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
57static struct proc_dir_entry *ext4_proc_root;
58static struct kset *ext4_kset;
59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
62static int ext4_mballoc_ready;
63
64static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65			     unsigned long journal_devnum);
66static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67static int ext4_commit_super(struct super_block *sb, int sync);
68static void ext4_mark_recovery_complete(struct super_block *sb,
69					struct ext4_super_block *es);
70static void ext4_clear_journal_err(struct super_block *sb,
71				   struct ext4_super_block *es);
72static int ext4_sync_fs(struct super_block *sb, int wait);
73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75static int ext4_unfreeze(struct super_block *sb);
76static int ext4_freeze(struct super_block *sb);
77static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78		       const char *dev_name, void *data);
79static inline int ext2_feature_set_ok(struct super_block *sb);
80static inline int ext3_feature_set_ok(struct super_block *sb);
81static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
84static void ext4_clear_request_list(void);
85static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
86
87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89	.owner		= THIS_MODULE,
90	.name		= "ext2",
91	.mount		= ext4_mount,
92	.kill_sb	= kill_block_super,
93	.fs_flags	= FS_REQUIRES_DEV,
94};
95MODULE_ALIAS_FS("ext2");
96MODULE_ALIAS("ext2");
97#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98#else
99#define IS_EXT2_SB(sb) (0)
100#endif
101
102
103#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104static struct file_system_type ext3_fs_type = {
105	.owner		= THIS_MODULE,
106	.name		= "ext3",
107	.mount		= ext4_mount,
108	.kill_sb	= kill_block_super,
109	.fs_flags	= FS_REQUIRES_DEV,
110};
111MODULE_ALIAS_FS("ext3");
112MODULE_ALIAS("ext3");
113#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114#else
115#define IS_EXT3_SB(sb) (0)
116#endif
117
118static int ext4_verify_csum_type(struct super_block *sb,
119				 struct ext4_super_block *es)
120{
121	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123		return 1;
124
125	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126}
127
128static __le32 ext4_superblock_csum(struct super_block *sb,
129				   struct ext4_super_block *es)
130{
131	struct ext4_sb_info *sbi = EXT4_SB(sb);
132	int offset = offsetof(struct ext4_super_block, s_checksum);
133	__u32 csum;
134
135	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136
137	return cpu_to_le32(csum);
138}
139
140static int ext4_superblock_csum_verify(struct super_block *sb,
141				       struct ext4_super_block *es)
142{
143	if (!ext4_has_metadata_csum(sb))
144		return 1;
145
146	return es->s_checksum == ext4_superblock_csum(sb, es);
147}
148
149void 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_metadata_csum(sb))
154		return;
155
156	es->s_checksum = ext4_superblock_csum(sb, es);
157}
158
159void *ext4_kvmalloc(size_t size, gfp_t flags)
160{
161	void *ret;
162
163	ret = kmalloc(size, flags | __GFP_NOWARN);
164	if (!ret)
165		ret = __vmalloc(size, flags, PAGE_KERNEL);
166	return ret;
167}
168
169void *ext4_kvzalloc(size_t size, gfp_t flags)
170{
171	void *ret;
172
173	ret = kzalloc(size, flags | __GFP_NOWARN);
174	if (!ret)
175		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176	return ret;
177}
178
179void ext4_kvfree(void *ptr)
180{
181	if (is_vmalloc_addr(ptr))
182		vfree(ptr);
183	else
184		kfree(ptr);
185
186}
187
188ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
189			       struct ext4_group_desc *bg)
190{
191	return le32_to_cpu(bg->bg_block_bitmap_lo) |
192		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
193		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
194}
195
196ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
197			       struct ext4_group_desc *bg)
198{
199	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
200		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
201		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
202}
203
204ext4_fsblk_t ext4_inode_table(struct super_block *sb,
205			      struct ext4_group_desc *bg)
206{
207	return le32_to_cpu(bg->bg_inode_table_lo) |
208		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
209		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
210}
211
212__u32 ext4_free_group_clusters(struct super_block *sb,
213			       struct ext4_group_desc *bg)
214{
215	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
216		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
217		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
218}
219
220__u32 ext4_free_inodes_count(struct super_block *sb,
221			      struct ext4_group_desc *bg)
222{
223	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
224		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
225		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
226}
227
228__u32 ext4_used_dirs_count(struct super_block *sb,
229			      struct ext4_group_desc *bg)
230{
231	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
232		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
233		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
234}
235
236__u32 ext4_itable_unused_count(struct super_block *sb,
237			      struct ext4_group_desc *bg)
238{
239	return le16_to_cpu(bg->bg_itable_unused_lo) |
240		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
241		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
242}
243
244void ext4_block_bitmap_set(struct super_block *sb,
245			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
246{
247	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
248	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
249		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
250}
251
252void ext4_inode_bitmap_set(struct super_block *sb,
253			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
254{
255	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
256	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
257		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
258}
259
260void ext4_inode_table_set(struct super_block *sb,
261			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
262{
263	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
264	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
265		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
266}
267
268void ext4_free_group_clusters_set(struct super_block *sb,
269				  struct ext4_group_desc *bg, __u32 count)
270{
271	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
272	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
273		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
274}
275
276void ext4_free_inodes_set(struct super_block *sb,
277			  struct ext4_group_desc *bg, __u32 count)
278{
279	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
280	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
281		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
282}
283
284void ext4_used_dirs_set(struct super_block *sb,
285			  struct ext4_group_desc *bg, __u32 count)
286{
287	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
288	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
289		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
290}
291
292void ext4_itable_unused_set(struct super_block *sb,
293			  struct ext4_group_desc *bg, __u32 count)
294{
295	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
296	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
297		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
298}
299
300
301static void __save_error_info(struct super_block *sb, const char *func,
302			    unsigned int line)
303{
304	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
305
306	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
307	if (bdev_read_only(sb->s_bdev))
308		return;
309	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
310	es->s_last_error_time = cpu_to_le32(get_seconds());
311	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
312	es->s_last_error_line = cpu_to_le32(line);
313	if (!es->s_first_error_time) {
314		es->s_first_error_time = es->s_last_error_time;
315		strncpy(es->s_first_error_func, func,
316			sizeof(es->s_first_error_func));
317		es->s_first_error_line = cpu_to_le32(line);
318		es->s_first_error_ino = es->s_last_error_ino;
319		es->s_first_error_block = es->s_last_error_block;
320	}
321	/*
322	 * Start the daily error reporting function if it hasn't been
323	 * started already
324	 */
325	if (!es->s_error_count)
326		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
327	le32_add_cpu(&es->s_error_count, 1);
328}
329
330static void save_error_info(struct super_block *sb, const char *func,
331			    unsigned int line)
332{
333	__save_error_info(sb, func, line);
334	ext4_commit_super(sb, 1);
335}
336
337/*
338 * The del_gendisk() function uninitializes the disk-specific data
339 * structures, including the bdi structure, without telling anyone
340 * else.  Once this happens, any attempt to call mark_buffer_dirty()
341 * (for example, by ext4_commit_super), will cause a kernel OOPS.
342 * This is a kludge to prevent these oops until we can put in a proper
343 * hook in del_gendisk() to inform the VFS and file system layers.
344 */
345static int block_device_ejected(struct super_block *sb)
346{
347	struct inode *bd_inode = sb->s_bdev->bd_inode;
348	struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
349
350	return bdi->dev == NULL;
351}
352
353static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
354{
355	struct super_block		*sb = journal->j_private;
356	struct ext4_sb_info		*sbi = EXT4_SB(sb);
357	int				error = is_journal_aborted(journal);
358	struct ext4_journal_cb_entry	*jce;
359
360	BUG_ON(txn->t_state == T_FINISHED);
361	spin_lock(&sbi->s_md_lock);
362	while (!list_empty(&txn->t_private_list)) {
363		jce = list_entry(txn->t_private_list.next,
364				 struct ext4_journal_cb_entry, jce_list);
365		list_del_init(&jce->jce_list);
366		spin_unlock(&sbi->s_md_lock);
367		jce->jce_func(sb, jce, error);
368		spin_lock(&sbi->s_md_lock);
369	}
370	spin_unlock(&sbi->s_md_lock);
371}
372
373/* Deal with the reporting of failure conditions on a filesystem such as
374 * inconsistencies detected or read IO failures.
375 *
376 * On ext2, we can store the error state of the filesystem in the
377 * superblock.  That is not possible on ext4, because we may have other
378 * write ordering constraints on the superblock which prevent us from
379 * writing it out straight away; and given that the journal is about to
380 * be aborted, we can't rely on the current, or future, transactions to
381 * write out the superblock safely.
382 *
383 * We'll just use the jbd2_journal_abort() error code to record an error in
384 * the journal instead.  On recovery, the journal will complain about
385 * that error until we've noted it down and cleared it.
386 */
387
388static void ext4_handle_error(struct super_block *sb)
389{
390	if (sb->s_flags & MS_RDONLY)
391		return;
392
393	if (!test_opt(sb, ERRORS_CONT)) {
394		journal_t *journal = EXT4_SB(sb)->s_journal;
395
396		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
397		if (journal)
398			jbd2_journal_abort(journal, -EIO);
399	}
400	if (test_opt(sb, ERRORS_RO)) {
401		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
402		/*
403		 * Make sure updated value of ->s_mount_flags will be visible
404		 * before ->s_flags update
405		 */
406		smp_wmb();
407		sb->s_flags |= MS_RDONLY;
408	}
409	if (test_opt(sb, ERRORS_PANIC))
410		panic("EXT4-fs (device %s): panic forced after error\n",
411			sb->s_id);
412}
413
414#define ext4_error_ratelimit(sb)					\
415		___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),	\
416			     "EXT4-fs error")
417
418void __ext4_error(struct super_block *sb, const char *function,
419		  unsigned int line, const char *fmt, ...)
420{
421	struct va_format vaf;
422	va_list args;
423
424	if (ext4_error_ratelimit(sb)) {
425		va_start(args, fmt);
426		vaf.fmt = fmt;
427		vaf.va = &args;
428		printk(KERN_CRIT
429		       "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
430		       sb->s_id, function, line, current->comm, &vaf);
431		va_end(args);
432	}
433	save_error_info(sb, function, line);
434	ext4_handle_error(sb);
435}
436
437void __ext4_error_inode(struct inode *inode, const char *function,
438			unsigned int line, ext4_fsblk_t block,
439			const char *fmt, ...)
440{
441	va_list args;
442	struct va_format vaf;
443	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
444
445	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
446	es->s_last_error_block = cpu_to_le64(block);
447	if (ext4_error_ratelimit(inode->i_sb)) {
448		va_start(args, fmt);
449		vaf.fmt = fmt;
450		vaf.va = &args;
451		if (block)
452			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
453			       "inode #%lu: block %llu: comm %s: %pV\n",
454			       inode->i_sb->s_id, function, line, inode->i_ino,
455			       block, current->comm, &vaf);
456		else
457			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
458			       "inode #%lu: comm %s: %pV\n",
459			       inode->i_sb->s_id, function, line, inode->i_ino,
460			       current->comm, &vaf);
461		va_end(args);
462	}
463	save_error_info(inode->i_sb, function, line);
464	ext4_handle_error(inode->i_sb);
465}
466
467void __ext4_error_file(struct file *file, const char *function,
468		       unsigned int line, ext4_fsblk_t block,
469		       const char *fmt, ...)
470{
471	va_list args;
472	struct va_format vaf;
473	struct ext4_super_block *es;
474	struct inode *inode = file_inode(file);
475	char pathname[80], *path;
476
477	es = EXT4_SB(inode->i_sb)->s_es;
478	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
479	if (ext4_error_ratelimit(inode->i_sb)) {
480		path = d_path(&(file->f_path), pathname, sizeof(pathname));
481		if (IS_ERR(path))
482			path = "(unknown)";
483		va_start(args, fmt);
484		vaf.fmt = fmt;
485		vaf.va = &args;
486		if (block)
487			printk(KERN_CRIT
488			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
489			       "block %llu: comm %s: path %s: %pV\n",
490			       inode->i_sb->s_id, function, line, inode->i_ino,
491			       block, current->comm, path, &vaf);
492		else
493			printk(KERN_CRIT
494			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
495			       "comm %s: path %s: %pV\n",
496			       inode->i_sb->s_id, function, line, inode->i_ino,
497			       current->comm, path, &vaf);
498		va_end(args);
499	}
500	save_error_info(inode->i_sb, function, line);
501	ext4_handle_error(inode->i_sb);
502}
503
504const char *ext4_decode_error(struct super_block *sb, int errno,
505			      char nbuf[16])
506{
507	char *errstr = NULL;
508
509	switch (errno) {
510	case -EIO:
511		errstr = "IO failure";
512		break;
513	case -ENOMEM:
514		errstr = "Out of memory";
515		break;
516	case -EROFS:
517		if (!sb || (EXT4_SB(sb)->s_journal &&
518			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
519			errstr = "Journal has aborted";
520		else
521			errstr = "Readonly filesystem";
522		break;
523	default:
524		/* If the caller passed in an extra buffer for unknown
525		 * errors, textualise them now.  Else we just return
526		 * NULL. */
527		if (nbuf) {
528			/* Check for truncated error codes... */
529			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
530				errstr = nbuf;
531		}
532		break;
533	}
534
535	return errstr;
536}
537
538/* __ext4_std_error decodes expected errors from journaling functions
539 * automatically and invokes the appropriate error response.  */
540
541void __ext4_std_error(struct super_block *sb, const char *function,
542		      unsigned int line, int errno)
543{
544	char nbuf[16];
545	const char *errstr;
546
547	/* Special case: if the error is EROFS, and we're not already
548	 * inside a transaction, then there's really no point in logging
549	 * an error. */
550	if (errno == -EROFS && journal_current_handle() == NULL &&
551	    (sb->s_flags & MS_RDONLY))
552		return;
553
554	if (ext4_error_ratelimit(sb)) {
555		errstr = ext4_decode_error(sb, errno, nbuf);
556		printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
557		       sb->s_id, function, line, errstr);
558	}
559
560	save_error_info(sb, function, line);
561	ext4_handle_error(sb);
562}
563
564/*
565 * ext4_abort is a much stronger failure handler than ext4_error.  The
566 * abort function may be used to deal with unrecoverable failures such
567 * as journal IO errors or ENOMEM at a critical moment in log management.
568 *
569 * We unconditionally force the filesystem into an ABORT|READONLY state,
570 * unless the error response on the fs has been set to panic in which
571 * case we take the easy way out and panic immediately.
572 */
573
574void __ext4_abort(struct super_block *sb, const char *function,
575		unsigned int line, const char *fmt, ...)
576{
577	va_list args;
578
579	save_error_info(sb, function, line);
580	va_start(args, fmt);
581	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
582	       function, line);
583	vprintk(fmt, args);
584	printk("\n");
585	va_end(args);
586
587	if ((sb->s_flags & MS_RDONLY) == 0) {
588		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
589		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
590		/*
591		 * Make sure updated value of ->s_mount_flags will be visible
592		 * before ->s_flags update
593		 */
594		smp_wmb();
595		sb->s_flags |= MS_RDONLY;
596		if (EXT4_SB(sb)->s_journal)
597			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
598		save_error_info(sb, function, line);
599	}
600	if (test_opt(sb, ERRORS_PANIC))
601		panic("EXT4-fs panic from previous error\n");
602}
603
604void __ext4_msg(struct super_block *sb,
605		const char *prefix, const char *fmt, ...)
606{
607	struct va_format vaf;
608	va_list args;
609
610	if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
611		return;
612
613	va_start(args, fmt);
614	vaf.fmt = fmt;
615	vaf.va = &args;
616	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
617	va_end(args);
618}
619
620void __ext4_warning(struct super_block *sb, const char *function,
621		    unsigned int line, const char *fmt, ...)
622{
623	struct va_format vaf;
624	va_list args;
625
626	if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
627			  "EXT4-fs warning"))
628		return;
629
630	va_start(args, fmt);
631	vaf.fmt = fmt;
632	vaf.va = &args;
633	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
634	       sb->s_id, function, line, &vaf);
635	va_end(args);
636}
637
638void __ext4_grp_locked_error(const char *function, unsigned int line,
639			     struct super_block *sb, ext4_group_t grp,
640			     unsigned long ino, ext4_fsblk_t block,
641			     const char *fmt, ...)
642__releases(bitlock)
643__acquires(bitlock)
644{
645	struct va_format vaf;
646	va_list args;
647	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
648
649	es->s_last_error_ino = cpu_to_le32(ino);
650	es->s_last_error_block = cpu_to_le64(block);
651	__save_error_info(sb, function, line);
652
653	if (ext4_error_ratelimit(sb)) {
654		va_start(args, fmt);
655		vaf.fmt = fmt;
656		vaf.va = &args;
657		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
658		       sb->s_id, function, line, grp);
659		if (ino)
660			printk(KERN_CONT "inode %lu: ", ino);
661		if (block)
662			printk(KERN_CONT "block %llu:",
663			       (unsigned long long) block);
664		printk(KERN_CONT "%pV\n", &vaf);
665		va_end(args);
666	}
667
668	if (test_opt(sb, ERRORS_CONT)) {
669		ext4_commit_super(sb, 0);
670		return;
671	}
672
673	ext4_unlock_group(sb, grp);
674	ext4_handle_error(sb);
675	/*
676	 * We only get here in the ERRORS_RO case; relocking the group
677	 * may be dangerous, but nothing bad will happen since the
678	 * filesystem will have already been marked read/only and the
679	 * journal has been aborted.  We return 1 as a hint to callers
680	 * who might what to use the return value from
681	 * ext4_grp_locked_error() to distinguish between the
682	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
683	 * aggressively from the ext4 function in question, with a
684	 * more appropriate error code.
685	 */
686	ext4_lock_group(sb, grp);
687	return;
688}
689
690void ext4_update_dynamic_rev(struct super_block *sb)
691{
692	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
693
694	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
695		return;
696
697	ext4_warning(sb,
698		     "updating to rev %d because of new feature flag, "
699		     "running e2fsck is recommended",
700		     EXT4_DYNAMIC_REV);
701
702	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
703	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
704	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
705	/* leave es->s_feature_*compat flags alone */
706	/* es->s_uuid will be set by e2fsck if empty */
707
708	/*
709	 * The rest of the superblock fields should be zero, and if not it
710	 * means they are likely already in use, so leave them alone.  We
711	 * can leave it up to e2fsck to clean up any inconsistencies there.
712	 */
713}
714
715/*
716 * Open the external journal device
717 */
718static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
719{
720	struct block_device *bdev;
721	char b[BDEVNAME_SIZE];
722
723	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
724	if (IS_ERR(bdev))
725		goto fail;
726	return bdev;
727
728fail:
729	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
730			__bdevname(dev, b), PTR_ERR(bdev));
731	return NULL;
732}
733
734/*
735 * Release the journal device
736 */
737static void ext4_blkdev_put(struct block_device *bdev)
738{
739	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
740}
741
742static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
743{
744	struct block_device *bdev;
745	bdev = sbi->journal_bdev;
746	if (bdev) {
747		ext4_blkdev_put(bdev);
748		sbi->journal_bdev = NULL;
749	}
750}
751
752static inline struct inode *orphan_list_entry(struct list_head *l)
753{
754	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
755}
756
757static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
758{
759	struct list_head *l;
760
761	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
762		 le32_to_cpu(sbi->s_es->s_last_orphan));
763
764	printk(KERN_ERR "sb_info orphan list:\n");
765	list_for_each(l, &sbi->s_orphan) {
766		struct inode *inode = orphan_list_entry(l);
767		printk(KERN_ERR "  "
768		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
769		       inode->i_sb->s_id, inode->i_ino, inode,
770		       inode->i_mode, inode->i_nlink,
771		       NEXT_ORPHAN(inode));
772	}
773}
774
775static void ext4_put_super(struct super_block *sb)
776{
777	struct ext4_sb_info *sbi = EXT4_SB(sb);
778	struct ext4_super_block *es = sbi->s_es;
779	int i, err;
780
781	ext4_unregister_li_request(sb);
782	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
783
784	flush_workqueue(sbi->rsv_conversion_wq);
785	destroy_workqueue(sbi->rsv_conversion_wq);
786
787	if (sbi->s_journal) {
788		err = jbd2_journal_destroy(sbi->s_journal);
789		sbi->s_journal = NULL;
790		if (err < 0)
791			ext4_abort(sb, "Couldn't clean up the journal");
792	}
793
794	ext4_es_unregister_shrinker(sbi);
795	del_timer_sync(&sbi->s_err_report);
796	ext4_release_system_zone(sb);
797	ext4_mb_release(sb);
798	ext4_ext_release(sb);
799	ext4_xattr_put_super(sb);
800
801	if (!(sb->s_flags & MS_RDONLY)) {
802		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
803		es->s_state = cpu_to_le16(sbi->s_mount_state);
804	}
805	if (!(sb->s_flags & MS_RDONLY))
806		ext4_commit_super(sb, 1);
807
808	if (sbi->s_proc) {
809		remove_proc_entry("options", sbi->s_proc);
810		remove_proc_entry(sb->s_id, ext4_proc_root);
811	}
812	kobject_del(&sbi->s_kobj);
813
814	for (i = 0; i < sbi->s_gdb_count; i++)
815		brelse(sbi->s_group_desc[i]);
816	ext4_kvfree(sbi->s_group_desc);
817	ext4_kvfree(sbi->s_flex_groups);
818	percpu_counter_destroy(&sbi->s_freeclusters_counter);
819	percpu_counter_destroy(&sbi->s_freeinodes_counter);
820	percpu_counter_destroy(&sbi->s_dirs_counter);
821	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
822	brelse(sbi->s_sbh);
823#ifdef CONFIG_QUOTA
824	for (i = 0; i < EXT4_MAXQUOTAS; i++)
825		kfree(sbi->s_qf_names[i]);
826#endif
827
828	/* Debugging code just in case the in-memory inode orphan list
829	 * isn't empty.  The on-disk one can be non-empty if we've
830	 * detected an error and taken the fs readonly, but the
831	 * in-memory list had better be clean by this point. */
832	if (!list_empty(&sbi->s_orphan))
833		dump_orphan_list(sb, sbi);
834	J_ASSERT(list_empty(&sbi->s_orphan));
835
836	invalidate_bdev(sb->s_bdev);
837	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
838		/*
839		 * Invalidate the journal device's buffers.  We don't want them
840		 * floating about in memory - the physical journal device may
841		 * hotswapped, and it breaks the `ro-after' testing code.
