xref: /openbmc/linux/fs/f2fs/super.c (revision 089a49b6)
1 /*
2  * fs/f2fs/super.c
3  *
4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5  *             http://www.samsung.com/
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License version 2 as
9  * published by the Free Software Foundation.
10  */
11 #include <linux/module.h>
12 #include <linux/init.h>
13 #include <linux/fs.h>
14 #include <linux/statfs.h>
15 #include <linux/buffer_head.h>
16 #include <linux/backing-dev.h>
17 #include <linux/kthread.h>
18 #include <linux/parser.h>
19 #include <linux/mount.h>
20 #include <linux/seq_file.h>
21 #include <linux/proc_fs.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/f2fs_fs.h>
26 #include <linux/sysfs.h>
27 
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "gc.h"
33 
34 #define CREATE_TRACE_POINTS
35 #include <trace/events/f2fs.h>
36 
37 static struct proc_dir_entry *f2fs_proc_root;
38 static struct kmem_cache *f2fs_inode_cachep;
39 static struct kset *f2fs_kset;
40 
41 enum {
42 	Opt_gc_background,
43 	Opt_disable_roll_forward,
44 	Opt_discard,
45 	Opt_noheap,
46 	Opt_nouser_xattr,
47 	Opt_noacl,
48 	Opt_active_logs,
49 	Opt_disable_ext_identify,
50 	Opt_inline_xattr,
51 	Opt_err,
52 };
53 
54 static match_table_t f2fs_tokens = {
55 	{Opt_gc_background, "background_gc=%s"},
56 	{Opt_disable_roll_forward, "disable_roll_forward"},
57 	{Opt_discard, "discard"},
58 	{Opt_noheap, "no_heap"},
59 	{Opt_nouser_xattr, "nouser_xattr"},
60 	{Opt_noacl, "noacl"},
61 	{Opt_active_logs, "active_logs=%u"},
62 	{Opt_disable_ext_identify, "disable_ext_identify"},
63 	{Opt_inline_xattr, "inline_xattr"},
64 	{Opt_err, NULL},
65 };
66 
67 /* Sysfs support for f2fs */
68 struct f2fs_attr {
69 	struct attribute attr;
70 	ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
71 	ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
72 			 const char *, size_t);
73 	int offset;
74 };
75 
76 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
77 			struct f2fs_sb_info *sbi, char *buf)
78 {
79 	struct f2fs_gc_kthread *gc_kth = sbi->gc_thread;
80 	unsigned int *ui;
81 
82 	if (!gc_kth)
83 		return -EINVAL;
84 
85 	ui = (unsigned int *)(((char *)gc_kth) + a->offset);
86 
87 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
88 }
89 
90 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
91 			struct f2fs_sb_info *sbi,
92 			const char *buf, size_t count)
93 {
94 	struct f2fs_gc_kthread *gc_kth = sbi->gc_thread;
95 	unsigned long t;
96 	unsigned int *ui;
97 	ssize_t ret;
98 
99 	if (!gc_kth)
100 		return -EINVAL;
101 
102 	ui = (unsigned int *)(((char *)gc_kth) + a->offset);
103 
104 	ret = kstrtoul(skip_spaces(buf), 0, &t);
105 	if (ret < 0)
106 		return ret;
107 	*ui = t;
108 	return count;
109 }
110 
111 static ssize_t f2fs_attr_show(struct kobject *kobj,
112 				struct attribute *attr, char *buf)
113 {
114 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
115 								s_kobj);
116 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
117 
118 	return a->show ? a->show(a, sbi, buf) : 0;
119 }
120 
121 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
122 						const char *buf, size_t len)
123 {
124 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
125 									s_kobj);
126 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
127 
128 	return a->store ? a->store(a, sbi, buf, len) : 0;
129 }
130 
131 static void f2fs_sb_release(struct kobject *kobj)
132 {
133 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
134 								s_kobj);
135 	complete(&sbi->s_kobj_unregister);
136 }
137 
138 #define F2FS_ATTR_OFFSET(_name, _mode, _show, _store, _elname) \
