(sb, KERN_ERR, "can't mount with "
3454 				 "both data=journal and delalloc");
3455 			goto failed_mount;
3456 		}
3457 		if (test_opt(sb, DELALLOC))
3458 			clear_opt(sb, DELALLOC);
3459 	}
3460 
3461 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3462 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3463 
3464 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3465 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3466 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3467 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3468 		ext4_msg(sb, KERN_WARNING,
3469 		       "feature flags set on rev 0 fs, "
3470 		       "running e2fsck is recommended");
3471 
3472 	if (IS_EXT2_SB(sb)) {
3473 		if (ext2_feature_set_ok(sb))
3474 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3475 				 "using the ext4 subsystem");
3476 		else {
3477 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3478 				 "to feature incompatibilities");
3479 			goto failed_mount;
3480 		}
3481 	}
3482 
3483 	if (IS_EXT3_SB(sb)) {
3484 		if (ext3_feature_set_ok(sb))
3485 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3486 				 "using the ext4 subsystem");
3487 		else {
3488 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3489 				 "to feature incompatibilities");
3490 			goto failed_mount;
3491 		}
3492 	}
3493 
3494 	/*
3495 	 * Check feature flags regardless of the revision level, since we
3496 	 * previously didn't change the revision level when setting the flags,
3497 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3498 	 */
3499 	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3500 		goto failed_mount;
3501 
3502 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3503 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3504 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3505 		ext4_msg(sb, KERN_ERR,
3506 		       "Unsupported filesystem blocksize %d", blocksize);
3507 		goto failed_mount;
3508 	}
3509 
3510 	if (sb->s_blocksize != blocksize) {
3511 		/* Validate the filesystem blocksize */
3512 		if (!sb_set_blocksize(sb, blocksize)) {
3513 			ext4_msg(sb, KERN_ERR, "bad block size %d",
3514 					blocksize);
3515 			goto failed_mount;
3516 		}
3517 
3518 		brelse(bh);
3519 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3520 		offset = do_div(logical_sb_block, blocksize);
3521 		bh = sb_bread(sb, logical_sb_block);
3522 		if (!bh) {
3523 			ext4_msg(sb, KERN_ERR,
3524 			       "Can't read superblock on 2nd try");
3525 			goto failed_mount;
3526 		}
3527 		es = (struct ext4_super_block *)(bh->b_data + offset);
3528 		sbi->s_es = es;
3529 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3530 			ext4_msg(sb, KERN_ERR,
3531 			       "Magic mismatch, very weird!");
3532 			goto failed_mount;
3533 		}
3534 	}
3535 
3536 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3537 				EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3538 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3539 						      has_huge_files);
3540 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3541 
3542 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3543 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3544 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3545 	} else {
3546 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3547 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3548 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3549 		    (!is_power_of_2(sbi->s_inode_size)) ||
3550 		    (sbi->s_inode_size > blocksize)) {
3551 			ext4_msg(sb, KERN_ERR,
3552 			       "unsupported inode size: %d",
3553 			       sbi->s_inode_size);
3554 			goto failed_mount;
3555 		}
3556 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3557 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3558 	}
3559 
3560 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3561 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3562 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3563 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3564 		    !is_power_of_2(sbi->s_desc_size)) {
3565 			ext4_msg(sb, KERN_ERR,
3566 			       "unsupported descriptor size %lu",
3567 			       sbi->s_desc_size);
3568 			goto failed_mount;
3569 		}
3570 	} else
3571 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3572 
3573 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3574 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3575 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3576 		goto cantfind_ext4;
3577 
3578 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3579 	if (sbi->s_inodes_per_block == 0)
3580 		goto cantfind_ext4;
3581 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3582 					sbi->s_inodes_per_block;
3583 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3584 	sbi->s_sbh = bh;
3585 	sbi->s_mount_state = le16_to_cpu(es->s_state);
3586 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3587 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3588 
3589 	/* Do we have standard group size of blocksize * 8 blocks ? */
3590 	if (sbi->s_blocks_per_group == blocksize << 3)
3591 		set_opt2(sb, STD_GROUP_SIZE);
3592 
3593 	for (i = 0; i < 4; i++)
3594 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3595 	sbi->s_def_hash_version = es->s_def_hash_version;
3596 	i = le32_to_cpu(es->s_flags);
3597 	if (i & EXT2_FLAGS_UNSIGNED_HASH)
3598 		sbi->s_hash_unsigned = 3;
3599 	else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3600 #ifdef __CHAR_UNSIGNED__
3601 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3602 		sbi->s_hash_unsigned = 3;
3603 #else
3604 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3605 #endif
3606 	}
3607 
3608 	/* Handle clustersize */
3609 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3610 	has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3611 				EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3612 	if (has_bigalloc) {
3613 		if (clustersize < blocksize) {
3614 			ext4_msg(sb, KERN_ERR,
3615 				 "cluster size (%d) smaller than "
3616 				 "block size (%d)", clustersize, blocksize);
3617 			goto failed_mount;
3618 		}
3619 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3620 			le32_to_cpu(es->s_log_block_size);
3621 		sbi->s_clusters_per_group =
3622 			le32_to_cpu(es->s_clusters_per_group);
3623 		if (sbi->s_clusters_per_group > blocksize * 8) {
3624 			ext4_msg(sb, KERN_ERR,
3625 				 "#clusters per group too big: %lu",
3626 				 sbi->s_clusters_per_group);
3627 			goto failed_mount;
3628 		}
3629 		if (sbi->s_blocks_per_group !=
3630 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3631 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3632 				 "clusters per group (%lu) inconsistent",
3633 				 sbi->s_blocks_per_group,
3634 				 sbi->s_clusters_per_group);
3635 			goto failed_mount;
3636 		}
3637 	} else {
3638 		if (clustersize != blocksize) {
3639 			ext4_warning(sb, "fragment/cluster size (%d) != "
3640 				     "block size (%d)", clustersize,
3641 				     blocksize);
3642 			clustersize = blocksize;
3643 		}
3644 		if (sbi->s_blocks_per_group > blocksize * 8) {
3645 			ext4_msg(sb, KERN_ERR,
3646 				 "#blocks per group too big: %lu",
3647 				 sbi->s_blocks_per_group);
3648 			goto failed_mount;
3649 		}
3650 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3651 		sbi->s_cluster_bits = 0;
3652 	}
3653 	sbi->s_cluster_ratio = clustersize / blocksize;
3654 
3655 	if (sbi->s_inodes_per_group > blocksize * 8) {
3656 		ext4_msg(sb, KERN_ERR,
3657 		       "#inodes per group too big: %lu",
3658 		       sbi->s_inodes_per_group);
3659 		goto failed_mount;
3660 	}
3661 
3662 	/*
3663 	 * Test whether we have more sectors than will fit in sector_t,
3664 	 * and whether the max offset is addressable by the page cache.
3665 	 */
3666 	err = generic_check_addressable(sb->s_blocksize_bits,
3667 					ext4_blocks_count(es));
3668 	if (err) {
3669 		ext4_msg(sb, KERN_ERR, "filesystem"
3670 			 " too large to mount safely on this system");
3671 		if (sizeof(sector_t) < 8)
3672 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3673 		goto failed_mount;
3674 	}
3675 
3676 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3677 		goto cantfind_ext4;
3678 
3679 	/* check blocks count against device size */
3680 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3681 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3682 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3683 		       "exceeds size of device (%llu blocks)",
3684 		       ext4_blocks_count(es), blocks_count);
3685 		goto failed_mount;
3686 	}
3687 
3688 	/*
3689 	 * It makes no sense for the first data block to be beyond the end
3690 	 * of the filesystem.
