xref: /openbmc/linux/fs/f2fs/super.c (revision d2999e1b)
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_inline_data,
54 	Opt_flush_merge,
55 	Opt_err,
56 };
57 
58 static match_table_t f2fs_tokens = {
59 	{Opt_gc_background, "background_gc=%s"},
60 	{Opt_disable_roll_forward, "disable_roll_forward"},
61 	{Opt_discard, "discard"},
62 	{Opt_noheap, "no_heap"},
63 	{Opt_user_xattr, "user_xattr"},
64 	{Opt_nouser_xattr, "nouser_xattr"},
65 	{Opt_acl, "acl"},
66 	{Opt_noacl, "noacl"},
67 	{Opt_active_logs, "active_logs=%u"},
68 	{Opt_disable_ext_identify, "disable_ext_identify"},
69 	{Opt_inline_xattr, "inline_xattr"},
70 	{Opt_inline_data, "inline_data"},
71 	{Opt_flush_merge, "flush_merge"},
72 	{Opt_err, NULL},
73 };
74 
75 /* Sysfs support for f2fs */
76 enum {
77 	GC_THREAD,	/* struct f2fs_gc_thread */
78 	SM_INFO,	/* struct f2fs_sm_info */
79 	NM_INFO,	/* struct f2fs_nm_info */
80 	F2FS_SBI,	/* struct f2fs_sb_info */
81 };
82 
83 struct f2fs_attr {
84 	struct attribute attr;
85 	ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
86 	ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
87 			 const char *, size_t);
88 	int struct_type;
89 	int offset;
90 };
91 
92 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
93 {
94 	if (struct_type == GC_THREAD)
95 		return (unsigned char *)sbi->gc_thread;
96 	else if (struct_type == SM_INFO)
97 		return (unsigned char *)SM_I(sbi);
98 	else if (struct_type == NM_INFO)
99 		return (unsigned char *)NM_I(sbi);
100 	else if (struct_type == F2FS_SBI)
101 		return (unsigned char *)sbi;
102 	return NULL;
103 }
104 
105 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
106 			struct f2fs_sb_info *sbi, char *buf)
107 {
108 	unsigned char *ptr = NULL;
109 	unsigned int *ui;
110 
111 	ptr = __struct_ptr(sbi, a->struct_type);
112 	if (!ptr)
113 		return -EINVAL;
114 
115 	ui = (unsigned int *)(ptr + a->offset);
116 
117 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
118 }
119 
120 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
121 			struct f2fs_sb_info *sbi,
122 			const char *buf, size_t count)
123 {
124 	unsigned char *ptr;
125 	unsigned long t;
126 	unsigned int *ui;
127 	ssize_t ret;
128 
129 	ptr = __struct_ptr(sbi, a->struct_type);
130 	if (!ptr)
131 		return -EINVAL;
132 
133 	ui = (unsigned int *)(ptr + a->offset);
134 
135 	ret = kstrtoul(skip_spaces(buf), 0, &t);
136 	if (ret < 0)
137 		return ret;
138 	*ui = t;
139 	return count;
140 }
141 
142 static ssize_t f2fs_attr_show(struct kobject *kobj,
143 				struct attribute *attr, char *buf)
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->show ? a->show(a, sbi, buf) : 0;
150 }
151 
152 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
153 						const char *buf, size_t len)
154 {
155 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
156 									s_kobj);
157 	struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
158 
159 	return a->store ? a->store(a, sbi, buf, len) : 0;
160 }
161 
162 static void f2fs_sb_release(struct kobject *kobj)
163 {
164 	struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
165 								s_kobj);
166 	complete(&sbi->s_kobj_unregister);
167 }
168 
169 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
170 static struct f2fs_attr f2fs_attr_##_name = {			\
171 	.attr = {.name = __stringify(_name), .mode = _mode },	\
172 	.show	= _show,					\
173 	.store	= _store,					\
174 	.struct_type = _struct_type,				\
175 	.offset = _offset					\
176 }
177 
178 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)	\
179 	F2FS_ATTR_OFFSET(struct_type, name, 0644,		\
180 		f2fs_sbi_show, f2fs_sbi_store,			\
181 		offsetof(struct struct_name, elname))
182 
183 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
184 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
185 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
186 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
187 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
188 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
189 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
190 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
191 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
192 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
193 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
194 
195 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
196 static struct attribute *f2fs_attrs[] = {
197 	ATTR_LIST(gc_min_sleep_time),
198 	ATTR_LIST(gc_max_sleep_time),
199 	ATTR_LIST(gc_no_gc_sleep_time),
200 	ATTR_LIST(gc_idle),
201 	ATTR_LIST(reclaim_segments),
202 	ATTR_LIST(max_small_discards),
203 	ATTR_LIST(ipu_policy),