139 static struct f2fs_attr f2fs_attr_##_name = {			\
140 	.attr = {.name = __stringify(_name), .mode = _mode },	\
141 	.show	= _show,					\
142 	.store	= _store,					\
143 	.offset = offsetof(struct f2fs_gc_kthread, _elname),	\
144 }
145 
146 #define F2FS_RW_ATTR(name, elname)	\
147 	F2FS_ATTR_OFFSET(name, 0644, f2fs_sbi_show, f2fs_sbi_store, elname)
148 
149 F2FS_RW_ATTR(gc_min_sleep_time, min_sleep_time);
150 F2FS_RW_ATTR(gc_max_sleep_time, max_sleep_time);
151 F2FS_RW_ATTR(gc_no_gc_sleep_time, no_gc_sleep_time);
152 F2FS_RW_ATTR(gc_idle, gc_idle);
153 
154 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
155 static struct attribute *f2fs_attrs[] = {
156 	ATTR_LIST(gc_min_sleep_time),
157 	ATTR_LIST(gc_max_sleep_time),
158 	ATTR_LIST(gc_no_gc_sleep_time),
159 	ATTR_LIST(gc_idle),
160 	NULL,
161 };
162 
163 static const struct sysfs_ops f2fs_attr_ops = {
164 	.show	= f2fs_attr_show,
165 	.store	= f2fs_attr_store,
166 };
167 
168 static struct kobj_type f2fs_ktype = {
169 	.default_attrs	= f2fs_attrs,
170 	.sysfs_ops	= &f2fs_attr_ops,
171 	.release	= f2fs_sb_release,
172 };
173 
174 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
175 {
176 	struct va_format vaf;
177 	va_list args;
178 
179 	va_start(args, fmt);
180 	vaf.fmt = fmt;
181 	vaf.va = &args;
182 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
183 	va_end(args);
184 }
185 
186 static void init_once(void *foo)
187 {
188 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
189 
190 	inode_init_once(&fi->vfs_inode);
191 }
192 
193 static int parse_options(struct super_block *sb, char *options)
194 {
195 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
196 	substring_t args[MAX_OPT_ARGS];
197 	char *p, *name;
198 	int arg = 0;
199 
200 	if (!options)
201 		return 0;
202 
203 	while ((p = strsep(&options, ",")) != NULL) {
204 		int token;
205 		if (!*p)
206 			continue;
207 		/*
208 		 * Initialize args struct so we know whether arg was
209 		 * found; some options take optional arguments.
210 		 */
211 		args[0].to = args[0].from = NULL;
212 		token = match_token(p, f2fs_tokens, args);
213 
214 		switch (token) {
215 		case Opt_gc_background:
216 			name = match_strdup(&args[0]);
217 
218 			if (!name)
219 				return -ENOMEM;
220 			if (!strncmp(name, "on", 2))
221 				set_opt(sbi, BG_GC);
222 			else if (!strncmp(name, "off", 3))
223 				clear_opt(sbi, BG_GC);
224 			else {
225 				kfree(name);
226 				return -EINVAL;
227 			}
228 			kfree(name);
229 			break;
230 		case Opt_disable_roll_forward:
231 			set_opt(sbi, DISABLE_ROLL_FORWARD);
232 			break;
233 		case Opt_discard:
234 			set_opt(sbi, DISCARD);
235 			break;
236 		case Opt_noheap:
237 			set_opt(sbi, NOHEAP);
238 			break;
239 #ifdef CONFIG_F2FS_FS_XATTR
240 		case Opt_nouser_xattr:
241 			clear_opt(sbi, XATTR_USER);
242 			break;
243 		case Opt_inline_xattr:
244 			set_opt(sbi, INLINE_XATTR);
245 			break;
246 #else
247 		case Opt_nouser_xattr:
248 			f2fs_msg(sb, KERN_INFO,
249 				"nouser_xattr options not supported");
250 			break;
251 		case Opt_inline_xattr:
252 			f2fs_msg(sb, KERN_INFO,
253 				"inline_xattr options not supported");
254 			break;
255 #endif
256 #ifdef CONFIG_F2FS_FS_POSIX_ACL
257 		case Opt_noacl:
258 			clear_opt(sbi, POSIX_ACL);
259 			break;
260 #else
261 		case Opt_noacl:
262 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
263 			break;
264 #endif
265 		case Opt_active_logs:
266 			if (args->from && match_int(args, &arg))
267 				return -EINVAL;
268 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
269 				return -EINVAL;
270 			sbi->active_logs = arg;
271 			break;
272 		case Opt_disable_ext_identify:
