xref: /linux/fs/f2fs/super.c (revision 4603f53a1dc3c76dfba841d123db9fa6204934f5)
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/random.h>
22 #include <linux/exportfs.h>
23 #include <linux/blkdev.h>
24 #include <linux/f2fs_fs.h>
25 
26 #include "f2fs.h"
27 #include "node.h"
28 #include "segment.h"
29 #include "xattr.h"
30 
31 #define CREATE_TRACE_POINTS
32 #include <trace/events/f2fs.h>
33 
34 static struct kmem_cache *f2fs_inode_cachep;
35 
36 enum {
37 	Opt_gc_background,
38 	Opt_disable_roll_forward,
39 	Opt_discard,
40 	Opt_noheap,
41 	Opt_nouser_xattr,
42 	Opt_noacl,
43 	Opt_active_logs,
44 	Opt_disable_ext_identify,
45 	Opt_err,
46 };
47 
48 static match_table_t f2fs_tokens = {
49 	{Opt_gc_background, "background_gc=%s"},
50 	{Opt_disable_roll_forward, "disable_roll_forward"},
51 	{Opt_discard, "discard"},
52 	{Opt_noheap, "no_heap"},
53 	{Opt_nouser_xattr, "nouser_xattr"},
54 	{Opt_noacl, "noacl"},
55 	{Opt_active_logs, "active_logs=%u"},
56 	{Opt_disable_ext_identify, "disable_ext_identify"},
57 	{Opt_err, NULL},
58 };
59 
60 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
61 {
62 	struct va_format vaf;
63 	va_list args;
64 
65 	va_start(args, fmt);
66 	vaf.fmt = fmt;
67 	vaf.va = &args;
68 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
69 	va_end(args);
70 }
71 
72 static void init_once(void *foo)
73 {
74 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
75 
76 	inode_init_once(&fi->vfs_inode);
77 }
78 
79 static int parse_options(struct super_block *sb, char *options)
80 {
81 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
82 	substring_t args[MAX_OPT_ARGS];
83 	char *p, *name;
84 	int arg = 0;
85 
86 	if (!options)
87 		return 0;
88 
89 	while ((p = strsep(&options, ",")) != NULL) {
90 		int token;
91 		if (!*p)
92 			continue;
93 		/*
94 		 * Initialize args struct so we know whether arg was
95 		 * found; some options take optional arguments.
96 		 */
97 		args[0].to = args[0].from = NULL;
98 		token = match_token(p, f2fs_tokens, args);
99 
100 		switch (token) {
101 		case Opt_gc_background:
102 			name = match_strdup(&args[0]);
103 
104 			if (!name)
105 				return -ENOMEM;
106 			if (!strncmp(name, "on", 2))
107 				set_opt(sbi, BG_GC);
108 			else if (!strncmp(name, "off", 3))
109 				clear_opt(sbi, BG_GC);
110 			else {
111 				kfree(name);
112 				return -EINVAL;
113 			}
114 			kfree(name);
115 			break;
116 		case Opt_disable_roll_forward:
117 			set_opt(sbi, DISABLE_ROLL_FORWARD);
118 			break;
119 		case Opt_discard:
120 			set_opt(sbi, DISCARD);
121 			break;
122 		case Opt_noheap:
123 			set_opt(sbi, NOHEAP);
124 			break;
125 #ifdef CONFIG_F2FS_FS_XATTR
126 		case Opt_nouser_xattr:
127 			clear_opt(sbi, XATTR_USER);
128 			break;
129 #else
130 		case Opt_nouser_xattr:
131 			f2fs_msg(sb, KERN_INFO,
132 				"nouser_xattr options not supported");
133 			break;
134 #endif
135 #ifdef CONFIG_F2FS_FS_POSIX_ACL
136 		case Opt_noacl:
137 			clear_opt(sbi, POSIX_ACL);
138 			break;
139 #else
140 		case Opt_noacl:
141 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
142 			break;
143 #endif
144 		case Opt_active_logs:
145 			if (args->from && match_int(args, &arg))
146 				return -EINVAL;
147 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
148 				return -EINVAL;
149 			sbi->active_logs = arg;
150 			break;
151 		case Opt_disable_ext_identify:
152 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
153 			break;
154 		default:
155 			f2fs_msg(sb, KERN_ERR,
156 				"Unrecognized mount option \"%s\" or missing value",
157 				p);
158 			return -EINVAL;
159 		}
160 	}
161 	return 0;
162 }
163 
164 static struct inode *f2fs_alloc_inode(struct super_block *sb)
165 {
166 	struct f2fs_inode_info *fi;
167 
168 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_NOFS | __GFP_ZERO);
169 	if (!fi)
170 		return NULL;
171 
172 	init_once((void *) fi);
173 
174 	/* Initialize f2fs-specific inode info */
175 	fi->vfs_inode.i_version = 1;
176 	atomic_set(&fi->dirty_dents, 0);
177 	fi->i_current_depth = 1;
178 	fi->i_advise = 0;
179 	rwlock_init(&fi->ext.ext_lock);
180 
181 	set_inode_flag(fi, FI_NEW_INODE);
182 
183 	return &fi->vfs_inode;
184 }
185 
186 static int f2fs_drop_inode(struct inode *inode)
187 {
188 	/*
189 	 * This is to avoid a deadlock condition like below.