842		 */
843		sync_blockdev(sbi->journal_bdev);
844		invalidate_bdev(sbi->journal_bdev);
845		ext4_blkdev_remove(sbi);
846	}
847	if (sbi->s_mb_cache) {
848		ext4_xattr_destroy_cache(sbi->s_mb_cache);
849		sbi->s_mb_cache = NULL;
850	}
851	if (sbi->s_mmp_tsk)
852		kthread_stop(sbi->s_mmp_tsk);
853	sb->s_fs_info = NULL;
854	/*
855	 * Now that we are completely done shutting down the
856	 * superblock, we need to actually destroy the kobject.
857	 */
858	kobject_put(&sbi->s_kobj);
859	wait_for_completion(&sbi->s_kobj_unregister);
860	if (sbi->s_chksum_driver)
861		crypto_free_shash(sbi->s_chksum_driver);
862	kfree(sbi->s_blockgroup_lock);
863	kfree(sbi);
864}
865
866static struct kmem_cache *ext4_inode_cachep;
867
868/*
869 * Called inside transaction, so use GFP_NOFS
870 */
871static struct inode *ext4_alloc_inode(struct super_block *sb)
872{
873	struct ext4_inode_info *ei;
874
875	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
876	if (!ei)
877		return NULL;
878
879	ei->vfs_inode.i_version = 1;
880	spin_lock_init(&ei->i_raw_lock);
881	INIT_LIST_HEAD(&ei->i_prealloc_list);
882	spin_lock_init(&ei->i_prealloc_lock);
883	ext4_es_init_tree(&ei->i_es_tree);
884	rwlock_init(&ei->i_es_lock);
885	INIT_LIST_HEAD(&ei->i_es_lru);
886	ei->i_es_all_nr = 0;
887	ei->i_es_lru_nr = 0;
888	ei->i_touch_when = 0;
889	ei->i_reserved_data_blocks = 0;
890	ei->i_reserved_meta_blocks = 0;
891	ei->i_allocated_meta_blocks = 0;
892	ei->i_da_metadata_calc_len = 0;
893	ei->i_da_metadata_calc_last_lblock = 0;
894	spin_lock_init(&(ei->i_block_reservation_lock));
895#ifdef CONFIG_QUOTA
896	ei->i_reserved_quota = 0;
897#endif
898	ei->jinode = NULL;
899	INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
900	spin_lock_init(&ei->i_completed_io_lock);
901	ei->i_sync_tid = 0;
902	ei->i_datasync_tid = 0;
903	atomic_set(&ei->i_ioend_count, 0);
904	atomic_set(&ei->i_unwritten, 0);
905	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
906
907	return &ei->vfs_inode;
908}
909
910static int ext4_drop_inode(struct inode *inode)
911{
912	int drop = generic_drop_inode(inode);
913
914	trace_ext4_drop_inode(inode, drop);
915	return drop;
916}
917
918static void ext4_i_callback(struct rcu_head *head)
919{
920	struct inode *inode = container_of(head, struct inode, i_rcu);
921	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
922}
923
924static void ext4_destroy_inode(struct inode *inode)
925{
926	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
927		ext4_msg(inode->i_sb, KERN_ERR,
928			 "Inode %lu (%p): orphan list check failed!",
929			 inode->i_ino, EXT4_I(inode));
930		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
931				EXT4_I(inode), sizeof(struct ext4_inode_info),
932				true);
933		dump_stack();
934	}
935	call_rcu(&inode->i_rcu, ext4_i_callback);
936}
937
938static void init_once(void *foo)
939{
940	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
941
942	INIT_LIST_HEAD(&ei->i_orphan);
943	init_rwsem(&ei->xattr_sem);
944	init_rwsem(&ei->i_data_sem);
945	inode_init_once(&ei->vfs_inode);
946}
947
948static int __init init_inodecache(void)
949{
950	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
951					     sizeof(struct ext4_inode_info),
952					     0, (SLAB_RECLAIM_ACCOUNT|
953						SLAB_MEM_SPREAD),
954					     init_once);
955	if (ext4_inode_cachep == NULL)
956		return -ENOMEM;
957	return 0;
958}
959
960static void destroy_inodecache(void)
961{
962	/*
963	 * Make sure all delayed rcu free inodes are flushed before we
964	 * destroy cache.
965	 */
966	rcu_barrier();
967	kmem_cache_destroy(ext4_inode_cachep);
968}
969
970void ext4_clear_inode(struct inode *inode)
971{
972	invalidate_inode_buffers(inode);
973	clear_inode(inode);
974	dquot_drop(inode);
975	ext4_discard_preallocations(inode);
976	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
977	ext4_es_lru_del(inode);
978	if (EXT4_I(inode)->jinode) {
979		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
980					       EXT4_I(inode)->jinode);
981		jbd2_free_inode(EXT4_I(inode)->jinode);
982		EXT4_I(inode)->jinode = NULL;
983	}
984}
985
986static struct inode *ext4_nfs_get_inode(struct super_block *sb,
987					u64 ino, u32 generation)
988{
989	struct inode *inode;
990
991	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
992		return ERR_PTR(-ESTALE);
993	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
994		return ERR_PTR(-ESTALE);
995
996	/* iget isn't really right if the inode is currently unallocated!!
997	 *
998	 * ext4_read_inode will return a bad_inode if the inode had been
999	 * deleted, so we should be safe.
1000	 *
1001	 * Currently we don't know the generation for parent directory, so
1002	 * a generation of 0 means "accept any"
1003	 */
1004	inode = ext4_iget_normal(sb, ino);
1005	if (IS_ERR(inode))
1006		return ERR_CAST(inode);
1007	if (generation && inode->i_generation != generation) {
1008		iput(inode);
1009		return ERR_PTR(-ESTALE);
1010	}
1011
1012	return inode;
1013}
1014
1015static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1016					int fh_len, int fh_type)
1017{
1018	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1019				    ext4_nfs_get_inode);
1020}
1021
1022static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1023					int fh_len, int fh_type)
1024{
1025	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1026				    ext4_nfs_get_inode);
1027}
1028
1029/*
1030 * Try to release metadata pages (indirect blocks, directories) which are
1031 * mapped via the block device.  Since these pages could have journal heads
1032 * which would prevent try_to_free_buffers() from freeing them, we must use
1033 * jbd2 layer's try_to_free_buffers() function to release them.
1034 */
1035static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1036				 gfp_t wait)
1037{
1038	journal_t *journal = EXT4_SB(sb)->s_journal;
1039
1040	WARN_ON(PageChecked(page));
1041	if (!page_has_buffers(page))
1042		return 0;
1043	if (journal)
1044		return jbd2_journal_try_to_free_buffers(journal, page,
1045							wait & ~__GFP_WAIT);
1046	return try_to_free_buffers(page);
1047}
1048
1049#ifdef CONFIG_QUOTA
1050#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1051#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1052
1053static int ext4_write_dquot(struct dquot *dquot);
1054static int ext4_acquire_dquot(struct dquot *dquot);
1055static int ext4_release_dquot(struct dquot *dquot);
1056static int ext4_mark_dquot_dirty(struct dquot *dquot);
1057static int ext4_write_info(struct super_block *sb, int type);
1058static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1059			 struct path *path);
1060static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1061				 int format_id);
1062static int ext4_quota_off(struct super_block *sb, int type);
1063static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1064static int ext4_quota_on_mount(struct super_block *sb, int type);
1065static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1066			       size_t len, loff_t off);
1067static ssize_t ext4_quota_write(struct super_block *sb, int type,
1068				const char *data, size_t len, loff_t off);
1069static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1070			     unsigned int flags);
1071static int ext4_enable_quotas(struct super_block *sb);
1072
1073static const struct dquot_operations ext4_quota_operations = {
1074	.get_reserved_space = ext4_get_reserved_space,
1075	.write_dquot	= ext4_write_dquot,
1076	.acquire_dquot	= ext4_acquire_dquot,
1077	.release_dquot	= ext4_release_dquot,
1078	.mark_dirty	= ext4_mark_dquot_dirty,
1079	.write_info	= ext4_write_info,
1080	.alloc_dquot	= dquot_alloc,
1081	.destroy_dquot	= dquot_destroy,
1082};
1083
1084static const struct quotactl_ops ext4_qctl_operations = {
1085	.quota_on	= ext4_quota_on,
1086	.quota_off	= ext4_quota_off,
1087	.quota_sync	= dquot_quota_sync,
1088	.get_info	= dquot_get_dqinfo,
1089	.set_info	= dquot_set_dqinfo,
1090	.get_dqblk	= dquot_get_dqblk,
1091	.set_dqblk	= dquot_set_dqblk
1092};
1093
1094static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1095	.quota_on_meta	= ext4_quota_on_sysfile,
1096	.quota_off	= ext4_quota_off_sysfile,
1097	.quota_sync	= dquot_quota_sync,
1098	.get_info	= dquot_get_dqinfo,
1099	.set_info	= dquot_set_dqinfo,
1100	.get_dqblk	= dquot_get_dqblk,
1101	.set_dqblk	= dquot_set_dqblk
1102};
1103#endif
1104
1105static const struct super_operations ext4_sops = {
1106	.alloc_inode	= ext4_alloc_inode,
1107	.destroy_inode	= ext4_destroy_inode,
1108	.write_inode	= ext4_write_inode,
1109	.dirty_inode	= ext4_dirty_inode,
1110	.drop_inode	= ext4_drop_inode,
1111	.evict_inode	= ext4_evict_inode,
1112	.put_super	= ext4_put_super,
1113	.sync_fs	= ext4_sync_fs,
1114	.freeze_fs	= ext4_freeze,
1115	.unfreeze_fs	= ext4_unfreeze,
1116	.statfs		= ext4_statfs,
1117	.remount_fs	= ext4_remount,
1118	.show_options	= ext4_show_options,
1119#ifdef CONFIG_QUOTA
1120	.quota_read	= ext4_quota_read,
1121	.quota_write	= ext4_quota_write,
1122#endif
1123	.bdev_try_to_free_page = bdev_try_to_free_page,
1124};
1125
1126static const struct export_operations ext4_export_ops = {
1127	.fh_to_dentry = ext4_fh_to_dentry,
1128	.fh_to_parent = ext4_fh_to_parent,
1129	.get_parent = ext4_get_parent,
1130};
1131
1132enum {
1133	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1134	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1135	Opt_nouid32, Opt_debug, Opt_removed,
1136	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1137	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1138	Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1139	Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1140	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1141	Opt_data_err_abort, Opt_data_err_ignore,
1142	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1143	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1144	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1145	Opt_usrquota, Opt_grpquota, Opt_i_version,
1146	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1147	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1148	Opt_inode_readahead_blks, Opt_journal_ioprio,
1149	Opt_dioread_nolock, Opt_dioread_lock,
1150	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1151	Opt_max_dir_size_kb,
1152};
1153
1154static const match_table_t tokens = {
1155	{Opt_bsd_df, "bsddf"},
1156	{Opt_minix_df, "minixdf"},
1157	{Opt_grpid, "grpid"},
1158	{Opt_grpid, "bsdgroups"},
1159	{Opt_nogrpid, "nogrpid"},
1160	{Opt_nogrpid, "sysvgroups"},
1161	{Opt_resgid, "resgid=%u"},
1162	{Opt_resuid, "resuid=%u"},
1163	{Opt_sb, "sb=%u"},
1164	{Opt_err_cont, "errors=continue"},
1165	{Opt_err_panic, "errors=panic"},
1166	{Opt_err_ro, "errors=remount-ro"},
1167	{Opt_nouid32, "nouid32"},
1168	{Opt_debug, "debug"},
1169	{Opt_removed, "oldalloc"},
1170	{Opt_removed, "orlov"},
1171	{Opt_user_xattr, "user_xattr"},
1172	{Opt_nouser_xattr, "nouser_xattr"},
1173	{Opt_acl, "acl"},
1174	{Opt_noacl, "noacl"},
1175	{Opt_noload, "norecovery"},
1176	{Opt_noload, "noload"},
1177	{Opt_removed, "nobh"},
1178	{Opt_removed, "bh"},
1179	{Opt_commit, "commit=%u"},
1180	{Opt_min_batch_time, "min_batch_time=%u"},
1181	{Opt_max_batch_time, "max_batch_time=%u"},
1182	{Opt_journal_dev, "journal_dev=%u"},
1183	{Opt_journal_path, "journal_path=%s"},
1184	{Opt_journal_checksum, "journal_checksum"},
1185	{Opt_journal_async_commit, "journal_async_commit"},
1186	{Opt_abort, "abort"},
1187	{Opt_data_journal, "data=journal"},
1188	{Opt_data_ordered, "data=ordered"},
1189	{Opt_data_writeback, "data=writeback"},
1190	{Opt_data_err_abort, "data_err=abort"},
1191	{Opt_data_err_ignore, "data_err=ignore"},
1192	{Opt_offusrjquota, "usrjquota="},
1193	{Opt_usrjquota, "usrjquota=%s"},
1194	{Opt_offgrpjquota, "grpjquota="},
1195	{Opt_grpjquota, "grpjquota=%s"},
1196	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1197	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1198	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1199	{Opt_grpquota, "grpquota"},
1200	{Opt_noquota, "noquota"},
1201	{Opt_quota, "quota"},
1202	{Opt_usrquota, "usrquota"},
1203	{Opt_barrier, "barrier=%u"},
1204	{Opt_barrier, "barrier"},
1205	{Opt_nobarrier, "nobarrier"},
1206	{Opt_i_version, "i_version"},
1207	{Opt_stripe, "stripe=%u"},
1208	{Opt_delalloc, "delalloc"},
1209	{Opt_nodelalloc, "nodelalloc"},
1210	{Opt_removed, "mblk_io_submit"},
1211	{Opt_removed, "nomblk_io_submit"},
1212	{Opt_block_validity, "block_validity"},
1213	{Opt_noblock_validity, "noblock_validity"},
1214	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1215	{Opt_journal_ioprio, "journal_ioprio=%u"},
1216	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1217	{Opt_auto_da_alloc, "auto_da_alloc"},
1218	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1219	{Opt_dioread_nolock, "dioread_nolock"},
1220	{Opt_dioread_lock, "dioread_lock"},
1221	{Opt_discard, "discard"},
1222	{Opt_nodiscard, "nodiscard"},
1223	{Opt_init_itable, "init_itable=%u"},
1224	{Opt_init_itable, "init_itable"},
1225	{Opt_noinit_itable, "noinit_itable"},
1226	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1227	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1228	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1229	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1230	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1231	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1232	{Opt_err, NULL},
1233};
1234
1235static ext4_fsblk_t get_sb_block(void **data)
1236{
1237	ext4_fsblk_t	sb_block;
1238	char		*options = (char *) *data;
1239
1240	if (!options || strncmp(options, "sb=", 3) != 0)
1241		return 1;	/* Default location */
1242
1243	options += 3;
1244	/* TODO: use simple_strtoll with >32bit ext4 */
1245	sb_block = simple_strtoul(options, &options, 0);
1246	if (*options && *options != ',') {
1247		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1248		       (char *) *data);
1249		return 1;
1250	}
1251	if (*options == ',')
1252		options++;
1253	*data = (void *) options;
1254
1255	return sb_block;
1256}
1257
1258#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1259static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1260	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1261
1262#ifdef CONFIG_QUOTA
1263static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1264{
1265	struct ext4_sb_info *sbi = EXT4_SB(sb);
1266	char *qname;
1267	int ret = -1;
1268
1269	if (sb_any_quota_loaded(sb) &&
1270		!sbi->s_qf_names[qtype]) {
1271		ext4_msg(sb, KERN_ERR,
1272			"Cannot change journaled "
1273			"quota options when quota turned on");
1274		return -1;
1275	}
1276	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1277		ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1278			 "when QUOTA feature is enabled");
1279		return -1;
1280	}
1281	qname = match_strdup(args);
1282	if (!qname) {
1283		ext4_msg(sb, KERN_ERR,
1284			"Not enough memory for storing quotafile name");
1285		return -1;
1286	}
1287	if (sbi->s_qf_names[qtype]) {
1288		if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1289			ret = 1;
1290		else
1291			ext4_msg(sb, KERN_ERR,
1292				 "%s quota file already specified",
1293				 QTYPE2NAME(qtype));
1294		goto errout;
1295	}
1296	if (strchr(qname, '/')) {
1297		ext4_msg(sb, KERN_ERR,
1298			"quotafile must be on filesystem root");
1299		goto errout;
1300	}
1301	sbi->s_qf_names[qtype] = qname;
1302	set_opt(sb, QUOTA);
1303	return 1;
1304errout:
1305	kfree(qname);
1306	return ret;
1307}
1308
1309static int clear_qf_name(struct super_block *sb, int qtype)
1310{
1311
1312	struct ext4_sb_info *sbi = EXT4_SB(sb);
1313
1314	if (sb_any_quota_loaded(sb) &&
1315		sbi->s_qf_names[qtype]) {
1316		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1317			" when quota turned on");
1318		return -1;
1319	}
1320	kfree(sbi->s_qf_names[qtype]);
1321	sbi->s_qf_names[qtype] = NULL;
1322	return 1;
1323}
1324#endif
1325
1326#define MOPT_SET	0x0001
1327#define MOPT_CLEAR	0x0002
1328#define MOPT_NOSUPPORT	0x0004
1329#define MOPT_EXPLICIT	0x0008
1330#define MOPT_CLEAR_ERR	0x0010
1331#define MOPT_GTE0	0x0020
1332#ifdef CONFIG_QUOTA
1333#define MOPT_Q		0
1334#define MOPT_QFMT	0x0040
1335#else
1336#define MOPT_Q		MOPT_NOSUPPORT
1337#define MOPT_QFMT	MOPT_NOSUPPORT
1338#endif
1339#define MOPT_DATAJ	0x0080
1340#define MOPT_NO_EXT2	0x0100
1341#define MOPT_NO_EXT3	0x0200
1342#define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1343#define MOPT_STRING	0x0400
1344
1345static const struct mount_opts {
1346	int	token;
1347	int	mount_opt;
1348	int	flags;
1349} ext4_mount_opts[] = {
1350	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1351	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1352	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1353	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1354	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1355	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1356	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1357	 MOPT_EXT4_ONLY | MOPT_SET},
1358	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1359	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1360	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1361	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1362	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1363	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1364	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1365	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1366	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1367	 MOPT_EXT4_ONLY | MOPT_SET},
1368	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1369				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1370	 MOPT_EXT4_ONLY | MOPT_SET},
1371	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1372	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1373	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1374	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1375	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1376	 MOPT_NO_EXT2 | MOPT_SET},
1377	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1378	 MOPT_NO_EXT2 | MOPT_CLEAR},
1379	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1380	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1381	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1382	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1383	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1384	{Opt_commit, 0, MOPT_GTE0},
1385	{Opt_max_batch_time, 0, MOPT_GTE0},
1386	{Opt_min_batch_time, 0, MOPT_GTE0},
1387	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1388	{Opt_init_itable, 0, MOPT_GTE0},
1389	{Opt_stripe, 0, MOPT_GTE0},
1390	{Opt_resuid, 0, MOPT_GTE0},
1391	{Opt_resgid, 0, MOPT_GTE0},
1392	{Opt_journal_dev, 0, MOPT_GTE0},
1393	{Opt_journal_path, 0, MOPT_STRING},
1394	{Opt_journal_ioprio, 0, MOPT_GTE0},
1395	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1396	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1397	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1398	 MOPT_NO_EXT2 | MOPT_DATAJ},
1399	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1400	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1401#ifdef CONFIG_EXT4_FS_POSIX_ACL
1402	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1403	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1404#else
1405	{Opt_acl, 0, MOPT_NOSUPPORT},
1406	{Opt_noacl, 0, MOPT_NOSUPPORT},
1407#endif
1408	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1409	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1410	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1411	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1412							MOPT_SET | MOPT_Q},
1413	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1414							MOPT_SET | MOPT_Q},
1415	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1416		       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1417	{Opt_usrjquota, 0, MOPT_Q},
1418	{Opt_grpjquota, 0, MOPT_Q},
1419	{Opt_offusrjquota, 0, MOPT_Q},
1420	{Opt_offgrpjquota, 0, MOPT_Q},
1421	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1422	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1423	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1424	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1425	{Opt_err, 0, 0}
1426};
1427
1428static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1429			    substring_t *args, unsigned long *journal_devnum,
1430			    unsigned int *journal_ioprio, int is_remount)
1431{
1432	struct ext4_sb_info *sbi = EXT4_SB(sb);
1433	const struct mount_opts *m;
1434	kuid_t uid;
1435	kgid_t gid;
1436	int arg = 0;
1437
1438#ifdef CONFIG_QUOTA
1439	if (token == Opt_usrjquota)
1440		return set_qf_name(sb, USRQUOTA, &args[0]);
1441	else if (token == Opt_grpjquota)
1442		return set_qf_name(sb, GRPQUOTA, &args[0]);
1443	else if (token == Opt_offusrjquota)
1444		return clear_qf_name(sb, USRQUOTA);
1445	else if (token == Opt_offgrpjquota)
1446		return clear_qf_name(sb, GRPQUOTA);
1447#endif
1448	switch (token) {
1449	case Opt_noacl:
1450	case Opt_nouser_xattr:
1451		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1452		break;
1453	case Opt_sb:
1454		return 1;	/* handled by get_sb_block() */
1455	case Opt_removed:
1456		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1457		return 1;
1458	case Opt_abort:
1459		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1460		return 1;
1461	case Opt_i_version:
1462		sb->s_flags |= MS_I_VERSION;
1463		return 1;
1464	}
1465
1466	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1467		if (token == m->token)
1468			break;
1469
1470	if (m->token == Opt_err) {
1471		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1472			 "or missing value", opt);
