xref: /openbmc/linux/fs/f2fs/super.c (revision 930beb5a)
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_user_xattr,
47 	Opt_nouser_xattr,
48 	Opt_acl,
49 	Opt_noacl,
50 	Opt_active_logs,
51 	Opt_disable_ext_identify,
52 	Opt_inline_xattr,
53 	Opt_err,
54 };
55 
56 static match_table_t f2fs_tokens = {
57 	{Opt_gc_background, "background_gc=%s"},
58 	{Opt_disable_roll_forward, "disable_roll_forward"},
59 	{Opt_discard, "discard"},
60 	{Opt_noheap, "no_heap"},
61 	{Opt_user_xattr, "user_xattr"},
62 	{Opt_nouser_xattr, "nouser_xattr"},
63 	{Opt_acl, "acl"},
64 	{Opt_noacl, "noacl"},
65 	{Opt_active_logs, "active_logs=%u"},
66 	{Opt_disable_ext_identify, "disable_ext_identify"},
67 	{Opt_inline_xattr, "inline_xattr"},
68 	{Opt_err, NULL},
69 };
70 
71 /* Sysfs support for f2fs */
72 enum {
73 	GC_THREAD,	/* struct f2fs_gc_thread */
74 	SM_INFO,	/* struct f2fs_sm_info */
75 };
76 
77 struct f2fs_attr {
78 	struct attribute attr;
79 	ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
80 	ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
81 			 const char *, size_t);
82 	int struct_type;
83 	int offset;
84 };
85 
86 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
87 {
88 	if (struct_type == GC_THREAD)
89 		return (unsigned char *)sbi->gc_thread;
90 	else if (struct_type == SM_INFO)
91 		return (unsigned char *)SM_I(sbi);
92 	return NULL;
93 }
94 
95 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
96 			struct f2fs_sb_info *sbi, char *buf)
97 {
98 	unsigned char *ptr = NULL;
99 	unsigned int *ui;
100 
101 	ptr = __struct_ptr(sbi, a->struct_type);
102 	if (!ptr)
103 		return -EINVAL;
104 
105 	ui = (unsigned int *)(ptr + a->offset);
106 
107 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
108 }
109 
110 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
111 			struct f2fs_sb_info *sbi,
112 			const char *buf, size_t count)
113 {
114 	unsigned char *ptr;
115 	unsigned long t;
116 	unsigned int *ui;
117 	ssize_t ret;
118 
119 	ptr = __struct_ptr(sbi, a->struct_type);
120 	if (!ptr)
121 		return -EINVAL;
122 
123 	ui = (unsigned int *)(ptr + a->offset);
124 
125 	ret = kstrtoul(skip_spaces(buf), 0, &t);
126 	if (ret < 0)
127 		return ret;
128 	*ui = t;
129 	return count;
130 }
131 
132 static ssize_t f2fs_attr_show(struct kobject *kobj,
133 				struct attribute *attr, char *buf)
134 {
135 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
136 								s_kobj);
137 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
138 
139 	return a->show ? a->show(a, sbi, buf) : 0;
140 }
141 
142 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
143 						const char *buf, size_t len)
144 {
145 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
146 									s_kobj);
147 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
148 
149 	return a->store ? a->store(a, sbi, buf, len) : 0;
150 }
151 
152 static void f2fs_sb_release(struct kobject *kobj)
153 {
154 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
155 								s_kobj);
156 	complete(&sbi->s_kobj_unregister);
157 }
158 
159 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
160 static struct f2fs_attr f2fs_attr_##_name = {			\
161 	.attr = {.name = __stringify(_name), .mode = _mode },	\
162 	.show	= _show,					\
163 	.store	= _store,					\
164 	.struct_type = _struct_type,				\
165 	.offset = _offset					\
166 }
167 
168 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)	\
169 	F2FS_ATTR_OFFSET(struct_type, name, 0644,		\
170 		f2fs_sbi_show, f2fs_sbi_store,			\
171 		offsetof(struct struct_name, elname))
172 
173 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
174 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
175 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
176 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
177 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
178 
179 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
180 static struct attribute *f2fs_attrs[] = {
181 	ATTR_LIST(gc_min_sleep_time),
182 	ATTR_LIST(gc_max_sleep_time),
183 	ATTR_LIST(gc_no_gc_sleep_time),
184 	ATTR_LIST(gc_idle),
185 	ATTR_LIST(reclaim_segments),
186 	NULL,
187 };
188 
189 static const struct sysfs_ops f2fs_attr_ops = {