273 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
274 			break;
275 		default:
276 			f2fs_msg(sb, KERN_ERR,
277 				"Unrecognized mount option \"%s\" or missing value",
278 				p);
279 			return -EINVAL;
280 		}
281 	}
282 	return 0;
283 }
284 
285 static struct inode *f2fs_alloc_inode(struct super_block *sb)
286 {
287 	struct f2fs_inode_info *fi;
288 
289 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_NOFS | __GFP_ZERO);
290 	if (!fi)
291 		return NULL;
292 
293 	init_once((void *) fi);
294 
295 	/* Initialize f2fs-specific inode info */
296 	fi->vfs_inode.i_version = 1;
297 	atomic_set(&fi->dirty_dents, 0);
298 	fi->i_current_depth = 1;
299 	fi->i_advise = 0;
300 	rwlock_init(&fi->ext.ext_lock);
301 
302 	set_inode_flag(fi, FI_NEW_INODE);
303 
304 	if (test_opt(F2FS_SB(sb), INLINE_XATTR))
305 		set_inode_flag(fi, FI_INLINE_XATTR);
306 
307 	return &fi->vfs_inode;
308 }
309 
310 static int f2fs_drop_inode(struct inode *inode)
311 {
312 	/*
313 	 * This is to avoid a deadlock condition like below.
314 	 * writeback_single_inode(inode)
315 	 *  - f2fs_write_data_page
316 	 *    - f2fs_gc -> iput -> evict
317 	 *       - inode_wait_for_writeback(inode)
318 	 */
319 	if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
320 		return 0;
321 	return generic_drop_inode(inode);
322 }
323 
324 /*
325  * f2fs_dirty_inode() is called from __mark_inode_dirty()
326  *
327  * We should call set_dirty_inode to write the dirty inode through write_inode.
328  */
329 static void f2fs_dirty_inode(struct inode *inode, int flags)
330 {
331 	set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
332 }
333 
334 static void f2fs_i_callback(struct rcu_head *head)
335 {
336 	struct inode *inode = container_of(head, struct inode, i_rcu);
337 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
338 }
339 
340 static void f2fs_destroy_inode(struct inode *inode)
341 {
342 	call_rcu(&inode->i_rcu, f2fs_i_callback);
343 }
344 
345 static void f2fs_put_super(struct super_block *sb)
346 {
347 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
348 
349 	if (sbi->s_proc) {
350 		remove_proc_entry("segment_info", sbi->s_proc);
351 		remove_proc_entry(sb->s_id, f2fs_proc_root);
352 	}
353 	kobject_del(&sbi->s_kobj);
354 
355 	f2fs_destroy_stats(sbi);
356 	stop_gc_thread(sbi);
357 
358 	write_checkpoint(sbi, true);
359 
360 	iput(sbi->node_inode);
361 	iput(sbi->meta_inode);
362 
363 	/* destroy f2fs internal modules */
364 	destroy_node_manager(sbi);
365 	destroy_segment_manager(sbi);
366 
367 	kfree(sbi->ckpt);
368 	kobject_put(&sbi->s_kobj);
369 	wait_for_completion(&sbi->s_kobj_unregister);
370 
371 	sb->s_fs_info = NULL;
372 	brelse(sbi->raw_super_buf);
373 	kfree(sbi);
374 }
375 
376 int f2fs_sync_fs(struct super_block *sb, int sync)
377 {
378 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
379 
380 	trace_f2fs_sync_fs(sb, sync);
381 
382 	if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
383 		return 0;
384 
385 	if (sync) {
386 		mutex_lock(&sbi->gc_mutex);
387 		write_checkpoint(sbi, false);
388 		mutex_unlock(&sbi->gc_mutex);
389 	} else {
390 		f2fs_balance_fs(sbi);
391 	}
392 
393 	return 0;
394 }
395 
396 static int f2fs_freeze(struct super_block *sb)
397 {
398 	int err;
399 
400 	if (f2fs_readonly(sb))
401 		return 0;
402 
403 	err = f2fs_sync_fs(sb, 1);
404 	return err;
405 }
406 
407 static int f2fs_unfreeze(struct super_block *sb)
408 {
409 	return 0;
410 }
411 
412 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
413 {
414 	struct super_block *sb = dentry->d_sb;
415 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
416 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