1473		return -1;
1474	}
1475
1476	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1477		ext4_msg(sb, KERN_ERR,
1478			 "Mount option \"%s\" incompatible with ext2", opt);
1479		return -1;
1480	}
1481	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1482		ext4_msg(sb, KERN_ERR,
1483			 "Mount option \"%s\" incompatible with ext3", opt);
1484		return -1;
1485	}
1486
1487	if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1488		return -1;
1489	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1490		return -1;
1491	if (m->flags & MOPT_EXPLICIT)
1492		set_opt2(sb, EXPLICIT_DELALLOC);
1493	if (m->flags & MOPT_CLEAR_ERR)
1494		clear_opt(sb, ERRORS_MASK);
1495	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1496		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1497			 "options when quota turned on");
1498		return -1;
1499	}
1500
1501	if (m->flags & MOPT_NOSUPPORT) {
1502		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1503	} else if (token == Opt_commit) {
1504		if (arg == 0)
1505			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1506		sbi->s_commit_interval = HZ * arg;
1507	} else if (token == Opt_max_batch_time) {
1508		sbi->s_max_batch_time = arg;
1509	} else if (token == Opt_min_batch_time) {
1510		sbi->s_min_batch_time = arg;
1511	} else if (token == Opt_inode_readahead_blks) {
1512		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1513			ext4_msg(sb, KERN_ERR,
1514				 "EXT4-fs: inode_readahead_blks must be "
1515				 "0 or a power of 2 smaller than 2^31");
1516			return -1;
1517		}
1518		sbi->s_inode_readahead_blks = arg;
1519	} else if (token == Opt_init_itable) {
1520		set_opt(sb, INIT_INODE_TABLE);
1521		if (!args->from)
1522			arg = EXT4_DEF_LI_WAIT_MULT;
1523		sbi->s_li_wait_mult = arg;
1524	} else if (token == Opt_max_dir_size_kb) {
1525		sbi->s_max_dir_size_kb = arg;
1526	} else if (token == Opt_stripe) {
1527		sbi->s_stripe = arg;
1528	} else if (token == Opt_resuid) {
1529		uid = make_kuid(current_user_ns(), arg);
1530		if (!uid_valid(uid)) {
1531			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1532			return -1;
1533		}
1534		sbi->s_resuid = uid;
1535	} else if (token == Opt_resgid) {
1536		gid = make_kgid(current_user_ns(), arg);
1537		if (!gid_valid(gid)) {
1538			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1539			return -1;
1540		}
1541		sbi->s_resgid = gid;
1542	} else if (token == Opt_journal_dev) {
1543		if (is_remount) {
1544			ext4_msg(sb, KERN_ERR,
1545				 "Cannot specify journal on remount");
1546			return -1;
1547		}
1548		*journal_devnum = arg;
1549	} else if (token == Opt_journal_path) {
1550		char *journal_path;
1551		struct inode *journal_inode;
1552		struct path path;
1553		int error;
1554
1555		if (is_remount) {
1556			ext4_msg(sb, KERN_ERR,
1557				 "Cannot specify journal on remount");
1558			return -1;
1559		}
1560		journal_path = match_strdup(&args[0]);
1561		if (!journal_path) {
1562			ext4_msg(sb, KERN_ERR, "error: could not dup "
1563				"journal device string");
1564			return -1;
1565		}
1566
1567		error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1568		if (error) {
1569			ext4_msg(sb, KERN_ERR, "error: could not find "
1570				"journal device path: error %d", error);
1571			kfree(journal_path);
1572			return -1;
1573		}
1574
1575		journal_inode = path.dentry->d_inode;
1576		if (!S_ISBLK(journal_inode->i_mode)) {
1577			ext4_msg(sb, KERN_ERR, "error: journal path %s "
1578				"is not a block device", journal_path);
1579			path_put(&path);
1580			kfree(journal_path);
1581			return -1;
1582		}
1583
1584		*journal_devnum = new_encode_dev(journal_inode->i_rdev);
1585		path_put(&path);
1586		kfree(journal_path);
1587	} else if (token == Opt_journal_ioprio) {
1588		if (arg > 7) {
1589			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1590				 " (must be 0-7)");
1591			return -1;
1592		}
1593		*journal_ioprio =
1594			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1595	} else if (m->flags & MOPT_DATAJ) {
1596		if (is_remount) {
1597			if (!sbi->s_journal)
1598				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1599			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1600				ext4_msg(sb, KERN_ERR,
1601					 "Cannot change data mode on remount");
1602				return -1;
1603			}
1604		} else {
1605			clear_opt(sb, DATA_FLAGS);
1606			sbi->s_mount_opt |= m->mount_opt;
1607		}
1608#ifdef CONFIG_QUOTA
1609	} else if (m->flags & MOPT_QFMT) {
1610		if (sb_any_quota_loaded(sb) &&
1611		    sbi->s_jquota_fmt != m->mount_opt) {
1612			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1613				 "quota options when quota turned on");
1614			return -1;
1615		}
1616		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1617					       EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1618			ext4_msg(sb, KERN_ERR,
1619				 "Cannot set journaled quota options "
1620				 "when QUOTA feature is enabled");
1621			return -1;
1622		}
1623		sbi->s_jquota_fmt = m->mount_opt;
1624#endif
1625	} else {
1626		if (!args->from)
1627			arg = 1;
1628		if (m->flags & MOPT_CLEAR)
1629			arg = !arg;
1630		else if (unlikely(!(m->flags & MOPT_SET))) {
1631			ext4_msg(sb, KERN_WARNING,
1632				 "buggy handling of option %s", opt);
1633			WARN_ON(1);
1634			return -1;
1635		}
1636		if (arg != 0)
1637			sbi->s_mount_opt |= m->mount_opt;
1638		else
1639			sbi->s_mount_opt &= ~m->mount_opt;
1640	}
1641	return 1;
1642}
1643
1644static int parse_options(char *options, struct super_block *sb,
1645			 unsigned long *journal_devnum,
1646			 unsigned int *journal_ioprio,
1647			 int is_remount)
1648{
1649	struct ext4_sb_info *sbi = EXT4_SB(sb);
1650	char *p;
1651	substring_t args[MAX_OPT_ARGS];
1652	int token;
1653
1654	if (!options)
1655		return 1;
1656
1657	while ((p = strsep(&options, ",")) != NULL) {
1658		if (!*p)
1659			continue;
1660		/*
1661		 * Initialize args struct so we know whether arg was
1662		 * found; some options take optional arguments.
1663		 */
1664		args[0].to = args[0].from = NULL;
1665		token = match_token(p, tokens, args);
1666		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1667				     journal_ioprio, is_remount) < 0)
1668			return 0;
1669	}
1670#ifdef CONFIG_QUOTA
1671	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1672	    (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1673		ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1674			 "feature is enabled");
1675		return 0;
1676	}
1677	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1678		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1679			clear_opt(sb, USRQUOTA);
1680
1681		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1682			clear_opt(sb, GRPQUOTA);
1683
1684		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1685			ext4_msg(sb, KERN_ERR, "old and new quota "
1686					"format mixing");
1687			return 0;
1688		}
1689
1690		if (!sbi->s_jquota_fmt) {
1691			ext4_msg(sb, KERN_ERR, "journaled quota format "
1692					"not specified");
1693			return 0;
1694		}
1695	}
1696#endif
1697	if (test_opt(sb, DIOREAD_NOLOCK)) {
1698		int blocksize =
1699			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1700
1701		if (blocksize < PAGE_CACHE_SIZE) {
1702			ext4_msg(sb, KERN_ERR, "can't mount with "
1703				 "dioread_nolock if block size != PAGE_SIZE");
1704			return 0;
1705		}
1706	}
1707	return 1;
1708}
1709
1710static inline void ext4_show_quota_options(struct seq_file *seq,
1711					   struct super_block *sb)
1712{
1713#if defined(CONFIG_QUOTA)
1714	struct ext4_sb_info *sbi = EXT4_SB(sb);
1715
1716	if (sbi->s_jquota_fmt) {
1717		char *fmtname = "";
1718
1719		switch (sbi->s_jquota_fmt) {
1720		case QFMT_VFS_OLD:
1721			fmtname = "vfsold";
1722			break;
1723		case QFMT_VFS_V0:
1724			fmtname = "vfsv0";
1725			break;
1726		case QFMT_VFS_V1:
1727			fmtname = "vfsv1";
1728			break;
1729		}
1730		seq_printf(seq, ",jqfmt=%s", fmtname);
1731	}
1732
1733	if (sbi->s_qf_names[USRQUOTA])
1734		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1735
1736	if (sbi->s_qf_names[GRPQUOTA])
1737		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1738#endif
1739}
1740
1741static const char *token2str(int token)
1742{
1743	const struct match_token *t;
1744
1745	for (t = tokens; t->token != Opt_err; t++)
1746		if (t->token == token && !strchr(t->pattern, '='))
1747			break;
1748	return t->pattern;
1749}
1750
1751/*
1752 * Show an option if
1753 *  - it's set to a non-default value OR
1754 *  - if the per-sb default is different from the global default
1755 */
1756static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1757			      int nodefs)
1758{
1759	struct ext4_sb_info *sbi = EXT4_SB(sb);
1760	struct ext4_super_block *es = sbi->s_es;
1761	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1762	const struct mount_opts *m;
1763	char sep = nodefs ? '\n' : ',';
1764
1765#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1766#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1767
1768	if (sbi->s_sb_block != 1)
1769		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1770
1771	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1772		int want_set = m->flags & MOPT_SET;
1773		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1774		    (m->flags & MOPT_CLEAR_ERR))
1775			continue;
1776		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1777			continue; /* skip if same as the default */
1778		if ((want_set &&
1779		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1780		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1781			continue; /* select Opt_noFoo vs Opt_Foo */
1782		SEQ_OPTS_PRINT("%s", token2str(m->token));
1783	}
1784
1785	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1786	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1787		SEQ_OPTS_PRINT("resuid=%u",
1788				from_kuid_munged(&init_user_ns, sbi->s_resuid));
1789	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1790	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1791		SEQ_OPTS_PRINT("resgid=%u",
1792				from_kgid_munged(&init_user_ns, sbi->s_resgid));
1793	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1794	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1795		SEQ_OPTS_PUTS("errors=remount-ro");
1796	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1797		SEQ_OPTS_PUTS("errors=continue");
1798	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1799		SEQ_OPTS_PUTS("errors=panic");
1800	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1801		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1802	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1803		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1804	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1805		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1806	if (sb->s_flags & MS_I_VERSION)
1807		SEQ_OPTS_PUTS("i_version");
1808	if (nodefs || sbi->s_stripe)
1809		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1810	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1811		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1812			SEQ_OPTS_PUTS("data=journal");
1813		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1814			SEQ_OPTS_PUTS("data=ordered");
1815		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1816			SEQ_OPTS_PUTS("data=writeback");
1817	}
1818	if (nodefs ||
1819	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1820		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1821			       sbi->s_inode_readahead_blks);
1822
1823	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1824		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1825		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1826	if (nodefs || sbi->s_max_dir_size_kb)
1827		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1828
1829	ext4_show_quota_options(seq, sb);
1830	return 0;
1831}
1832
1833static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1834{
1835	return _ext4_show_options(seq, root->d_sb, 0);
1836}
1837
1838static int options_seq_show(struct seq_file *seq, void *offset)
1839{
1840	struct super_block *sb = seq->private;
1841	int rc;
1842
1843	seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1844	rc = _ext4_show_options(seq, sb, 1);
1845	seq_puts(seq, "\n");
1846	return rc;
1847}
1848
1849static int options_open_fs(struct inode *inode, struct file *file)
1850{
1851	return single_open(file, options_seq_show, PDE_DATA(inode));
1852}
1853
1854static const struct file_operations ext4_seq_options_fops = {
1855	.owner = THIS_MODULE,
1856	.open = options_open_fs,
1857	.read = seq_read,
1858	.llseek = seq_lseek,
1859	.release = single_release,
1860};
1861
1862static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1863			    int read_only)
1864{
1865	struct ext4_sb_info *sbi = EXT4_SB(sb);
1866	int res = 0;
1867
1868	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1869		ext4_msg(sb, KERN_ERR, "revision level too high, "
1870			 "forcing read-only mode");
1871		res = MS_RDONLY;
1872	}
1873	if (read_only)
1874		goto done;
1875	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1876		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1877			 "running e2fsck is recommended");
1878	else if (sbi->s_mount_state & EXT4_ERROR_FS)
1879		ext4_msg(sb, KERN_WARNING,
1880			 "warning: mounting fs with errors, "
1881			 "running e2fsck is recommended");
1882	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1883		 le16_to_cpu(es->s_mnt_count) >=
1884		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1885		ext4_msg(sb, KERN_WARNING,
1886			 "warning: maximal mount count reached, "
1887			 "running e2fsck is recommended");
1888	else if (le32_to_cpu(es->s_checkinterval) &&
1889		(le32_to_cpu(es->s_lastcheck) +
1890			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1891		ext4_msg(sb, KERN_WARNING,
1892			 "warning: checktime reached, "
1893			 "running e2fsck is recommended");
1894	if (!sbi->s_journal)
1895		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1896	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1897		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1898	le16_add_cpu(&es->s_mnt_count, 1);
1899	es->s_mtime = cpu_to_le32(get_seconds());
1900	ext4_update_dynamic_rev(sb);
1901	if (sbi->s_journal)
1902		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1903
1904	ext4_commit_super(sb, 1);
1905done:
1906	if (test_opt(sb, DEBUG))
1907		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1908				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1909			sb->s_blocksize,
1910			sbi->s_groups_count,
1911			EXT4_BLOCKS_PER_GROUP(sb),
1912			EXT4_INODES_PER_GROUP(sb),
1913			sbi->s_mount_opt, sbi->s_mount_opt2);
1914
1915	cleancache_init_fs(sb);
1916	return res;
1917}
1918
1919int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1920{
1921	struct ext4_sb_info *sbi = EXT4_SB(sb);
1922	struct flex_groups *new_groups;
1923	int size;
1924
1925	if (!sbi->s_log_groups_per_flex)
1926		return 0;
1927
1928	size = ext4_flex_group(sbi, ngroup - 1) + 1;
1929	if (size <= sbi->s_flex_groups_allocated)
1930		return 0;
1931
1932	size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1933	new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1934	if (!new_groups) {
1935		ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1936			 size / (int) sizeof(struct flex_groups));
1937		return -ENOMEM;
1938	}
1939
1940	if (sbi->s_flex_groups) {
1941		memcpy(new_groups, sbi->s_flex_groups,
1942		       (sbi->s_flex_groups_allocated *
1943			sizeof(struct flex_groups)));
1944		ext4_kvfree(sbi->s_flex_groups);
1945	}
1946	sbi->s_flex_groups = new_groups;
1947	sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1948	return 0;
1949}
1950
1951static int ext4_fill_flex_info(struct super_block *sb)
1952{
1953	struct ext4_sb_info *sbi = EXT4_SB(sb);
1954	struct ext4_group_desc *gdp = NULL;
1955	ext4_group_t flex_group;
1956	int i, err;
1957
1958	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1959	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1960		sbi->s_log_groups_per_flex = 0;
1961		return 1;
1962	}
1963
1964	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1965	if (err)
1966		goto failed;
1967
1968	for (i = 0; i < sbi->s_groups_count; i++) {
1969		gdp = ext4_get_group_desc(sb, i, NULL);
1970
1971		flex_group = ext4_flex_group(sbi, i);
1972		atomic_add(ext4_free_inodes_count(sb, gdp),
1973			   &sbi->s_flex_groups[flex_group].free_inodes);
1974		atomic64_add(ext4_free_group_clusters(sb, gdp),
1975			     &sbi->s_flex_groups[flex_group].free_clusters);
1976		atomic_add(ext4_used_dirs_count(sb, gdp),
1977			   &sbi->s_flex_groups[flex_group].used_dirs);
1978	}
1979
1980	return 1;
1981failed:
1982	return 0;
1983}
1984
1985static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1986				   struct ext4_group_desc *gdp)
1987{
1988	int offset;
1989	__u16 crc = 0;
1990	__le32 le_group = cpu_to_le32(block_group);
1991
1992	if (ext4_has_metadata_csum(sbi->s_sb)) {
1993		/* Use new metadata_csum algorithm */
1994		__le16 save_csum;
1995		__u32 csum32;
1996
1997		save_csum = gdp->bg_checksum;
1998		gdp->bg_checksum = 0;
1999		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2000				     sizeof(le_group));
2001		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2002				     sbi->s_desc_size);
2003		gdp->bg_checksum = save_csum;
2004
2005		crc = csum32 & 0xFFFF;
2006		goto out;
2007	}
2008
2009	/* old crc16 code */
2010	if (!(sbi->s_es->s_feature_ro_compat &
2011	      cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2012		return 0;
2013
2014	offset = offsetof(struct ext4_group_desc, bg_checksum);
2015
2016	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2017	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2018	crc = crc16(crc, (__u8 *)gdp, offset);
2019	offset += sizeof(gdp->bg_checksum); /* skip checksum */
2020	/* for checksum of struct ext4_group_desc do the rest...*/
2021	if ((sbi->s_es->s_feature_incompat &
2022	     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2023	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
2024		crc = crc16(crc, (__u8 *)gdp + offset,
2025			    le16_to_cpu(sbi->s_es->s_desc_size) -
2026				offset);
2027
2028out:
2029	return cpu_to_le16(crc);
2030}
2031
2032int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2033				struct ext4_group_desc *gdp)
2034{
2035	if (ext4_has_group_desc_csum(sb) &&
2036	    (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2037						      block_group, gdp)))
2038		return 0;
2039
2040	return 1;
2041}
2042
2043void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2044			      struct ext4_group_desc *gdp)
2045{
2046	if (!ext4_has_group_desc_csum(sb))
2047		return;
2048	gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2049}
2050
2051/* Called at mount-time, super-block is locked */
2052static int ext4_check_descriptors(struct super_block *sb,
2053				  ext4_group_t *first_not_zeroed)
2054{
2055	struct ext4_sb_info *sbi = EXT4_SB(sb);
2056	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2057	ext4_fsblk_t last_block;
2058	ext4_fsblk_t block_bitmap;
2059	ext4_fsblk_t inode_bitmap;
2060	ext4_fsblk_t inode_table;
2061	int flexbg_flag = 0;
2062	ext4_group_t i, grp = sbi->s_groups_count;
2063
2064	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2065		flexbg_flag = 1;
2066
2067	ext4_debug("Checking group descriptors");
2068
2069	for (i = 0; i < sbi->s_groups_count; i++) {
2070		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2071
2072		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2073			last_block = ext4_blocks_count(sbi->s_es) - 1;
2074		else
2075			last_block = first_block +
2076				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2077
2078		if ((grp == sbi->s_groups_count) &&
2079		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2080			grp = i;
2081
2082		block_bitmap = ext4_block_bitmap(sb, gdp);
2083		if (block_bitmap < first_block || block_bitmap > last_block) {
2084			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2085			       "Block bitmap for group %u not in group "
2086			       "(block %llu)!", i, block_bitmap);
2087			return 0;
2088		}
2089		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2090		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2091			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2092			       "Inode bitmap for group %u not in group "
2093			       "(block %llu)!", i, inode_bitmap);
2094			return 0;
2095		}
2096		inode_table = ext4_inode_table(sb, gdp);
2097		if (inode_table < first_block ||
2098		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2099			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2100			       "Inode table for group %u not in group "
2101			       "(block %llu)!", i, inode_table);
2102			return 0;
2103		}
2104		ext4_lock_group(sb, i);
2105		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2106			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2107				 "Checksum for group %u failed (%u!=%u)",
2108				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2109				     gdp)), le16_to_cpu(gdp->bg_checksum));
2110			if (!(sb->s_flags & MS_RDONLY)) {
2111				ext4_unlock_group(sb, i);
2112				return 0;
2113			}
2114		}
2115		ext4_unlock_group(sb, i);
2116		if (!flexbg_flag)
2117			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2118	}
2119	if (NULL != first_not_zeroed)
2120		*first_not_zeroed = grp;
2121	return 1;
2122}
2123
2124/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2125 * the superblock) which were deleted from all directories, but held open by
2126 * a process at the time of a crash.  We walk the list and try to delete these
2127 * inodes at recovery time (only with a read-write filesystem).