190 	 * writeback_single_inode(inode)
191 	 *  - f2fs_write_data_page
192 	 *    - f2fs_gc -> iput -> evict
193 	 *       - inode_wait_for_writeback(inode)
194 	 */
195 	if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
196 		return 0;
197 	return generic_drop_inode(inode);
198 }
199 
200 /*
201  * f2fs_dirty_inode() is called from __mark_inode_dirty()
202  *
203  * We should call set_dirty_inode to write the dirty inode through write_inode.
204  */
205 static void f2fs_dirty_inode(struct inode *inode, int flags)
206 {
207 	set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
208 	return;
209 }
210 
211 static void f2fs_i_callback(struct rcu_head *head)
212 {
213 	struct inode *inode = container_of(head, struct inode, i_rcu);
214 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
215 }
216 
217 static void f2fs_destroy_inode(struct inode *inode)
218 {
219 	call_rcu(&inode->i_rcu, f2fs_i_callback);
220 }
221 
222 static void f2fs_put_super(struct super_block *sb)
223 {
224 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
225 
226 	f2fs_destroy_stats(sbi);
227 	stop_gc_thread(sbi);
228 
229 	write_checkpoint(sbi, true);
230 
231 	iput(sbi->node_inode);
232 	iput(sbi->meta_inode);
233 
234 	/* destroy f2fs internal modules */
235 	destroy_node_manager(sbi);
236 	destroy_segment_manager(sbi);
237 
238 	kfree(sbi->ckpt);
239 
240 	sb->s_fs_info = NULL;
241 	brelse(sbi->raw_super_buf);
242 	kfree(sbi);
243 }
244 
245 int f2fs_sync_fs(struct super_block *sb, int sync)
246 {
247 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
248 
249 	trace_f2fs_sync_fs(sb, sync);
250 
251 	if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
252 		return 0;
253 
254 	if (sync) {
255 		mutex_lock(&sbi->gc_mutex);
256 		write_checkpoint(sbi, false);
257 		mutex_unlock(&sbi->gc_mutex);
258 	} else {
259 		f2fs_balance_fs(sbi);
260 	}
261 
262 	return 0;
263 }
264 
265 static int f2fs_freeze(struct super_block *sb)
266 {
267 	int err;
268 
269 	if (f2fs_readonly(sb))
270 		return 0;
271 
272 	err = f2fs_sync_fs(sb, 1);
273 	return err;
274 }
275 
276 static int f2fs_unfreeze(struct super_block *sb)
277 {
278 	return 0;
279 }
280 
281 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
282 {
283 	struct super_block *sb = dentry->d_sb;
284 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
285 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
286 	block_t total_count, user_block_count, start_count, ovp_count;
287 
288 	total_count = le64_to_cpu(sbi->raw_super->block_count);
289 	user_block_count = sbi->user_block_count;
290 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
291 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
292 	buf->f_type = F2FS_SUPER_MAGIC;
293 	buf->f_bsize = sbi->blocksize;
294 
295 	buf->f_blocks = total_count - start_count;
296 	buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
297 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
298 
299 	buf->f_files = sbi->total_node_count;
300 	buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
301 
302 	buf->f_namelen = F2FS_NAME_LEN;
303 	buf->f_fsid.val[0] = (u32)id;
304 	buf->f_fsid.val[1] = (u32)(id >> 32);
305 
306 	return 0;
307 }
308 
309 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
310 {
311 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
312 
313 	if (!(root->d_sb->s_flags & MS_RDONLY) && test_opt(sbi, BG_GC))
314 		seq_printf(seq, ",background_gc=%s", "on");
315 	else
316 		seq_printf(seq, ",background_gc=%s", "off");
317 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
318 		seq_puts(seq, ",disable_roll_forward");
319 	if (test_opt(sbi, DISCARD))
320 		seq_puts(seq, ",discard");
321 	if (test_opt(sbi, NOHEAP))
322 		seq_puts(seq, ",no_heap_alloc");
323 #ifdef CONFIG_F2FS_FS_XATTR
324 	if (test_opt(sbi, XATTR_USER))
325 		seq_puts(seq, ",user_xattr");
326 	else
327 		seq_puts(seq, ",nouser_xattr");
328 #endif
329 #ifdef CONFIG_F2FS_FS_POSIX_ACL
330 	if (test_opt(sbi, POSIX_ACL))
331 		seq_puts(seq, ",acl");
332 	else
333 		seq_puts(seq, ",noacl");
334 #endif
335 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
336 		seq_puts(seq, ",disable_ext_identify");
337 
338 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
339 
340 	return 0;
341 }
342 
343 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
344 {
345 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
346 	struct f2fs_mount_info org_mount_opt;
347 	int err, active_logs;
348 
349 	/*
350 	 * Save the old mount options in case we
351 	 * need to restore them.