3691 	 */
3692 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3693 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3694 			 "block %u is beyond end of filesystem (%llu)",
3695 			 le32_to_cpu(es->s_first_data_block),
3696 			 ext4_blocks_count(es));
3697 		goto failed_mount;
3698 	}
3699 	blocks_count = (ext4_blocks_count(es) -
3700 			le32_to_cpu(es->s_first_data_block) +
3701 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3702 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3703 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3704 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3705 		       "(block count %llu, first data block %u, "
3706 		       "blocks per group %lu)", sbi->s_groups_count,
3707 		       ext4_blocks_count(es),
3708 		       le32_to_cpu(es->s_first_data_block),
3709 		       EXT4_BLOCKS_PER_GROUP(sb));
3710 		goto failed_mount;
3711 	}
3712 	sbi->s_groups_count = blocks_count;
3713 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3714 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3715 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3716 		   EXT4_DESC_PER_BLOCK(sb);
3717 	sbi->s_group_desc = ext4_kvmalloc(db_count *
3718 					  sizeof(struct buffer_head *),
3719 					  GFP_KERNEL);
3720 	if (sbi->s_group_desc == NULL) {
3721 		ext4_msg(sb, KERN_ERR, "not enough memory");
3722 		ret = -ENOMEM;
3723 		goto failed_mount;
3724 	}
3725 
3726 	if (ext4_proc_root)
3727 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3728 
3729 	if (sbi->s_proc)
3730 		proc_create_data("options", S_IRUGO, sbi->s_proc,
3731 				 &ext4_seq_options_fops, sb);
3732 
3733 	bgl_lock_init(sbi->s_blockgroup_lock);
3734 
3735 	for (i = 0; i < db_count; i++) {
3736 		block = descriptor_loc(sb, logical_sb_block, i);
3737 		sbi->s_group_desc[i] = sb_bread(sb, block);
3738 		if (!sbi->s_group_desc[i]) {
3739 			ext4_msg(sb, KERN_ERR,
3740 			       "can't read group descriptor %d", i);
3741 			db_count = i;
3742 			goto failed_mount2;
3743 		}
3744 	}
3745 	if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3746 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3747 		goto failed_mount2;
3748 	}
3749 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3750 		if (!ext4_fill_flex_info(sb)) {
3751 			ext4_msg(sb, KERN_ERR,
3752 			       "unable to initialize "
3753 			       "flex_bg meta info!");
3754 			goto failed_mount2;
3755 		}
3756 
3757 	sbi->s_gdb_count = db_count;
3758 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3759 	spin_lock_init(&sbi->s_next_gen_lock);
3760 
3761 	init_timer(&sbi->s_err_report);
3762 	sbi->s_err_report.function = print_daily_error_info;
3763 	sbi->s_err_report.data = (unsigned long) sb;
3764 
3765 	/* Register extent status tree shrinker */
3766 	ext4_es_register_shrinker(sb);
3767 
3768 	err = percpu_counter_init(&sbi->s_freeclusters_counter,
3769 			ext4_count_free_clusters(sb));
3770 	if (!err) {
3771 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
3772 				ext4_count_free_inodes(sb));
3773 	}
3774 	if (!err) {
3775 		err = percpu_counter_init(&sbi->s_dirs_counter,
3776 				ext4_count_dirs(sb));
3777 	}
3778 	if (!err) {
3779 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3780 	}
3781 	if (!err) {
3782 		err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3783 	}
3784 	if (err) {
3785 		ext4_msg(sb, KERN_ERR, "insufficient memory");
3786 		goto failed_mount3;
3787 	}
3788 
3789 	sbi->s_stripe = ext4_get_stripe_size(sbi);
3790 	sbi->s_max_writeback_mb_bump = 128;
3791 	sbi->s_extent_max_zeroout_kb = 32;
3792 
3793 	/*
3794 	 * set up enough so that it can read an inode
3795 	 */
3796 	if (!test_opt(sb, NOLOAD) &&
3797 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3798 		sb->s_op = &ext4_sops;
3799 	else
3800 		sb->s_op = &ext4_nojournal_sops;
3801 	sb->s_export_op = &ext4_export_ops;
3802 	sb->s_xattr = ext4_xattr_handlers;
3803 #ifdef CONFIG_QUOTA
3804 	sb->dq_op = &ext4_quota_operations;
3805 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3806 		sb->s_qcop = &ext4_qctl_sysfile_operations;
3807 	else
3808 		sb->s_qcop = &ext4_qctl_operations;
3809 #endif
3810 	memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3811 
3812 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3813 	mutex_init(&sbi->s_orphan_lock);
3814 
3815 	sb->s_root = NULL;
3816 
3817 	needs_recovery = (es->s_last_orphan != 0 ||
3818 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
3819 				    EXT4_FEATURE_INCOMPAT_RECOVER));
3820 
3821 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3822 	    !(sb->s_flags & MS_RDONLY))
3823 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3824 			goto failed_mount3;
3825 
3826 	/*
3827 	 * The first inode we look at is the journal inode.  Don't try
3828 	 * root first: it may be modified in the journal!
3829 	 */
3830 	if (!test_opt(sb, NOLOAD) &&
3831 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3832 		if (ext4_load_journal(sb, es, journal_devnum))
3833 			goto failed_mount3;
3834 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3835 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3836 		ext4_msg(sb, KERN_ERR, "required journal recovery "
3837 		       "suppressed and not mounted read-only");
3838 		goto failed_mount_wq;
3839 	} else {
3840 		clear_opt(sb, DATA_FLAGS);
3841 		sbi->s_journal = NULL;
3842 		needs_recovery = 0;
3843 		goto no_journal;
3844 	}
3845 
3846 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3847 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3848 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
3849 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3850 		goto failed_mount_wq;
3851 	}
3852 
3853 	if (!set_journal_csum_feature_set(sb)) {
3854 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3855 			 "feature set");
3856 		goto failed_mount_wq;
3857 	}
3858 
3859 	/* We have now updated the journal if required, so we can
3860 	 * validate the data journaling mode. */
3861 	switch (test_opt(sb, DATA_FLAGS)) {
3862 	case 0:
3863 		/* No mode set, assume a default based on the journal
3864 		 * capabilities: ORDERED_DATA if the journal can
3865 		 * cope, else JOURNAL_DATA
3866 		 */
3867 		if (jbd2_journal_check_available_features
3868 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3869 			set_opt(sb, ORDERED_DATA);
3870 		else
3871 			set_opt(sb, JOURNAL_DATA);
3872 		break;
3873 
3874 	case EXT4_MOUNT_ORDERED_DATA:
3875 	case EXT4_MOUNT_WRITEBACK_DATA:
3876 		if (!jbd2_journal_check_available_features
3877 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3878 			ext4_msg(sb, KERN_ERR, "Journal does not support "
3879 			       "requested data journaling mode");
3880 			goto failed_mount_wq;
3881 		}
3882 	default:
3883 		break;
3884 	}
3885 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3886 
3887 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3888 
3889 	/*
3890 	 * The journal may have updated the bg summary counts, so we
3891 	 * need to update the global counters.
3892 	 */
3893 	percpu_counter_set(&sbi->s_freeclusters_counter,
3894 			   ext4_count_free_clusters(sb));
3895 	percpu_counter_set(&sbi->s_freeinodes_counter,
3896 			   ext4_count_free_inodes(sb));
3897 	percpu_counter_set(&sbi->s_dirs_counter,
3898 			   ext4_count_dirs(sb));
3899 	percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3900 
3901 no_journal:
3902 	/*
3903 	 * Get the # of file system overhead blocks from the
3904 	 * superblock if present.
3905 	 */
3906 	if (es->s_overhead_clusters)
3907 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3908 	else {
3909 		err = ext4_calculate_overhead(sb);
3910 		if (err)
3911 			goto failed_mount_wq;
3912 	}
3913 
3914 	/*
3915 	 * The maximum number of concurrent works can be high and
3916 	 * concurrency isn't really necessary.  Limit it to 1.
3917 	 */
3918 	EXT4_SB(sb)->dio_unwritten_wq =
3919 		alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3920 	if (!EXT4_SB(sb)->dio_unwritten_wq) {
3921 		printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3922 		ret = -ENOMEM;
3923 		goto failed_mount_wq;
3924 	}
3925 
3926 	/*
3927 	 * The jbd2_journal_load will have done any necessary log recovery,
3928 	 * so we can safely mount the rest of the filesystem now.