190 	.show	= f2fs_attr_show,
191 	.store	= f2fs_attr_store,
192 };
193 
194 static struct kobj_type f2fs_ktype = {
195 	.default_attrs	= f2fs_attrs,
196 	.sysfs_ops	= &f2fs_attr_ops,
197 	.release	= f2fs_sb_release,
198 };
199 
200 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
201 {
202 	struct va_format vaf;
203 	va_list args;
204 
205 	va_start(args, fmt);
206 	vaf.fmt = fmt;
207 	vaf.va = &args;
208 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
209 	va_end(args);
210 }
211 
212 static void init_once(void *foo)
213 {
214 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
215 
216 	inode_init_once(&fi->vfs_inode);
217 }
218 
219 static int parse_options(struct super_block *sb, char *options)
220 {
221 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
222 	substring_t args[MAX_OPT_ARGS];
223 	char *p, *name;
224 	int arg = 0;
225 
226 	if (!options)
227 		return 0;
228 
229 	while ((p = strsep(&options, ",")) != NULL) {
230 		int token;
231 		if (!*p)
232 			continue;
233 		/*
234 		 * Initialize args struct so we know whether arg was
235 		 * found; some options take optional arguments.
236 		 */
237 		args[0].to = args[0].from = NULL;
238 		token = match_token(p, f2fs_tokens, args);
239 
240 		switch (token) {
241 		case Opt_gc_background:
242 			name = match_strdup(&args[0]);
243 
244 			if (!name)
245 				return -ENOMEM;
246 			if (!strncmp(name, "on", 2))
247 				set_opt(sbi, BG_GC);
248 			else if (!strncmp(name, "off", 3))
249 				clear_opt(sbi, BG_GC);
250 			else {
251 				kfree(name);
252 				return -EINVAL;
253 			}
254 			kfree(name);
255 			break;
256 		case Opt_disable_roll_forward:
257 			set_opt(sbi, DISABLE_ROLL_FORWARD);
258 			break;
259 		case Opt_discard:
260 			set_opt(sbi, DISCARD);
261 			break;
262 		case Opt_noheap:
263 			set_opt(sbi, NOHEAP);
264 			break;
265 #ifdef CONFIG_F2FS_FS_XATTR
266 		case Opt_user_xattr:
267 			set_opt(sbi, XATTR_USER);
268 			break;
269 		case Opt_nouser_xattr:
270 			clear_opt(sbi, XATTR_USER);
271 			break;
272 		case Opt_inline_xattr:
273 			set_opt(sbi, INLINE_XATTR);
274 			break;
275 #else
276 		case Opt_user_xattr:
277 			f2fs_msg(sb, KERN_INFO,
278 				"user_xattr options not supported");
279 			break;
280 		case Opt_nouser_xattr:
281 			f2fs_msg(sb, KERN_INFO,
282 				"nouser_xattr options not supported");
283 			break;
284 		case Opt_inline_xattr:
285 			f2fs_msg(sb, KERN_INFO,
286 				"inline_xattr options not supported");
287 			break;
288 #endif
289 #ifdef CONFIG_F2FS_FS_POSIX_ACL
290 		case Opt_acl:
291 			set_opt(sbi, POSIX_ACL);
292 			break;
293 		case Opt_noacl:
294 			clear_opt(sbi, POSIX_ACL);
295 			break;
296 #else
297 		case Opt_acl:
298 			f2fs_msg(sb, KERN_INFO, "acl options not supported");
299 			break;
300 		case Opt_noacl:
301 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
302 			break;
303 #endif
304 		case Opt_active_logs:
305 			if (args->from && match_int(args, &arg))
306 				return -EINVAL;
307 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
308 				return -EINVAL;
309 			sbi->active_logs = arg;
310 			break;
311 		case Opt_disable_ext_identify:
312 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
313 			break;
314 		default:
315 			f2fs_msg(sb, KERN_ERR,
316 				"Unrecognized mount option \"%s\" or missing value",
317 				p);
318 			return -EINVAL;
319 		}
320 	}
321 	return 0;
322 }
323 
324 static struct inode *f2fs_alloc_inode(struct super_block *sb)
325 {
326 	struct f2fs_inode_info *fi;
327 
328 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_NOFS | __GFP_ZERO);
329 	if (!fi)
330 		return NULL;
331 
332 	init_once((void *) fi);
333 
334 	/* Initialize f2fs-specific inode info */
335 	fi->vfs_inode.i_version = 1;
336 	atomic_set(&fi->dirty_dents, 0);
337 	fi->i_current_depth = 1;
338 	fi->i_advise = 0;
339 	rwlock_init(&fi->ext.ext_lock);
340 
341 	set_inode_flag(fi, FI_NEW_INODE);
342 
343 	if (test_opt(F2FS_SB(sb), INLINE_XATTR))
344 		set_inode_flag(fi, FI_INLINE_XATTR);
345 
346 	return &fi->vfs_inode;
347 }
348 
349 static int f2fs_drop_inode(struct inode *inode)
350 {
351 	/*
352 	 * This is to avoid a deadlock condition like below.