352 	 */
353 	org_mount_opt = sbi->mount_opt;
354 	active_logs = sbi->active_logs;
355 
356 	/* parse mount options */
357 	err = parse_options(sb, data);
358 	if (err)
359 		goto restore_opts;
360 
361 	/*
362 	 * Previous and new state of filesystem is RO,
363 	 * so no point in checking GC conditions.
364 	 */
365 	if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
366 		goto skip;
367 
368 	/*
369 	 * We stop the GC thread if FS is mounted as RO
370 	 * or if background_gc = off is passed in mount
371 	 * option. Also sync the filesystem.
372 	 */
373 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
374 		if (sbi->gc_thread) {
375 			stop_gc_thread(sbi);
376 			f2fs_sync_fs(sb, 1);
377 		}
378 	} else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
379 		err = start_gc_thread(sbi);
380 		if (err)
381 			goto restore_opts;
382 	}
383 skip:
384 	/* Update the POSIXACL Flag */
385 	 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
386 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
387 	return 0;
388 
389 restore_opts:
390 	sbi->mount_opt = org_mount_opt;
391 	sbi->active_logs = active_logs;
392 	return err;
393 }
394 
395 static struct super_operations f2fs_sops = {
396 	.alloc_inode	= f2fs_alloc_inode,
397 	.drop_inode	= f2fs_drop_inode,
398 	.destroy_inode	= f2fs_destroy_inode,
399 	.write_inode	= f2fs_write_inode,
400 	.dirty_inode	= f2fs_dirty_inode,
401 	.show_options	= f2fs_show_options,
402 	.evict_inode	= f2fs_evict_inode,
403 	.put_super	= f2fs_put_super,
404 	.sync_fs	= f2fs_sync_fs,
405 	.freeze_fs	= f2fs_freeze,
406 	.unfreeze_fs	= f2fs_unfreeze,
407 	.statfs		= f2fs_statfs,
408 	.remount_fs	= f2fs_remount,
409 };
410 
411 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
412 		u64 ino, u32 generation)
413 {
414 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
415 	struct inode *inode;
416 
417 	if (ino < F2FS_ROOT_INO(sbi))
418 		return ERR_PTR(-ESTALE);
419 
420 	/*
421 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
422 	 * However f2fs_iget currently does appropriate checks to handle stale
423 	 * inodes so everything is OK.