3929 	 */
3930 
3931 	root = ext4_iget(sb, EXT4_ROOT_INO);
3932 	if (IS_ERR(root)) {
3933 		ext4_msg(sb, KERN_ERR, "get root inode failed");
3934 		ret = PTR_ERR(root);
3935 		root = NULL;
3936 		goto failed_mount4;
3937 	}
3938 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3939 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3940 		iput(root);
3941 		goto failed_mount4;
3942 	}
3943 	sb->s_root = d_make_root(root);
3944 	if (!sb->s_root) {
3945 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
3946 		ret = -ENOMEM;
3947 		goto failed_mount4;
3948 	}
3949 
3950 	if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3951 		sb->s_flags |= MS_RDONLY;
3952 
3953 	/* determine the minimum size of new large inodes, if present */
3954 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3955 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3956 						     EXT4_GOOD_OLD_INODE_SIZE;
3957 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3958 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3959 			if (sbi->s_want_extra_isize <
3960 			    le16_to_cpu(es->s_want_extra_isize))
3961 				sbi->s_want_extra_isize =
3962 					le16_to_cpu(es->s_want_extra_isize);
3963 			if (sbi->s_want_extra_isize <
3964 			    le16_to_cpu(es->s_min_extra_isize))
3965 				sbi->s_want_extra_isize =
3966 					le16_to_cpu(es->s_min_extra_isize);
3967 		}
3968 	}
3969 	/* Check if enough inode space is available */
3970 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3971 							sbi->s_inode_size) {
3972 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3973 						       EXT4_GOOD_OLD_INODE_SIZE;
3974 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
3975 			 "available");
3976 	}
3977 
3978 	err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi));
3979 	if (err) {
3980 		ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
3981 			 "reserved pool", ext4_calculate_resv_clusters(sbi));
3982 		goto failed_mount4a;
3983 	}
3984 
3985 	err = ext4_setup_system_zone(sb);
3986 	if (err) {
3987 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
3988 			 "zone (%d)", err);
3989 		goto failed_mount4a;
3990 	}
3991 
3992 	ext4_ext_init(sb);
3993 	err = ext4_mb_init(sb);
3994 	if (err) {
3995 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3996 			 err);
3997 		goto failed_mount5;
3998 	}
3999 
4000 	err = ext4_register_li_request(sb, first_not_zeroed);
4001 	if (err)
4002 		goto failed_mount6;
4003 
4004 	sbi->s_kobj.kset = ext4_kset;
4005 	init_completion(&sbi->s_kobj_unregister);
4006 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4007 				   "%s", sb->s_id);
4008 	if (err)
4009 		goto failed_mount7;
4010 
4011 #ifdef CONFIG_QUOTA
4012 	/* Enable quota usage during mount. */
4013 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4014 	    !(sb->s_flags & MS_RDONLY)) {
4015 		err = ext4_enable_quotas(sb);
4016 		if (err)
4017 			goto failed_mount8;
4018 	}
4019 #endif  /* CONFIG_QUOTA */
4020 
4021 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4022 	ext4_orphan_cleanup(sb, es);
4023 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4024 	if (needs_recovery) {
4025 		ext4_msg(sb, KERN_INFO, "recovery complete");
4026 		ext4_mark_recovery_complete(sb, es);
4027 	}
4028 	if (EXT4_SB(sb)->s_journal) {
4029 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4030 			descr = " journalled data mode";
4031 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4032 			descr = " ordered data mode";
4033 		else
4034 			descr = " writeback data mode";
4035 	} else
4036 		descr = "out journal";
4037 
4038 	if (test_opt(sb, DISCARD)) {
4039 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
4040 		if (!blk_queue_discard(q))
4041 			ext4_msg(sb, KERN_WARNING,
4042 				 "mounting with \"discard\" option, but "
4043 				 "the device does not support discard");
4044 	}
4045 
4046 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4047 		 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4048 		 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4049 
4050 	if (es->s_error_count)
4051 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4052 
4053 	kfree(orig_data);
4054 	return 0;
4055 
4056 cantfind_ext4:
4057 	if (!silent)
4058 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4059 	goto failed_mount;
4060 
4061 #ifdef CONFIG_QUOTA
4062 failed_mount8:
4063 	kobject_del(&sbi->s_kobj);
4064 #endif
4065 failed_mount7:
4066 	ext4_unregister_li_request(sb);
4067 failed_mount6:
4068 	ext4_mb_release(sb);
4069 failed_mount5:
4070 	ext4_ext_release(sb);
4071 	ext4_release_system_zone(sb);
4072 failed_mount4a:
4073 	dput(sb->s_root);
4074 	sb->s_root = NULL;
4075 failed_mount4:
4076 	ext4_msg(sb, KERN_ERR, "mount failed");
4077 	destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4078 failed_mount_wq:
4079 	if (sbi->s_journal) {
4080 		jbd2_journal_destroy(sbi->s_journal);
4081 		sbi->s_journal = NULL;
4082 	}
4083 failed_mount3:
4084 	ext4_es_unregister_shrinker(sb);
4085 	del_timer(&sbi->s_err_report);
4086 	if (sbi->s_flex_groups)
4087 		ext4_kvfree(sbi->s_flex_groups);
4088 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
4089 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
4090 	percpu_counter_destroy(&sbi->s_dirs_counter);
4091 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4092 	percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4093 	if (sbi->s_mmp_tsk)
4094 		kthread_stop(sbi->s_mmp_tsk);
4095 failed_mount2:
4096 	for (i = 0; i < db_count; i++)
4097 		brelse(sbi->s_group_desc[i]);
4098 	ext4_kvfree(sbi->s_group_desc);
4099 failed_mount:
4100 	if (sbi->s_chksum_driver)
4101 		crypto_free_shash(sbi->s_chksum_driver);
4102 	if (sbi->s_proc) {
4103 		remove_proc_entry("options", sbi->s_proc);
4104 		remove_proc_entry(sb->s_id, ext4_proc_root);
4105 	}
4106 #ifdef CONFIG_QUOTA
4107 	for (i = 0; i < MAXQUOTAS; i++)
4108 		kfree(sbi->s_qf_names[i]);
4109 #endif
4110 	ext4_blkdev_remove(sbi);
4111 	brelse(bh);
4112 out_fail:
4113 	sb->s_fs_info = NULL;
4114 	kfree(sbi->s_blockgroup_lock);
4115 	kfree(sbi);
4116 out_free_orig:
4117 	kfree(orig_data);
4118 	return err ? err : ret;
4119 }
4120 
4121 /*
4122  * Setup any per-fs journal parameters now.  We'll do this both on
4123  * initial mount, once the journal has been initialised but before we've
4124  * done any recovery; and again on any subsequent remount.