204 	ATTR_LIST(min_ipu_util),
205 	ATTR_LIST(max_victim_search),
206 	ATTR_LIST(dir_level),
207 	ATTR_LIST(ram_thresh),
208 	NULL,
209 };
210 
211 static const struct sysfs_ops f2fs_attr_ops = {
212 	.show	= f2fs_attr_show,
213 	.store	= f2fs_attr_store,
214 };
215 
216 static struct kobj_type f2fs_ktype = {
217 	.default_attrs	= f2fs_attrs,
218 	.sysfs_ops	= &f2fs_attr_ops,
219 	.release	= f2fs_sb_release,
220 };
221 
222 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
223 {
224 	struct va_format vaf;
225 	va_list args;
226 
227 	va_start(args, fmt);
228 	vaf.fmt = fmt;
229 	vaf.va = &args;
230 	printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
231 	va_end(args);
232 }
233 
234 static void init_once(void *foo)
235 {
236 	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
237 
238 	inode_init_once(&fi->vfs_inode);
239 }
240 
241 static int parse_options(struct super_block *sb, char *options)
242 {
243 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
244 	substring_t args[MAX_OPT_ARGS];
245 	char *p, *name;
246 	int arg = 0;
247 
248 	if (!options)
249 		return 0;
250 
251 	while ((p = strsep(&options, ",")) != NULL) {
252 		int token;
253 		if (!*p)
254 			continue;
255 		/*
256 		 * Initialize args struct so we know whether arg was
257 		 * found; some options take optional arguments.
258 		 */
259 		args[0].to = args[0].from = NULL;
260 		token = match_token(p, f2fs_tokens, args);
261 
262 		switch (token) {
263 		case Opt_gc_background:
264 			name = match_strdup(&args[0]);
265 
266 			if (!name)
267 				return -ENOMEM;
268 			if (strlen(name) == 2 && !strncmp(name, "on", 2))
269 				set_opt(sbi, BG_GC);
270 			else if (strlen(name) == 3 && !strncmp(name, "off", 3))
271 				clear_opt(sbi, BG_GC);
272 			else {
273 				kfree(name);
274 				return -EINVAL;
275 			}
276 			kfree(name);
277 			break;
278 		case Opt_disable_roll_forward:
279 			set_opt(sbi, DISABLE_ROLL_FORWARD);
280 			break;
281 		case Opt_discard:
282 			set_opt(sbi, DISCARD);
283 			break;
284 		case Opt_noheap:
285 			set_opt(sbi, NOHEAP);
286 			break;
287 #ifdef CONFIG_F2FS_FS_XATTR
288 		case Opt_user_xattr:
289 			set_opt(sbi, XATTR_USER);
290 			break;
291 		case Opt_nouser_xattr:
292 			clear_opt(sbi, XATTR_USER);
293 			break;
294 		case Opt_inline_xattr:
295 			set_opt(sbi, INLINE_XATTR);
296 			break;
297 #else
298 		case Opt_user_xattr:
299 			f2fs_msg(sb, KERN_INFO,
300 				"user_xattr options not supported");
301 			break;
302 		case Opt_nouser_xattr:
303 			f2fs_msg(sb, KERN_INFO,
304 				"nouser_xattr options not supported");
305 			break;
306 		case Opt_inline_xattr:
307 			f2fs_msg(sb, KERN_INFO,
308 				"inline_xattr options not supported");
309 			break;
310 #endif
311 #ifdef CONFIG_F2FS_FS_POSIX_ACL
312 		case Opt_acl:
313 			set_opt(sbi, POSIX_ACL);
314 			break;
315 		case Opt_noacl:
316 			clear_opt(sbi, POSIX_ACL);
317 			break;
318 #else
319 		case Opt_acl:
320 			f2fs_msg(sb, KERN_INFO, "acl options not supported");
321 			break;
322 		case Opt_noacl:
323 			f2fs_msg(sb, KERN_INFO, "noacl options not supported");
324 			break;
325 #endif
326 		case Opt_active_logs:
327 			if (args->from && match_int(args, &arg))
328 				return -EINVAL;
329 			if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
330 				return -EINVAL;
331 			sbi->active_logs = arg;
332 			break;
333 		case Opt_disable_ext_identify:
334 			set_opt(sbi, DISABLE_EXT_IDENTIFY);
335 			break;
336 		case Opt_inline_data:
337 			set_opt(sbi, INLINE_DATA);
338 			break;
339 		case Opt_flush_merge:
340 			set_opt(sbi, FLUSH_MERGE);
341 			break;
342 		default:
343 			f2fs_msg(sb, KERN_ERR,
344 				"Unrecognized mount option \"%s\" or missing value",
345 				p);
346 			return -EINVAL;
347 		}
348 	}
349 	return 0;
350 }
351 
352 static struct inode *f2fs_alloc_inode(struct super_block *sb)
353 {
354 	struct f2fs_inode_info *fi;
355 
356 	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
357 	if (!fi)
358 		return NULL;
359 
360 	init_once((void *) fi);
361 
362 	/* Initialize f2fs-specific inode info */
363 	fi->vfs_inode.i_version = 1;
364 	atomic_set(&fi->dirty_dents, 0);
365 	fi->i_current_depth = 1;
366 	fi->i_advise = 0;
367 	rwlock_init(&fi->ext.ext_lock);
368 	init_rwsem(&fi->i_sem);
369 
370 	set_inode_flag(fi, FI_NEW_INODE);
371 
372 	if (test_opt(F2FS_SB(sb), INLINE_XATTR))
373 		set_inode_flag(fi, FI_INLINE_XATTR);
374 
375 	/* Will be used by directory only */
376 	fi->i_dir_level = F2FS_SB(sb)->dir_level;
377 
378 	return &fi->vfs_inode;
379 }
380 
381 static int f2fs_drop_inode(struct inode *inode)
382 {
383 	/*
384 	 * This is to avoid a deadlock condition like below.