424 	 */
425 	inode = f2fs_iget(sb, ino);
426 	if (IS_ERR(inode))
427 		return ERR_CAST(inode);
428 	if (generation && inode->i_generation != generation) {
429 		/* we didn't find the right inode.. */
430 		iput(inode);
431 		return ERR_PTR(-ESTALE);
432 	}
433 	return inode;
434 }
435 
436 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
437 		int fh_len, int fh_type)
438 {
439 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
440 				    f2fs_nfs_get_inode);
441 }
442 
443 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
444 		int fh_len, int fh_type)
445 {
446 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
447 				    f2fs_nfs_get_inode);
448 }
449 
450 static const struct export_operations f2fs_export_ops = {
451 	.fh_to_dentry = f2fs_fh_to_dentry,
452 	.fh_to_parent = f2fs_fh_to_parent,
453 	.get_parent = f2fs_get_parent,
454 };
455 
456 static loff_t max_file_size(unsigned bits)
457 {
458 	loff_t result = ADDRS_PER_INODE;
459 	loff_t leaf_count = ADDRS_PER_BLOCK;
460 
461 	/* two direct node blocks */
462 	result += (leaf_count * 2);
463 
464 	/* two indirect node blocks */
465 	leaf_count *= NIDS_PER_BLOCK;
466 	result += (leaf_count * 2);
467 
468 	/* one double indirect node block */
469 	leaf_count *= NIDS_PER_BLOCK;
470 	result += leaf_count;
471 
472 	result <<= bits;
473 	return result;
474 }
475 
476 static int sanity_check_raw_super(struct super_block *sb,
477 			struct f2fs_super_block *raw_super)
478 {
479 	unsigned int blocksize;
480 
481 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
482 		f2fs_msg(sb, KERN_INFO,
483 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
484 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
485 		return 1;
486 	}
487 
488 	/* Currently, support only 4KB page cache size */
489 	if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
490 		f2fs_msg(sb, KERN_INFO,
491 			"Invalid page_cache_size (%lu), supports only 4KB\n",
492 			PAGE_CACHE_SIZE);
493 		return 1;
494 	}
495 
496 	/* Currently, support only 4KB block size */
497 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
498 	if (blocksize != F2FS_BLKSIZE) {
499 		f2fs_msg(sb, KERN_INFO,
500 			"Invalid blocksize (%u), supports only 4KB\n",
501 			blocksize);
502 		return 1;
503 	}
504 
505 	if (le32_to_cpu(raw_super->log_sectorsize) !=
506 					F2FS_LOG_SECTOR_SIZE) {
507 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
508 		return 1;
509 	}
510 	if (le32_to_cpu(raw_super->log_sectors_per_block) !=
511 					F2FS_LOG_SECTORS_PER_BLOCK) {
512 		f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
513 		return 1;
514 	}
515 	return 0;
516 }
517 
518 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
519 {
520 	unsigned int total, fsmeta;
521 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
522 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
523 
524 	total = le32_to_cpu(raw_super->segment_count);
525 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
526 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
527 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
528 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
529 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
530 
531 	if (fsmeta >= total)
532 		return 1;
533 
534 	if (is_set_ckpt_flags(ckpt, CP_ERROR_FLAG)) {
535 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
536 		return 1;
537 	}
538 	return 0;
539 }
540 
541 static void init_sb_info(struct f2fs_sb_info *sbi)
542 {
543 	struct f2fs_super_block *raw_super = sbi->raw_super;
544 	int i;
545 
546 	sbi->log_sectors_per_block =
547 		le32_to_cpu(raw_super->log_sectors_per_block);
548 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
549 	sbi->blocksize = 1 << sbi->log_blocksize;
550 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
551 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
552 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
553 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
554 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
555 	sbi->total_node_count =
556 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
557 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
558 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
559 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
560 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
561 	sbi->cur_victim_sec = NULL_SECNO;
562 
563 	for (i = 0; i < NR_COUNT_TYPE; i++)
564 		atomic_set(&sbi->nr_pages[i], 0);
565 }
566 
567 static int validate_superblock(struct super_block *sb,
568 		struct f2fs_super_block **raw_super,
569 		struct buffer_head **raw_super_buf, sector_t block)
570 {
571 	const char *super = (block == 0 ? "first" : "second");
572 
573 	/* read f2fs raw super block */
574 	*raw_super_buf = sb_bread(sb, block);
575 	if (!*raw_super_buf) {
576 		f2fs_msg(sb, KERN_ERR, "unable to read %s superblock",
577 				super);
578 		return -EIO;
579 	}
580 
581 	*raw_super = (struct f2fs_super_block *)
582 		((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
583 
584 	/* sanity checking of raw super */
585 	if (!sanity_check_raw_super(sb, *raw_super))
586 		return 0;
587 