4125  */
4126 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4127 {
4128 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4129 
4130 	journal->j_commit_interval = sbi->s_commit_interval;
4131 	journal->j_min_batch_time = sbi->s_min_batch_time;
4132 	journal->j_max_batch_time = sbi->s_max_batch_time;
4133 
4134 	write_lock(&journal->j_state_lock);
4135 	if (test_opt(sb, BARRIER))
4136 		journal->j_flags |= JBD2_BARRIER;
4137 	else
4138 		journal->j_flags &= ~JBD2_BARRIER;
4139 	if (test_opt(sb, DATA_ERR_ABORT))
4140 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4141 	else
4142 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4143 	write_unlock(&journal->j_state_lock);
4144 }
4145 
4146 static journal_t *ext4_get_journal(struct super_block *sb,
4147 				   unsigned int journal_inum)
4148 {
4149 	struct inode *journal_inode;
4150 	journal_t *journal;
4151 
4152 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4153 
4154 	/* First, test for the existence of a valid inode on disk.  Bad
4155 	 * things happen if we iget() an unused inode, as the subsequent
4156 	 * iput() will try to delete it. */
4157 
4158 	journal_inode = ext4_iget(sb, journal_inum);
4159 	if (IS_ERR(journal_inode)) {
4160 		ext4_msg(sb, KERN_ERR, "no journal found");
4161 		return NULL;
4162 	}
4163 	if (!journal_inode->i_nlink) {
4164 		make_bad_inode(journal_inode);
4165 		iput(journal_inode);
4166 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4167 		return NULL;
4168 	}
4169 
4170 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4171 		  journal_inode, journal_inode->i_size);
4172 	if (!S_ISREG(journal_inode->i_mode)) {
4173 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4174 		iput(journal_inode);
4175 		return NULL;
4176 	}
4177 
4178 	journal = jbd2_journal_init_inode(journal_inode);
4179 	if (!journal) {
4180 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4181 		iput(journal_inode);
4182 		return NULL;
4183 	}
4184 	journal->j_private = sb;
4185 	ext4_init_journal_params(sb, journal);
4186 	return journal;
4187 }
4188 
4189 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4190 				       dev_t j_dev)
4191 {
4192 	struct buffer_head *bh;
4193 	journal_t *journal;
4194 	ext4_fsblk_t start;
4195 	ext4_fsblk_t len;
4196 	int hblock, blocksize;
4197 	ext4_fsblk_t sb_block;
4198 	unsigned long offset;
4199 	struct ext4_super_block *es;
4200 	struct block_device *bdev;
4201 
4202 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4203 
4204 	bdev = ext4_blkdev_get(j_dev, sb);
4205 	if (bdev == NULL)
4206 		return NULL;
4207 
4208 	blocksize = sb->s_blocksize;
4209 	hblock = bdev_logical_block_size(bdev);
4210 	if (blocksize < hblock) {
4211 		ext4_msg(sb, KERN_ERR,
4212 			"blocksize too small for journal device");
4213 		goto out_bdev;
4214 	}
4215 
4216 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4217 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4218 	set_blocksize(bdev, blocksize);
4219 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4220 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4221 		       "external journal");
4222 		goto out_bdev;
4223 	}
4224 
4225 	es = (struct ext4_super_block *) (bh->b_data + offset);
4226 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4227 	    !(le32_to_cpu(es->s_feature_incompat) &
4228 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4229 		ext4_msg(sb, KERN_ERR, "external journal has "
4230 					"bad superblock");
4231 		brelse(bh);
4232 		goto out_bdev;
4233 	}
4234 
4235 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4236 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4237 		brelse(bh);
4238 		goto out_bdev;
4239 	}
4240 
4241 	len = ext4_blocks_count(es);
4242 	start = sb_block + 1;
4243 	brelse(bh);	/* we're done with the superblock */
4244 
4245 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4246 					start, len, blocksize);
4247 	if (!journal) {
4248 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4249 		goto out_bdev;
4250 	}
4251 	journal->j_private = sb;
4252 	ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4253 	wait_on_buffer(journal->j_sb_buffer);
4254 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4255 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4256 		goto out_journal;
4257 	}
4258 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4259 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4260 					"user (unsupported) - %d",
4261 			be32_to_cpu(journal->j_superblock->s_nr_users));
4262 		goto out_journal;
4263 	}
4264 	EXT4_SB(sb)->journal_bdev = bdev;
4265 	ext4_init_journal_params(sb, journal);
4266 	return journal;
4267 
4268 out_journal:
4269 	jbd2_journal_destroy(journal);
4270 out_bdev:
4271 	ext4_blkdev_put(bdev);
4272 	return NULL;
4273 }
4274 
4275 static int ext4_load_journal(struct super_block *sb,
4276 			     struct ext4_super_block *es,
4277 			     unsigned long journal_devnum)
4278 {
4279 	journal_t *journal;
4280 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4281 	dev_t journal_dev;
4282 	int err = 0;
4283 	int really_read_only;
4284 
4285 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4286 
4287 	if (journal_devnum &&
4288 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4289 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4290 			"numbers have changed");
4291 		journal_dev = new_decode_dev(journal_devnum);
4292 	} else
4293 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4294 
4295 	really_read_only = bdev_read_only(sb->s_bdev);
4296 
4297 	/*
4298 	 * Are we loading a blank journal or performing recovery after a
4299 	 * crash?  For recovery, we need to check in advance whether we
4300 	 * can get read-write access to the device.
4301 	 */
4302 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4303 		if (sb->s_flags & MS_RDONLY) {
4304 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4305 					"required on readonly filesystem");
4306 			if (really_read_only) {
4307 				ext4_msg(sb, KERN_ERR, "write access "
4308 					"unavailable, cannot proceed");
4309 				return -EROFS;
4310 			}
4311 			ext4_msg(sb, KERN_INFO, "write access will "
4312 			       "be enabled during recovery");
4313 		}
4314 	}
4315 
4316 	if (journal_inum && journal_dev) {
4317 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4318 		       "and inode journals!");
4319 		return -EINVAL;
4320 	}
4321 
4322 	if (journal_inum) {
4323 		if (!(journal = ext4_get_journal(sb, journal_inum)))
4324 			return -EINVAL;
4325 	} else {
4326 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4327 			return -EINVAL;
4328 	}
4329 
4330 	if (!(journal->j_flags & JBD2_BARRIER))
4331 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4332 
4333 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4334 		err = jbd2_journal_wipe(journal, !really_read_only);
4335 	if (!err) {
4336 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4337 		if (save)
4338 			memcpy(save, ((char *) es) +
4339 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4340 		err = jbd2_journal_load(journal);
4341 		if (save)
4342 			memcpy(((char *) es) + EXT4_S_ERR_START,
4343 			       save, EXT4_S_ERR_LEN);
4344 		kfree(save);
4345 	}
4346 
4347 	if (err) {
4348 		ext4_msg(sb, KERN_ERR, "error loading journal");
4349 		jbd2_journal_destroy(journal);
4350 		return err;
4351 	}
4352 
4353 	EXT4_SB(sb)->s_journal = journal;
4354 	ext4_clear_journal_err(sb, es);
4355 
4356 	if (!really_read_only && journal_devnum &&
4357 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4358 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4359 
4360 		/* Make sure we flush the recovery flag to disk. */
4361 		ext4_commit_super(sb, 1);
4362 	}
4363 
4364 	return 0;
4365 }
4366 
4367 static int ext4_commit_super(struct super_block *sb, int sync)
4368 {
4369 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4370 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4371 	int error = 0;
4372 
4373 	if (!sbh || block_device_ejected(sb))
4374 		return error;
4375 	if (buffer_write_io_error(sbh)) {
4376 		/*
4377 		 * Oh, dear.  A previous attempt to write the
4378 		 * superblock failed.  This could happen because the
4379 		 * USB device was yanked out.  Or it could happen to
4380 		 * be a transient write error and maybe the block will
4381 		 * be remapped.  Nothing we can do but to retry the
4382 		 * write and hope for the best.
4383 		 */
4384 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4385 		       "superblock detected");
4386 		clear_buffer_write_io_error(sbh);
4387 		set_buffer_uptodate(sbh);
4388 	}
4389 	/*
4390 	 * If the file system is mounted read-only, don't update the
4391 	 * superblock write time.  This avoids updating the superblock
4392 	 * write time when we are mounting the root file system
4393 	 * read/only but we need to replay the journal; at that point,
4394 	 * for people who are east of GMT and who make their clock
4395 	 * tick in localtime for Windows bug-for-bug compatibility,
4396 	 * the clock is set in the future, and this will cause e2fsck
4397 	 * to complain and force a full file system check.