353 	 * writeback_single_inode(inode)
354 	 *  - f2fs_write_data_page
355 	 *    - f2fs_gc -> iput -> evict
356 	 *       - inode_wait_for_writeback(inode)
357 	 */
358 	if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
359 		return 0;
360 	return generic_drop_inode(inode);
361 }
362 
363 /*
364  * f2fs_dirty_inode() is called from __mark_inode_dirty()
365  *
366  * We should call set_dirty_inode to write the dirty inode through write_inode.
367  */
368 static void f2fs_dirty_inode(struct inode *inode, int flags)
369 {
370 	set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
371 }
372 
373 static void f2fs_i_callback(struct rcu_head *head)
374 {
375 	struct inode *inode = container_of(head, struct inode, i_rcu);
376 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
377 }
378 
379 static void f2fs_destroy_inode(struct inode *inode)
380 {
381 	call_rcu(&inode->i_rcu, f2fs_i_callback);
382 }
383 
384 static void f2fs_put_super(struct super_block *sb)
385 {
386 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
387 
388 	if (sbi->s_proc) {
389 		remove_proc_entry("segment_info", sbi->s_proc);
390 		remove_proc_entry(sb->s_id, f2fs_proc_root);
391 	}
392 	kobject_del(&sbi->s_kobj);
393 
394 	f2fs_destroy_stats(sbi);
395 	stop_gc_thread(sbi);
396 
397 	/* We don't need to do checkpoint when it's clean */
398 	if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
399 		write_checkpoint(sbi, true);
400 
401 	iput(sbi->node_inode);
402 	iput(sbi->meta_inode);
403 
404 	/* destroy f2fs internal modules */
405 	destroy_node_manager(sbi);
406 	destroy_segment_manager(sbi);
407 
408 	kfree(sbi->ckpt);
409 	kobject_put(&sbi->s_kobj);
410 	wait_for_completion(&sbi->s_kobj_unregister);
411 
412 	sb->s_fs_info = NULL;
413 	brelse(sbi->raw_super_buf);
414 	kfree(sbi);
415 }
416 
417 int f2fs_sync_fs(struct super_block *sb, int sync)
418 {
419 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
420 
421 	trace_f2fs_sync_fs(sb, sync);
422 
423 	if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
424 		return 0;
425 
426 	if (sync) {
427 		mutex_lock(&sbi->gc_mutex);
428 		write_checkpoint(sbi, false);
429 		mutex_unlock(&sbi->gc_mutex);
430 	} else {
431 		f2fs_balance_fs(sbi);
432 	}
433 
434 	return 0;
435 }
436 
437 static int f2fs_freeze(struct super_block *sb)
438 {
439 	int err;
440 
441 	if (f2fs_readonly(sb))
442 		return 0;
443 
444 	err = f2fs_sync_fs(sb, 1);
445 	return err;
446 }
447 
448 static int f2fs_unfreeze(struct super_block *sb)
449 {
450 	return 0;
451 }
452 
453 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
454 {
455 	struct super_block *sb = dentry->d_sb;
456 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
457 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
458 	block_t total_count, user_block_count, start_count, ovp_count;
459 
460 	total_count = le64_to_cpu(sbi->raw_super->block_count);
461 	user_block_count = sbi->user_block_count;
462 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
463 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
464 	buf->f_type = F2FS_SUPER_MAGIC;
465 	buf->f_bsize = sbi->blocksize;
466 
467 	buf->f_blocks = total_count - start_count;
468 	buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