2128 *
2129 * In order to keep the orphan inode chain consistent during traversal (in
2130 * case of crash during recovery), we link each inode into the superblock
2131 * orphan list_head and handle it the same way as an inode deletion during
2132 * normal operation (which journals the operations for us).
2133 *
2134 * We only do an iget() and an iput() on each inode, which is very safe if we
2135 * accidentally point at an in-use or already deleted inode.  The worst that
2136 * can happen in this case is that we get a "bit already cleared" message from
2137 * ext4_free_inode().  The only reason we would point at a wrong inode is if
2138 * e2fsck was run on this filesystem, and it must have already done the orphan
2139 * inode cleanup for us, so we can safely abort without any further action.
2140 */
2141static void ext4_orphan_cleanup(struct super_block *sb,
2142				struct ext4_super_block *es)
2143{
2144	unsigned int s_flags = sb->s_flags;
2145	int nr_orphans = 0, nr_truncates = 0;
2146#ifdef CONFIG_QUOTA
2147	int i;
2148#endif
2149	if (!es->s_last_orphan) {
2150		jbd_debug(4, "no orphan inodes to clean up\n");
2151		return;
2152	}
2153
2154	if (bdev_read_only(sb->s_bdev)) {
2155		ext4_msg(sb, KERN_ERR, "write access "
2156			"unavailable, skipping orphan cleanup");
2157		return;
2158	}
2159
2160	/* Check if feature set would not allow a r/w mount */
2161	if (!ext4_feature_set_ok(sb, 0)) {
2162		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2163			 "unknown ROCOMPAT features");
2164		return;
2165	}
2166
2167	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2168		/* don't clear list on RO mount w/ errors */
2169		if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2170			ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2171				  "clearing orphan list.\n");
2172			es->s_last_orphan = 0;
2173		}
2174		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2175		return;
2176	}
2177
2178	if (s_flags & MS_RDONLY) {
2179		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2180		sb->s_flags &= ~MS_RDONLY;
2181	}
2182#ifdef CONFIG_QUOTA
2183	/* Needed for iput() to work correctly and not trash data */
2184	sb->s_flags |= MS_ACTIVE;
2185	/* Turn on quotas so that they are updated correctly */
2186	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2187		if (EXT4_SB(sb)->s_qf_names[i]) {
2188			int ret = ext4_quota_on_mount(sb, i);
2189			if (ret < 0)
2190				ext4_msg(sb, KERN_ERR,
2191					"Cannot turn on journaled "
2192					"quota: error %d", ret);
2193		}
2194	}
2195#endif
2196
2197	while (es->s_last_orphan) {
2198		struct inode *inode;
2199
2200		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2201		if (IS_ERR(inode)) {
2202			es->s_last_orphan = 0;
2203			break;
2204		}
2205
2206		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2207		dquot_initialize(inode);
2208		if (inode->i_nlink) {
2209			if (test_opt(sb, DEBUG))
2210				ext4_msg(sb, KERN_DEBUG,
2211					"%s: truncating inode %lu to %lld bytes",
2212					__func__, inode->i_ino, inode->i_size);
2213			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2214				  inode->i_ino, inode->i_size);
2215			mutex_lock(&inode->i_mutex);
2216			truncate_inode_pages(inode->i_mapping, inode->i_size);
2217			ext4_truncate(inode);
2218			mutex_unlock(&inode->i_mutex);
2219			nr_truncates++;
2220		} else {
2221			if (test_opt(sb, DEBUG))
2222				ext4_msg(sb, KERN_DEBUG,
2223					"%s: deleting unreferenced inode %lu",
2224					__func__, inode->i_ino);
2225			jbd_debug(2, "deleting unreferenced inode %lu\n",
2226				  inode->i_ino);
2227			nr_orphans++;
2228		}
2229		iput(inode);  /* The delete magic happens here! */
2230	}
2231
2232#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2233
2234	if (nr_orphans)
2235		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2236		       PLURAL(nr_orphans));
2237	if (nr_truncates)
2238		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2239		       PLURAL(nr_truncates));
2240#ifdef CONFIG_QUOTA
2241	/* Turn quotas off */
2242	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2243		if (sb_dqopt(sb)->files[i])
2244			dquot_quota_off(sb, i);
2245	}
2246#endif
2247	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2248}
2249
2250/*
2251 * Maximal extent format file size.
2252 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2253 * extent format containers, within a sector_t, and within i_blocks
2254 * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2255 * so that won't be a limiting factor.
2256 *
2257 * However there is other limiting factor. We do store extents in the form
2258 * of starting block and length, hence the resulting length of the extent
2259 * covering maximum file size must fit into on-disk format containers as
2260 * well. Given that length is always by 1 unit bigger than max unit (because
2261 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2262 *
2263 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2264 */
2265static loff_t ext4_max_size(int blkbits, int has_huge_files)
2266{
2267	loff_t res;
2268	loff_t upper_limit = MAX_LFS_FILESIZE;
2269
2270	/* small i_blocks in vfs inode? */
2271	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2272		/*
2273		 * CONFIG_LBDAF is not enabled implies the inode
2274		 * i_block represent total blocks in 512 bytes
2275		 * 32 == size of vfs inode i_blocks * 8
2276		 */
2277		upper_limit = (1LL << 32) - 1;
2278
2279		/* total blocks in file system block size */
2280		upper_limit >>= (blkbits - 9);
2281		upper_limit <<= blkbits;
2282	}
2283
2284	/*
2285	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2286	 * by one fs block, so ee_len can cover the extent of maximum file
2287	 * size
2288	 */
2289	res = (1LL << 32) - 1;
2290	res <<= blkbits;
2291
2292	/* Sanity check against vm- & vfs- imposed limits */
2293	if (res > upper_limit)
2294		res = upper_limit;
2295
2296	return res;
2297}
2298
2299/*
2300 * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2301 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2302 * We need to be 1 filesystem block less than the 2^48 sector limit.
2303 */
2304static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2305{
2306	loff_t res = EXT4_NDIR_BLOCKS;
2307	int meta_blocks;
2308	loff_t upper_limit;
2309	/* This is calculated to be the largest file size for a dense, block
2310	 * mapped file such that the file's total number of 512-byte sectors,
2311	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2312	 *
2313	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2314	 * number of 512-byte sectors of the file.
2315	 */
2316
2317	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2318		/*
2319		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2320		 * the inode i_block field represents total file blocks in
2321		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2322		 */
2323		upper_limit = (1LL << 32) - 1;
2324
2325		/* total blocks in file system block size */
2326		upper_limit >>= (bits - 9);
2327
2328	} else {
2329		/*
2330		 * We use 48 bit ext4_inode i_blocks
2331		 * With EXT4_HUGE_FILE_FL set the i_blocks
2332		 * represent total number of blocks in
2333		 * file system block size
2334		 */
2335		upper_limit = (1LL << 48) - 1;
2336
2337	}
2338
2339	/* indirect blocks */
2340	meta_blocks = 1;
2341	/* double indirect blocks */
2342	meta_blocks += 1 + (1LL << (bits-2));
2343	/* tripple indirect blocks */
2344	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2345
2346	upper_limit -= meta_blocks;
2347	upper_limit <<= bits;
2348
2349	res += 1LL << (bits-2);
2350	res += 1LL << (2*(bits-2));
2351	res += 1LL << (3*(bits-2));
2352	res <<= bits;
2353	if (res > upper_limit)
2354		res = upper_limit;
2355
2356	if (res > MAX_LFS_FILESIZE)
2357		res = MAX_LFS_FILESIZE;
2358
2359	return res;
2360}
2361
2362static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2363				   ext4_fsblk_t logical_sb_block, int nr)
2364{
2365	struct ext4_sb_info *sbi = EXT4_SB(sb);
2366	ext4_group_t bg, first_meta_bg;
2367	int has_super = 0;
2368
2369	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2370
2371	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2372	    nr < first_meta_bg)
2373		return logical_sb_block + nr + 1;
2374	bg = sbi->s_desc_per_block * nr;
2375	if (ext4_bg_has_super(sb, bg))
2376		has_super = 1;
2377
2378	/*
2379	 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2380	 * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2381	 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2382	 * compensate.
2383	 */
2384	if (sb->s_blocksize == 1024 && nr == 0 &&
2385	    le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2386		has_super++;
2387
2388	return (has_super + ext4_group_first_block_no(sb, bg));
2389}
2390
2391/**
2392 * ext4_get_stripe_size: Get the stripe size.
2393 * @sbi: In memory super block info
2394 *
2395 * If we have specified it via mount option, then
2396 * use the mount option value. If the value specified at mount time is
2397 * greater than the blocks per group use the super block value.
2398 * If the super block value is greater than blocks per group return 0.
2399 * Allocator needs it be less than blocks per group.
2400 *
2401 */
2402static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2403{
2404	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2405	unsigned long stripe_width =
2406			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2407	int ret;
2408
2409	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2410		ret = sbi->s_stripe;
2411	else if (stripe_width <= sbi->s_blocks_per_group)
2412		ret = stripe_width;
2413	else if (stride <= sbi->s_blocks_per_group)
2414		ret = stride;
2415	else
2416		ret = 0;
2417
2418	/*
2419	 * If the stripe width is 1, this makes no sense and
2420	 * we set it to 0 to turn off stripe handling code.
2421	 */
2422	if (ret <= 1)
2423		ret = 0;
2424
2425	return ret;
2426}
2427
2428/* sysfs supprt */
2429
2430struct ext4_attr {
2431	struct attribute attr;
2432	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2433	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2434			 const char *, size_t);
2435	union {
2436		int offset;
2437		int deprecated_val;
2438	} u;
2439};
2440
2441static int parse_strtoull(const char *buf,
2442		unsigned long long max, unsigned long long *value)
2443{
2444	int ret;
2445
2446	ret = kstrtoull(skip_spaces(buf), 0, value);
2447	if (!ret && *value > max)
2448		ret = -EINVAL;
2449	return ret;
2450}
2451
2452static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2453					      struct ext4_sb_info *sbi,
2454					      char *buf)
2455{
2456	return snprintf(buf, PAGE_SIZE, "%llu\n",
2457		(s64) EXT4_C2B(sbi,
2458			percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2459}
2460
2461static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2462					 struct ext4_sb_info *sbi, char *buf)
2463{
2464	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2465
2466	if (!sb->s_bdev->bd_part)
2467		return snprintf(buf, PAGE_SIZE, "0\n");
2468	return snprintf(buf, PAGE_SIZE, "%lu\n",
2469			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2470			 sbi->s_sectors_written_start) >> 1);
2471}
2472
2473static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2474					  struct ext4_sb_info *sbi, char *buf)
2475{
2476	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2477
2478	if (!sb->s_bdev->bd_part)
2479		return snprintf(buf, PAGE_SIZE, "0\n");
2480	return snprintf(buf, PAGE_SIZE, "%llu\n",
2481			(unsigned long long)(sbi->s_kbytes_written +
2482			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2483			  EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2484}
2485
2486static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2487					  struct ext4_sb_info *sbi,
2488					  const char *buf, size_t count)
2489{
2490	unsigned long t;
2491	int ret;
2492
2493	ret = kstrtoul(skip_spaces(buf), 0, &t);
2494	if (ret)
2495		return ret;
2496
2497	if (t && (!is_power_of_2(t) || t > 0x40000000))
2498		return -EINVAL;
2499
2500	sbi->s_inode_readahead_blks = t;
2501	return count;
2502}
2503
2504static ssize_t sbi_ui_show(struct ext4_attr *a,
2505			   struct ext4_sb_info *sbi, char *buf)
2506{
2507	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2508
2509	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2510}
2511
2512static ssize_t sbi_ui_store(struct ext4_attr *a,
2513			    struct ext4_sb_info *sbi,
2514			    const char *buf, size_t count)
2515{
2516	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2517	unsigned long t;
2518	int ret;
2519
2520	ret = kstrtoul(skip_spaces(buf), 0, &t);
2521	if (ret)
2522		return ret;
2523	*ui = t;
2524	return count;
2525}
2526
2527static ssize_t es_ui_show(struct ext4_attr *a,
2528			   struct ext4_sb_info *sbi, char *buf)
2529{
2530
2531	unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2532			   a->u.offset);
2533
2534	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2535}
2536
2537static ssize_t reserved_clusters_show(struct ext4_attr *a,
2538				  struct ext4_sb_info *sbi, char *buf)
2539{
2540	return snprintf(buf, PAGE_SIZE, "%llu\n",
2541		(unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2542}
2543
2544static ssize_t reserved_clusters_store(struct ext4_attr *a,
2545				   struct ext4_sb_info *sbi,
2546				   const char *buf, size_t count)
2547{
2548	unsigned long long val;
2549	int ret;
2550
2551	if (parse_strtoull(buf, -1ULL, &val))
2552		return -EINVAL;
2553	ret = ext4_reserve_clusters(sbi, val);
2554
2555	return ret ? ret : count;
2556}
2557
2558static ssize_t trigger_test_error(struct ext4_attr *a,
2559				  struct ext4_sb_info *sbi,
2560				  const char *buf, size_t count)
2561{
2562	int len = count;
2563
2564	if (!capable(CAP_SYS_ADMIN))
2565		return -EPERM;
2566
2567	if (len && buf[len-1] == '\n')
2568		len--;
2569
2570	if (len)
2571		ext4_error(sbi->s_sb, "%.*s", len, buf);
2572	return count;
2573}
2574
2575static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2576				   struct ext4_sb_info *sbi, char *buf)
2577{
2578	return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2579}
2580
2581#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2582static struct ext4_attr ext4_attr_##_name = {			\
2583	.attr = {.name = __stringify(_name), .mode = _mode },	\
2584	.show	= _show,					\
2585	.store	= _store,					\
2586	.u = {							\
2587		.offset = offsetof(struct ext4_sb_info, _elname),\
2588	},							\
2589}
2590
2591#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname)		\
2592static struct ext4_attr ext4_attr_##_name = {				\
2593	.attr = {.name = __stringify(_name), .mode = _mode },		\
2594	.show	= _show,						\
2595	.store	= _store,						\
2596	.u = {								\
2597		.offset = offsetof(struct ext4_super_block, _elname),	\
2598	},								\
2599}
2600
2601#define EXT4_ATTR(name, mode, show, store) \
2602static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2603
2604#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2605#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2606#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2607
2608#define EXT4_RO_ATTR_ES_UI(name, elname)	\
2609	EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2610#define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2611	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2612
2613#define ATTR_LIST(name) &ext4_attr_##name.attr
2614#define EXT4_DEPRECATED_ATTR(_name, _val)	\
2615static struct ext4_attr ext4_attr_##_name = {			\
2616	.attr = {.name = __stringify(_name), .mode = 0444 },	\
2617	.show	= sbi_deprecated_show,				\
2618	.u = {							\
2619		.deprecated_val = _val,				\
2620	},							\
2621}
2622
2623EXT4_RO_ATTR(delayed_allocation_blocks);
2624EXT4_RO_ATTR(session_write_kbytes);
2625EXT4_RO_ATTR(lifetime_write_kbytes);
2626EXT4_RW_ATTR(reserved_clusters);
2627EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2628		 inode_readahead_blks_store, s_inode_readahead_blks);
2629EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2630EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2631EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2632EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2633EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2634EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2635EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2636EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2637EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2638EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2639EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2640EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2641EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2642EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2643EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2644EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2645EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2646EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2647EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2648
2649static struct attribute *ext4_attrs[] = {
2650	ATTR_LIST(delayed_allocation_blocks),
2651	ATTR_LIST(session_write_kbytes),
2652	ATTR_LIST(lifetime_write_kbytes),
2653	ATTR_LIST(reserved_clusters),
2654	ATTR_LIST(inode_readahead_blks),
2655	ATTR_LIST(inode_goal),
2656	ATTR_LIST(mb_stats),
2657	ATTR_LIST(mb_max_to_scan),
2658	ATTR_LIST(mb_min_to_scan),
2659	ATTR_LIST(mb_order2_req),
2660	ATTR_LIST(mb_stream_req),
2661	ATTR_LIST(mb_group_prealloc),
2662	ATTR_LIST(max_writeback_mb_bump),
2663	ATTR_LIST(extent_max_zeroout_kb),
2664	ATTR_LIST(trigger_fs_error),
2665	ATTR_LIST(err_ratelimit_interval_ms),
2666	ATTR_LIST(err_ratelimit_burst),
2667	ATTR_LIST(warning_ratelimit_interval_ms),
2668	ATTR_LIST(warning_ratelimit_burst),
2669	ATTR_LIST(msg_ratelimit_interval_ms),
2670	ATTR_LIST(msg_ratelimit_burst),
2671	ATTR_LIST(errors_count),
2672	ATTR_LIST(first_error_time),
2673	ATTR_LIST(last_error_time),
2674	NULL,
2675};
2676
2677/* Features this copy of ext4 supports */
2678EXT4_INFO_ATTR(lazy_itable_init);
2679EXT4_INFO_ATTR(batched_discard);
2680EXT4_INFO_ATTR(meta_bg_resize);
2681
2682static struct attribute *ext4_feat_attrs[] = {
2683	ATTR_LIST(lazy_itable_init),
2684	ATTR_LIST(batched_discard),
2685	ATTR_LIST(meta_bg_resize),
2686	NULL,
2687};
2688
2689static ssize_t ext4_attr_show(struct kobject *kobj,
2690			      struct attribute *attr, char *buf)
2691{
2692	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2693						s_kobj);
2694	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2695
2696	return a->show ? a->show(a, sbi, buf) : 0;
2697}
2698
2699static ssize_t ext4_attr_store(struct kobject *kobj,
2700			       struct attribute *attr,
2701			       const char *buf, size_t len)
2702{
2703	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2704						s_kobj);
2705	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2706
2707	return a->store ? a->store(a, sbi, buf, len) : 0;
2708}
2709
2710static void ext4_sb_release(struct kobject *kobj)
2711{
2712	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2713						s_kobj);
2714	complete(&sbi->s_kobj_unregister);
2715}
2716
2717static const struct sysfs_ops ext4_attr_ops = {
2718	.show	= ext4_attr_show,
2719	.store	= ext4_attr_store,
2720};
2721
2722static struct kobj_type ext4_ktype = {
2723	.default_attrs	= ext4_attrs,
2724	.sysfs_ops	= &ext4_attr_ops,
2725	.release	= ext4_sb_release,
2726};
2727
2728static void ext4_feat_release(struct kobject *kobj)
2729{
2730	complete(&ext4_feat->f_kobj_unregister);
2731}
2732
2733static ssize_t ext4_feat_show(struct kobject *kobj,
2734			      struct attribute *attr, char *buf)
2735{
2736	return snprintf(buf, PAGE_SIZE, "supported\n");
2737}
2738
2739/*
2740 * We can not use ext4_attr_show/store because it relies on the kobject
2741 * being embedded in the ext4_sb_info structure which is definitely not
2742 * true in this case.
2743 */
2744static const struct sysfs_ops ext4_feat_ops = {
2745	.show	= ext4_feat_show,
2746	.store	= NULL,
2747};
2748
2749static struct kobj_type ext4_feat_ktype = {
2750	.default_attrs	= ext4_feat_attrs,
2751	.sysfs_ops	= &ext4_feat_ops,
2752	.release	= ext4_feat_release,
2753};
2754
2755/*
2756 * Check whether this filesystem can be mounted based on
2757 * the features present and the RDONLY/RDWR mount requested.
2758 * Returns 1 if this filesystem can be mounted as requested,
2759 * 0 if it cannot be.