4398 	 */
4399 	if (!(sb->s_flags & MS_RDONLY))
4400 		es->s_wtime = cpu_to_le32(get_seconds());
4401 	if (sb->s_bdev->bd_part)
4402 		es->s_kbytes_written =
4403 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4404 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4405 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4406 	else
4407 		es->s_kbytes_written =
4408 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4409 	ext4_free_blocks_count_set(es,
4410 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4411 				&EXT4_SB(sb)->s_freeclusters_counter)));
4412 	es->s_free_inodes_count =
4413 		cpu_to_le32(percpu_counter_sum_positive(
4414 				&EXT4_SB(sb)->s_freeinodes_counter));
4415 	BUFFER_TRACE(sbh, "marking dirty");
4416 	ext4_superblock_csum_set(sb);
4417 	mark_buffer_dirty(sbh);
4418 	if (sync) {
4419 		error = sync_dirty_buffer(sbh);
4420 		if (error)
4421 			return error;
4422 
4423 		error = buffer_write_io_error(sbh);
4424 		if (error) {
4425 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4426 			       "superblock");
4427 			clear_buffer_write_io_error(sbh);
4428 			set_buffer_uptodate(sbh);
4429 		}
4430 	}
4431 	return error;
4432 }
4433 
4434 /*
4435  * Have we just finished recovery?  If so, and if we are mounting (or
4436  * remounting) the filesystem readonly, then we will end up with a
4437  * consistent fs on disk.  Record that fact.
4438  */
4439 static void ext4_mark_recovery_complete(struct super_block *sb,
4440 					struct ext4_super_block *es)
4441 {
4442 	journal_t *journal = EXT4_SB(sb)->s_journal;
4443 
4444 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4445 		BUG_ON(journal != NULL);
4446 		return;
4447 	}
4448 	jbd2_journal_lock_updates(journal);
4449 	if (jbd2_journal_flush(journal) < 0)
4450 		goto out;
4451 
4452 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4453 	    sb->s_flags & MS_RDONLY) {
4454 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4455 		ext4_commit_super(sb, 1);
4456 	}
4457 
4458 out:
4459 	jbd2_journal_unlock_updates(journal);
4460 }
4461 
4462 /*
4463  * If we are mounting (or read-write remounting) a filesystem whose journal
4464  * has recorded an error from a previous lifetime, move that error to the
4465  * main filesystem now.
4466  */
4467 static void ext4_clear_journal_err(struct super_block *sb,
4468 				   struct ext4_super_block *es)
4469 {
4470 	journal_t *journal;
4471 	int j_errno;
4472 	const char *errstr;
4473 
4474 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4475 
4476 	journal = EXT4_SB(sb)->s_journal;
4477 
4478 	/*
4479 	 * Now check for any error status which may have been recorded in the
4480 	 * journal by a prior ext4_error() or ext4_abort()
4481 	 */
4482 
4483 	j_errno = jbd2_journal_errno(journal);
4484 	if (j_errno) {
4485 		char nbuf[16];
4486 
4487 		errstr = ext4_decode_error(sb, j_errno, nbuf);
4488 		ext4_warning(sb, "Filesystem error recorded "
4489 			     "from previous mount: %s", errstr);
4490 		ext4_warning(sb, "Marking fs in need of filesystem check.");
4491 
4492 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4493 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4494 		ext4_commit_super(sb, 1);
4495 
4496 		jbd2_journal_clear_err(journal);
4497 		jbd2_journal_update_sb_errno(journal);
4498 	}
4499 }
4500 
4501 /*
4502  * Force the running and committing transactions to commit,
4503  * and wait on the commit.
4504  */
4505 int ext4_force_commit(struct super_block *sb)
4506 {
4507 	journal_t *journal;
4508 
4509 	if (sb->s_flags & MS_RDONLY)
4510 		return 0;
4511 
4512 	journal = EXT4_SB(sb)->s_journal;
4513 	return ext4_journal_force_commit(journal);
4514 }
4515 
4516 static int ext4_sync_fs(struct super_block *sb, int wait)
4517 {
4518 	int ret = 0;
4519 	tid_t target;
4520 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4521 
4522 	trace_ext4_sync_fs(sb, wait);
4523 	flush_workqueue(sbi->dio_unwritten_wq);
4524 	/*
4525 	 * Writeback quota in non-journalled quota case - journalled quota has
4526 	 * no dirty dquots
4527 	 */
4528 	dquot_writeback_dquots(sb, -1);
4529 	if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4530 		if (wait)
4531 			jbd2_log_wait_commit(sbi->s_journal, target);
4532 	}
4533 	return ret;
4534 }
4535 
4536 /*
4537  * LVM calls this function before a (read-only) snapshot is created.  This
4538  * gives us a chance to flush the journal completely and mark the fs clean.
4539  *
4540  * Note that only this function cannot bring a filesystem to be in a clean
4541  * state independently. It relies on upper layer to stop all data & metadata
4542  * modifications.
4543  */
4544 static int ext4_freeze(struct super_block *sb)
4545 {
4546 	int error = 0;
4547 	journal_t *journal;
4548 
4549 	if (sb->s_flags & MS_RDONLY)
4550 		return 0;
4551 
4552 	journal = EXT4_SB(sb)->s_journal;
4553 
4554 	/* Now we set up the journal barrier. */
4555 	jbd2_journal_lock_updates(journal);
4556 
4557 	/*
4558 	 * Don't clear the needs_recovery flag if we failed to flush
4559 	 * the journal.
4560 	 */
4561 	error = jbd2_journal_flush(journal);
4562 	if (error < 0)
4563 		goto out;
4564 
4565 	/* Journal blocked and flushed, clear needs_recovery flag. */
4566 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4567 	error = ext4_commit_super(sb, 1);
4568 out:
4569 	/* we rely on upper layer to stop further updates */
4570 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4571 	return error;
4572 }
4573 
4574 /*
4575  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4576  * flag here, even though the filesystem is not technically dirty yet.
4577  */
4578 static int ext4_unfreeze(struct super_block *sb)
4579 {
4580 	if (sb->s_flags & MS_RDONLY)
4581 		return 0;
4582 
4583 	/* Reset the needs_recovery flag before the fs is unlocked. */
4584 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4585 	ext4_commit_super(sb, 1);
4586 	return 0;
4587 }
4588 
4589 /*
4590  * Structure to save mount options for ext4_remount's benefit
4591  */
4592 struct ext4_mount_options {
4593 	unsigned long s_mount_opt;
4594 	unsigned long s_mount_opt2;
4595 	kuid_t s_resuid;
4596 	kgid_t s_resgid;
4597 	unsigned long s_commit_interval;
4598 	u32 s_min_batch_time, s_max_batch_time;
4599 #ifdef CONFIG_QUOTA
4600 	int s_jquota_fmt;
4601 	char *s_qf_names[MAXQUOTAS];
4602 #endif
4603 };
4604 
4605 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4606 {
4607 	struct ext4_super_block *es;
4608 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4609 	unsigned long old_sb_flags;
4610 	struct ext4_mount_options old_opts;
4611 	int enable_quota = 0;
4612 	ext4_group_t g;
4613 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4614 	int err = 0;
4615 #ifdef CONFIG_QUOTA
4616 	int i, j;
4617 #endif
4618 	char *orig_data = kstrdup(data, GFP_KERNEL);
4619 
4620 	/* Store the original options */
4621 	old_sb_flags = sb->s_flags;
4622 	old_opts.s_mount_opt = sbi->s_mount_opt;
4623 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4624 	old_opts.s_resuid = sbi->s_resuid;
4625 	old_opts.s_resgid = sbi->s_resgid;
4626 	old_opts.s_commit_interval = sbi->s_commit_interval;
4627 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4628 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4629 #ifdef CONFIG_QUOTA
4630 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4631 	for (i = 0; i < MAXQUOTAS; i++)
4632 		if (sbi->s_qf_names[i]) {
4633 			old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4634 							 GFP_KERNEL);
4635 			if (!old_opts.s_qf_names[i]) {
4636 				for (j = 0; j < i; j++)
4637 					kfree(old_opts.s_qf_names[j]);
4638 				kfree(orig_data);
4639 				return -ENOMEM;
4640 			}
4641 		} else
4642 			old_opts.s_qf_names[i] = NULL;
4643 #endif
4644 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4645 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4646 
4647 	/*
4648 	 * Allow the "check" option to be passed as a remount option.