469 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
470 
471 	buf->f_files = sbi->total_node_count;
472 	buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
473 
474 	buf->f_namelen = F2FS_NAME_LEN;
475 	buf->f_fsid.val[0] = (u32)id;
476 	buf->f_fsid.val[1] = (u32)(id >> 32);
477 
478 	return 0;
479 }
480 
481 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
482 {
483 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
484 
485 	if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
486 		seq_printf(seq, ",background_gc=%s", "on");
487 	else
488 		seq_printf(seq, ",background_gc=%s", "off");
489 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
490 		seq_puts(seq, ",disable_roll_forward");
491 	if (test_opt(sbi, DISCARD))
492 		seq_puts(seq, ",discard");
493 	if (test_opt(sbi, NOHEAP))
494 		seq_puts(seq, ",no_heap_alloc");
495 #ifdef CONFIG_F2FS_FS_XATTR
496 	if (test_opt(sbi, XATTR_USER))
497 		seq_puts(seq, ",user_xattr");
498 	else
499 		seq_puts(seq, ",nouser_xattr");
500 	if (test_opt(sbi, INLINE_XATTR))
501 		seq_puts(seq, ",inline_xattr");
502 #endif
503 #ifdef CONFIG_F2FS_FS_POSIX_ACL
504 	if (test_opt(sbi, POSIX_ACL))
505 		seq_puts(seq, ",acl");
506 	else
507 		seq_puts(seq, ",noacl");
508 #endif
509 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
510 		seq_puts(seq, ",disable_ext_identify");
511 
512 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
513 
514 	return 0;
515 }
516 
517 static int segment_info_seq_show(struct seq_file *seq, void *offset)
518 {
519 	struct super_block *sb = seq->private;
520 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
521 	unsigned int total_segs = le32_to_cpu(sbi->raw_super->segment_count_main);
522 	int i;
523 
524 	for (i = 0; i < total_segs; i++) {
525 		seq_printf(seq, "%u", get_valid_blocks(sbi, i, 1));
526 		if (i != 0 && (i % 10) == 0)
527 			seq_puts(seq, "\n");
528 		else
529 			seq_puts(seq, " ");
530 	}
531 	return 0;
532 }
533 
534 static int segment_info_open_fs(struct inode *inode, struct file *file)
535 {
536 	return single_open(file, segment_info_seq_show, PDE_DATA(inode));
537 }
538 
539 static const struct file_operations f2fs_seq_segment_info_fops = {
540 	.owner = THIS_MODULE,
541 	.open = segment_info_open_fs,
542 	.read = seq_read,
543 	.llseek = seq_lseek,
544 	.release = single_release,
545 };
546 
547 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
548 {
549 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
550 	struct f2fs_mount_info org_mount_opt;
551 	int err, active_logs;
552 
553 	/*
554 	 * Save the old mount options in case we
555 	 * need to restore them.
556 	 */
557 	org_mount_opt = sbi->mount_opt;
558 	active_logs = sbi->active_logs;
559 
560 	/* parse mount options */
561 	err = parse_options(sb, data);
562 	if (err)
563 		goto restore_opts;
564 
565 	/*
566 	 * Previous and new state of filesystem is RO,
567 	 * so no point in checking GC conditions.
568 	 */
569 	if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
570 		goto skip;
571 
572 	/*
573 	 * We stop the GC thread if FS is mounted as RO
574 	 * or if background_gc = off is passed in mount
575 	 * option. Also sync the filesystem.