417 	block_t total_count, user_block_count, start_count, ovp_count;
418 
419 	total_count = le64_to_cpu(sbi->raw_super->block_count);
420 	user_block_count = sbi->user_block_count;
421 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
422 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
423 	buf->f_type = F2FS_SUPER_MAGIC;
424 	buf->f_bsize = sbi->blocksize;
425 
426 	buf->f_blocks = total_count - start_count;
427 	buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
428 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
429 
430 	buf->f_files = sbi->total_node_count;
431 	buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
432 
433 	buf->f_namelen = F2FS_NAME_LEN;
434 	buf->f_fsid.val[0] = (u32)id;
435 	buf->f_fsid.val[1] = (u32)(id >> 32);
436 
437 	return 0;
438 }
439 
440 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
441 {
442 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
443 
444 	if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
445 		seq_printf(seq, ",background_gc=%s", "on");
446 	else
447 		seq_printf(seq, ",background_gc=%s", "off");
448 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
449 		seq_puts(seq, ",disable_roll_forward");
450 	if (test_opt(sbi, DISCARD))
451 		seq_puts(seq, ",discard");
452 	if (test_opt(sbi, NOHEAP))
453 		seq_puts(seq, ",no_heap_alloc");
454 #ifdef CONFIG_F2FS_FS_XATTR
455 	if (test_opt(sbi, XATTR_USER))
456 		seq_puts(seq, ",user_xattr");
457 	else
458 		seq_puts(seq, ",nouser_xattr");
459 	if (test_opt(sbi, INLINE_XATTR))
460 		seq_puts(seq, ",inline_xattr");
461 #endif
462 #ifdef CONFIG_F2FS_FS_POSIX_ACL
463 	if (test_opt(sbi, POSIX_ACL))
464 		seq_puts(seq, ",acl");
465 	else
466 		seq_puts(seq, ",noacl");
467 #endif
468 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
469 		seq_puts(seq, ",disable_ext_identify");
470 
471 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
472 
473 	return 0;
474 }
475 
476 static int segment_info_seq_show(struct seq_file *seq, void *offset)
477 {
478 	struct super_block *sb = seq->private;
479 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
480 	unsigned int total_segs = le32_to_cpu(sbi->raw_super->segment_count_main);
481 	int i;
482 
483 	for (i = 0; i < total_segs; i++) {
484 		seq_printf(seq, "%u", get_valid_blocks(sbi, i, 1));
485 		if (i != 0 && (i % 10) == 0)
486 			seq_puts(seq, "\n");
487 		else
488 			seq_puts(seq, " ");
489 	}
490 	return 0;
491 }
492 
493 static int segment_info_open_fs(struct inode *inode, struct file *file)
494 {
495 	return single_open(file, segment_info_seq_show, PDE_DATA(inode));
496 }
497 
498 static const struct file_operations f2fs_seq_segment_info_fops = {
499 	.owner = THIS_MODULE,
500 	.open = segment_info_open_fs,
501 	.read = seq_read,
502 	.llseek = seq_lseek,
503 	.release = single_release,
504 };
505 
506 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
507 {
508 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
509 	struct f2fs_mount_info org_mount_opt;
510 	int err, active_logs;
511 
512 	/*
513 	 * Save the old mount options in case we
514 	 * need to restore them.
515 	 */
516 	org_mount_opt = sbi->mount_opt;
517 	active_logs = sbi->active_logs;
518 
519 	/* parse mount options */
520 	err = parse_options(sb, data);
521 	if (err)
522 		goto restore_opts;
523 
524 	/*
525 	 * Previous and new state of filesystem is RO,
526 	 * so no point in checking GC conditions.
527 	 */
528 	if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
529 		goto skip;
530 
531 	/*
532 	 * We stop the GC thread if FS is mounted as RO
533 	 * or if background_gc = off is passed in mount
534 	 * option. Also sync the filesystem.