2760 */
2761static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2762{
2763	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2764		ext4_msg(sb, KERN_ERR,
2765			"Couldn't mount because of "
2766			"unsupported optional features (%x)",
2767			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2768			~EXT4_FEATURE_INCOMPAT_SUPP));
2769		return 0;
2770	}
2771
2772	if (readonly)
2773		return 1;
2774
2775	/* Check that feature set is OK for a read-write mount */
2776	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2777		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2778			 "unsupported optional features (%x)",
2779			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2780				~EXT4_FEATURE_RO_COMPAT_SUPP));
2781		return 0;
2782	}
2783	/*
2784	 * Large file size enabled file system can only be mounted
2785	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2786	 */
2787	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2788		if (sizeof(blkcnt_t) < sizeof(u64)) {
2789			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2790				 "cannot be mounted RDWR without "
2791				 "CONFIG_LBDAF");
2792			return 0;
2793		}
2794	}
2795	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2796	    !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2797		ext4_msg(sb, KERN_ERR,
2798			 "Can't support bigalloc feature without "
2799			 "extents feature\n");
2800		return 0;
2801	}
2802
2803#ifndef CONFIG_QUOTA
2804	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2805	    !readonly) {
2806		ext4_msg(sb, KERN_ERR,
2807			 "Filesystem with quota feature cannot be mounted RDWR "
2808			 "without CONFIG_QUOTA");
2809		return 0;
2810	}
2811#endif  /* CONFIG_QUOTA */
2812	return 1;
2813}
2814
2815/*
2816 * This function is called once a day if we have errors logged
2817 * on the file system
2818 */
2819static void print_daily_error_info(unsigned long arg)
2820{
2821	struct super_block *sb = (struct super_block *) arg;
2822	struct ext4_sb_info *sbi;
2823	struct ext4_super_block *es;
2824
2825	sbi = EXT4_SB(sb);
2826	es = sbi->s_es;
2827
2828	if (es->s_error_count)
2829		/* fsck newer than v1.41.13 is needed to clean this condition. */
2830		ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2831			 le32_to_cpu(es->s_error_count));
2832	if (es->s_first_error_time) {
2833		printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2834		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2835		       (int) sizeof(es->s_first_error_func),
2836		       es->s_first_error_func,
2837		       le32_to_cpu(es->s_first_error_line));
2838		if (es->s_first_error_ino)
2839			printk(": inode %u",
2840			       le32_to_cpu(es->s_first_error_ino));
2841		if (es->s_first_error_block)
2842			printk(": block %llu", (unsigned long long)
2843			       le64_to_cpu(es->s_first_error_block));
2844		printk("\n");
2845	}
2846	if (es->s_last_error_time) {
2847		printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2848		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2849		       (int) sizeof(es->s_last_error_func),
2850		       es->s_last_error_func,
2851		       le32_to_cpu(es->s_last_error_line));
2852		if (es->s_last_error_ino)
2853			printk(": inode %u",
2854			       le32_to_cpu(es->s_last_error_ino));
2855		if (es->s_last_error_block)
2856			printk(": block %llu", (unsigned long long)
2857			       le64_to_cpu(es->s_last_error_block));
2858		printk("\n");
2859	}
2860	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2861}
2862
2863/* Find next suitable group and run ext4_init_inode_table */
2864static int ext4_run_li_request(struct ext4_li_request *elr)
2865{
2866	struct ext4_group_desc *gdp = NULL;
2867	ext4_group_t group, ngroups;
2868	struct super_block *sb;
2869	unsigned long timeout = 0;
2870	int ret = 0;
2871
2872	sb = elr->lr_super;
2873	ngroups = EXT4_SB(sb)->s_groups_count;
2874
2875	sb_start_write(sb);
2876	for (group = elr->lr_next_group; group < ngroups; group++) {
2877		gdp = ext4_get_group_desc(sb, group, NULL);
2878		if (!gdp) {
2879			ret = 1;
2880			break;
2881		}
2882
2883		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2884			break;
2885	}
2886
2887	if (group >= ngroups)
2888		ret = 1;
2889
2890	if (!ret) {
2891		timeout = jiffies;
2892		ret = ext4_init_inode_table(sb, group,
2893					    elr->lr_timeout ? 0 : 1);
2894		if (elr->lr_timeout == 0) {
2895			timeout = (jiffies - timeout) *
2896				  elr->lr_sbi->s_li_wait_mult;
2897			elr->lr_timeout = timeout;
2898		}
2899		elr->lr_next_sched = jiffies + elr->lr_timeout;
2900		elr->lr_next_group = group + 1;
2901	}
2902	sb_end_write(sb);
2903
2904	return ret;
2905}
2906
2907/*
2908 * Remove lr_request from the list_request and free the
2909 * request structure. Should be called with li_list_mtx held
2910 */
2911static void ext4_remove_li_request(struct ext4_li_request *elr)
2912{
2913	struct ext4_sb_info *sbi;
2914
2915	if (!elr)
2916		return;
2917
2918	sbi = elr->lr_sbi;
2919
2920	list_del(&elr->lr_request);
2921	sbi->s_li_request = NULL;
2922	kfree(elr);
2923}
2924
2925static void ext4_unregister_li_request(struct super_block *sb)
2926{
2927	mutex_lock(&ext4_li_mtx);
2928	if (!ext4_li_info) {
2929		mutex_unlock(&ext4_li_mtx);
2930		return;
2931	}
2932
2933	mutex_lock(&ext4_li_info->li_list_mtx);
2934	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2935	mutex_unlock(&ext4_li_info->li_list_mtx);
2936	mutex_unlock(&ext4_li_mtx);
2937}
2938
2939static struct task_struct *ext4_lazyinit_task;
2940
2941/*
2942 * This is the function where ext4lazyinit thread lives. It walks
2943 * through the request list searching for next scheduled filesystem.
2944 * When such a fs is found, run the lazy initialization request
2945 * (ext4_rn_li_request) and keep track of the time spend in this
2946 * function. Based on that time we compute next schedule time of
2947 * the request. When walking through the list is complete, compute
2948 * next waking time and put itself into sleep.
2949 */
2950static int ext4_lazyinit_thread(void *arg)
2951{
2952	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2953	struct list_head *pos, *n;
2954	struct ext4_li_request *elr;
2955	unsigned long next_wakeup, cur;
2956
2957	BUG_ON(NULL == eli);
2958
2959cont_thread:
2960	while (true) {
2961		next_wakeup = MAX_JIFFY_OFFSET;
2962
2963		mutex_lock(&eli->li_list_mtx);
2964		if (list_empty(&eli->li_request_list)) {
2965			mutex_unlock(&eli->li_list_mtx);
2966			goto exit_thread;
2967		}
2968
2969		list_for_each_safe(pos, n, &eli->li_request_list) {
2970			elr = list_entry(pos, struct ext4_li_request,
2971					 lr_request);
2972
2973			if (time_after_eq(jiffies, elr->lr_next_sched)) {
2974				if (ext4_run_li_request(elr) != 0) {
2975					/* error, remove the lazy_init job */
2976					ext4_remove_li_request(elr);
2977					continue;
2978				}
2979			}
2980
2981			if (time_before(elr->lr_next_sched, next_wakeup))
2982				next_wakeup = elr->lr_next_sched;
2983		}
2984		mutex_unlock(&eli->li_list_mtx);
2985
2986		try_to_freeze();
2987
2988		cur = jiffies;
2989		if ((time_after_eq(cur, next_wakeup)) ||
2990		    (MAX_JIFFY_OFFSET == next_wakeup)) {
2991			cond_resched();
2992			continue;
2993		}
2994
2995		schedule_timeout_interruptible(next_wakeup - cur);
2996
2997		if (kthread_should_stop()) {
2998			ext4_clear_request_list();
2999			goto exit_thread;
3000		}
3001	}
3002
3003exit_thread:
3004	/*
3005	 * It looks like the request list is empty, but we need
3006	 * to check it under the li_list_mtx lock, to prevent any
3007	 * additions into it, and of course we should lock ext4_li_mtx
3008	 * to atomically free the list and ext4_li_info, because at
3009	 * this point another ext4 filesystem could be registering
3010	 * new one.
3011	 */
3012	mutex_lock(&ext4_li_mtx);
3013	mutex_lock(&eli->li_list_mtx);
3014	if (!list_empty(&eli->li_request_list)) {
3015		mutex_unlock(&eli->li_list_mtx);
3016		mutex_unlock(&ext4_li_mtx);
3017		goto cont_thread;
3018	}
3019	mutex_unlock(&eli->li_list_mtx);
3020	kfree(ext4_li_info);
3021	ext4_li_info = NULL;
3022	mutex_unlock(&ext4_li_mtx);
3023
3024	return 0;
3025}
3026
3027static void ext4_clear_request_list(void)
3028{
3029	struct list_head *pos, *n;
3030	struct ext4_li_request *elr;
3031
3032	mutex_lock(&ext4_li_info->li_list_mtx);
3033	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3034		elr = list_entry(pos, struct ext4_li_request,
3035				 lr_request);
3036		ext4_remove_li_request(elr);
3037	}
3038	mutex_unlock(&ext4_li_info->li_list_mtx);
3039}
3040
3041static int ext4_run_lazyinit_thread(void)
3042{
3043	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3044					 ext4_li_info, "ext4lazyinit");
3045	if (IS_ERR(ext4_lazyinit_task)) {
3046		int err = PTR_ERR(ext4_lazyinit_task);
3047		ext4_clear_request_list();
3048		kfree(ext4_li_info);
3049		ext4_li_info = NULL;
3050		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3051				 "initialization thread\n",
3052				 err);
3053		return err;
3054	}
3055	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3056	return 0;
3057}
3058
3059/*
3060 * Check whether it make sense to run itable init. thread or not.
3061 * If there is at least one uninitialized inode table, return
3062 * corresponding group number, else the loop goes through all
3063 * groups and return total number of groups.
3064 */
3065static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3066{
3067	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3068	struct ext4_group_desc *gdp = NULL;
3069
3070	for (group = 0; group < ngroups; group++) {
3071		gdp = ext4_get_group_desc(sb, group, NULL);
3072		if (!gdp)
3073			continue;
3074
3075		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3076			break;
3077	}
3078
3079	return group;
3080}
3081
3082static int ext4_li_info_new(void)
3083{
3084	struct ext4_lazy_init *eli = NULL;
3085
3086	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3087	if (!eli)
3088		return -ENOMEM;
3089
3090	INIT_LIST_HEAD(&eli->li_request_list);
3091	mutex_init(&eli->li_list_mtx);
3092
3093	eli->li_state |= EXT4_LAZYINIT_QUIT;
3094
3095	ext4_li_info = eli;
3096
3097	return 0;
3098}
3099
3100static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3101					    ext4_group_t start)
3102{
3103	struct ext4_sb_info *sbi = EXT4_SB(sb);
3104	struct ext4_li_request *elr;
3105
3106	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3107	if (!elr)
3108		return NULL;
3109
3110	elr->lr_super = sb;
3111	elr->lr_sbi = sbi;
3112	elr->lr_next_group = start;
3113
3114	/*
3115	 * Randomize first schedule time of the request to
3116	 * spread the inode table initialization requests
3117	 * better.
3118	 */
3119	elr->lr_next_sched = jiffies + (prandom_u32() %
3120				(EXT4_DEF_LI_MAX_START_DELAY * HZ));
3121	return elr;
3122}
3123
3124int ext4_register_li_request(struct super_block *sb,
3125			     ext4_group_t first_not_zeroed)
3126{
3127	struct ext4_sb_info *sbi = EXT4_SB(sb);
3128	struct ext4_li_request *elr = NULL;
3129	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3130	int ret = 0;
3131
3132	mutex_lock(&ext4_li_mtx);
3133	if (sbi->s_li_request != NULL) {
3134		/*
3135		 * Reset timeout so it can be computed again, because
3136		 * s_li_wait_mult might have changed.
3137		 */
3138		sbi->s_li_request->lr_timeout = 0;
3139		goto out;
3140	}
3141
3142	if (first_not_zeroed == ngroups ||
3143	    (sb->s_flags & MS_RDONLY) ||
3144	    !test_opt(sb, INIT_INODE_TABLE))
3145		goto out;
3146
3147	elr = ext4_li_request_new(sb, first_not_zeroed);
3148	if (!elr) {
3149		ret = -ENOMEM;
3150		goto out;
3151	}
3152
3153	if (NULL == ext4_li_info) {
3154		ret = ext4_li_info_new();
3155		if (ret)
3156			goto out;
3157	}
3158
3159	mutex_lock(&ext4_li_info->li_list_mtx);
3160	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3161	mutex_unlock(&ext4_li_info->li_list_mtx);
3162
3163	sbi->s_li_request = elr;
3164	/*
3165	 * set elr to NULL here since it has been inserted to
3166	 * the request_list and the removal and free of it is
3167	 * handled by ext4_clear_request_list from now on.
3168	 */
3169	elr = NULL;
3170
3171	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3172		ret = ext4_run_lazyinit_thread();
3173		if (ret)
3174			goto out;
3175	}
3176out:
3177	mutex_unlock(&ext4_li_mtx);
3178	if (ret)
3179		kfree(elr);
3180	return ret;
3181}
3182
3183/*
3184 * We do not need to lock anything since this is called on
3185 * module unload.
3186 */
3187static void ext4_destroy_lazyinit_thread(void)
3188{
3189	/*
3190	 * If thread exited earlier
3191	 * there's nothing to be done.
3192	 */
3193	if (!ext4_li_info || !ext4_lazyinit_task)
3194		return;
3195
3196	kthread_stop(ext4_lazyinit_task);
3197}
3198
3199static int set_journal_csum_feature_set(struct super_block *sb)
3200{
3201	int ret = 1;
3202	int compat, incompat;
3203	struct ext4_sb_info *sbi = EXT4_SB(sb);
3204
3205	if (ext4_has_metadata_csum(sb)) {
3206		/* journal checksum v3 */
3207		compat = 0;
3208		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3209	} else {
3210		/* journal checksum v1 */
3211		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3212		incompat = 0;
3213	}
3214
3215	jbd2_journal_clear_features(sbi->s_journal,
3216			JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3217			JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3218			JBD2_FEATURE_INCOMPAT_CSUM_V2);
3219	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3220		ret = jbd2_journal_set_features(sbi->s_journal,
3221				compat, 0,
3222				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3223				incompat);
3224	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3225		ret = jbd2_journal_set_features(sbi->s_journal,
3226				compat, 0,
3227				incompat);
3228		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3229				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3230	} else {
3231		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3232				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3233	}
3234
3235	return ret;
3236}
3237
3238/*
3239 * Note: calculating the overhead so we can be compatible with
3240 * historical BSD practice is quite difficult in the face of
3241 * clusters/bigalloc.  This is because multiple metadata blocks from
3242 * different block group can end up in the same allocation cluster.
3243 * Calculating the exact overhead in the face of clustered allocation
3244 * requires either O(all block bitmaps) in memory or O(number of block
3245 * groups**2) in time.  We will still calculate the superblock for
3246 * older file systems --- and if we come across with a bigalloc file
3247 * system with zero in s_overhead_clusters the estimate will be close to
3248 * correct especially for very large cluster sizes --- but for newer
3249 * file systems, it's better to calculate this figure once at mkfs
3250 * time, and store it in the superblock.  If the superblock value is
3251 * present (even for non-bigalloc file systems), we will use it.
3252 */
3253static int count_overhead(struct super_block *sb, ext4_group_t grp,
3254			  char *buf)
3255{
3256	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3257	struct ext4_group_desc	*gdp;
3258	ext4_fsblk_t		first_block, last_block, b;
3259	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3260	int			s, j, count = 0;
3261
3262	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3263		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3264			sbi->s_itb_per_group + 2);
3265
3266	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3267		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3268	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3269	for (i = 0; i < ngroups; i++) {
3270		gdp = ext4_get_group_desc(sb, i, NULL);
3271		b = ext4_block_bitmap(sb, gdp);
3272		if (b >= first_block && b <= last_block) {
3273			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3274			count++;
3275		}
3276		b = ext4_inode_bitmap(sb, gdp);
3277		if (b >= first_block && b <= last_block) {
3278			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3279			count++;
3280		}
3281		b = ext4_inode_table(sb, gdp);
3282		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3283			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3284				int c = EXT4_B2C(sbi, b - first_block);
3285				ext4_set_bit(c, buf);
3286				count++;
3287			}
3288		if (i != grp)
3289			continue;
3290		s = 0;
3291		if (ext4_bg_has_super(sb, grp)) {
3292			ext4_set_bit(s++, buf);
3293			count++;
3294		}
3295		for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3296			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3297			count++;
3298		}
3299	}
3300	if (!count)
3301		return 0;
3302	return EXT4_CLUSTERS_PER_GROUP(sb) -
3303		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3304}
3305
3306/*
3307 * Compute the overhead and stash it in sbi->s_overhead
3308 */
3309int ext4_calculate_overhead(struct super_block *sb)
3310{
3311	struct ext4_sb_info *sbi = EXT4_SB(sb);
3312	struct ext4_super_block *es = sbi->s_es;
3313	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3314	ext4_fsblk_t overhead = 0;
3315	char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3316
3317	if (!buf)
3318		return -ENOMEM;
3319
3320	/*
3321	 * Compute the overhead (FS structures).  This is constant
3322	 * for a given filesystem unless the number of block groups
3323	 * changes so we cache the previous value until it does.
3324	 */
3325
3326	/*
3327	 * All of the blocks before first_data_block are overhead
3328	 */
3329	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3330
3331	/*
3332	 * Add the overhead found in each block group
3333	 */
3334	for (i = 0; i < ngroups; i++) {
3335		int blks;
3336
3337		blks = count_overhead(sb, i, buf);
3338		overhead += blks;
3339		if (blks)
3340			memset(buf, 0, PAGE_SIZE);
3341		cond_resched();
3342	}
3343	/* Add the journal blocks as well */
3344	if (sbi->s_journal)
3345		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3346
3347	sbi->s_overhead = overhead;
3348	smp_wmb();
3349	free_page((unsigned long) buf);
3350	return 0;
3351}
3352
3353
3354static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3355{
3356	ext4_fsblk_t resv_clusters;
3357
3358	/*
3359	 * There's no need to reserve anything when we aren't using extents.
3360	 * The space estimates are exact, there are no unwritten extents,
3361	 * hole punching doesn't need new metadata... This is needed especially
3362	 * to keep ext2/3 backward compatibility.
3363	 */
3364	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3365		return 0;
3366	/*
3367	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3368	 * This should cover the situations where we can not afford to run
3369	 * out of space like for example punch hole, or converting
3370	 * unwritten extents in delalloc path. In most cases such
3371	 * allocation would require 1, or 2 blocks, higher numbers are
3372	 * very rare.
3373	 */
3374	resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3375			EXT4_SB(sb)->s_cluster_bits;
3376
3377	do_div(resv_clusters, 50);
3378	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3379
3380	return resv_clusters;
3381}
3382
3383
3384static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3385{
3386	ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3387				sbi->s_cluster_bits;
3388
3389	if (count >= clusters)
3390		return -EINVAL;
3391
3392	atomic64_set(&sbi->s_resv_clusters, count);
3393	return 0;
3394}
3395
3396static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3397{
3398	char *orig_data = kstrdup(data, GFP_KERNEL);
3399	struct buffer_head *bh;
3400	struct ext4_super_block *es = NULL;
3401	struct ext4_sb_info *sbi;
3402	ext4_fsblk_t block;
3403	ext4_fsblk_t sb_block = get_sb_block(&data);
3404	ext4_fsblk_t logical_sb_block;
3405	unsigned long offset = 0;
3406	unsigned long journal_devnum = 0;
3407	unsigned long def_mount_opts;
3408	struct inode *root;
3409	char *cp;
3410	const char *descr;
3411	int ret = -ENOMEM;
3412	int blocksize, clustersize;
3413	unsigned int db_count;
3414	unsigned int i;
3415	int needs_recovery, has_huge_files, has_bigalloc;
3416	__u64 blocks_count;
3417	int err = 0;
3418	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3419	ext4_group_t first_not_zeroed;
3420
3421	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3422	if (!sbi)
3423		goto out_free_orig;
3424
3425	sbi->s_blockgroup_lock =
3426		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3427	if (!sbi->s_blockgroup_lock) {
3428		kfree(sbi);
3429		goto out_free_orig;
3430	}
3431	sb->s_fs_info = sbi;
3432	sbi->s_sb = sb;
3433	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3434	sbi->s_sb_block = sb_block;
3435	if (sb->s_bdev->bd_part)
3436		sbi->s_sectors_written_start =
3437			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3438
3439	/* Cleanup superblock name */
3440	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3441		*cp = '!';
3442
3443	/* -EINVAL is default */
3444	ret = -EINVAL;
3445	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3446	if (!blocksize) {
3447		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3448		goto out_fail;
3449	}
3450
3451	/*
3452	 * The ext4 superblock will not be buffer aligned for other than 1kB
3453	 * block sizes.  We need to calculate the offset from buffer start.
3454	 */
3455	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3456		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3457		offset = do_div(logical_sb_block, blocksize);
3458	} else {
3459		logical_sb_block = sb_block;
3460	}
3461
3462	if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3463		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3464		goto out_fail;
3465	}
3466	/*
3467	 * Note: s_es must be initialized as soon as possible because
3468	 *       some ext4 macro-instructions depend on its value
3469	 */
3470	es = (struct ext4_super_block *) (bh->b_data + offset);
3471	sbi->s_es = es;
3472	sb->s_magic = le16_to_cpu(es->s_magic);
3473	if (sb->s_magic != EXT4_SUPER_MAGIC)
3474		goto cantfind_ext4;
3475	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3476
3477	/* Warn if metadata_csum and gdt_csum are both set. */
3478	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3479				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3480	    EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3481		ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3482			     "redundant flags; please run fsck.");
3483
3484	/* Check for a known checksum algorithm */
3485	if (!ext4_verify_csum_type(sb, es)) {
3486		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3487			 "unknown checksum algorithm.");
3488		silent = 1;
3489		goto cantfind_ext4;
3490	}
3491
3492	/* Load the checksum driver */
3493	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3494				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3495		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3496		if (IS_ERR(sbi->s_chksum_driver)) {
3497			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3498			ret = PTR_ERR(sbi->s_chksum_driver);
3499			sbi->s_chksum_driver = NULL;
3500			goto failed_mount;
3501		}
3502	}
3503
3504	/* Check superblock checksum */
3505	if (!ext4_superblock_csum_verify(sb, es)) {
3506		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3507			 "invalid superblock checksum.  Run e2fsck?");
3508		silent = 1;
3509		goto cantfind_ext4;
3510	}
3511
3512	/* Precompute checksum seed for all metadata */
3513	if (ext4_has_metadata_csum(sb))
3514		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3515					       sizeof(es->s_uuid));
3516
3517	/* Set defaults before we parse the mount options */
3518	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3519	set_opt(sb, INIT_INODE_TABLE);
3520	if (def_mount_opts & EXT4_DEFM_DEBUG)
3521		set_opt(sb, DEBUG);
3522	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3523		set_opt(sb, GRPID);
3524	if (def_mount_opts & EXT4_DEFM_UID16)
3525		set_opt(sb, NO_UID32);
3526	/* xattr user namespace & acls are now defaulted on */
3527	set_opt(sb, XATTR_USER);
3528#ifdef CONFIG_EXT4_FS_POSIX_ACL
3529	set_opt(sb, POSIX_ACL);
3530#endif
3531	/* don't forget to enable journal_csum when metadata_csum is enabled. */
3532	if (ext4_has_metadata_csum(sb))
3533		set_opt(sb, JOURNAL_CHECKSUM);
3534
3535	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3536		set_opt(sb, JOURNAL_DATA);
3537	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3538		set_opt(sb, ORDERED_DATA);
3539	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3540		set_opt(sb, WRITEBACK_DATA);
3541
3542	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3543		set_opt(sb, ERRORS_PANIC);
3544	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3545		set_opt(sb, ERRORS_CONT);
3546	else
3547		set_opt(sb, ERRORS_RO);
3548	/* block_validity enabled by default; disable with noblock_validity */
3549	set_opt(sb, BLOCK_VALIDITY);
3550	if (def_mount_opts & EXT4_DEFM_DISCARD)
3551		set_opt(sb, DISCARD);
3552
3553	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3554	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3555	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3556	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3557	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3558
3559	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3560		set_opt(sb, BARRIER);
3561
3562	/*
3563	 * enable delayed allocation by default
3564	 * Use -o nodelalloc to turn it off
3565	 */
3566	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3567	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3568		set_opt(sb, DELALLOC);
3569
3570	/*
3571	 * set default s_li_wait_mult for lazyinit, for the case there is
3572	 * no mount option specified.