576 	 */
577 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
578 		if (sbi->gc_thread) {
579 			stop_gc_thread(sbi);
580 			f2fs_sync_fs(sb, 1);
581 		}
582 	} else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
583 		err = start_gc_thread(sbi);
584 		if (err)
585 			goto restore_opts;
586 	}
587 skip:
588 	/* Update the POSIXACL Flag */
589 	 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
590 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
591 	return 0;
592 
593 restore_opts:
594 	sbi->mount_opt = org_mount_opt;
595 	sbi->active_logs = active_logs;
596 	return err;
597 }
598 
599 static struct super_operations f2fs_sops = {
600 	.alloc_inode	= f2fs_alloc_inode,
601 	.drop_inode	= f2fs_drop_inode,
602 	.destroy_inode	= f2fs_destroy_inode,
603 	.write_inode	= f2fs_write_inode,
604 	.dirty_inode	= f2fs_dirty_inode,
605 	.show_options	= f2fs_show_options,
606 	.evict_inode	= f2fs_evict_inode,
607 	.put_super	= f2fs_put_super,
608 	.sync_fs	= f2fs_sync_fs,
609 	.freeze_fs	= f2fs_freeze,
610 	.unfreeze_fs	= f2fs_unfreeze,
611 	.statfs		= f2fs_statfs,
612 	.remount_fs	= f2fs_remount,
613 };
614 
615 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
616 		u64 ino, u32 generation)
617 {
618 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
619 	struct inode *inode;
620 
621 	if (ino < F2FS_ROOT_INO(sbi))
622 		return ERR_PTR(-ESTALE);
623 
624 	/*
625 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
626 	 * However f2fs_iget currently does appropriate checks to handle stale
627 	 * inodes so everything is OK.
628 	 */
629 	inode = f2fs_iget(sb, ino);
630 	if (IS_ERR(inode))
631 		return ERR_CAST(inode);
632 	if (generation && inode->i_generation != generation) {
633 		/* we didn't find the right inode.. */
634 		iput(inode);
635 		return ERR_PTR(-ESTALE);
636 	}
637 	return inode;
638 }
639 
640 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
641 		int fh_len, int fh_type)
642 {
643 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
644 				    f2fs_nfs_get_inode);
645 }
646 
647 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
648 		int fh_len, int fh_type)
649 {
650 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
651 				    f2fs_nfs_get_inode);
652 }
653 
654 static const struct export_operations f2fs_export_ops = {
655 	.fh_to_dentry = f2fs_fh_to_dentry,
656 	.fh_to_parent = f2fs_fh_to_parent,
657 	.get_parent = f2fs_get_parent,
658 };
659 
660 static loff_t max_file_size(unsigned bits)
661 {
662 	loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
663 	loff_t leaf_count = ADDRS_PER_BLOCK;
664 
665 	/* two direct node blocks */
666 	result += (leaf_count * 2);
667 
668 	/* two indirect node blocks */
669 	leaf_count *= NIDS_PER_BLOCK;
670 	result += (leaf_count * 2);
671 
672 	/* one double indirect node block */
673 	leaf_count *= NIDS_PER_BLOCK;
674 	result += leaf_count;
675 
676 	result <<= bits;
677 	return result;
678 }
679 
680 static int sanity_check_raw_super(struct super_block *sb,
681 			struct f2fs_super_block *raw_super)
682 {
683 	unsigned int blocksize;
684 
685 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
686 		f2fs_msg(sb, KERN_INFO,
687 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
688 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
689 		return 1;
690 	}
691 
692 	/* Currently, support only 4KB page cache size */
693 	if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
694 		f2fs_msg(sb, KERN_INFO,
695 			"Invalid page_cache_size (%lu), supports only 4KB\n",
696 			PAGE_CACHE_SIZE);
697 		return 1;
698 	}
699 
700 	/* Currently, support only 4KB block size */
701 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
702 	if (blocksize != F2FS_BLKSIZE) {
703 		f2fs_msg(sb, KERN_INFO,
704 			"Invalid blocksize (%u), supports only 4KB\n",
705 			blocksize);
706 		return 1;
707 	}
708 
709 	if (le32_to_cpu(raw_super->log_sectorsize) !=
710 					F2FS_LOG_SECTOR_SIZE) {
711 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
712 		return 1;
713 	}
714 	if (le32_to_cpu(raw_super->log_sectors_per_block) !=
715 					F2FS_LOG_SECTORS_PER_BLOCK) {
716 		f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
717 		return 1;
718 	}
719 	return 0;
720 }
721 
722 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
723 {
724 	unsigned int total, fsmeta;
725 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
726 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
727 
728 	total = le32_to_cpu(raw_super->segment_count);
729 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
730 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
731 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
732 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
733 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
734 
735 	if (fsmeta >= total)
736 		return 1;
737 
738 	if (is_set_ckpt_flags(ckpt, CP_ERROR_FLAG)) {
739 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
740 		return 1;
741 	}
742 	return 0;
743 }
744 
745 static void init_sb_info(struct f2fs_sb_info *sbi)
746 {
747 	struct f2fs_super_block *raw_super = sbi->raw_super;
748 	int i;
749 
750 	sbi->log_sectors_per_block =
751 		le32_to_cpu(raw_super->log_sectors_per_block);
752 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
753 	sbi->blocksize = 1 << sbi->log_blocksize;
754 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
755 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
756 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
757 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
758 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
759 	sbi->total_node_count =
760 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
761 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
762 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
763 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
764 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
765 	sbi->cur_victim_sec = NULL_SECNO;
766 
767 	for (i = 0; i < NR_COUNT_TYPE; i++)
768 		atomic_set(&sbi->nr_pages[i], 0);
769 }
770 
771 /*
772  * Read f2fs raw super block.