385 	 * writeback_single_inode(inode)
386 	 *  - f2fs_write_data_page
387 	 *    - f2fs_gc -> iput -> evict
388 	 *       - inode_wait_for_writeback(inode)
389 	 */
390 	if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
391 		return 0;
392 	return generic_drop_inode(inode);
393 }
394 
395 /*
396  * f2fs_dirty_inode() is called from __mark_inode_dirty()
397  *
398  * We should call set_dirty_inode to write the dirty inode through write_inode.
399  */
400 static void f2fs_dirty_inode(struct inode *inode, int flags)
401 {
402 	set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
403 }
404 
405 static void f2fs_i_callback(struct rcu_head *head)
406 {
407 	struct inode *inode = container_of(head, struct inode, i_rcu);
408 	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
409 }
410 
411 static void f2fs_destroy_inode(struct inode *inode)
412 {
413 	call_rcu(&inode->i_rcu, f2fs_i_callback);
414 }
415 
416 static void f2fs_put_super(struct super_block *sb)
417 {
418 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
419 
420 	if (sbi->s_proc) {
421 		remove_proc_entry("segment_info", sbi->s_proc);
422 		remove_proc_entry(sb->s_id, f2fs_proc_root);
423 	}
424 	kobject_del(&sbi->s_kobj);
425 
426 	f2fs_destroy_stats(sbi);
427 	stop_gc_thread(sbi);
428 
429 	/* We don't need to do checkpoint when it's clean */
430 	if (sbi->s_dirty && get_pages(sbi, F2FS_DIRTY_NODES))
431 		write_checkpoint(sbi, true);
432 
433 	iput(sbi->node_inode);
434 	iput(sbi->meta_inode);
435 
436 	/* destroy f2fs internal modules */
437 	destroy_node_manager(sbi);
438 	destroy_segment_manager(sbi);
439 
440 	kfree(sbi->ckpt);
441 	kobject_put(&sbi->s_kobj);
442 	wait_for_completion(&sbi->s_kobj_unregister);
443 
444 	sb->s_fs_info = NULL;
445 	brelse(sbi->raw_super_buf);
446 	kfree(sbi);
447 }
448 
449 int f2fs_sync_fs(struct super_block *sb, int sync)
450 {
451 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
452 
453 	trace_f2fs_sync_fs(sb, sync);
454 
455 	if (!sbi->s_dirty && !get_pages(sbi, F2FS_DIRTY_NODES))
456 		return 0;
457 
458 	if (sync) {
459 		mutex_lock(&sbi->gc_mutex);
460 		write_checkpoint(sbi, false);
461 		mutex_unlock(&sbi->gc_mutex);
462 	} else {
463 		f2fs_balance_fs(sbi);
464 	}
465 
466 	return 0;
467 }
468 
469 static int f2fs_freeze(struct super_block *sb)
470 {
471 	int err;
472 
473 	if (f2fs_readonly(sb))
474 		return 0;
475 
476 	err = f2fs_sync_fs(sb, 1);
477 	return err;
478 }
479 
480 static int f2fs_unfreeze(struct super_block *sb)
481 {
482 	return 0;
483 }
484 
485 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
486 {
487 	struct super_block *sb = dentry->d_sb;
488 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
489 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
490 	block_t total_count, user_block_count, start_count, ovp_count;
491 
492 	total_count = le64_to_cpu(sbi->raw_super->block_count);
493 	user_block_count = sbi->user_block_count;
494 	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
495 	ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
496 	buf->f_type = F2FS_SUPER_MAGIC;
497 	buf->f_bsize = sbi->blocksize;
498 
499 	buf->f_blocks = total_count - start_count;
500 	buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
501 	buf->f_bavail = user_block_count - valid_user_blocks(sbi);
502 
503 	buf->f_files = sbi->total_node_count;
504 	buf->f_ffree = sbi->total_node_count - valid_inode_count(sbi);
505 
506 	buf->f_namelen = F2FS_NAME_LEN;
507 	buf->f_fsid.val[0] = (u32)id;
508 	buf->f_fsid.val[1] = (u32)(id >> 32);
509 
510 	return 0;
511 }
512 
513 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
514 {
515 	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
516 
517 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