3573	 */
3574	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3575
3576	if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3577			   &journal_devnum, &journal_ioprio, 0)) {
3578		ext4_msg(sb, KERN_WARNING,
3579			 "failed to parse options in superblock: %s",
3580			 sbi->s_es->s_mount_opts);
3581	}
3582	sbi->s_def_mount_opt = sbi->s_mount_opt;
3583	if (!parse_options((char *) data, sb, &journal_devnum,
3584			   &journal_ioprio, 0))
3585		goto failed_mount;
3586
3587	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3588		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3589			    "with data=journal disables delayed "
3590			    "allocation and O_DIRECT support!\n");
3591		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3592			ext4_msg(sb, KERN_ERR, "can't mount with "
3593				 "both data=journal and delalloc");
3594			goto failed_mount;
3595		}
3596		if (test_opt(sb, DIOREAD_NOLOCK)) {
3597			ext4_msg(sb, KERN_ERR, "can't mount with "
3598				 "both data=journal and dioread_nolock");
3599			goto failed_mount;
3600		}
3601		if (test_opt(sb, DELALLOC))
3602			clear_opt(sb, DELALLOC);
3603	}
3604
3605	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3606		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3607
3608	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3609	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3610	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3611	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3612		ext4_msg(sb, KERN_WARNING,
3613		       "feature flags set on rev 0 fs, "
3614		       "running e2fsck is recommended");
3615
3616	if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3617		set_opt2(sb, HURD_COMPAT);
3618		if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3619					      EXT4_FEATURE_INCOMPAT_64BIT)) {
3620			ext4_msg(sb, KERN_ERR,
3621				 "The Hurd can't support 64-bit file systems");
3622			goto failed_mount;
3623		}
3624	}
3625
3626	if (IS_EXT2_SB(sb)) {
3627		if (ext2_feature_set_ok(sb))
3628			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3629				 "using the ext4 subsystem");
3630		else {
3631			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3632				 "to feature incompatibilities");
3633			goto failed_mount;
3634		}
3635	}
3636
3637	if (IS_EXT3_SB(sb)) {
3638		if (ext3_feature_set_ok(sb))
3639			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3640				 "using the ext4 subsystem");
3641		else {
3642			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3643				 "to feature incompatibilities");
3644			goto failed_mount;
3645		}
3646	}
3647
3648	/*
3649	 * Check feature flags regardless of the revision level, since we
3650	 * previously didn't change the revision level when setting the flags,
3651	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3652	 */
3653	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3654		goto failed_mount;
3655
3656	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3657	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3658	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3659		ext4_msg(sb, KERN_ERR,
3660		       "Unsupported filesystem blocksize %d", blocksize);
3661		goto failed_mount;
3662	}
3663
3664	if (sb->s_blocksize != blocksize) {
3665		/* Validate the filesystem blocksize */
3666		if (!sb_set_blocksize(sb, blocksize)) {
3667			ext4_msg(sb, KERN_ERR, "bad block size %d",
3668					blocksize);
3669			goto failed_mount;
3670		}
3671
3672		brelse(bh);
3673		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3674		offset = do_div(logical_sb_block, blocksize);
3675		bh = sb_bread_unmovable(sb, logical_sb_block);
3676		if (!bh) {
3677			ext4_msg(sb, KERN_ERR,
3678			       "Can't read superblock on 2nd try");
3679			goto failed_mount;
3680		}
3681		es = (struct ext4_super_block *)(bh->b_data + offset);
3682		sbi->s_es = es;
3683		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3684			ext4_msg(sb, KERN_ERR,
3685			       "Magic mismatch, very weird!");
3686			goto failed_mount;
3687		}
3688	}
3689
3690	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3691				EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3692	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3693						      has_huge_files);
3694	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3695
3696	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3697		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3698		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3699	} else {
3700		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3701		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3702		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3703		    (!is_power_of_2(sbi->s_inode_size)) ||
3704		    (sbi->s_inode_size > blocksize)) {
3705			ext4_msg(sb, KERN_ERR,
3706			       "unsupported inode size: %d",
3707			       sbi->s_inode_size);
3708			goto failed_mount;
3709		}
3710		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3711			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3712	}
3713
3714	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3715	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3716		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3717		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3718		    !is_power_of_2(sbi->s_desc_size)) {
3719			ext4_msg(sb, KERN_ERR,
3720			       "unsupported descriptor size %lu",
3721			       sbi->s_desc_size);
3722			goto failed_mount;
3723		}
3724	} else
3725		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3726
3727	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3728	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3729	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3730		goto cantfind_ext4;
3731
3732	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3733	if (sbi->s_inodes_per_block == 0)
3734		goto cantfind_ext4;
3735	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3736					sbi->s_inodes_per_block;
3737	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3738	sbi->s_sbh = bh;
3739	sbi->s_mount_state = le16_to_cpu(es->s_state);
3740	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3741	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3742
3743	for (i = 0; i < 4; i++)
3744		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3745	sbi->s_def_hash_version = es->s_def_hash_version;
3746	if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3747		i = le32_to_cpu(es->s_flags);
3748		if (i & EXT2_FLAGS_UNSIGNED_HASH)
3749			sbi->s_hash_unsigned = 3;
3750		else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3751#ifdef __CHAR_UNSIGNED__
3752			if (!(sb->s_flags & MS_RDONLY))
3753				es->s_flags |=
3754					cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3755			sbi->s_hash_unsigned = 3;
3756#else
3757			if (!(sb->s_flags & MS_RDONLY))
3758				es->s_flags |=
3759					cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3760#endif
3761		}
3762	}
3763
3764	/* Handle clustersize */
3765	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3766	has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3767				EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3768	if (has_bigalloc) {
3769		if (clustersize < blocksize) {
3770			ext4_msg(sb, KERN_ERR,
3771				 "cluster size (%d) smaller than "
3772				 "block size (%d)", clustersize, blocksize);
3773			goto failed_mount;
3774		}
3775		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3776			le32_to_cpu(es->s_log_block_size);
3777		sbi->s_clusters_per_group =
3778			le32_to_cpu(es->s_clusters_per_group);
3779		if (sbi->s_clusters_per_group > blocksize * 8) {
3780			ext4_msg(sb, KERN_ERR,
3781				 "#clusters per group too big: %lu",
3782				 sbi->s_clusters_per_group);
3783			goto failed_mount;
3784		}
3785		if (sbi->s_blocks_per_group !=
3786		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3787			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3788				 "clusters per group (%lu) inconsistent",
3789				 sbi->s_blocks_per_group,
3790				 sbi->s_clusters_per_group);
3791			goto failed_mount;
3792		}
3793	} else {
3794		if (clustersize != blocksize) {
3795			ext4_warning(sb, "fragment/cluster size (%d) != "
3796				     "block size (%d)", clustersize,
3797				     blocksize);
3798			clustersize = blocksize;
3799		}
3800		if (sbi->s_blocks_per_group > blocksize * 8) {
3801			ext4_msg(sb, KERN_ERR,
3802				 "#blocks per group too big: %lu",
3803				 sbi->s_blocks_per_group);
3804			goto failed_mount;
3805		}
3806		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3807		sbi->s_cluster_bits = 0;
3808	}
3809	sbi->s_cluster_ratio = clustersize / blocksize;
3810
3811	if (sbi->s_inodes_per_group > blocksize * 8) {
3812		ext4_msg(sb, KERN_ERR,
3813		       "#inodes per group too big: %lu",
3814		       sbi->s_inodes_per_group);
3815		goto failed_mount;
3816	}
3817
3818	/* Do we have standard group size of clustersize * 8 blocks ? */
3819	if (sbi->s_blocks_per_group == clustersize << 3)
3820		set_opt2(sb, STD_GROUP_SIZE);
3821
3822	/*
3823	 * Test whether we have more sectors than will fit in sector_t,
3824	 * and whether the max offset is addressable by the page cache.
3825	 */
3826	err = generic_check_addressable(sb->s_blocksize_bits,
3827					ext4_blocks_count(es));
3828	if (err) {
3829		ext4_msg(sb, KERN_ERR, "filesystem"
3830			 " too large to mount safely on this system");
3831		if (sizeof(sector_t) < 8)
3832			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3833		goto failed_mount;
3834	}
3835
3836	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3837		goto cantfind_ext4;
3838
3839	/* check blocks count against device size */
3840	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3841	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3842		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3843		       "exceeds size of device (%llu blocks)",
3844		       ext4_blocks_count(es), blocks_count);
3845		goto failed_mount;
3846	}
3847
3848	/*
3849	 * It makes no sense for the first data block to be beyond the end
3850	 * of the filesystem.
3851	 */
3852	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3853		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3854			 "block %u is beyond end of filesystem (%llu)",
3855			 le32_to_cpu(es->s_first_data_block),
3856			 ext4_blocks_count(es));
3857		goto failed_mount;
3858	}
3859	blocks_count = (ext4_blocks_count(es) -
3860			le32_to_cpu(es->s_first_data_block) +
3861			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3862	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3863	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3864		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3865		       "(block count %llu, first data block %u, "
3866		       "blocks per group %lu)", sbi->s_groups_count,
3867		       ext4_blocks_count(es),
3868		       le32_to_cpu(es->s_first_data_block),
3869		       EXT4_BLOCKS_PER_GROUP(sb));
3870		goto failed_mount;
3871	}
3872	sbi->s_groups_count = blocks_count;
3873	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3874			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3875	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3876		   EXT4_DESC_PER_BLOCK(sb);
3877	sbi->s_group_desc = ext4_kvmalloc(db_count *
3878					  sizeof(struct buffer_head *),
3879					  GFP_KERNEL);
3880	if (sbi->s_group_desc == NULL) {
3881		ext4_msg(sb, KERN_ERR, "not enough memory");
3882		ret = -ENOMEM;
3883		goto failed_mount;
3884	}
3885
3886	if (ext4_proc_root)
3887		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3888
3889	if (sbi->s_proc)
3890		proc_create_data("options", S_IRUGO, sbi->s_proc,
3891				 &ext4_seq_options_fops, sb);
3892
3893	bgl_lock_init(sbi->s_blockgroup_lock);
3894
3895	for (i = 0; i < db_count; i++) {
3896		block = descriptor_loc(sb, logical_sb_block, i);
3897		sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3898		if (!sbi->s_group_desc[i]) {
3899			ext4_msg(sb, KERN_ERR,
3900			       "can't read group descriptor %d", i);
3901			db_count = i;
3902			goto failed_mount2;
3903		}
3904	}
3905	if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3906		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3907		goto failed_mount2;
3908	}
3909
3910	sbi->s_gdb_count = db_count;
3911	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3912	spin_lock_init(&sbi->s_next_gen_lock);
3913
3914	init_timer(&sbi->s_err_report);
3915	sbi->s_err_report.function = print_daily_error_info;
3916	sbi->s_err_report.data = (unsigned long) sb;
3917
3918	/* Register extent status tree shrinker */
3919	if (ext4_es_register_shrinker(sbi))
3920		goto failed_mount3;
3921
3922	sbi->s_stripe = ext4_get_stripe_size(sbi);
3923	sbi->s_extent_max_zeroout_kb = 32;
3924
3925	/*
3926	 * set up enough so that it can read an inode
3927	 */
3928	sb->s_op = &ext4_sops;
3929	sb->s_export_op = &ext4_export_ops;
3930	sb->s_xattr = ext4_xattr_handlers;
3931#ifdef CONFIG_QUOTA
3932	sb->dq_op = &ext4_quota_operations;
3933	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3934		sb->s_qcop = &ext4_qctl_sysfile_operations;
3935	else
3936		sb->s_qcop = &ext4_qctl_operations;
3937#endif
3938	memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3939
3940	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3941	mutex_init(&sbi->s_orphan_lock);
3942
3943	sb->s_root = NULL;
3944
3945	needs_recovery = (es->s_last_orphan != 0 ||
3946			  EXT4_HAS_INCOMPAT_FEATURE(sb,
3947				    EXT4_FEATURE_INCOMPAT_RECOVER));
3948
3949	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3950	    !(sb->s_flags & MS_RDONLY))
3951		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3952			goto failed_mount3a;
3953
3954	/*
3955	 * The first inode we look at is the journal inode.  Don't try
3956	 * root first: it may be modified in the journal!
3957	 */
3958	if (!test_opt(sb, NOLOAD) &&
3959	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3960		if (ext4_load_journal(sb, es, journal_devnum))
3961			goto failed_mount3a;
3962	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3963	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3964		ext4_msg(sb, KERN_ERR, "required journal recovery "
3965		       "suppressed and not mounted read-only");
3966		goto failed_mount_wq;
3967	} else {
3968		clear_opt(sb, DATA_FLAGS);
3969		sbi->s_journal = NULL;
3970		needs_recovery = 0;
3971		goto no_journal;
3972	}
3973
3974	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3975	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3976				       JBD2_FEATURE_INCOMPAT_64BIT)) {
3977		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3978		goto failed_mount_wq;
3979	}
3980
3981	if (!set_journal_csum_feature_set(sb)) {
3982		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3983			 "feature set");
3984		goto failed_mount_wq;
3985	}
3986
3987	/* We have now updated the journal if required, so we can
3988	 * validate the data journaling mode. */
3989	switch (test_opt(sb, DATA_FLAGS)) {
3990	case 0:
3991		/* No mode set, assume a default based on the journal
3992		 * capabilities: ORDERED_DATA if the journal can
3993		 * cope, else JOURNAL_DATA
3994		 */
3995		if (jbd2_journal_check_available_features
3996		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3997			set_opt(sb, ORDERED_DATA);
3998		else
3999			set_opt(sb, JOURNAL_DATA);
4000		break;
4001
4002	case EXT4_MOUNT_ORDERED_DATA:
4003	case EXT4_MOUNT_WRITEBACK_DATA:
4004		if (!jbd2_journal_check_available_features
4005		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4006			ext4_msg(sb, KERN_ERR, "Journal does not support "
4007			       "requested data journaling mode");
4008			goto failed_mount_wq;
4009		}
4010	default:
4011		break;
4012	}
4013	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4014
4015	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4016
4017no_journal:
4018	if (ext4_mballoc_ready) {
4019		sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4020		if (!sbi->s_mb_cache) {
4021			ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4022			goto failed_mount_wq;
4023		}
4024	}
4025
4026	/*
4027	 * Get the # of file system overhead blocks from the
4028	 * superblock if present.
4029	 */
4030	if (es->s_overhead_clusters)
4031		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4032	else {
4033		err = ext4_calculate_overhead(sb);
4034		if (err)
4035			goto failed_mount_wq;
4036	}
4037
4038	/*
4039	 * The maximum number of concurrent works can be high and
4040	 * concurrency isn't really necessary.  Limit it to 1.
4041	 */
4042	EXT4_SB(sb)->rsv_conversion_wq =
4043		alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4044	if (!EXT4_SB(sb)->rsv_conversion_wq) {
4045		printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4046		ret = -ENOMEM;
4047		goto failed_mount4;
4048	}
4049
4050	/*
4051	 * The jbd2_journal_load will have done any necessary log recovery,
4052	 * so we can safely mount the rest of the filesystem now.
4053	 */
4054
4055	root = ext4_iget(sb, EXT4_ROOT_INO);
4056	if (IS_ERR(root)) {
4057		ext4_msg(sb, KERN_ERR, "get root inode failed");
4058		ret = PTR_ERR(root);
4059		root = NULL;
4060		goto failed_mount4;
4061	}
4062	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4063		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4064		iput(root);
4065		goto failed_mount4;
4066	}
4067	sb->s_root = d_make_root(root);
4068	if (!sb->s_root) {
4069		ext4_msg(sb, KERN_ERR, "get root dentry failed");
4070		ret = -ENOMEM;
4071		goto failed_mount4;
4072	}
4073
4074	if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4075		sb->s_flags |= MS_RDONLY;
4076
4077	/* determine the minimum size of new large inodes, if present */
4078	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4079		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4080						     EXT4_GOOD_OLD_INODE_SIZE;
4081		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4082				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4083			if (sbi->s_want_extra_isize <
4084			    le16_to_cpu(es->s_want_extra_isize))
4085				sbi->s_want_extra_isize =
4086					le16_to_cpu(es->s_want_extra_isize);
4087			if (sbi->s_want_extra_isize <
4088			    le16_to_cpu(es->s_min_extra_isize))
4089				sbi->s_want_extra_isize =
4090					le16_to_cpu(es->s_min_extra_isize);
4091		}
4092	}
4093	/* Check if enough inode space is available */
4094	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4095							sbi->s_inode_size) {
4096		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4097						       EXT4_GOOD_OLD_INODE_SIZE;
4098		ext4_msg(sb, KERN_INFO, "required extra inode space not"
4099			 "available");
4100	}
4101
4102	err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4103	if (err) {
4104		ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4105			 "reserved pool", ext4_calculate_resv_clusters(sb));
4106		goto failed_mount4a;
4107	}
4108
4109	err = ext4_setup_system_zone(sb);
4110	if (err) {
4111		ext4_msg(sb, KERN_ERR, "failed to initialize system "
4112			 "zone (%d)", err);
4113		goto failed_mount4a;
4114	}
4115
4116	ext4_ext_init(sb);
4117	err = ext4_mb_init(sb);
4118	if (err) {
4119		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4120			 err);
4121		goto failed_mount5;
4122	}
4123
4124	block = ext4_count_free_clusters(sb);
4125	ext4_free_blocks_count_set(sbi->s_es,
4126				   EXT4_C2B(sbi, block));
4127	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4128				  GFP_KERNEL);
4129	if (!err) {
4130		unsigned long freei = ext4_count_free_inodes(sb);
4131		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4132		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4133					  GFP_KERNEL);
4134	}
4135	if (!err)
4136		err = percpu_counter_init(&sbi->s_dirs_counter,
4137					  ext4_count_dirs(sb), GFP_KERNEL);
4138	if (!err)
4139		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4140					  GFP_KERNEL);
4141	if (err) {
4142		ext4_msg(sb, KERN_ERR, "insufficient memory");
4143		goto failed_mount6;
4144	}
4145
4146	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4147		if (!ext4_fill_flex_info(sb)) {
4148			ext4_msg(sb, KERN_ERR,
4149			       "unable to initialize "
4150			       "flex_bg meta info!");
4151			goto failed_mount6;
4152		}
4153
4154	err = ext4_register_li_request(sb, first_not_zeroed);
4155	if (err)
4156		goto failed_mount6;
4157
4158	sbi->s_kobj.kset = ext4_kset;
4159	init_completion(&sbi->s_kobj_unregister);
4160	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4161				   "%s", sb->s_id);
4162	if (err)
4163		goto failed_mount7;
4164
4165#ifdef CONFIG_QUOTA
4166	/* Enable quota usage during mount. */
4167	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4168	    !(sb->s_flags & MS_RDONLY)) {
4169		err = ext4_enable_quotas(sb);
4170		if (err)
4171			goto failed_mount8;
4172	}
4173#endif  /* CONFIG_QUOTA */
4174
4175	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4176	ext4_orphan_cleanup(sb, es);
4177	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4178	if (needs_recovery) {
4179		ext4_msg(sb, KERN_INFO, "recovery complete");
4180		ext4_mark_recovery_complete(sb, es);
4181	}
4182	if (EXT4_SB(sb)->s_journal) {
4183		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4184			descr = " journalled data mode";
4185		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4186			descr = " ordered data mode";
4187		else
4188			descr = " writeback data mode";
4189	} else
4190		descr = "out journal";
4191
4192	if (test_opt(sb, DISCARD)) {
4193		struct request_queue *q = bdev_get_queue(sb->s_bdev);
4194		if (!blk_queue_discard(q))
4195			ext4_msg(sb, KERN_WARNING,
4196				 "mounting with \"discard\" option, but "
4197				 "the device does not support discard");
4198	}
4199
4200	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4201		 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4202		 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4203
4204	if (es->s_error_count)
4205		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4206
4207	/* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4208	ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4209	ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4210	ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4211
4212	kfree(orig_data);
4213	return 0;
4214
4215cantfind_ext4:
4216	if (!silent)
4217		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4218	goto failed_mount;
4219
4220#ifdef CONFIG_QUOTA
4221failed_mount8:
4222	kobject_del(&sbi->s_kobj);
4223#endif
4224failed_mount7:
4225	ext4_unregister_li_request(sb);
4226failed_mount6:
4227	ext4_mb_release(sb);
4228	if (sbi->s_flex_groups)
4229		ext4_kvfree(sbi->s_flex_groups);
4230	percpu_counter_destroy(&sbi->s_freeclusters_counter);
4231	percpu_counter_destroy(&sbi->s_freeinodes_counter);
4232	percpu_counter_destroy(&sbi->s_dirs_counter);
4233	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4234failed_mount5:
4235	ext4_ext_release(sb);
4236	ext4_release_system_zone(sb);
4237failed_mount4a:
4238	dput(sb->s_root);
4239	sb->s_root = NULL;
4240failed_mount4:
4241	ext4_msg(sb, KERN_ERR, "mount failed");
4242	if (EXT4_SB(sb)->rsv_conversion_wq)
4243		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4244failed_mount_wq:
4245	if (sbi->s_journal) {
4246		jbd2_journal_destroy(sbi->s_journal);
4247		sbi->s_journal = NULL;
4248	}
4249failed_mount3a:
4250	ext4_es_unregister_shrinker(sbi);
4251failed_mount3:
4252	del_timer_sync(&sbi->s_err_report);
4253	if (sbi->s_mmp_tsk)
4254		kthread_stop(sbi->s_mmp_tsk);
4255failed_mount2:
4256	for (i = 0; i < db_count; i++)
4257		brelse(sbi->s_group_desc[i]);
4258	ext4_kvfree(sbi->s_group_desc);
4259failed_mount:
4260	if (sbi->s_chksum_driver)
4261		crypto_free_shash(sbi->s_chksum_driver);
4262	if (sbi->s_proc) {
4263		remove_proc_entry("options", sbi->s_proc);
4264		remove_proc_entry(sb->s_id, ext4_proc_root);
4265	}
4266#ifdef CONFIG_QUOTA
4267	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4268		kfree(sbi->s_qf_names[i]);
4269#endif
4270	ext4_blkdev_remove(sbi);
4271	brelse(bh);
4272out_fail:
4273	sb->s_fs_info = NULL;
4274	kfree(sbi->s_blockgroup_lock);
4275	kfree(sbi);
4276out_free_orig:
4277	kfree(orig_data);
4278	return err ? err : ret;
4279}
4280
4281/*
4282 * Setup any per-fs journal parameters now.  We'll do this both on
4283 * initial mount, once the journal has been initialised but before we've
4284 * done any recovery; and again on any subsequent remount.