773  * Because we have two copies of super block, so read the first one at first,
774  * if the first one is invalid, move to read the second one.
775  */
776 static int read_raw_super_block(struct super_block *sb,
777 			struct f2fs_super_block **raw_super,
778 			struct buffer_head **raw_super_buf)
779 {
780 	int block = 0;
781 
782 retry:
783 	*raw_super_buf = sb_bread(sb, block);
784 	if (!*raw_super_buf) {
785 		f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
786 				block + 1);
787 		if (block == 0) {
788 			block++;
789 			goto retry;
790 		} else {
791 			return -EIO;
792 		}
793 	}
794 
795 	*raw_super = (struct f2fs_super_block *)
796 		((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
797 
798 	/* sanity checking of raw super */
799 	if (sanity_check_raw_super(sb, *raw_super)) {
800 		brelse(*raw_super_buf);
801 		f2fs_msg(sb, KERN_ERR, "Can't find a valid F2FS filesystem "
802 				"in %dth superblock", block + 1);
803 		if(block == 0) {
804 			block++;
805 			goto retry;
806 		} else {
807 			return -EINVAL;
808 		}
809 	}
810 
811 	return 0;
812 }
813 
814 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
815 {
816 	struct f2fs_sb_info *sbi;
817 	struct f2fs_super_block *raw_super;
818 	struct buffer_head *raw_super_buf;
819 	struct inode *root;
820 	long err = -EINVAL;
821 
822 	/* allocate memory for f2fs-specific super block info */
823 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
824 	if (!sbi)
825 		return -ENOMEM;
826 
827 	/* set a block size */
828 	if (!sb_set_blocksize(sb, F2FS_BLKSIZE)) {
829 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
830 		goto free_sbi;
831 	}
832 
833 	err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
834 	if (err)
835 		goto free_sbi;
836 
837 	sb->s_fs_info = sbi;
838 	/* init some FS parameters */
839 	sbi->active_logs = NR_CURSEG_TYPE;
840 
841 	set_opt(sbi, BG_GC);
842 
843 #ifdef CONFIG_F2FS_FS_XATTR
844 	set_opt(sbi, XATTR_USER);
845 #endif
846 #ifdef CONFIG_F2FS_FS_POSIX_ACL
847 	set_opt(sbi, POSIX_ACL);
848 #endif
849 	/* parse mount options */
850 	err = parse_options(sb, (char *)data);
851 	if (err)
852 		goto free_sb_buf;
853 
854 	sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
855 	sb->s_max_links = F2FS_LINK_MAX;
856 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
857 
858 	sb->s_op = &f2fs_sops;
859 	sb->s_xattr = f2fs_xattr_handlers;
860 	sb->s_export_op = &f2fs_export_ops;
861 	sb->s_magic = F2FS_SUPER_MAGIC;
862 	sb->s_time_gran = 1;
863 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
864 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
865 	memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
866 
867 	/* init f2fs-specific super block info */
868 	sbi->sb = sb;
869 	sbi->raw_super = raw_super;
870 	sbi->raw_super_buf = raw_super_buf;
871 	mutex_init(&sbi->gc_mutex);
872 	mutex_init(&sbi->writepages);
873 	mutex_init(&sbi->cp_mutex);
874 	mutex_init(&sbi->node_write);