518 		seq_printf(seq, ",background_gc=%s", "on");
519 	else
520 		seq_printf(seq, ",background_gc=%s", "off");
521 	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
522 		seq_puts(seq, ",disable_roll_forward");
523 	if (test_opt(sbi, DISCARD))
524 		seq_puts(seq, ",discard");
525 	if (test_opt(sbi, NOHEAP))
526 		seq_puts(seq, ",no_heap_alloc");
527 #ifdef CONFIG_F2FS_FS_XATTR
528 	if (test_opt(sbi, XATTR_USER))
529 		seq_puts(seq, ",user_xattr");
530 	else
531 		seq_puts(seq, ",nouser_xattr");
532 	if (test_opt(sbi, INLINE_XATTR))
533 		seq_puts(seq, ",inline_xattr");
534 #endif
535 #ifdef CONFIG_F2FS_FS_POSIX_ACL
536 	if (test_opt(sbi, POSIX_ACL))
537 		seq_puts(seq, ",acl");
538 	else
539 		seq_puts(seq, ",noacl");
540 #endif
541 	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
542 		seq_puts(seq, ",disable_ext_identify");
543 	if (test_opt(sbi, INLINE_DATA))
544 		seq_puts(seq, ",inline_data");
545 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
546 		seq_puts(seq, ",flush_merge");
547 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
548 
549 	return 0;
550 }
551 
552 static int segment_info_seq_show(struct seq_file *seq, void *offset)
553 {
554 	struct super_block *sb = seq->private;
555 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
556 	unsigned int total_segs =
557 			le32_to_cpu(sbi->raw_super->segment_count_main);
558 	int i;
559 
560 	seq_puts(seq, "format: segment_type|valid_blocks\n"
561 		"segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
562 
563 	for (i = 0; i < total_segs; i++) {
564 		struct seg_entry *se = get_seg_entry(sbi, i);
565 
566 		if ((i % 10) == 0)
567 			seq_printf(seq, "%-5d", i);
568 		seq_printf(seq, "%d|%-3u", se->type,
569 					get_valid_blocks(sbi, i, 1));
570 		if ((i % 10) == 9 || i == (total_segs - 1))
571 			seq_putc(seq, '\n');
572 		else
573 			seq_putc(seq, ' ');
574 	}
575 
576 	return 0;
577 }
578 
579 static int segment_info_open_fs(struct inode *inode, struct file *file)
580 {
581 	return single_open(file, segment_info_seq_show, PDE_DATA(inode));
582 }
583 
584 static const struct file_operations f2fs_seq_segment_info_fops = {
585 	.owner = THIS_MODULE,
586 	.open = segment_info_open_fs,
587 	.read = seq_read,
588 	.llseek = seq_lseek,
589 	.release = single_release,
590 };
591 
592 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
593 {
594 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
595 	struct f2fs_mount_info org_mount_opt;
596 	int err, active_logs;
597 	bool need_restart_gc = false;
598 	bool need_stop_gc = false;
599 
600 	sync_filesystem(sb);
601 
602 	/*
603 	 * Save the old mount options in case we
604 	 * need to restore them.
605 	 */
606 	org_mount_opt = sbi->mount_opt;
607 	active_logs = sbi->active_logs;
608 
609 	/* parse mount options */
610 	err = parse_options(sb, data);
611 	if (err)
612 		goto restore_opts;
613 
614 	/*
615 	 * Previous and new state of filesystem is RO,
616 	 * so skip checking GC and FLUSH_MERGE conditions.
617 	 */
618 	if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
619 		goto skip;
620 
621 	/*
622 	 * We stop the GC thread if FS is mounted as RO
623 	 * or if background_gc = off is passed in mount
624 	 * option. Also sync the filesystem.
625 	 */
626 	if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
627 		if (sbi->gc_thread) {
628 			stop_gc_thread(sbi);
629 			f2fs_sync_fs(sb, 1);
630 			need_restart_gc = true;
631 		}
632 	} else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
633 		err = start_gc_thread(sbi);
634 		if (err)
635 			goto restore_opts;
636 		need_stop_gc = true;
637 	}
638 
639 	/*
640 	 * We stop issue flush thread if FS is mounted as RO
641 	 * or if flush_merge is not passed in mount option.