4285 */
4286static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4287{
4288	struct ext4_sb_info *sbi = EXT4_SB(sb);
4289
4290	journal->j_commit_interval = sbi->s_commit_interval;
4291	journal->j_min_batch_time = sbi->s_min_batch_time;
4292	journal->j_max_batch_time = sbi->s_max_batch_time;
4293
4294	write_lock(&journal->j_state_lock);
4295	if (test_opt(sb, BARRIER))
4296		journal->j_flags |= JBD2_BARRIER;
4297	else
4298		journal->j_flags &= ~JBD2_BARRIER;
4299	if (test_opt(sb, DATA_ERR_ABORT))
4300		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4301	else
4302		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4303	write_unlock(&journal->j_state_lock);
4304}
4305
4306static journal_t *ext4_get_journal(struct super_block *sb,
4307				   unsigned int journal_inum)
4308{
4309	struct inode *journal_inode;
4310	journal_t *journal;
4311
4312	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4313
4314	/* First, test for the existence of a valid inode on disk.  Bad
4315	 * things happen if we iget() an unused inode, as the subsequent
4316	 * iput() will try to delete it. */
4317
4318	journal_inode = ext4_iget(sb, journal_inum);
4319	if (IS_ERR(journal_inode)) {
4320		ext4_msg(sb, KERN_ERR, "no journal found");
4321		return NULL;
4322	}
4323	if (!journal_inode->i_nlink) {
4324		make_bad_inode(journal_inode);
4325		iput(journal_inode);
4326		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4327		return NULL;
4328	}
4329
4330	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4331		  journal_inode, journal_inode->i_size);
4332	if (!S_ISREG(journal_inode->i_mode)) {
4333		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4334		iput(journal_inode);
4335		return NULL;
4336	}
4337
4338	journal = jbd2_journal_init_inode(journal_inode);
4339	if (!journal) {
4340		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4341		iput(journal_inode);
4342		return NULL;
4343	}
4344	journal->j_private = sb;
4345	ext4_init_journal_params(sb, journal);
4346	return journal;
4347}
4348
4349static journal_t *ext4_get_dev_journal(struct super_block *sb,
4350				       dev_t j_dev)
4351{
4352	struct buffer_head *bh;
4353	journal_t *journal;
4354	ext4_fsblk_t start;
4355	ext4_fsblk_t len;
4356	int hblock, blocksize;
4357	ext4_fsblk_t sb_block;
4358	unsigned long offset;
4359	struct ext4_super_block *es;
4360	struct block_device *bdev;
4361
4362	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4363
4364	bdev = ext4_blkdev_get(j_dev, sb);
4365	if (bdev == NULL)
4366		return NULL;
4367
4368	blocksize = sb->s_blocksize;
4369	hblock = bdev_logical_block_size(bdev);
4370	if (blocksize < hblock) {
4371		ext4_msg(sb, KERN_ERR,
4372			"blocksize too small for journal device");
4373		goto out_bdev;
4374	}
4375
4376	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4377	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4378	set_blocksize(bdev, blocksize);
4379	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4380		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4381		       "external journal");
4382		goto out_bdev;
4383	}
4384
4385	es = (struct ext4_super_block *) (bh->b_data + offset);
4386	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4387	    !(le32_to_cpu(es->s_feature_incompat) &
4388	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4389		ext4_msg(sb, KERN_ERR, "external journal has "
4390					"bad superblock");
4391		brelse(bh);
4392		goto out_bdev;
4393	}
4394
4395	if ((le32_to_cpu(es->s_feature_ro_compat) &
4396	     EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4397	    es->s_checksum != ext4_superblock_csum(sb, es)) {
4398		ext4_msg(sb, KERN_ERR, "external journal has "
4399				       "corrupt superblock");
4400		brelse(bh);
4401		goto out_bdev;
4402	}
4403
4404	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4405		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4406		brelse(bh);
4407		goto out_bdev;
4408	}
4409
4410	len = ext4_blocks_count(es);
4411	start = sb_block + 1;
4412	brelse(bh);	/* we're done with the superblock */
4413
4414	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4415					start, len, blocksize);
4416	if (!journal) {
4417		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4418		goto out_bdev;
4419	}
4420	journal->j_private = sb;
4421	ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4422	wait_on_buffer(journal->j_sb_buffer);
4423	if (!buffer_uptodate(journal->j_sb_buffer)) {
4424		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4425		goto out_journal;
4426	}
4427	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4428		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4429					"user (unsupported) - %d",
4430			be32_to_cpu(journal->j_superblock->s_nr_users));
4431		goto out_journal;
4432	}
4433	EXT4_SB(sb)->journal_bdev = bdev;
4434	ext4_init_journal_params(sb, journal);
4435	return journal;
4436
4437out_journal:
4438	jbd2_journal_destroy(journal);
4439out_bdev:
4440	ext4_blkdev_put(bdev);
4441	return NULL;
4442}
4443
4444static int ext4_load_journal(struct super_block *sb,
4445			     struct ext4_super_block *es,
4446			     unsigned long journal_devnum)
4447{
4448	journal_t *journal;
4449	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4450	dev_t journal_dev;
4451	int err = 0;
4452	int really_read_only;
4453
4454	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4455
4456	if (journal_devnum &&
4457	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4458		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4459			"numbers have changed");
4460		journal_dev = new_decode_dev(journal_devnum);
4461	} else
4462		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4463
4464	really_read_only = bdev_read_only(sb->s_bdev);
4465
4466	/*
4467	 * Are we loading a blank journal or performing recovery after a
4468	 * crash?  For recovery, we need to check in advance whether we
4469	 * can get read-write access to the device.
4470	 */
4471	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4472		if (sb->s_flags & MS_RDONLY) {
4473			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4474					"required on readonly filesystem");
4475			if (really_read_only) {
4476				ext4_msg(sb, KERN_ERR, "write access "
4477					"unavailable, cannot proceed");
4478				return -EROFS;
4479			}
4480			ext4_msg(sb, KERN_INFO, "write access will "
4481			       "be enabled during recovery");
4482		}
4483	}
4484
4485	if (journal_inum && journal_dev) {
4486		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4487		       "and inode journals!");
4488		return -EINVAL;
4489	}
4490
4491	if (journal_inum) {
4492		if (!(journal = ext4_get_journal(sb, journal_inum)))
4493			return -EINVAL;
4494	} else {
4495		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4496			return -EINVAL;
4497	}
4498
4499	if (!(journal->j_flags & JBD2_BARRIER))
4500		ext4_msg(sb, KERN_INFO, "barriers disabled");
4501
4502	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4503		err = jbd2_journal_wipe(journal, !really_read_only);
4504	if (!err) {
4505		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4506		if (save)
4507			memcpy(save, ((char *) es) +
4508			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4509		err = jbd2_journal_load(journal);
4510		if (save)
4511			memcpy(((char *) es) + EXT4_S_ERR_START,
4512			       save, EXT4_S_ERR_LEN);
4513		kfree(save);
4514	}
4515
4516	if (err) {
4517		ext4_msg(sb, KERN_ERR, "error loading journal");
4518		jbd2_journal_destroy(journal);
4519		return err;
4520	}
4521
4522	EXT4_SB(sb)->s_journal = journal;
4523	ext4_clear_journal_err(sb, es);
4524
4525	if (!really_read_only && journal_devnum &&
4526	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4527		es->s_journal_dev = cpu_to_le32(journal_devnum);
4528
4529		/* Make sure we flush the recovery flag to disk. */
4530		ext4_commit_super(sb, 1);
4531	}
4532
4533	return 0;
4534}
4535
4536static int ext4_commit_super(struct super_block *sb, int sync)
4537{
4538	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4539	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4540	int error = 0;
4541
4542	if (!sbh || block_device_ejected(sb))
4543		return error;
4544	if (buffer_write_io_error(sbh)) {
4545		/*
4546		 * Oh, dear.  A previous attempt to write the
4547		 * superblock failed.  This could happen because the
4548		 * USB device was yanked out.  Or it could happen to
4549		 * be a transient write error and maybe the block will
4550		 * be remapped.  Nothing we can do but to retry the
4551		 * write and hope for the best.
4552		 */
4553		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4554		       "superblock detected");
4555		clear_buffer_write_io_error(sbh);
4556		set_buffer_uptodate(sbh);
4557	}
4558	/*
4559	 * If the file system is mounted read-only, don't update the
4560	 * superblock write time.  This avoids updating the superblock
4561	 * write time when we are mounting the root file system
4562	 * read/only but we need to replay the journal; at that point,
4563	 * for people who are east of GMT and who make their clock
4564	 * tick in localtime for Windows bug-for-bug compatibility,
4565	 * the clock is set in the future, and this will cause e2fsck
4566	 * to complain and force a full file system check.
4567	 */
4568	if (!(sb->s_flags & MS_RDONLY))
4569		es->s_wtime = cpu_to_le32(get_seconds());
4570	if (sb->s_bdev->bd_part)
4571		es->s_kbytes_written =
4572			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4573			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4574			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4575	else
4576		es->s_kbytes_written =
4577			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4578	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4579		ext4_free_blocks_count_set(es,
4580			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4581				&EXT4_SB(sb)->s_freeclusters_counter)));
4582	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4583		es->s_free_inodes_count =
4584			cpu_to_le32(percpu_counter_sum_positive(
4585				&EXT4_SB(sb)->s_freeinodes_counter));
4586	BUFFER_TRACE(sbh, "marking dirty");
4587	ext4_superblock_csum_set(sb);
4588	mark_buffer_dirty(sbh);
4589	if (sync) {
4590		error = sync_dirty_buffer(sbh);
4591		if (error)
4592			return error;
4593
4594		error = buffer_write_io_error(sbh);
4595		if (error) {
4596			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4597			       "superblock");
4598			clear_buffer_write_io_error(sbh);
4599			set_buffer_uptodate(sbh);
4600		}
4601	}
4602	return error;
4603}
4604
4605/*
4606 * Have we just finished recovery?  If so, and if we are mounting (or
4607 * remounting) the filesystem readonly, then we will end up with a
4608 * consistent fs on disk.  Record that fact.
4609 */
4610static void ext4_mark_recovery_complete(struct super_block *sb,
4611					struct ext4_super_block *es)
4612{
4613	journal_t *journal = EXT4_SB(sb)->s_journal;
4614
4615	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4616		BUG_ON(journal != NULL);
4617		return;
4618	}
4619	jbd2_journal_lock_updates(journal);
4620	if (jbd2_journal_flush(journal) < 0)
4621		goto out;
4622
4623	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4624	    sb->s_flags & MS_RDONLY) {
4625		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4626		ext4_commit_super(sb, 1);
4627	}
4628
4629out:
4630	jbd2_journal_unlock_updates(journal);
4631}
4632
4633/*
4634 * If we are mounting (or read-write remounting) a filesystem whose journal
4635 * has recorded an error from a previous lifetime, move that error to the
4636 * main filesystem now.
4637 */
4638static void ext4_clear_journal_err(struct super_block *sb,
4639				   struct ext4_super_block *es)
4640{
4641	journal_t *journal;
4642	int j_errno;
4643	const char *errstr;
4644
4645	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4646
4647	journal = EXT4_SB(sb)->s_journal;
4648
4649	/*
4650	 * Now check for any error status which may have been recorded in the
4651	 * journal by a prior ext4_error() or ext4_abort()
4652	 */
4653
4654	j_errno = jbd2_journal_errno(journal);
4655	if (j_errno) {
4656		char nbuf[16];
4657
4658		errstr = ext4_decode_error(sb, j_errno, nbuf);
4659		ext4_warning(sb, "Filesystem error recorded "
4660			     "from previous mount: %s", errstr);
4661		ext4_warning(sb, "Marking fs in need of filesystem check.");
4662
4663		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4664		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4665		ext4_commit_super(sb, 1);
4666
4667		jbd2_journal_clear_err(journal);
4668		jbd2_journal_update_sb_errno(journal);
4669	}
4670}
4671
4672/*
4673 * Force the running and committing transactions to commit,
4674 * and wait on the commit.
4675 */
4676int ext4_force_commit(struct super_block *sb)
4677{
4678	journal_t *journal;
4679
4680	if (sb->s_flags & MS_RDONLY)
4681		return 0;
4682
4683	journal = EXT4_SB(sb)->s_journal;
4684	return ext4_journal_force_commit(journal);
4685}
4686
4687static int ext4_sync_fs(struct super_block *sb, int wait)
4688{
4689	int ret = 0;
4690	tid_t target;
4691	bool needs_barrier = false;
4692	struct ext4_sb_info *sbi = EXT4_SB(sb);
4693
4694	trace_ext4_sync_fs(sb, wait);
4695	flush_workqueue(sbi->rsv_conversion_wq);
4696	/*
4697	 * Writeback quota in non-journalled quota case - journalled quota has
4698	 * no dirty dquots
4699	 */
4700	dquot_writeback_dquots(sb, -1);
4701	/*
4702	 * Data writeback is possible w/o journal transaction, so barrier must
4703	 * being sent at the end of the function. But we can skip it if
4704	 * transaction_commit will do it for us.
4705	 */
4706	if (sbi->s_journal) {
4707		target = jbd2_get_latest_transaction(sbi->s_journal);
4708		if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4709		    !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4710			needs_barrier = true;
4711
4712		if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4713			if (wait)
4714				ret = jbd2_log_wait_commit(sbi->s_journal,
4715							   target);
4716		}
4717	} else if (wait && test_opt(sb, BARRIER))
4718		needs_barrier = true;
4719	if (needs_barrier) {
4720		int err;
4721		err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4722		if (!ret)
4723			ret = err;
4724	}
4725
4726	return ret;
4727}
4728
4729/*
4730 * LVM calls this function before a (read-only) snapshot is created.  This
4731 * gives us a chance to flush the journal completely and mark the fs clean.
4732 *
4733 * Note that only this function cannot bring a filesystem to be in a clean
4734 * state independently. It relies on upper layer to stop all data & metadata
4735 * modifications.
4736 */
4737static int ext4_freeze(struct super_block *sb)
4738{
4739	int error = 0;
4740	journal_t *journal;
4741
4742	if (sb->s_flags & MS_RDONLY)
4743		return 0;
4744
4745	journal = EXT4_SB(sb)->s_journal;
4746
4747	if (journal) {
4748		/* Now we set up the journal barrier. */
4749		jbd2_journal_lock_updates(journal);
4750
4751		/*
4752		 * Don't clear the needs_recovery flag if we failed to
4753		 * flush the journal.
4754		 */
4755		error = jbd2_journal_flush(journal);
4756		if (error < 0)
4757			goto out;
4758	}
4759
4760	/* Journal blocked and flushed, clear needs_recovery flag. */
4761	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4762	error = ext4_commit_super(sb, 1);
4763out:
4764	if (journal)
4765		/* we rely on upper layer to stop further updates */
4766		jbd2_journal_unlock_updates(journal);
4767	return error;
4768}
4769
4770/*
4771 * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4772 * flag here, even though the filesystem is not technically dirty yet.
4773 */
4774static int ext4_unfreeze(struct super_block *sb)
4775{
4776	if (sb->s_flags & MS_RDONLY)
4777		return 0;
4778
4779	/* Reset the needs_recovery flag before the fs is unlocked. */
4780	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4781	ext4_commit_super(sb, 1);
4782	return 0;
4783}
4784
4785/*
4786 * Structure to save mount options for ext4_remount's benefit
4787 */
4788struct ext4_mount_options {
4789	unsigned long s_mount_opt;
4790	unsigned long s_mount_opt2;
4791	kuid_t s_resuid;
4792	kgid_t s_resgid;
4793	unsigned long s_commit_interval;
4794	u32 s_min_batch_time, s_max_batch_time;
4795#ifdef CONFIG_QUOTA
4796	int s_jquota_fmt;
4797	char *s_qf_names[EXT4_MAXQUOTAS];
4798#endif
4799};
4800
4801static int ext4_remount(struct super_block *sb, int *flags, char *data)
4802{
4803	struct ext4_super_block *es;
4804	struct ext4_sb_info *sbi = EXT4_SB(sb);
4805	unsigned long old_sb_flags;
4806	struct ext4_mount_options old_opts;
4807	int enable_quota = 0;
4808	ext4_group_t g;
4809	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4810	int err = 0;
4811#ifdef CONFIG_QUOTA
4812	int i, j;
4813#endif
4814	char *orig_data = kstrdup(data, GFP_KERNEL);
4815
4816	/* Store the original options */
4817	old_sb_flags = sb->s_flags;
4818	old_opts.s_mount_opt = sbi->s_mount_opt;
4819	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4820	old_opts.s_resuid = sbi->s_resuid;
4821	old_opts.s_resgid = sbi->s_resgid;
4822	old_opts.s_commit_interval = sbi->s_commit_interval;
4823	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4824	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4825#ifdef CONFIG_QUOTA
4826	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4827	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4828		if (sbi->s_qf_names[i]) {
4829			old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4830							 GFP_KERNEL);
4831			if (!old_opts.s_qf_names[i]) {
4832				for (j = 0; j < i; j++)
4833					kfree(old_opts.s_qf_names[j]);
4834				kfree(orig_data);
4835				return -ENOMEM;
4836			}
4837		} else
4838			old_opts.s_qf_names[i] = NULL;
4839#endif
4840	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4841		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4842
4843	/*
4844	 * Allow the "check" option to be passed as a remount option.
4845	 */
4846	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4847		err = -EINVAL;
4848		goto restore_opts;
4849	}
4850
4851	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4852	    test_opt(sb, JOURNAL_CHECKSUM)) {
4853		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4854			 "during remount not supported");
4855		err = -EINVAL;
4856		goto restore_opts;
4857	}
4858
4859	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4860		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4861			ext4_msg(sb, KERN_ERR, "can't mount with "
4862				 "both data=journal and delalloc");
4863			err = -EINVAL;
4864			goto restore_opts;
4865		}
4866		if (test_opt(sb, DIOREAD_NOLOCK)) {
4867			ext4_msg(sb, KERN_ERR, "can't mount with "
4868				 "both data=journal and dioread_nolock");
4869			err = -EINVAL;
4870			goto restore_opts;
4871		}
4872	}
4873
4874	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4875		ext4_abort(sb, "Abort forced by user");
4876
4877	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4878		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4879
4880	es = sbi->s_es;
4881
4882	if (sbi->s_journal) {
4883		ext4_init_journal_params(sb, sbi->s_journal);
4884		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4885	}
4886
4887	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4888		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4889			err = -EROFS;
4890			goto restore_opts;
4891		}
4892
4893		if (*flags & MS_RDONLY) {
4894			err = sync_filesystem(sb);
4895			if (err < 0)
4896				goto restore_opts;
4897			err = dquot_suspend(sb, -1);
4898			if (err < 0)
4899				goto restore_opts;
4900
4901			/*
4902			 * First of all, the unconditional stuff we have to do
4903			 * to disable replay of the journal when we next remount
4904			 */
4905			sb->s_flags |= MS_RDONLY;
4906
4907			/*
4908			 * OK, test if we are remounting a valid rw partition
4909			 * readonly, and if so set the rdonly flag and then
4910			 * mark the partition as valid again.