535 	 */
536 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
537 		if (sbi->gc_thread) {
538 			stop_gc_thread(sbi);
539 			f2fs_sync_fs(sb, 1);
540 		}
541 	} else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
542 		err = start_gc_thread(sbi);
543 		if (err)
544 			goto restore_opts;
545 	}
546 skip:
547 	/* Update the POSIXACL Flag */
548 	 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
549 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
550 	return 0;
551 
552 restore_opts:
553 	sbi->mount_opt = org_mount_opt;
554 	sbi->active_logs = active_logs;
555 	return err;
556 }
557 
558 static struct super_operations f2fs_sops = {
559 	.alloc_inode	= f2fs_alloc_inode,
560 	.drop_inode	= f2fs_drop_inode,
561 	.destroy_inode	= f2fs_destroy_inode,
562 	.write_inode	= f2fs_write_inode,
563 	.dirty_inode	= f2fs_dirty_inode,
564 	.show_options	= f2fs_show_options,
565 	.evict_inode	= f2fs_evict_inode,
566 	.put_super	= f2fs_put_super,
567 	.sync_fs	= f2fs_sync_fs,
568 	.freeze_fs	= f2fs_freeze,
569 	.unfreeze_fs	= f2fs_unfreeze,
570 	.statfs		= f2fs_statfs,
571 	.remount_fs	= f2fs_remount,
572 };
573 
574 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
575 		u64 ino, u32 generation)
576 {
577 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
578 	struct inode *inode;
579 
580 	if (ino < F2FS_ROOT_INO(sbi))
581 		return ERR_PTR(-ESTALE);
582 
583 	/*
584 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
585 	 * However f2fs_iget currently does appropriate checks to handle stale
586 	 * inodes so everything is OK.
587 	 */
588 	inode = f2fs_iget(sb, ino);
589 	if (IS_ERR(inode))
590 		return ERR_CAST(inode);
591 	if (generation && inode->i_generation != generation) {
592 		/* we didn't find the right inode.. */
593 		iput(inode);
594 		return ERR_PTR(-ESTALE);
595 	}
596 	return inode;
597 }
598 
599 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
600 		int fh_len, int fh_type)
601 {
602 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
603 				    f2fs_nfs_get_inode);
604 }
605 
606 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
607 		int fh_len, int fh_type)
608 {
609 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
610 				    f2fs_nfs_get_inode);
611 }
612 
613 static const struct export_operations f2fs_export_ops = {
614 	.fh_to_dentry = f2fs_fh_to_dentry,
615 	.fh_to_parent = f2fs_fh_to_parent,
616 	.get_parent = f2fs_get_parent,
617 };
618 
619 static loff_t max_file_size(unsigned bits)
620 {
621 	loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
622 	loff_t leaf_count = ADDRS_PER_BLOCK;
623 
624 	/* two direct node blocks */
625 	result += (leaf_count * 2);
626 
627 	/* two indirect node blocks */
628 	leaf_count *= NIDS_PER_BLOCK;
629 	result += (leaf_count * 2);
630 
631 	/* one double indirect node block */
632 	leaf_count *= NIDS_PER_BLOCK;
633 	result += leaf_count;
634 
635 	result <<= bits;
636 	return result;
637 }
638 
639 static int sanity_check_raw_super(struct super_block *sb,
640 			struct f2fs_super_block *raw_super)
641 {
642 	unsigned int blocksize;
643 
644 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
645 		f2fs_msg(sb, KERN_INFO,
646 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
647 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
648 		return 1;
649 	}
650 
651 	/* Currently, support only 4KB page cache size */
652 	if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
653 		f2fs_msg(sb, KERN_INFO,
654 			"Invalid page_cache_size (%lu), supports only 4KB\n",
655 			PAGE_CACHE_SIZE);
656 		return 1;
657 	}
658 
659 	/* Currently, support only 4KB block size */
660 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
661 	if (blocksize != F2FS_BLKSIZE) {
662 		f2fs_msg(sb, KERN_INFO,
663 			"Invalid blocksize (%u), supports only 4KB\n",
664 			blocksize);
665 		return 1;
666 	}
667 
668 	if (le32_to_cpu(raw_super->log_sectorsize) !=
669 					F2FS_LOG_SECTOR_SIZE) {
670 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