642 	 */
643 	if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
644 		destroy_flush_cmd_control(sbi);
645 	} else if (test_opt(sbi, FLUSH_MERGE) &&
646 					!sbi->sm_info->cmd_control_info) {
647 		err = create_flush_cmd_control(sbi);
648 		if (err)
649 			goto restore_gc;
650 	}
651 skip:
652 	/* Update the POSIXACL Flag */
653 	 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
654 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
655 	return 0;
656 restore_gc:
657 	if (need_restart_gc) {
658 		if (start_gc_thread(sbi))
659 			f2fs_msg(sbi->sb, KERN_WARNING,
660 				"background gc thread is stop");
661 	} else if (need_stop_gc) {
662 		stop_gc_thread(sbi);
663 	}
664 restore_opts:
665 	sbi->mount_opt = org_mount_opt;
666 	sbi->active_logs = active_logs;
667 	return err;
668 }
669 
670 static struct super_operations f2fs_sops = {
671 	.alloc_inode	= f2fs_alloc_inode,
672 	.drop_inode	= f2fs_drop_inode,
673 	.destroy_inode	= f2fs_destroy_inode,
674 	.write_inode	= f2fs_write_inode,
675 	.dirty_inode	= f2fs_dirty_inode,
676 	.show_options	= f2fs_show_options,
677 	.evict_inode	= f2fs_evict_inode,
678 	.put_super	= f2fs_put_super,
679 	.sync_fs	= f2fs_sync_fs,
680 	.freeze_fs	= f2fs_freeze,
681 	.unfreeze_fs	= f2fs_unfreeze,
682 	.statfs		= f2fs_statfs,
683 	.remount_fs	= f2fs_remount,
684 };
685 
686 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
687 		u64 ino, u32 generation)
688 {
689 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
690 	struct inode *inode;
691 
692 	if (unlikely(ino < F2FS_ROOT_INO(sbi)))
693 		return ERR_PTR(-ESTALE);
694 	if (unlikely(ino >= NM_I(sbi)->max_nid))
695 		return ERR_PTR(-ESTALE);
696 
697 	/*
698 	 * f2fs_iget isn't quite right if the inode is currently unallocated!
699 	 * However f2fs_iget currently does appropriate checks to handle stale
700 	 * inodes so everything is OK.
701 	 */
702 	inode = f2fs_iget(sb, ino);
703 	if (IS_ERR(inode))
704 		return ERR_CAST(inode);
705 	if (unlikely(generation && inode->i_generation != generation)) {
706 		/* we didn't find the right inode.. */
707 		iput(inode);
708 		return ERR_PTR(-ESTALE);
709 	}
710 	return inode;
711 }
712 
713 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
714 		int fh_len, int fh_type)
715 {
716 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
717 				    f2fs_nfs_get_inode);
718 }
719 
720 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
721 		int fh_len, int fh_type)
722 {
723 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
724 				    f2fs_nfs_get_inode);
725 }
726 
727 static const struct export_operations f2fs_export_ops = {
728 	.fh_to_dentry = f2fs_fh_to_dentry,
729 	.fh_to_parent = f2fs_fh_to_parent,
730 	.get_parent = f2fs_get_parent,
731 };
732 
733 static loff_t max_file_size(unsigned bits)
734 {
735 	loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
736 	loff_t leaf_count = ADDRS_PER_BLOCK;
737 
738 	/* two direct node blocks */
739 	result += (leaf_count * 2);
740 
741 	/* two indirect node blocks */
742 	leaf_count *= NIDS_PER_BLOCK;
743 	result += (leaf_count * 2);
744 
745 	/* one double indirect node block */
746 	leaf_count *= NIDS_PER_BLOCK;
747 	result += leaf_count;
748 
749 	result <<= bits;
750 	return result;
751 }
752 
753 static int sanity_check_raw_super(struct super_block *sb,
754 			struct f2fs_super_block *raw_super)
755 {
756 	unsigned int blocksize;
757 
758 	if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
759 		f2fs_msg(sb, KERN_INFO,
760 			"Magic Mismatch, valid(0x%x) - read(0x%x)",
761 			F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
762 		return 1;
763 	}
764 
765 	/* Currently, support only 4KB page cache size */
766 	if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
767 		f2fs_msg(sb, KERN_INFO,
768 			"Invalid page_cache_size (%lu), supports only 4KB\n",
769 			PAGE_CACHE_SIZE);
770 		return 1;
771 	}
772 
773 	/* Currently, support only 4KB block size */
774 	blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
775 	if (blocksize != F2FS_BLKSIZE) {
776 		f2fs_msg(sb, KERN_INFO,
777 			"Invalid blocksize (%u), supports only 4KB\n",
778 			blocksize);
779 		return 1;
780 	}
781 
782 	if (le32_to_cpu(raw_super->log_sectorsize) !=
783 					F2FS_LOG_SECTOR_SIZE) {
784 		f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
785 		return 1;
786 	}
787 	if (le32_to_cpu(raw_super->log_sectors_per_block) !=
788 					F2FS_LOG_SECTORS_PER_BLOCK) {
789 		f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
790 		return 1;
791 	}
792 	return 0;
793 }
794 