875 	sbi->por_doing = false;
876 	spin_lock_init(&sbi->stat_lock);
877 	init_rwsem(&sbi->bio_sem);
878 	init_rwsem(&sbi->cp_rwsem);
879 	init_waitqueue_head(&sbi->cp_wait);
880 	init_sb_info(sbi);
881 
882 	/* get an inode for meta space */
883 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
884 	if (IS_ERR(sbi->meta_inode)) {
885 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
886 		err = PTR_ERR(sbi->meta_inode);
887 		goto free_sb_buf;
888 	}
889 
890 	err = get_valid_checkpoint(sbi);
891 	if (err) {
892 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
893 		goto free_meta_inode;
894 	}
895 
896 	/* sanity checking of checkpoint */
897 	err = -EINVAL;
898 	if (sanity_check_ckpt(sbi)) {
899 		f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
900 		goto free_cp;
901 	}
902 
903 	sbi->total_valid_node_count =
904 				le32_to_cpu(sbi->ckpt->valid_node_count);
905 	sbi->total_valid_inode_count =
906 				le32_to_cpu(sbi->ckpt->valid_inode_count);
907 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
908 	sbi->total_valid_block_count =
909 				le64_to_cpu(sbi->ckpt->valid_block_count);
910 	sbi->last_valid_block_count = sbi->total_valid_block_count;
911 	sbi->alloc_valid_block_count = 0;
912 	INIT_LIST_HEAD(&sbi->dir_inode_list);
913 	spin_lock_init(&sbi->dir_inode_lock);
914 
915 	init_orphan_info(sbi);
916 
917 	/* setup f2fs internal modules */
918 	err = build_segment_manager(sbi);
919 	if (err) {
920 		f2fs_msg(sb, KERN_ERR,
921 			"Failed to initialize F2FS segment manager");
922 		goto free_sm;
923 	}
924 	err = build_node_manager(sbi);
925 	if (err) {
926 		f2fs_msg(sb, KERN_ERR,
927 			"Failed to initialize F2FS node manager");
928 		goto free_nm;
929 	}
930 
931 	build_gc_manager(sbi);
932 
933 	/* get an inode for node space */
934 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
935 	if (IS_ERR(sbi->node_inode)) {
936 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
937 		err = PTR_ERR(sbi->node_inode);
938 		goto free_nm;
939 	}
940 
941 	/* if there are nt orphan nodes free them */
942 	err = -EINVAL;
943 	if (recover_orphan_inodes(sbi))
944 		goto free_node_inode;
945 
946 	/* read root inode and dentry */
947 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
948 	if (IS_ERR(root)) {
949 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
950 		err = PTR_ERR(root);
951 		goto free_node_inode;
952 	}
953 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size)
954 		goto free_root_inode;
955 
956 	sb->s_root = d_make_root(root); /* allocate root dentry */
957 	if (!sb->s_root) {
958 		err = -ENOMEM;
959 		goto free_root_inode;
960 	}
961 
962 	/* recover fsynced data */
963 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
964 		err = recover_fsync_data(sbi);
965 		if (err)
966 			f2fs_msg(sb, KERN_ERR,
967 				"Cannot recover all fsync data errno=%ld", err);
968 	}
969 
970 	/*
971 	 * If filesystem is not mounted as read-only then
972 	 * do start the gc_thread.