671 		return 1;
672 	}
673 	if (le32_to_cpu(raw_super->log_sectors_per_block) !=
674 					F2FS_LOG_SECTORS_PER_BLOCK) {
675 		f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
676 		return 1;
677 	}
678 	return 0;
679 }
680 
681 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
682 {
683 	unsigned int total, fsmeta;
684 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
685 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
686 
687 	total = le32_to_cpu(raw_super->segment_count);
688 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
689 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
690 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
691 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
692 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
693 
694 	if (fsmeta >= total)
695 		return 1;
696 
697 	if (is_set_ckpt_flags(ckpt, CP_ERROR_FLAG)) {
698 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
699 		return 1;
700 	}
701 	return 0;
702 }
703 
704 static void init_sb_info(struct f2fs_sb_info *sbi)
705 {
706 	struct f2fs_super_block *raw_super = sbi->raw_super;
707 	int i;
708 
709 	sbi->log_sectors_per_block =
710 		le32_to_cpu(raw_super->log_sectors_per_block);
711 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
712 	sbi->blocksize = 1 << sbi->log_blocksize;
713 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
714 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
715 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
716 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
717 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
718 	sbi->total_node_count =
719 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
720 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
721 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
722 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
723 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
724 	sbi->cur_victim_sec = NULL_SECNO;
725 
726 	for (i = 0; i < NR_COUNT_TYPE; i++)
727 		atomic_set(&sbi->nr_pages[i], 0);
728 }
729 
730 static int validate_superblock(struct super_block *sb,
731 		struct f2fs_super_block **raw_super,
732 		struct buffer_head **raw_super_buf, sector_t block)
733 {
734 	const char *super = (block == 0 ? "first" : "second");
735 
736 	/* read f2fs raw super block */
737 	*raw_super_buf = sb_bread(sb, block);
738 	if (!*raw_super_buf) {
739 		f2fs_msg(sb, KERN_ERR, "unable to read %s superblock",
740 				super);
741 		return -EIO;
742 	}
743 
744 	*raw_super = (struct f2fs_super_block *)
745 		((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
746 
747 	/* sanity checking of raw super */
748 	if (!sanity_check_raw_super(sb, *raw_super))
749 		return 0;
750 
751 	f2fs_msg(sb, KERN_ERR, "Can't find a valid F2FS filesystem "
752 				"in %s superblock", super);
753 	return -EINVAL;
754 }
755 
756 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
757 {
758 	struct f2fs_sb_info *sbi;
759 	struct f2fs_super_block *raw_super;
760 	struct buffer_head *raw_super_buf;
761 	struct inode *root;
762 	long err = -EINVAL;
763 	int i;
764 
765 	/* allocate memory for f2fs-specific super block info */
766 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
767 	if (!sbi)
768 		return -ENOMEM;
769 
770 	/* set a block size */
771 	if (!sb_set_blocksize(sb, F2FS_BLKSIZE)) {
772 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
773 		goto free_sbi;
774 	}
775 
776 	err = validate_superblock(sb, &raw_super, &raw_super_buf, 0);
777 	if (err) {
778 		brelse(raw_super_buf);
779 		/* check secondary superblock when primary failed */
780 		err = validate_superblock(sb, &raw_super, &raw_super_buf, 1);
781 		if (err)
782 			goto free_sb_buf;
783 	}
784 	sb->s_fs_info = sbi;
785 	/* init some FS parameters */
786 	sbi->active_logs = NR_CURSEG_TYPE;
787 
788 	set_opt(sbi, BG_GC);
789 
790 #ifdef CONFIG_F2FS_FS_XATTR
791 	set_opt(sbi, XATTR_USER);