795 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
796 {
797 	unsigned int total, fsmeta;
798 	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
799 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
800 
801 	total = le32_to_cpu(raw_super->segment_count);
802 	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
803 	fsmeta += le32_to_cpu(raw_super->segment_count_sit);
804 	fsmeta += le32_to_cpu(raw_super->segment_count_nat);
805 	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
806 	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
807 
808 	if (unlikely(fsmeta >= total))
809 		return 1;
810 
811 	if (unlikely(is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
812 		f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
813 		return 1;
814 	}
815 	return 0;
816 }
817 
818 static void init_sb_info(struct f2fs_sb_info *sbi)
819 {
820 	struct f2fs_super_block *raw_super = sbi->raw_super;
821 	int i;
822 
823 	sbi->log_sectors_per_block =
824 		le32_to_cpu(raw_super->log_sectors_per_block);
825 	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
826 	sbi->blocksize = 1 << sbi->log_blocksize;
827 	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
828 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
829 	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
830 	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
831 	sbi->total_sections = le32_to_cpu(raw_super->section_count);
832 	sbi->total_node_count =
833 		(le32_to_cpu(raw_super->segment_count_nat) / 2)
834 			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
835 	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
836 	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
837 	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
838 	sbi->cur_victim_sec = NULL_SECNO;
839 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
840 
841 	for (i = 0; i < NR_COUNT_TYPE; i++)
842 		atomic_set(&sbi->nr_pages[i], 0);
843 
844 	sbi->dir_level = DEF_DIR_LEVEL;
845 }
846 
847 /*
848  * Read f2fs raw super block.
849  * Because we have two copies of super block, so read the first one at first,
850  * if the first one is invalid, move to read the second one.
851  */
852 static int read_raw_super_block(struct super_block *sb,
853 			struct f2fs_super_block **raw_super,
854 			struct buffer_head **raw_super_buf)
855 {
856 	int block = 0;
857 
858 retry:
859 	*raw_super_buf = sb_bread(sb, block);
860 	if (!*raw_super_buf) {
861 		f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
862 				block + 1);
863 		if (block == 0) {
864 			block++;
865 			goto retry;
866 		} else {
867 			return -EIO;
868 		}
869 	}
870 
871 	*raw_super = (struct f2fs_super_block *)
872 		((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
873 
874 	/* sanity checking of raw super */
875 	if (sanity_check_raw_super(sb, *raw_super)) {
876 		brelse(*raw_super_buf);
877 		f2fs_msg(sb, KERN_ERR,
878 			"Can't find valid F2FS filesystem in %dth superblock",
879 								block + 1);
880 		if (block == 0) {
881 			block++;
882 			goto retry;
883 		} else {
884 			return -EINVAL;
885 		}
886 	}
887 
888 	return 0;
889 }
890 
891 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
892 {
893 	struct f2fs_sb_info *sbi;
894 	struct f2fs_super_block *raw_super;
895 	struct buffer_head *raw_super_buf;
896 	struct inode *root;
897 	long err = -EINVAL;
898 	int i;
899 
900 	/* allocate memory for f2fs-specific super block info */
901 	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
902 	if (!sbi)
903 		return -ENOMEM;
904 
905 	/* set a block size */
906 	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
907 		f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
908 		goto free_sbi;
909 	}
910 
911 	err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
912 	if (err)
913 		goto free_sbi;
914 
915 	sb->s_fs_info = sbi;
916 	/* init some FS parameters */
917 	sbi->active_logs = NR_CURSEG_TYPE;
918 
919 	set_opt(sbi, BG_GC);
920 
921 #ifdef CONFIG_F2FS_FS_XATTR
922 	set_opt(sbi, XATTR_USER);
923 #endif
924 #ifdef CONFIG_F2FS_FS_POSIX_ACL
925 	set_opt(sbi, POSIX_ACL);
926 #endif
927 	/* parse mount options */
928 	err = parse_options(sb, (char *)data);
929 	if (err)
930 		goto free_sb_buf;
931 
932 	sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
933 	sb->s_max_links = F2FS_LINK_MAX;
934 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
935 
936 	sb->s_op = &f2fs_sops;
937 	sb->s_xattr = f2fs_xattr_handlers;
938 	sb->s_export_op = &f2fs_export_ops;
939 	sb->s_magic = F2FS_SUPER_MAGIC;