588 	f2fs_msg(sb, KERN_ERR, "Can't find a valid F2FS filesystem "
589 				"in %s superblock", super);
590 	return -EINVAL;
591 }
592 
593 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
594 {
595 	struct f2fs_sb_info *sbi;
596 	struct f2fs_super_block *raw_super;
597 	struct buffer_head *raw_super_buf;
598 	struct inode *root;
599 	long err = -EINVAL;
600 	int i;
601 
602 	/* allocate memory for f2fs-specific super block info */
603 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
604 	if (!sbi)
605 		return -ENOMEM;
606 
607 	/* set a block size */
608 	if (!sb_set_blocksize(sb, F2FS_BLKSIZE)) {
609 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
610 		goto free_sbi;
611 	}
612 
613 	err = validate_superblock(sb, &raw_super, &raw_super_buf, 0);
614 	if (err) {
615 		brelse(raw_super_buf);
616 		/* check secondary superblock when primary failed */
617 		err = validate_superblock(sb, &raw_super, &raw_super_buf, 1);
618 		if (err)
619 			goto free_sb_buf;
620 	}
621 	sb->s_fs_info = sbi;
622 	/* init some FS parameters */
623 	sbi->active_logs = NR_CURSEG_TYPE;
624 
625 	set_opt(sbi, BG_GC);
626 
627 #ifdef CONFIG_F2FS_FS_XATTR
628 	set_opt(sbi, XATTR_USER);
629 #endif
630 #ifdef CONFIG_F2FS_FS_POSIX_ACL
631 	set_opt(sbi, POSIX_ACL);
632 #endif
633 	/* parse mount options */
634 	err = parse_options(sb, (char *)data);
635 	if (err)
636 		goto free_sb_buf;
637 
638 	sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
639 	sb->s_max_links = F2FS_LINK_MAX;
640 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
641 
642 	sb->s_op = &f2fs_sops;
643 	sb->s_xattr = f2fs_xattr_handlers;
644 	sb->s_export_op = &f2fs_export_ops;
645 	sb->s_magic = F2FS_SUPER_MAGIC;
646 	sb->s_time_gran = 1;
647 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
648 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
649 	memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
650 
651 	/* init f2fs-specific super block info */
652 	sbi->sb = sb;
653 	sbi->raw_super = raw_super;
654 	sbi->raw_super_buf = raw_super_buf;
655 	mutex_init(&sbi->gc_mutex);
656 	mutex_init(&sbi->writepages);
657 	mutex_init(&sbi->cp_mutex);
658 	for (i = 0; i < NR_GLOBAL_LOCKS; i++)
659 		mutex_init(&sbi->fs_lock[i]);
660 	mutex_init(&sbi->node_write);
661 	sbi->por_doing = 0;
662 	spin_lock_init(&sbi->stat_lock);
663 	init_rwsem(&sbi->bio_sem);
664 	init_sb_info(sbi);
665 
666 	/* get an inode for meta space */
667 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
668 	if (IS_ERR(sbi->meta_inode)) {
669 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
670 		err = PTR_ERR(sbi->meta_inode);
671 		goto free_sb_buf;
672 	}
673 
674 	err = get_valid_checkpoint(sbi);
675 	if (err) {
676 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
677 		goto free_meta_inode;
678 	}
679 
680 	/* sanity checking of checkpoint */
681 	err = -EINVAL;
682 	if (sanity_check_ckpt(sbi)) {
683 		f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
684 		goto free_cp;
685 	}
686 
687 	sbi->total_valid_node_count =
688 				le32_to_cpu(sbi->ckpt->valid_node_count);
689 	sbi->total_valid_inode_count =
690 				le32_to_cpu(sbi->ckpt->valid_inode_count);
691 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
692 	sbi->total_valid_block_count =
693 				le64_to_cpu(sbi->ckpt->valid_block_count);
694 	sbi->last_valid_block_count = sbi->total_valid_block_count;
695 	sbi->alloc_valid_block_count = 0;
696 	INIT_LIST_HEAD(&sbi->dir_inode_list);
697 	spin_lock_init(&sbi->dir_inode_lock);
698 
699 	init_orphan_info(sbi);
700 
701 	/* setup f2fs internal modules */
702 	err = build_segment_manager(sbi);
703 	if (err) {
704 		f2fs_msg(sb, KERN_ERR,
705 			"Failed to initialize F2FS segment manager");
706 		goto free_sm;
707 	}
708 	err = build_node_manager(sbi);
709 	if (err) {
710 		f2fs_msg(sb, KERN_ERR,
711 			"Failed to initialize F2FS node manager");
712 		goto free_nm;
713 	}
714 
715 	build_gc_manager(sbi);
716 
717 	/* get an inode for node space */
718 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
719 	if (IS_ERR(sbi->node_inode)) {
720 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
721 		err = PTR_ERR(sbi->node_inode);
722 		goto free_nm;
723 	}
724 
725 	/* if there are nt orphan nodes free them */
726 	err = -EINVAL;
727 	if (recover_orphan_inodes(sbi))
728 		goto free_node_inode;
729 
730 	/* read root inode and dentry */
731 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
732 	if (IS_ERR(root)) {
733 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
734 		err = PTR_ERR(root);
735 		goto free_node_inode;
736 	}
737 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size)
738 		goto free_root_inode;
739 
740 	sb->s_root = d_make_root(root); /* allocate root dentry */
741 	if (!sb->s_root) {
742 		err = -ENOMEM;
743 		goto free_root_inode;
744 	}
745 
746 	/* recover fsynced data */
747 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
748 		err = recover_fsync_data(sbi);
749 		if (err)
750 			f2fs_msg(sb, KERN_ERR,
751 				"Cannot recover all fsync data errno=%ld", err);
752 	}
753 
754 	/*
755 	 * If filesystem is not mounted as read-only then
756 	 * do start the gc_thread.