792 #endif
793 #ifdef CONFIG_F2FS_FS_POSIX_ACL
794 	set_opt(sbi, POSIX_ACL);
795 #endif
796 	/* parse mount options */
797 	err = parse_options(sb, (char *)data);
798 	if (err)
799 		goto free_sb_buf;
800 
801 	sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
802 	sb->s_max_links = F2FS_LINK_MAX;
803 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
804 
805 	sb->s_op = &f2fs_sops;
806 	sb->s_xattr = f2fs_xattr_handlers;
807 	sb->s_export_op = &f2fs_export_ops;
808 	sb->s_magic = F2FS_SUPER_MAGIC;
809 	sb->s_time_gran = 1;
810 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
811 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
812 	memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
813 
814 	/* init f2fs-specific super block info */
815 	sbi->sb = sb;
816 	sbi->raw_super = raw_super;
817 	sbi->raw_super_buf = raw_super_buf;
818 	mutex_init(&sbi->gc_mutex);
819 	mutex_init(&sbi->writepages);
820 	mutex_init(&sbi->cp_mutex);
821 	for (i = 0; i < NR_GLOBAL_LOCKS; i++)
822 		mutex_init(&sbi->fs_lock[i]);
823 	mutex_init(&sbi->node_write);
824 	sbi->por_doing = 0;
825 	spin_lock_init(&sbi->stat_lock);
826 	init_rwsem(&sbi->bio_sem);
827 	init_sb_info(sbi);
828 
829 	/* get an inode for meta space */
830 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
831 	if (IS_ERR(sbi->meta_inode)) {
832 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
833 		err = PTR_ERR(sbi->meta_inode);
834 		goto free_sb_buf;
835 	}
836 
837 	err = get_valid_checkpoint(sbi);
838 	if (err) {
839 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
840 		goto free_meta_inode;
841 	}
842 
843 	/* sanity checking of checkpoint */
844 	err = -EINVAL;
845 	if (sanity_check_ckpt(sbi)) {
846 		f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
847 		goto free_cp;
848 	}
849 
850 	sbi->total_valid_node_count =
851 				le32_to_cpu(sbi->ckpt->valid_node_count);
852 	sbi->total_valid_inode_count =
853 				le32_to_cpu(sbi->ckpt->valid_inode_count);
854 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
855 	sbi->total_valid_block_count =
856 				le64_to_cpu(sbi->ckpt->valid_block_count);
857 	sbi->last_valid_block_count = sbi->total_valid_block_count;
858 	sbi->alloc_valid_block_count = 0;
859 	INIT_LIST_HEAD(&sbi->dir_inode_list);
860 	spin_lock_init(&sbi->dir_inode_lock);
861 
862 	init_orphan_info(sbi);
863 
864 	/* setup f2fs internal modules */
865 	err = build_segment_manager(sbi);
866 	if (err) {
867 		f2fs_msg(sb, KERN_ERR,
868 			"Failed to initialize F2FS segment manager");
869 		goto free_sm;
870 	}
871 	err = build_node_manager(sbi);
872 	if (err) {
873 		f2fs_msg(sb, KERN_ERR,
874 			"Failed to initialize F2FS node manager");
875 		goto free_nm;
876 	}
877 
878 	build_gc_manager(sbi);
879 
880 	/* get an inode for node space */
881 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
882 	if (IS_ERR(sbi->node_inode)) {
883 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
884 		err = PTR_ERR(sbi->node_inode);
885 		goto free_nm;
886 	}
887 
888 	/* if there are nt orphan nodes free them */
889 	err = -EINVAL;
890 	if (recover_orphan_inodes(sbi))
891 		goto free_node_inode;
892 
893 	/* read root inode and dentry */
894 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
895 	if (IS_ERR(root)) {
896 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
897 		err = PTR_ERR(root);
898 		goto free_node_inode;
899 	}
900 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size)
901 		goto free_root_inode;
902 
903 	sb->s_root = d_make_root(root); /* allocate root dentry */
904 	if (!sb->s_root) {
905 		err = -ENOMEM;
906 		goto free_root_inode;
907 	}
908 
909 	/* recover fsynced data */
910 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
911 		err = recover_fsync_data(sbi);
912 		if (err)
913 			f2fs_msg(sb, KERN_ERR,
914 				"Cannot recover all fsync data errno=%ld", err);
915 	}
916 
917 	/*
918 	 * If filesystem is not mounted as read-only then
919 	 * do start the gc_thread.