940 	sb->s_time_gran = 1;
941 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
942 		(test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
943 	memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
944 
945 	/* init f2fs-specific super block info */
946 	sbi->sb = sb;
947 	sbi->raw_super = raw_super;
948 	sbi->raw_super_buf = raw_super_buf;
949 	mutex_init(&sbi->gc_mutex);
950 	mutex_init(&sbi->writepages);
951 	mutex_init(&sbi->cp_mutex);
952 	mutex_init(&sbi->node_write);
953 	sbi->por_doing = false;
954 	spin_lock_init(&sbi->stat_lock);
955 
956 	init_rwsem(&sbi->read_io.io_rwsem);
957 	sbi->read_io.sbi = sbi;
958 	sbi->read_io.bio = NULL;
959 	for (i = 0; i < NR_PAGE_TYPE; i++) {
960 		init_rwsem(&sbi->write_io[i].io_rwsem);
961 		sbi->write_io[i].sbi = sbi;
962 		sbi->write_io[i].bio = NULL;
963 	}
964 
965 	init_rwsem(&sbi->cp_rwsem);
966 	init_waitqueue_head(&sbi->cp_wait);
967 	init_sb_info(sbi);
968 
969 	/* get an inode for meta space */
970 	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
971 	if (IS_ERR(sbi->meta_inode)) {
972 		f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
973 		err = PTR_ERR(sbi->meta_inode);
974 		goto free_sb_buf;
975 	}
976 
977 	err = get_valid_checkpoint(sbi);
978 	if (err) {
979 		f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
980 		goto free_meta_inode;
981 	}
982 
983 	/* sanity checking of checkpoint */
984 	err = -EINVAL;
985 	if (sanity_check_ckpt(sbi)) {
986 		f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
987 		goto free_cp;
988 	}
989 
990 	sbi->total_valid_node_count =
991 				le32_to_cpu(sbi->ckpt->valid_node_count);
992 	sbi->total_valid_inode_count =
993 				le32_to_cpu(sbi->ckpt->valid_inode_count);
994 	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
995 	sbi->total_valid_block_count =
996 				le64_to_cpu(sbi->ckpt->valid_block_count);
997 	sbi->last_valid_block_count = sbi->total_valid_block_count;
998 	sbi->alloc_valid_block_count = 0;
999 	INIT_LIST_HEAD(&sbi->dir_inode_list);
1000 	spin_lock_init(&sbi->dir_inode_lock);
1001 
1002 	init_orphan_info(sbi);
1003 
1004 	/* setup f2fs internal modules */
1005 	err = build_segment_manager(sbi);
1006 	if (err) {
1007 		f2fs_msg(sb, KERN_ERR,
1008 			"Failed to initialize F2FS segment manager");
1009 		goto free_sm;
1010 	}
1011 	err = build_node_manager(sbi);
1012 	if (err) {
1013 		f2fs_msg(sb, KERN_ERR,
1014 			"Failed to initialize F2FS node manager");
1015 		goto free_nm;
1016 	}
1017 
1018 	build_gc_manager(sbi);
1019 
1020 	/* get an inode for node space */
1021 	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1022 	if (IS_ERR(sbi->node_inode)) {
1023 		f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1024 		err = PTR_ERR(sbi->node_inode);
1025 		goto free_nm;
1026 	}
1027 
1028 	/* if there are nt orphan nodes free them */
1029 	recover_orphan_inodes(sbi);
1030 
1031 	/* read root inode and dentry */
1032 	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1033 	if (IS_ERR(root)) {
1034 		f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1035 		err = PTR_ERR(root);
1036 		goto free_node_inode;
1037 	}
1038 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1039 		err = -EINVAL;
1040 		goto free_root_inode;
1041 	}
1042 
1043 	sb->s_root = d_make_root(root); /* allocate root dentry */
1044 	if (!sb->s_root) {
1045 		err = -ENOMEM;
1046 		goto free_root_inode;
1047 	}
1048 
1049 	err = f2fs_build_stats(sbi);
1050 	if (err)
1051 		goto free_root_inode;
1052 
1053 	if (f2fs_proc_root)
1054 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1055 
1056 	if (sbi->s_proc)
1057 		proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1058 				 &f2fs_seq_segment_info_fops, sb);
1059 
1060 	if (test_opt(sbi, DISCARD)) {
1061 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
1062 		if (!blk_queue_discard(q))
1063 			f2fs_msg(sb, KERN_WARNING,
1064 					"mounting with \"discard\" option, but "
1065 					"the device does not support discard");
1066 	}
1067 
1068 	sbi->s_kobj.kset = f2fs_kset;
1069 	init_completion(&sbi->s_kobj_unregister);
1070 	err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1071 							"%s", sb->s_id);
1072 	if (err)
1073 		goto free_proc;
1074 
1075 	/* recover fsynced data */
1076 	if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1077 		err = recover_fsync_data(sbi);
1078 		if (err)
1079 			f2fs_msg(sb, KERN_ERR,
1080 				"Cannot recover all fsync data errno=%ld", err);
1081 	}
1082 
1083 	/*
1084 	 * If filesystem is not mounted as read-only then
1085 	 * do start the gc_thread.