4649 	 */
4650 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4651 		err = -EINVAL;
4652 		goto restore_opts;
4653 	}
4654 
4655 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4656 		ext4_abort(sb, "Abort forced by user");
4657 
4658 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4659 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4660 
4661 	es = sbi->s_es;
4662 
4663 	if (sbi->s_journal) {
4664 		ext4_init_journal_params(sb, sbi->s_journal);
4665 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4666 	}
4667 
4668 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4669 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4670 			err = -EROFS;
4671 			goto restore_opts;
4672 		}
4673 
4674 		if (*flags & MS_RDONLY) {
4675 			err = dquot_suspend(sb, -1);
4676 			if (err < 0)
4677 				goto restore_opts;
4678 
4679 			/*
4680 			 * First of all, the unconditional stuff we have to do
4681 			 * to disable replay of the journal when we next remount
4682 			 */
4683 			sb->s_flags |= MS_RDONLY;
4684 
4685 			/*
4686 			 * OK, test if we are remounting a valid rw partition
4687 			 * readonly, and if so set the rdonly flag and then
4688 			 * mark the partition as valid again.
4689 			 */
4690 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4691 			    (sbi->s_mount_state & EXT4_VALID_FS))
4692 				es->s_state = cpu_to_le16(sbi->s_mount_state);
4693 
4694 			if (sbi->s_journal)
4695 				ext4_mark_recovery_complete(sb, es);
4696 		} else {
4697 			/* Make sure we can mount this feature set readwrite */
4698 			if (!ext4_feature_set_ok(sb, 0)) {
4699 				err = -EROFS;
4700 				goto restore_opts;
4701 			}
4702 			/*
4703 			 * Make sure the group descriptor checksums
4704 			 * are sane.  If they aren't, refuse to remount r/w.
4705 			 */
4706 			for (g = 0; g < sbi->s_groups_count; g++) {
4707 				struct ext4_group_desc *gdp =
4708 					ext4_get_group_desc(sb, g, NULL);
4709 
4710 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4711 					ext4_msg(sb, KERN_ERR,
4712 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
4713 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4714 					       le16_to_cpu(gdp->bg_checksum));
4715 					err = -EINVAL;
4716 					goto restore_opts;
4717 				}
4718 			}
4719 
4720 			/*
4721 			 * If we have an unprocessed orphan list hanging
4722 			 * around from a previously readonly bdev mount,
4723 			 * require a full umount/remount for now.
4724 			 */
4725 			if (es->s_last_orphan) {
4726 				ext4_msg(sb, KERN_WARNING, "Couldn't "
4727 				       "remount RDWR because of unprocessed "
4728 				       "orphan inode list.  Please "
4729 				       "umount/remount instead");
4730 				err = -EINVAL;
4731 				goto restore_opts;
4732 			}
4733 
4734 			/*
4735 			 * Mounting a RDONLY partition read-write, so reread
4736 			 * and store the current valid flag.  (It may have
4737 			 * been changed by e2fsck since we originally mounted
4738 			 * the partition.)
4739 			 */
4740 			if (sbi->s_journal)
4741 				ext4_clear_journal_err(sb, es);
4742 			sbi->s_mount_state = le16_to_cpu(es->s_state);
4743 			if (!ext4_setup_super(sb, es, 0))
4744 				sb->s_flags &= ~MS_RDONLY;
4745 			if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4746 						     EXT4_FEATURE_INCOMPAT_MMP))
4747 				if (ext4_multi_mount_protect(sb,
4748 						le64_to_cpu(es->s_mmp_block))) {
4749 					err = -EROFS;
4750 					goto restore_opts;
4751 				}
4752 			enable_quota = 1;
4753 		}
4754 	}
4755 
4756 	/*
4757 	 * Reinitialize lazy itable initialization thread based on
4758 	 * current settings
4759 	 */
4760 	if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4761 		ext4_unregister_li_request(sb);
4762 	else {
4763 		ext4_group_t first_not_zeroed;
4764 		first_not_zeroed = ext4_has_uninit_itable(sb);
4765 		ext4_register_li_request(sb, first_not_zeroed);
4766 	}
4767 
4768 	ext4_setup_system_zone(sb);
4769 	if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4770 		ext4_commit_super(sb, 1);
4771 
4772 #ifdef CONFIG_QUOTA
4773 	/* Release old quota file names */
4774 	for (i = 0; i < MAXQUOTAS; i++)
4775 		kfree(old_opts.s_qf_names[i]);
4776 	if (enable_quota) {
4777 		if (sb_any_quota_suspended(sb))
4778 			dquot_resume(sb, -1);
4779 		else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4780 					EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4781 			err = ext4_enable_quotas(sb);
4782 			if (err)
4783 				goto restore_opts;
4784 		}
4785 	}
4786 #endif
4787 
4788 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4789 	kfree(orig_data);
4790 	return 0;
4791 
4792 restore_opts:
4793 	sb->s_flags = old_sb_flags;
4794 	sbi->s_mount_opt = old_opts.s_mount_opt;
4795 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4796 	sbi->s_resuid = old_opts.s_resuid;
4797 	sbi->s_resgid = old_opts.s_resgid;
4798 	sbi->s_commit_interval = old_opts.s_commit_interval;
4799 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
4800 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
4801 #ifdef CONFIG_QUOTA
4802 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4803 	for (i = 0; i < MAXQUOTAS; i++) {
4804 		kfree(sbi->s_qf_names[i]);
4805 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4806 	}
4807 #endif
4808 	kfree(orig_data);
4809 	return err;
4810 }
4811 
4812 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4813 {
4814 	struct super_block *sb = dentry->d_sb;
4815 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4816 	struct ext4_super_block *es = sbi->s_es;
4817 	ext4_fsblk_t overhead = 0, resv_blocks;
4818 	u64 fsid;
4819 	s64 bfree;
4820 	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4821 
4822 	if (!test_opt(sb, MINIX_DF))
4823 		overhead = sbi->s_overhead;
4824 
4825 	buf->f_type = EXT4_SUPER_MAGIC;
4826 	buf->f_bsize = sb->s_blocksize;
4827 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4828 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4829 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4830 	/* prevent underflow in case that few free space is available */
4831 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4832 	buf->f_bavail = buf->f_bfree -
4833 			(ext4_r_blocks_count(es) + resv_blocks);
4834 	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4835 		buf->f_bavail = 0;
4836 	buf->f_files = le32_to_cpu(es->s_inodes_count);
4837 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4838 	buf->f_namelen = EXT4_NAME_LEN;
4839 	fsid = le64_to_cpup((void *)es->s_uuid) ^
4840 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4841 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4842 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4843 
4844 	return 0;
4845 }
4846 
4847 /* Helper function for writing quotas on sync - we need to start transaction
4848  * before quota file is locked for write. Otherwise the are possible deadlocks:
4849  * Process 1                         Process 2
4850  * ext4_create()                     quota_sync()
4851  *   jbd2_journal_start()                  write_dquot()
4852  *   dquot_initialize()                         down(dqio_mutex)
4853  *     down(dqio_mutex)                    jbd2_journal_start()
4854  *
4855  */
4856 
4857 #ifdef CONFIG_QUOTA
4858 
4859 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4860 {
4861 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4862 }
4863 
4864 static int ext4_write_dquot(struct dquot *dquot)
4865 {
4866 	int ret, err;
4867 	handle_t *handle;
4868 	struct inode *inode;
4869 
4870 	inode = dquot_to_inode(dquot);
4871 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4872 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4873 	if (IS_ERR(handle))
4874 		return PTR_ERR(handle);
4875 	ret = dquot_commit(dquot);
4876 	err = ext4_journal_stop(handle);
4877 	if (!ret)
4878 		ret = err;
4879 	return ret;
4880 }
4881 
4882 static int ext4_acquire_dquot(struct dquot *dquot)
4883 {
4884 	int ret, err;
4885 	handle_t *handle;
4886 
4887 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4888 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4889 	if (IS_ERR(handle))
4890 		return PTR_ERR(handle);
4891 	ret = dquot_acquire(dquot);
4892 	err = ext4_journal_stop(handle);
4893 	if (!ret)
4894 		ret = err;
4895 	return ret;
4896 }
4897 
4898 static int ext4_release_dquot(struct dquot *dquot)
4899 {
4900 	int ret, err;
4901 	handle_t *handle;
4902 
4903 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4904 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4905 	if (IS_ERR(handle)) {
4906 		/* Release dquot anyway to avoid endless cycle in dqput() */
4907 		dquot_release(dquot);
4908 		return PTR_ERR(handle);
4909 	}
4910 	ret = dquot_release(dquot);
4911 	err = ext4_journal_stop(handle);
4912 	if (!ret)
4913 		ret = err;
4914 	return ret;
4915 }
4916 
4917 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4918 {
4919 	struct super_block *sb = dquot->dq_sb;
4920 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4921 
4922 	/* Are we journaling quotas? */
4923 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4924 	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4925 		dquot_mark_dquot_dirty(dquot);
4926 		return ext4_write_dquot(dquot);
4927 	} else {
4928 		return dquot_mark_dquot_dirty(dquot);
4929 	}
4930 }
4931 
4932 static int ext4_write_info(struct super_block *sb, int type)
4933 {
4934 	int ret, err;
4935 	handle_t *handle;
4936 
4937 	/* Data block + inode block */
4938 	handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
4939 	if (IS_ERR(handle))
4940 		return PTR_ERR(handle);
4941 	ret = dquot_commit_info(sb, type);
4942 	err = ext4_journal_stop(handle);
4943 	if (!ret)
4944 		ret = err;
4945 	return ret;
4946 }
4947 
4948 /*
4949  * Turn on quotas during mount time - we need to find
4950  * the quota file and such...