920 	 */
921 	if (!(sb->s_flags & MS_RDONLY)) {
922 		/* After POR, we can run background GC thread.*/
923 		err = start_gc_thread(sbi);
924 		if (err)
925 			goto fail;
926 	}
927 
928 	err = f2fs_build_stats(sbi);
929 	if (err)
930 		goto fail;
931 
932 	if (f2fs_proc_root)
933 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
934 
935 	if (sbi->s_proc)
936 		proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
937 				 &f2fs_seq_segment_info_fops, sb);
938 
939 	if (test_opt(sbi, DISCARD)) {
940 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
941 		if (!blk_queue_discard(q))
942 			f2fs_msg(sb, KERN_WARNING,
943 					"mounting with \"discard\" option, but "
944 					"the device does not support discard");
945 	}
946 
947 	sbi->s_kobj.kset = f2fs_kset;
948 	init_completion(&sbi->s_kobj_unregister);
949 	err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
950 							"%s", sb->s_id);
951 	if (err)
952 		goto fail;
953 
954 	return 0;
955 fail:
956 	stop_gc_thread(sbi);
957 free_root_inode:
958 	dput(sb->s_root);
959 	sb->s_root = NULL;
960 free_node_inode:
961 	iput(sbi->node_inode);
962 free_nm:
963 	destroy_node_manager(sbi);
964 free_sm:
965 	destroy_segment_manager(sbi);
966 free_cp:
967 	kfree(sbi->ckpt);
968 free_meta_inode:
969 	make_bad_inode(sbi->meta_inode);
970 	iput(sbi->meta_inode);
971 free_sb_buf:
972 	brelse(raw_super_buf);
973 free_sbi:
974 	kfree(sbi);
975 	return err;
976 }
977 
978 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
979 			const char *dev_name, void *data)
980 {
981 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
982 }
983 
984 static struct file_system_type f2fs_fs_type = {
985 	.owner		= THIS_MODULE,
986 	.name		= "f2fs",
987 	.mount		= f2fs_mount,
988 	.kill_sb	= kill_block_super,
989 	.fs_flags	= FS_REQUIRES_DEV,
990 };
991 MODULE_ALIAS_FS("f2fs");
992 
993 static int __init init_inodecache(void)
994 {
995 	f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
996 			sizeof(struct f2fs_inode_info), NULL);
997 	if (f2fs_inode_cachep == NULL)
998 		return -ENOMEM;
999 	return 0;
1000 }
1001 
1002 static void destroy_inodecache(void)
1003 {
1004 	/*
1005 	 * Make sure all delayed rcu free inodes are flushed before we
1006 	 * destroy cache.
1007 	 */
1008 	rcu_barrier();
1009 	kmem_cache_destroy(f2fs_inode_cachep);
1010 }
1011 
1012 static int __init init_f2fs_fs(void)
1013 {
1014 	int err;
1015 
1016 	err = init_inodecache();
1017 	if (err)
1018 		goto fail;
1019 	err = create_node_manager_caches();
1020 	if (err)
1021 		goto free_inodecache;
1022 	err = create_gc_caches();
1023 	if (err)
1024 		goto free_node_manager_caches;
1025 	err = create_checkpoint_caches();
1026 	if (err)
1027 		goto free_gc_caches;
1028 	f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1029 	if (!f2fs_kset) {
1030 		err = -ENOMEM;
1031 		goto free_checkpoint_caches;
1032 	}
1033 	err = register_filesystem(&f2fs_fs_type);
1034 	if (err)
1035 		goto free_kset;
1036 	f2fs_create_root_stats();
1037 	f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1038 	return 0;
1039 
1040 free_kset:
1041 	kset_unregister(f2fs_kset);
1042 free_checkpoint_caches:
1043 	destroy_checkpoint_caches();
1044 free_gc_caches:
1045 	destroy_gc_caches();
1046 free_node_manager_caches:
1047 	destroy_node_manager_caches();
1048 free_inodecache:
1049 	destroy_inodecache();
1050 fail:
1051 	return err;
1052 }
1053 
1054 static void __exit exit_f2fs_fs(void)
1055 {
1056 	remove_proc_entry("fs/f2fs", NULL);
1057 	f2fs_destroy_root_stats();
1058 	unregister_filesystem(&f2fs_fs_type);
1059 	destroy_checkpoint_caches();
1060 	destroy_gc_caches();
1061 	destroy_node_manager_caches();
1062 	destroy_inodecache();
1063 	kset_unregister(f2fs_kset);
1064 }
1065 
1066 module_init(init_f2fs_fs)
1067 module_exit(exit_f2fs_fs)
1068 
1069 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1070 MODULE_DESCRIPTION("Flash Friendly File System");
1071 MODULE_LICENSE("GPL");
1072