1086 	 */
1087 	if (!(sb->s_flags & MS_RDONLY)) {
1088 		/* After POR, we can run background GC thread.*/
1089 		err = start_gc_thread(sbi);
1090 		if (err)
1091 			goto free_kobj;
1092 	}
1093 	return 0;
1094 
1095 free_kobj:
1096 	kobject_del(&sbi->s_kobj);
1097 free_proc:
1098 	if (sbi->s_proc) {
1099 		remove_proc_entry("segment_info", sbi->s_proc);
1100 		remove_proc_entry(sb->s_id, f2fs_proc_root);
1101 	}
1102 	f2fs_destroy_stats(sbi);
1103 free_root_inode:
1104 	dput(sb->s_root);
1105 	sb->s_root = NULL;
1106 free_node_inode:
1107 	iput(sbi->node_inode);
1108 free_nm:
1109 	destroy_node_manager(sbi);
1110 free_sm:
1111 	destroy_segment_manager(sbi);
1112 free_cp:
1113 	kfree(sbi->ckpt);
1114 free_meta_inode:
1115 	make_bad_inode(sbi->meta_inode);
1116 	iput(sbi->meta_inode);
1117 free_sb_buf:
1118 	brelse(raw_super_buf);
1119 free_sbi:
1120 	kfree(sbi);
1121 	return err;
1122 }
1123 
1124 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1125 			const char *dev_name, void *data)
1126 {
1127 	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1128 }
1129 
1130 static struct file_system_type f2fs_fs_type = {
1131 	.owner		= THIS_MODULE,
1132 	.name		= "f2fs",
1133 	.mount		= f2fs_mount,
1134 	.kill_sb	= kill_block_super,
1135 	.fs_flags	= FS_REQUIRES_DEV,
1136 };
1137 MODULE_ALIAS_FS("f2fs");
1138 
1139 static int __init init_inodecache(void)
1140 {
1141 	f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1142 			sizeof(struct f2fs_inode_info));
1143 	if (!f2fs_inode_cachep)
1144 		return -ENOMEM;
1145 	return 0;
1146 }
1147 
1148 static void destroy_inodecache(void)
1149 {
1150 	/*
1151 	 * Make sure all delayed rcu free inodes are flushed before we
1152 	 * destroy cache.
1153 	 */
1154 	rcu_barrier();
1155 	kmem_cache_destroy(f2fs_inode_cachep);
1156 }
1157 
1158 static int __init init_f2fs_fs(void)
1159 {
1160 	int err;
1161 
1162 	err = init_inodecache();
1163 	if (err)
1164 		goto fail;
1165 	err = create_node_manager_caches();
1166 	if (err)
1167 		goto free_inodecache;
1168 	err = create_segment_manager_caches();
1169 	if (err)
1170 		goto free_node_manager_caches;
1171 	err = create_gc_caches();
1172 	if (err)
1173 		goto free_segment_manager_caches;
1174 	err = create_checkpoint_caches();
1175 	if (err)
1176 		goto free_gc_caches;
1177 	f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1178 	if (!f2fs_kset) {
1179 		err = -ENOMEM;
1180 		goto free_checkpoint_caches;
1181 	}
1182 	err = register_filesystem(&f2fs_fs_type);
1183 	if (err)
1184 		goto free_kset;
1185 	f2fs_create_root_stats();
1186 	f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1187 	return 0;
1188 
1189 free_kset:
1190 	kset_unregister(f2fs_kset);
1191 free_checkpoint_caches:
1192 	destroy_checkpoint_caches();
1193 free_gc_caches:
1194 	destroy_gc_caches();
1195 free_segment_manager_caches:
1196 	destroy_segment_manager_caches();
1197 free_node_manager_caches:
1198 	destroy_node_manager_caches();
1199 free_inodecache:
1200 	destroy_inodecache();
1201 fail:
1202 	return err;
1203 }
1204 
1205 static void __exit exit_f2fs_fs(void)
1206 {
1207 	remove_proc_entry("fs/f2fs", NULL);
1208 	f2fs_destroy_root_stats();
1209 	unregister_filesystem(&f2fs_fs_type);
1210 	destroy_checkpoint_caches();
1211 	destroy_gc_caches();
1212 	destroy_segment_manager_caches();
1213 	destroy_node_manager_caches();
1214 	destroy_inodecache();
1215 	kset_unregister(f2fs_kset);
1216 }
1217 
1218 module_init(init_f2fs_fs)
1219 module_exit(exit_f2fs_fs)
1220 
1221 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1222 MODULE_DESCRIPTION("Flash Friendly File System");
1223 MODULE_LICENSE("GPL");
1224