4951  */
4952 static int ext4_quota_on_mount(struct super_block *sb, int type)
4953 {
4954 	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4955 					EXT4_SB(sb)->s_jquota_fmt, type);
4956 }
4957 
4958 /*
4959  * Standard function to be called on quota_on
4960  */
4961 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4962 			 struct path *path)
4963 {
4964 	int err;
4965 
4966 	if (!test_opt(sb, QUOTA))
4967 		return -EINVAL;
4968 
4969 	/* Quotafile not on the same filesystem? */
4970 	if (path->dentry->d_sb != sb)
4971 		return -EXDEV;
4972 	/* Journaling quota? */
4973 	if (EXT4_SB(sb)->s_qf_names[type]) {
4974 		/* Quotafile not in fs root? */
4975 		if (path->dentry->d_parent != sb->s_root)
4976 			ext4_msg(sb, KERN_WARNING,
4977 				"Quota file not on filesystem root. "
4978 				"Journaled quota will not work");
4979 	}
4980 
4981 	/*
4982 	 * When we journal data on quota file, we have to flush journal to see
4983 	 * all updates to the file when we bypass pagecache...
4984 	 */
4985 	if (EXT4_SB(sb)->s_journal &&
4986 	    ext4_should_journal_data(path->dentry->d_inode)) {
4987 		/*
4988 		 * We don't need to lock updates but journal_flush() could
4989 		 * otherwise be livelocked...
4990 		 */
4991 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4992 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4993 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4994 		if (err)
4995 			return err;
4996 	}
4997 
4998 	return dquot_quota_on(sb, type, format_id, path);
4999 }
5000 
5001 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5002 			     unsigned int flags)
5003 {
5004 	int err;
5005 	struct inode *qf_inode;
5006 	unsigned long qf_inums[MAXQUOTAS] = {
5007 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5008 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5009 	};
5010 
5011 	BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5012 
5013 	if (!qf_inums[type])
5014 		return -EPERM;
5015 
5016 	qf_inode = ext4_iget(sb, qf_inums[type]);
5017 	if (IS_ERR(qf_inode)) {
5018 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5019 		return PTR_ERR(qf_inode);
5020 	}
5021 
5022 	/* Don't account quota for quota files to avoid recursion */
5023 	qf_inode->i_flags |= S_NOQUOTA;
5024 	err = dquot_enable(qf_inode, type, format_id, flags);
5025 	iput(qf_inode);
5026 
5027 	return err;
5028 }
5029 
5030 /* Enable usage tracking for all quota types. */
5031 static int ext4_enable_quotas(struct super_block *sb)
5032 {
5033 	int type, err = 0;
5034 	unsigned long qf_inums[MAXQUOTAS] = {
5035 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5036 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5037 	};
5038 
5039 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5040 	for (type = 0; type < MAXQUOTAS; type++) {
5041 		if (qf_inums[type]) {
5042 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5043 						DQUOT_USAGE_ENABLED);
5044 			if (err) {
5045 				ext4_warning(sb,
5046 					"Failed to enable quota tracking "
5047 					"(type=%d, err=%d). Please run "
5048 					"e2fsck to fix.", type, err);
5049 				return err;
5050 			}
5051 		}
5052 	}
5053 	return 0;
5054 }
5055 
5056 /*
5057  * quota_on function that is used when QUOTA feature is set.
5058  */
5059 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5060 				 int format_id)
5061 {
5062 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5063 		return -EINVAL;
5064 
5065 	/*
5066 	 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5067 	 */
5068 	return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5069 }
5070 
5071 static int ext4_quota_off(struct super_block *sb, int type)
5072 {
5073 	struct inode *inode = sb_dqopt(sb)->files[type];
5074 	handle_t *handle;
5075 
5076 	/* Force all delayed allocation blocks to be allocated.
5077 	 * Caller already holds s_umount sem */
5078 	if (test_opt(sb, DELALLOC))
5079 		sync_filesystem(sb);
5080 
5081 	if (!inode)
5082 		goto out;
5083 
5084 	/* Update modification times of quota files when userspace can
5085 	 * start looking at them */
5086 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5087 	if (IS_ERR(handle))
5088 		goto out;
5089 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5090 	ext4_mark_inode_dirty(handle, inode);
5091 	ext4_journal_stop(handle);
5092 
5093 out:
5094 	return dquot_quota_off(sb, type);
5095 }
5096 
5097 /*
5098  * quota_off function that is used when QUOTA feature is set.
5099  */
5100 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5101 {
5102 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5103 		return -EINVAL;
5104 
5105 	/* Disable only the limits. */
5106 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5107 }
5108 
5109 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5110  * acquiring the locks... As quota files are never truncated and quota code
5111  * itself serializes the operations (and no one else should touch the files)
5112  * we don't have to be afraid of races */
5113 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5114 			       size_t len, loff_t off)
5115 {
5116 	struct inode *inode = sb_dqopt(sb)->files[type];
5117 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5118 	int err = 0;
5119 	int offset = off & (sb->s_blocksize - 1);
5120 	int tocopy;
5121 	size_t toread;
5122 	struct buffer_head *bh;
5123 	loff_t i_size = i_size_read(inode);
5124 
5125 	if (off > i_size)
5126 		return 0;
5127 	if (off+len > i_size)
5128 		len = i_size-off;
5129 	toread = len;
5130 	while (toread > 0) {
5131 		tocopy = sb->s_blocksize - offset < toread ?