973 	 */
974 	if (!(sb->s_flags & MS_RDONLY)) {
975 		/* After POR, we can run background GC thread.*/
976 		err = start_gc_thread(sbi);
977 		if (err)
978 			goto free_gc;
979 	}
980 
981 	err = f2fs_build_stats(sbi);
982 	if (err)
983 		goto free_gc;
984 
985 	if (f2fs_proc_root)
986 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
987 
988 	if (sbi->s_proc)
989 		proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
990 				 &f2fs_seq_segment_info_fops, sb);
991 
992 	if (test_opt(sbi, DISCARD)) {
993 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
994 		if (!blk_queue_discard(q))
995 			f2fs_msg(sb, KERN_WARNING,
996 					"mounting with \"discard\" option, but "
997 					"the device does not support discard");
998 	}
999 
1000 	sbi->s_kobj.kset = f2fs_kset;
1001 	init_completion(&sbi->s_kobj_unregister);
1002 	err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1003 							"%s", sb->s_id);
1004 	if (err)
1005 		goto fail;
1006 
1007 	return 0;
1008 fail:
1009 	if (sbi->s_proc) {
1010 		remove_proc_entry("segment_info", sbi->s_proc);
1011 		remove_proc_entry(sb->s_id, f2fs_proc_root);
1012 	}
1013 	f2fs_destroy_stats(sbi);
1014 free_gc:
1015 	stop_gc_thread(sbi);
1016 free_root_inode:
1017 	dput(sb->s_root);
1018 	sb->s_root = NULL;
1019 free_node_inode:
1020 	iput(sbi->node_inode);
1021 free_nm:
1022 	destroy_node_manager(sbi);
1023 free_sm:
1024 	destroy_segment_manager(sbi);
1025 free_cp:
1026 	kfree(sbi->ckpt);
1027 free_meta_inode:
1028 	make_bad_inode(sbi->meta_inode);
1029 	iput(sbi->meta_inode);
1030 free_sb_buf:
1031 	brelse(raw_super_buf);
1032 free_sbi:
1033 	kfree(sbi);
1034 	return err;
1035 }
1036 
1037 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1038 			const char *dev_name, void *data)
1039 {
1040 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1041 }
1042 
1043 static struct file_system_type f2fs_fs_type = {
1044 	.owner		= THIS_MODULE,
1045 	.name		= "f2fs",
1046 	.mount		= f2fs_mount,
1047 	.kill_sb	= kill_block_super,
1048 	.fs_flags	= FS_REQUIRES_DEV,
1049 };
1050 MODULE_ALIAS_FS("f2fs");
1051 
1052 static int __init init_inodecache(void)
1053 {
1054 	f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1055 			sizeof(struct f2fs_inode_info), NULL);
1056 	if (f2fs_inode_cachep == NULL)
1057 		return -ENOMEM;
1058 	return 0;
1059 }
1060 
1061 static void destroy_inodecache(void)
1062 {
1063 	/*
1064 	 * Make sure all delayed rcu free inodes are flushed before we
1065 	 * destroy cache.
1066 	 */
1067 	rcu_barrier();
1068 	kmem_cache_destroy(f2fs_inode_cachep);
1069 }
1070 
1071 static int __init init_f2fs_fs(void)
1072 {
1073 	int err;
1074 
1075 	err = init_inodecache();
1076 	if (err)
1077 		goto fail;
1078 	err = create_node_manager_caches();
1079 	if (err)
1080 		goto free_inodecache;
1081 	err = create_gc_caches();
1082 	if (err)
1083 		goto free_node_manager_caches;
1084 	err = create_checkpoint_caches();
1085 	if (err)
1086 		goto free_gc_caches;
1087 	f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1088 	if (!f2fs_kset) {
1089 		err = -ENOMEM;
1090 		goto free_checkpoint_caches;
1091 	}
1092 	err = register_filesystem(&f2fs_fs_type);
1093 	if (err)
1094 		goto free_kset;
1095 	f2fs_create_root_stats();
1096 	f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1097 	return 0;
1098 
1099 free_kset:
1100 	kset_unregister(f2fs_kset);
1101 free_checkpoint_caches:
1102 	destroy_checkpoint_caches();
1103 free_gc_caches:
1104 	destroy_gc_caches();
1105 free_node_manager_caches:
1106 	destroy_node_manager_caches();
1107 free_inodecache:
1108 	destroy_inodecache();
1109 fail:
1110 	return err;
1111 }
1112 
1113 static void __exit exit_f2fs_fs(void)
1114 {
1115 	remove_proc_entry("fs/f2fs", NULL);
1116 	f2fs_destroy_root_stats();
1117 	unregister_filesystem(&f2fs_fs_type);
1118 	destroy_checkpoint_caches();
1119 	destroy_gc_caches();
1120 	destroy_node_manager_caches();
1121 	destroy_inodecache();
1122 	kset_unregister(f2fs_kset);
1123 }
1124 
1125 module_init(init_f2fs_fs)
1126 module_exit(exit_f2fs_fs)
1127 
1128 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1129 MODULE_DESCRIPTION("Flash Friendly File System");
1130 MODULE_LICENSE("GPL");
1131