757 	 */
758 	if (!(sb->s_flags & MS_RDONLY)) {
759 		/* After POR, we can run background GC thread.*/
760 		err = start_gc_thread(sbi);
761 		if (err)
762 			goto fail;
763 	}
764 
765 	err = f2fs_build_stats(sbi);
766 	if (err)
767 		goto fail;
768 
769 	if (test_opt(sbi, DISCARD)) {
770 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
771 		if (!blk_queue_discard(q))
772 			f2fs_msg(sb, KERN_WARNING,
773 					"mounting with \"discard\" option, but "
774 					"the device does not support discard");
775 	}
776 
777 	return 0;
778 fail:
779 	stop_gc_thread(sbi);
780 free_root_inode:
781 	dput(sb->s_root);
782 	sb->s_root = NULL;
783 free_node_inode:
784 	iput(sbi->node_inode);
785 free_nm:
786 	destroy_node_manager(sbi);
787 free_sm:
788 	destroy_segment_manager(sbi);
789 free_cp:
790 	kfree(sbi->ckpt);
791 free_meta_inode:
792 	make_bad_inode(sbi->meta_inode);
793 	iput(sbi->meta_inode);
794 free_sb_buf:
795 	brelse(raw_super_buf);
796 free_sbi:
797 	kfree(sbi);
798 	return err;
799 }
800 
801 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
802 			const char *dev_name, void *data)
803 {
804 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
805 }
806 
807 static struct file_system_type f2fs_fs_type = {
808 	.owner		= THIS_MODULE,
809 	.name		= "f2fs",
810 	.mount		= f2fs_mount,
811 	.kill_sb	= kill_block_super,
812 	.fs_flags	= FS_REQUIRES_DEV,
813 };
814 MODULE_ALIAS_FS("f2fs");
815 
816 static int __init init_inodecache(void)
817 {
818 	f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
819 			sizeof(struct f2fs_inode_info), NULL);
820 	if (f2fs_inode_cachep == NULL)
821 		return -ENOMEM;
822 	return 0;
823 }
824 
825 static void destroy_inodecache(void)
826 {
827 	/*
828 	 * Make sure all delayed rcu free inodes are flushed before we
829 	 * destroy cache.
830 	 */
831 	rcu_barrier();
832 	kmem_cache_destroy(f2fs_inode_cachep);
833 }
834 
835 static int __init init_f2fs_fs(void)
836 {
837 	int err;
838 
839 	err = init_inodecache();
840 	if (err)
841 		goto fail;
842 	err = create_node_manager_caches();
843 	if (err)
844 		goto fail;
845 	err = create_gc_caches();
846 	if (err)
847 		goto fail;
848 	err = create_checkpoint_caches();
849 	if (err)
850 		goto fail;
851 	err = register_filesystem(&f2fs_fs_type);
852 	if (err)
853 		goto fail;
854 	f2fs_create_root_stats();
855 fail:
856 	return err;
857 }
858 
859 static void __exit exit_f2fs_fs(void)
860 {
861 	f2fs_destroy_root_stats();
862 	unregister_filesystem(&f2fs_fs_type);
863 	destroy_checkpoint_caches();
864 	destroy_gc_caches();
865 	destroy_node_manager_caches();
866 	destroy_inodecache();
867 }
868 
869 module_init(init_f2fs_fs)
870 module_exit(exit_f2fs_fs)
871 
872 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
873 MODULE_DESCRIPTION("Flash Friendly File System");
874 MODULE_LICENSE("GPL");
875