4911			 */
4912			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4913			    (sbi->s_mount_state & EXT4_VALID_FS))
4914				es->s_state = cpu_to_le16(sbi->s_mount_state);
4915
4916			if (sbi->s_journal)
4917				ext4_mark_recovery_complete(sb, es);
4918		} else {
4919			/* Make sure we can mount this feature set readwrite */
4920			if (!ext4_feature_set_ok(sb, 0)) {
4921				err = -EROFS;
4922				goto restore_opts;
4923			}
4924			/*
4925			 * Make sure the group descriptor checksums
4926			 * are sane.  If they aren't, refuse to remount r/w.
4927			 */
4928			for (g = 0; g < sbi->s_groups_count; g++) {
4929				struct ext4_group_desc *gdp =
4930					ext4_get_group_desc(sb, g, NULL);
4931
4932				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4933					ext4_msg(sb, KERN_ERR,
4934	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
4935		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4936					       le16_to_cpu(gdp->bg_checksum));
4937					err = -EINVAL;
4938					goto restore_opts;
4939				}
4940			}
4941
4942			/*
4943			 * If we have an unprocessed orphan list hanging
4944			 * around from a previously readonly bdev mount,
4945			 * require a full umount/remount for now.
4946			 */
4947			if (es->s_last_orphan) {
4948				ext4_msg(sb, KERN_WARNING, "Couldn't "
4949				       "remount RDWR because of unprocessed "
4950				       "orphan inode list.  Please "
4951				       "umount/remount instead");
4952				err = -EINVAL;
4953				goto restore_opts;
4954			}
4955
4956			/*
4957			 * Mounting a RDONLY partition read-write, so reread
4958			 * and store the current valid flag.  (It may have
4959			 * been changed by e2fsck since we originally mounted
4960			 * the partition.)
4961			 */
4962			if (sbi->s_journal)
4963				ext4_clear_journal_err(sb, es);
4964			sbi->s_mount_state = le16_to_cpu(es->s_state);
4965			if (!ext4_setup_super(sb, es, 0))
4966				sb->s_flags &= ~MS_RDONLY;
4967			if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4968						     EXT4_FEATURE_INCOMPAT_MMP))
4969				if (ext4_multi_mount_protect(sb,
4970						le64_to_cpu(es->s_mmp_block))) {
4971					err = -EROFS;
4972					goto restore_opts;
4973				}
4974			enable_quota = 1;
4975		}
4976	}
4977
4978	/*
4979	 * Reinitialize lazy itable initialization thread based on
4980	 * current settings
4981	 */
4982	if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4983		ext4_unregister_li_request(sb);
4984	else {
4985		ext4_group_t first_not_zeroed;
4986		first_not_zeroed = ext4_has_uninit_itable(sb);
4987		ext4_register_li_request(sb, first_not_zeroed);
4988	}
4989
4990	ext4_setup_system_zone(sb);
4991	if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4992		ext4_commit_super(sb, 1);
4993
4994#ifdef CONFIG_QUOTA
4995	/* Release old quota file names */
4996	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4997		kfree(old_opts.s_qf_names[i]);
4998	if (enable_quota) {
4999		if (sb_any_quota_suspended(sb))
5000			dquot_resume(sb, -1);
5001		else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5002					EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5003			err = ext4_enable_quotas(sb);
5004			if (err)
5005				goto restore_opts;
5006		}
5007	}
5008#endif
5009
5010	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5011	kfree(orig_data);
5012	return 0;
5013
5014restore_opts:
5015	sb->s_flags = old_sb_flags;
5016	sbi->s_mount_opt = old_opts.s_mount_opt;
5017	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5018	sbi->s_resuid = old_opts.s_resuid;
5019	sbi->s_resgid = old_opts.s_resgid;
5020	sbi->s_commit_interval = old_opts.s_commit_interval;
5021	sbi->s_min_batch_time = old_opts.s_min_batch_time;
5022	sbi->s_max_batch_time = old_opts.s_max_batch_time;
5023#ifdef CONFIG_QUOTA
5024	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5025	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5026		kfree(sbi->s_qf_names[i]);
5027		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5028	}
5029#endif
5030	kfree(orig_data);
5031	return err;
5032}
5033
5034static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5035{
5036	struct super_block *sb = dentry->d_sb;
5037	struct ext4_sb_info *sbi = EXT4_SB(sb);
5038	struct ext4_super_block *es = sbi->s_es;
5039	ext4_fsblk_t overhead = 0, resv_blocks;
5040	u64 fsid;
5041	s64 bfree;
5042	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5043
5044	if (!test_opt(sb, MINIX_DF))
5045		overhead = sbi->s_overhead;
5046
5047	buf->f_type = EXT4_SUPER_MAGIC;
5048	buf->f_bsize = sb->s_blocksize;
5049	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5050	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5051		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5052	/* prevent underflow in case that few free space is available */
5053	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5054	buf->f_bavail = buf->f_bfree -
5055			(ext4_r_blocks_count(es) + resv_blocks);
5056	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5057		buf->f_bavail = 0;
5058	buf->f_files = le32_to_cpu(es->s_inodes_count);
5059	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5060	buf->f_namelen = EXT4_NAME_LEN;
5061	fsid = le64_to_cpup((void *)es->s_uuid) ^
5062	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5063	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5064	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5065
5066	return 0;
5067}
5068
5069/* Helper function for writing quotas on sync - we need to start transaction
5070 * before quota file is locked for write. Otherwise the are possible deadlocks:
5071 * Process 1                         Process 2
5072 * ext4_create()                     quota_sync()
5073 *   jbd2_journal_start()                  write_dquot()
5074 *   dquot_initialize()                         down(dqio_mutex)
5075 *     down(dqio_mutex)                    jbd2_journal_start()
5076 *
5077 */
5078
5079#ifdef CONFIG_QUOTA
5080
5081static inline struct inode *dquot_to_inode(struct dquot *dquot)
5082{
5083	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5084}
5085
5086static int ext4_write_dquot(struct dquot *dquot)
5087{
5088	int ret, err;
5089	handle_t *handle;
5090	struct inode *inode;
5091
5092	inode = dquot_to_inode(dquot);
5093	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5094				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5095	if (IS_ERR(handle))
5096		return PTR_ERR(handle);
5097	ret = dquot_commit(dquot);
5098	err = ext4_journal_stop(handle);
5099	if (!ret)
5100		ret = err;
5101	return ret;
5102}
5103
5104static int ext4_acquire_dquot(struct dquot *dquot)
5105{
5106	int ret, err;
5107	handle_t *handle;
5108
5109	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5110				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5111	if (IS_ERR(handle))
5112		return PTR_ERR(handle);
5113	ret = dquot_acquire(dquot);
5114	err = ext4_journal_stop(handle);
5115	if (!ret)
5116		ret = err;
5117	return ret;
5118}
5119
5120static int ext4_release_dquot(struct dquot *dquot)
5121{
5122	int ret, err;
5123	handle_t *handle;
5124
5125	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5126				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5127	if (IS_ERR(handle)) {
5128		/* Release dquot anyway to avoid endless cycle in dqput() */
5129		dquot_release(dquot);
5130		return PTR_ERR(handle);
5131	}
5132	ret = dquot_release(dquot);
5133	err = ext4_journal_stop(handle);
5134	if (!ret)
5135		ret = err;
5136	return ret;
5137}
5138
5139static int ext4_mark_dquot_dirty(struct dquot *dquot)
5140{
5141	struct super_block *sb = dquot->dq_sb;
5142	struct ext4_sb_info *sbi = EXT4_SB(sb);
5143
5144	/* Are we journaling quotas? */
5145	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5146	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5147		dquot_mark_dquot_dirty(dquot);
5148		return ext4_write_dquot(dquot);
5149	} else {
5150		return dquot_mark_dquot_dirty(dquot);
5151	}
5152}
5153
5154static int ext4_write_info(struct super_block *sb, int type)
5155{
5156	int ret, err;
5157	handle_t *handle;
5158
5159	/* Data block + inode block */
5160	handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5161	if (IS_ERR(handle))
5162		return PTR_ERR(handle);
5163	ret = dquot_commit_info(sb, type);
5164	err = ext4_journal_stop(handle);
5165	if (!ret)
5166		ret = err;
5167	return ret;
5168}
5169
5170/*
5171 * Turn on quotas during mount time - we need to find
5172 * the quota file and such...
5173 */
5174static int ext4_quota_on_mount(struct super_block *sb, int type)
5175{
5176	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5177					EXT4_SB(sb)->s_jquota_fmt, type);
5178}
5179
5180/*
5181 * Standard function to be called on quota_on
5182 */
5183static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5184			 struct path *path)
5185{
5186	int err;
5187
5188	if (!test_opt(sb, QUOTA))
5189		return -EINVAL;
5190
5191	/* Quotafile not on the same filesystem? */
5192	if (path->dentry->d_sb != sb)
5193		return -EXDEV;
5194	/* Journaling quota? */
5195	if (EXT4_SB(sb)->s_qf_names[type]) {
5196		/* Quotafile not in fs root? */
5197		if (path->dentry->d_parent != sb->s_root)
5198			ext4_msg(sb, KERN_WARNING,
5199				"Quota file not on filesystem root. "
5200				"Journaled quota will not work");
5201	}
5202
5203	/*
5204	 * When we journal data on quota file, we have to flush journal to see
5205	 * all updates to the file when we bypass pagecache...
5206	 */
5207	if (EXT4_SB(sb)->s_journal &&
5208	    ext4_should_journal_data(path->dentry->d_inode)) {
5209		/*
5210		 * We don't need to lock updates but journal_flush() could
5211		 * otherwise be livelocked...
5212		 */
5213		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5214		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5215		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5216		if (err)
5217			return err;
5218	}
5219
5220	return dquot_quota_on(sb, type, format_id, path);
5221}
5222
5223static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5224			     unsigned int flags)
5225{
5226	int err;
5227	struct inode *qf_inode;
5228	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5229		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5230		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5231	};
5232
5233	BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5234
5235	if (!qf_inums[type])
5236		return -EPERM;
5237
5238	qf_inode = ext4_iget(sb, qf_inums[type]);
5239	if (IS_ERR(qf_inode)) {
5240		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5241		return PTR_ERR(qf_inode);
5242	}
5243
5244	/* Don't account quota for quota files to avoid recursion */
5245	qf_inode->i_flags |= S_NOQUOTA;
5246	err = dquot_enable(qf_inode, type, format_id, flags);
5247	iput(qf_inode);
5248
5249	return err;
5250}
5251
5252/* Enable usage tracking for all quota types. */
5253static int ext4_enable_quotas(struct super_block *sb)
5254{
5255	int type, err = 0;
5256	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5257		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5258		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5259	};
5260
5261	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5262	for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5263		if (qf_inums[type]) {
5264			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5265						DQUOT_USAGE_ENABLED);
5266			if (err) {
5267				ext4_warning(sb,
5268					"Failed to enable quota tracking "
5269					"(type=%d, err=%d). Please run "
5270					"e2fsck to fix.", type, err);
5271				return err;
5272			}
5273		}
5274	}
5275	return 0;
5276}
5277
5278/*
5279 * quota_on function that is used when QUOTA feature is set.
5280 */
5281static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5282				 int format_id)
5283{
5284	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5285		return -EINVAL;
5286
5287	/*
5288	 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5289	 */
5290	return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5291}
5292
5293static int ext4_quota_off(struct super_block *sb, int type)
5294{
5295	struct inode *inode = sb_dqopt(sb)->files[type];
5296	handle_t *handle;
5297
5298	/* Force all delayed allocation blocks to be allocated.
5299	 * Caller already holds s_umount sem */
5300	if (test_opt(sb, DELALLOC))
5301		sync_filesystem(sb);
5302
5303	if (!inode)
5304		goto out;
5305
5306	/* Update modification times of quota files when userspace can
5307	 * start looking at them */
5308	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5309	if (IS_ERR(handle))
5310		goto out;
5311	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5312	ext4_mark_inode_dirty(handle, inode);
5313	ext4_journal_stop(handle);
5314
5315out:
5316	return dquot_quota_off(sb, type);
5317}
5318
5319/*
5320 * quota_off function that is used when QUOTA feature is set.
5321 */
5322static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5323{
5324	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5325		return -EINVAL;
5326
5327	/* Disable only the limits. */
5328	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5329}
5330
5331/* Read data from quotafile - avoid pagecache and such because we cannot afford
5332 * acquiring the locks... As quota files are never truncated and quota code
5333 * itself serializes the operations (and no one else should touch the files)
5334 * we don't have to be afraid of races */
5335static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5336			       size_t len, loff_t off)
5337{
5338	struct inode *inode = sb_dqopt(sb)->files[type];
5339	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5340	int offset = off & (sb->s_blocksize - 1);
5341	int tocopy;
5342	size_t toread;
5343	struct buffer_head *bh;
5344	loff_t i_size = i_size_read(inode);
5345
5346	if (off > i_size)
5347		return 0;
5348	if (off+len > i_size)
5349		len = i_size-off;
5350	toread = len;
5351	while (toread > 0) {
5352		tocopy = sb->s_blocksize - offset < toread ?
5353				sb->s_blocksize - offset : toread;
5354		bh = ext4_bread(NULL, inode, blk, 0);
5355		if (IS_ERR(bh))
5356			return PTR_ERR(bh);
5357		if (!bh)	/* A hole? */
5358			memset(data, 0, tocopy);
5359		else
5360			memcpy(data, bh->b_data+offset, tocopy);
5361		brelse(bh);
5362		offset = 0;
5363		toread -= tocopy;
5364		data += tocopy;
5365		blk++;
5366	}
5367	return len;
5368}
5369
5370/* Write to quotafile (we know the transaction is already started and has
5371 * enough credits) */
5372static ssize_t ext4_quota_write(struct super_block *sb, int type,
5373				const char *data, size_t len, loff_t off)
5374{
5375	struct inode *inode = sb_dqopt(sb)->files[type];
5376	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5377	int err, offset = off & (sb->s_blocksize - 1);
5378	struct buffer_head *bh;
5379	handle_t *handle = journal_current_handle();
5380
5381	if (EXT4_SB(sb)->s_journal && !handle) {
5382		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5383			" cancelled because transaction is not started",
5384			(unsigned long long)off, (unsigned long long)len);
5385		return -EIO;
5386	}
5387	/*
5388	 * Since we account only one data block in transaction credits,
5389	 * then it is impossible to cross a block boundary.
5390	 */
5391	if (sb->s_blocksize - offset < len) {
5392		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5393			" cancelled because not block aligned",
5394			(unsigned long long)off, (unsigned long long)len);
5395		return -EIO;
5396	}
5397
5398	bh = ext4_bread(handle, inode, blk, 1);
5399	if (IS_ERR(bh))
5400		return PTR_ERR(bh);
5401	if (!bh)
5402		goto out;
5403	BUFFER_TRACE(bh, "get write access");
5404	err = ext4_journal_get_write_access(handle, bh);
5405	if (err) {
5406		brelse(bh);
5407		return err;
5408	}
5409	lock_buffer(bh);
5410	memcpy(bh->b_data+offset, data, len);
5411	flush_dcache_page(bh->b_page);
5412	unlock_buffer(bh);
5413	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5414	brelse(bh);
5415out:
5416	if (inode->i_size < off + len) {
5417		i_size_write(inode, off + len);
5418		EXT4_I(inode)->i_disksize = inode->i_size;
5419		ext4_mark_inode_dirty(handle, inode);
5420	}
5421	return len;
5422}
5423
5424#endif
5425
5426static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5427		       const char *dev_name, void *data)
5428{
5429	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5430}
5431
5432#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5433static inline void register_as_ext2(void)
5434{
5435	int err = register_filesystem(&ext2_fs_type);
5436	if (err)
5437		printk(KERN_WARNING
5438		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5439}
5440
5441static inline void unregister_as_ext2(void)
5442{
5443	unregister_filesystem(&ext2_fs_type);
5444}
5445
5446static inline int ext2_feature_set_ok(struct super_block *sb)
5447{
5448	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5449		return 0;
5450	if (sb->s_flags & MS_RDONLY)
5451		return 1;
5452	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5453		return 0;
5454	return 1;
5455}
5456#else
5457static inline void register_as_ext2(void) { }
5458static inline void unregister_as_ext2(void) { }
5459static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5460#endif
5461
5462#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5463static inline void register_as_ext3(void)
5464{
5465	int err = register_filesystem(&ext3_fs_type);
5466	if (err)
5467		printk(KERN_WARNING
5468		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5469}
5470
5471static inline void unregister_as_ext3(void)
5472{
5473	unregister_filesystem(&ext3_fs_type);
5474}
5475
5476static inline int ext3_feature_set_ok(struct super_block *sb)
5477{
5478	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5479		return 0;
5480	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5481		return 0;
5482	if (sb->s_flags & MS_RDONLY)
5483		return 1;
5484	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5485		return 0;
5486	return 1;
5487}
5488#else
5489static inline void register_as_ext3(void) { }
5490static inline void unregister_as_ext3(void) { }
5491static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5492#endif
5493
5494static struct file_system_type ext4_fs_type = {
5495	.owner		= THIS_MODULE,
5496	.name		= "ext4",
5497	.mount		= ext4_mount,
5498	.kill_sb	= kill_block_super,
5499	.fs_flags	= FS_REQUIRES_DEV,
5500};
5501MODULE_ALIAS_FS("ext4");
5502
5503static int __init ext4_init_feat_adverts(void)
5504{
5505	struct ext4_features *ef;
5506	int ret = -ENOMEM;
5507
5508	ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5509	if (!ef)
5510		goto out;
5511
5512	ef->f_kobj.kset = ext4_kset;
5513	init_completion(&ef->f_kobj_unregister);
5514	ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5515				   "features");
5516	if (ret) {
5517		kfree(ef);
5518		goto out;
5519	}
5520
5521	ext4_feat = ef;
5522	ret = 0;
5523out:
5524	return ret;
5525}
5526
5527static void ext4_exit_feat_adverts(void)
5528{
5529	kobject_put(&ext4_feat->f_kobj);
5530	wait_for_completion(&ext4_feat->f_kobj_unregister);
5531	kfree(ext4_feat);
5532}
5533
5534/* Shared across all ext4 file systems */
5535wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5536struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5537
5538static int __init ext4_init_fs(void)
5539{
5540	int i, err;
5541
5542	ext4_li_info = NULL;
5543	mutex_init(&ext4_li_mtx);
5544
5545	/* Build-time check for flags consistency */
5546	ext4_check_flag_values();
5547
5548	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5549		mutex_init(&ext4__aio_mutex[i]);
5550		init_waitqueue_head(&ext4__ioend_wq[i]);
5551	}
5552
5553	err = ext4_init_es();
5554	if (err)
5555		return err;
5556
5557	err = ext4_init_pageio();
5558	if (err)
5559		goto out7;
5560
5561	err = ext4_init_system_zone();
5562	if (err)
5563		goto out6;
5564	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5565	if (!ext4_kset) {
5566		err = -ENOMEM;
5567		goto out5;
5568	}
5569	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5570
5571	err = ext4_init_feat_adverts();
5572	if (err)
5573		goto out4;
5574
5575	err = ext4_init_mballoc();
5576	if (err)
5577		goto out2;
5578	else
5579		ext4_mballoc_ready = 1;
5580	err = init_inodecache();
5581	if (err)
5582		goto out1;
5583	register_as_ext3();
5584	register_as_ext2();
5585	err = register_filesystem(&ext4_fs_type);
5586	if (err)
5587		goto out;
5588
5589	return 0;
5590out:
5591	unregister_as_ext2();
5592	unregister_as_ext3();
5593	destroy_inodecache();
5594out1:
5595	ext4_mballoc_ready = 0;
5596	ext4_exit_mballoc();
5597out2:
5598	ext4_exit_feat_adverts();
5599out4:
5600	if (ext4_proc_root)
5601		remove_proc_entry("fs/ext4", NULL);
5602	kset_unregister(ext4_kset);
5603out5:
5604	ext4_exit_system_zone();
5605out6:
5606	ext4_exit_pageio();
5607out7:
5608	ext4_exit_es();
5609
5610	return err;
5611}
5612
5613static void __exit ext4_exit_fs(void)
5614{
5615	ext4_destroy_lazyinit_thread();
5616	unregister_as_ext2();
5617	unregister_as_ext3();
5618	unregister_filesystem(&ext4_fs_type);
5619	destroy_inodecache();
5620	ext4_exit_mballoc();
5621	ext4_exit_feat_adverts();
5622	remove_proc_entry("fs/ext4", NULL);
5623	kset_unregister(ext4_kset);
5624	ext4_exit_system_zone();
5625	ext4_exit_pageio();
5626	ext4_exit_es();
5627}
5628
5629MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5630MODULE_DESCRIPTION("Fourth Extended Filesystem");
5631MODULE_LICENSE("GPL");
5632module_init(ext4_init_fs)
5633module_exit(ext4_exit_fs)
5634