5132 				sb->s_blocksize - offset : toread;
5133 		bh = ext4_bread(NULL, inode, blk, 0, &err);
5134 		if (err)
5135 			return err;
5136 		if (!bh)	/* A hole? */
5137 			memset(data, 0, tocopy);
5138 		else
5139 			memcpy(data, bh->b_data+offset, tocopy);
5140 		brelse(bh);
5141 		offset = 0;
5142 		toread -= tocopy;
5143 		data += tocopy;
5144 		blk++;
5145 	}
5146 	return len;
5147 }
5148 
5149 /* Write to quotafile (we know the transaction is already started and has
5150  * enough credits) */
5151 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5152 				const char *data, size_t len, loff_t off)
5153 {
5154 	struct inode *inode = sb_dqopt(sb)->files[type];
5155 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5156 	int err = 0;
5157 	int offset = off & (sb->s_blocksize - 1);
5158 	struct buffer_head *bh;
5159 	handle_t *handle = journal_current_handle();
5160 
5161 	if (EXT4_SB(sb)->s_journal && !handle) {
5162 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5163 			" cancelled because transaction is not started",
5164 			(unsigned long long)off, (unsigned long long)len);
5165 		return -EIO;
5166 	}
5167 	/*
5168 	 * Since we account only one data block in transaction credits,
5169 	 * then it is impossible to cross a block boundary.
5170 	 */
5171 	if (sb->s_blocksize - offset < len) {
5172 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5173 			" cancelled because not block aligned",
5174 			(unsigned long long)off, (unsigned long long)len);
5175 		return -EIO;
5176 	}
5177 
5178 	bh = ext4_bread(handle, inode, blk, 1, &err);
5179 	if (!bh)
5180 		goto out;
5181 	err = ext4_journal_get_write_access(handle, bh);
5182 	if (err) {
5183 		brelse(bh);
5184 		goto out;
5185 	}
5186 	lock_buffer(bh);
5187 	memcpy(bh->b_data+offset, data, len);
5188 	flush_dcache_page(bh->b_page);
5189 	unlock_buffer(bh);
5190 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5191 	brelse(bh);
5192 out:
5193 	if (err)
5194 		return err;
5195 	if (inode->i_size < off + len) {
5196 		i_size_write(inode, off + len);
5197 		EXT4_I(inode)->i_disksize = inode->i_size;
5198 		ext4_mark_inode_dirty(handle, inode);
5199 	}
5200 	return len;
5201 }
5202 
5203 #endif
5204 
5205 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5206 		       const char *dev_name, void *data)
5207 {
5208 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5209 }
5210 
5211 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5212 static inline void register_as_ext2(void)
5213 {
5214 	int err = register_filesystem(&ext2_fs_type);
5215 	if (err)
5216 		printk(KERN_WARNING
5217 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5218 }
5219 
5220 static inline void unregister_as_ext2(void)
5221 {
5222 	unregister_filesystem(&ext2_fs_type);
5223 }
5224 
5225 static inline int ext2_feature_set_ok(struct super_block *sb)
5226 {
5227 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5228 		return 0;
5229 	if (sb->s_flags & MS_RDONLY)
5230 		return 1;
5231 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5232 		return 0;
5233 	return 1;
5234 }
5235 #else
5236 static inline void register_as_ext2(void) { }
5237 static inline void unregister_as_ext2(void) { }
5238 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5239 #endif
5240 
5241 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5242 static inline void register_as_ext3(void)
5243 {
5244 	int err = register_filesystem(&ext3_fs_type);
5245 	if (err)
5246 		printk(KERN_WARNING
5247 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5248 }
5249 
5250 static inline void unregister_as_ext3(void)
5251 {
5252 	unregister_filesystem(&ext3_fs_type);
5253 }
5254 
5255 static inline int ext3_feature_set_ok(struct super_block *sb)
5256 {
5257 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5258 		return 0;
5259 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5260 		return 0;
5261 	if (sb->s_flags & MS_RDONLY)
5262 		return 1;
5263 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5264 		return 0;
5265 	return 1;
5266 }
5267 #else
5268 static inline void register_as_ext3(void) { }
5269 static inline void unregister_as_ext3(void) { }
5270 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5271 #endif
5272 
5273 static struct file_system_type ext4_fs_type = {
5274 	.owner		= THIS_MODULE,
5275 	.name		= "ext4",
5276 	.mount		= ext4_mount,
5277 	.kill_sb	= kill_block_super,
5278 	.fs_flags	= FS_REQUIRES_DEV,
5279 };
5280 MODULE_ALIAS_FS("ext4");
5281 
5282 static int __init ext4_init_feat_adverts(void)
5283 {
5284 	struct ext4_features *ef;
5285 	int ret = -ENOMEM;
5286 
5287 	ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5288 	if (!ef)
5289 		goto out;
5290 
5291 	ef->f_kobj.kset = ext4_kset;
5292 	init_completion(&ef->f_kobj_unregister);
5293 	ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5294 				   "features");
5295 	if (ret) {
5296 		kfree(ef);
5297 		goto out;
5298 	}
5299 
5300 	ext4_feat = ef;
5301 	ret = 0;
5302 out:
5303 	return ret;
5304 }
5305 
5306 static void ext4_exit_feat_adverts(void)
5307 {
5308 	kobject_put(&ext4_feat->f_kobj);
5309 	wait_for_completion(&ext4_feat->f_kobj_unregister);
5310 	kfree(ext4_feat);
5311 }
5312 
5313 /* Shared across all ext4 file systems */
5314 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5315 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5316 
5317 static int __init ext4_init_fs(void)
5318 {
5319 	int i, err;
5320 
5321 	ext4_li_info = NULL;
5322 	mutex_init(&ext4_li_mtx);
5323 
5324 	/* Build-time check for flags consistency */
5325 	ext4_check_flag_values();
5326 
5327 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5328 		mutex_init(&ext4__aio_mutex[i]);
5329 		init_waitqueue_head(&ext4__ioend_wq[i]);
5330 	}
5331 
5332 	err = ext4_init_es();
5333 	if (err)
5334 		return err;
5335 
5336 	err = ext4_init_pageio();
5337 	if (err)
5338 		goto out7;
5339 
5340 	err = ext4_init_system_zone();
5341 	if (err)
5342 		goto out6;
5343 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5344 	if (!ext4_kset) {
5345 		err = -ENOMEM;
5346 		goto out5;
5347 	}
5348 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5349 
5350 	err = ext4_init_feat_adverts();
5351 	if (err)
5352 		goto out4;
5353 
5354 	err = ext4_init_mballoc();
5355 	if (err)
5356 		goto out3;
5357 
5358 	err = ext4_init_xattr();
5359 	if (err)
5360 		goto out2;
5361 	err = init_inodecache();
5362 	if (err)
5363 		goto out1;
5364 	register_as_ext3();
5365 	register_as_ext2();
5366 	err = register_filesystem(&ext4_fs_type);
5367 	if (err)
5368 		goto out;
5369 
5370 	return 0;
5371 out:
5372 	unregister_as_ext2();
5373 	unregister_as_ext3();
5374 	destroy_inodecache();
5375 out1:
5376 	ext4_exit_xattr();
5377 out2:
5378 	ext4_exit_mballoc();
5379 out3:
5380 	ext4_exit_feat_adverts();
5381 out4:
5382 	if (ext4_proc_root)
5383 		remove_proc_entry("fs/ext4", NULL);
5384 	kset_unregister(ext4_kset);
5385 out5:
5386 	ext4_exit_system_zone();
5387 out6:
5388 	ext4_exit_pageio();
5389 out7:
5390 	ext4_exit_es();
5391 
5392 	return err;
5393 }
5394 
5395 static void __exit ext4_exit_fs(void)
5396 {
5397 	ext4_destroy_lazyinit_thread();
5398 	unregister_as_ext2();
5399 	unregister_as_ext3();
5400 	unregister_filesystem(&ext4_fs_type);
5401 	destroy_inodecache();
5402 	ext4_exit_xattr();
5403 	ext4_exit_mballoc();
5404 	ext4_exit_feat_adverts();
5405 	remove_proc_entry("fs/ext4", NULL);
5406 	kset_unregister(ext4_kset);
5407 	ext4_exit_system_zone();
5408 	ext4_exit_pageio();
5409 }
5410 
5411 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5412 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5413 MODULE_LICENSE("GPL");
5414 module_init(ext4_init_fs)
5415 module_exit(ext4_exit_fs)
5416