GNU Linux-libre 4.4.288-gnu1
[releases.git] / fs / f2fs / super.c
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 #include "trace.h"
34
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
37
38 static struct proc_dir_entry *f2fs_proc_root;
39 static struct kmem_cache *f2fs_inode_cachep;
40 static struct kset *f2fs_kset;
41
42 /* f2fs-wide shrinker description */
43 static struct shrinker f2fs_shrinker_info = {
44         .scan_objects = f2fs_shrink_scan,
45         .count_objects = f2fs_shrink_count,
46         .seeks = DEFAULT_SEEKS,
47 };
48
49 enum {
50         Opt_gc_background,
51         Opt_disable_roll_forward,
52         Opt_norecovery,
53         Opt_discard,
54         Opt_noheap,
55         Opt_user_xattr,
56         Opt_nouser_xattr,
57         Opt_acl,
58         Opt_noacl,
59         Opt_active_logs,
60         Opt_disable_ext_identify,
61         Opt_inline_xattr,
62         Opt_inline_data,
63         Opt_inline_dentry,
64         Opt_flush_merge,
65         Opt_nobarrier,
66         Opt_fastboot,
67         Opt_extent_cache,
68         Opt_noextent_cache,
69         Opt_noinline_data,
70         Opt_err,
71 };
72
73 static match_table_t f2fs_tokens = {
74         {Opt_gc_background, "background_gc=%s"},
75         {Opt_disable_roll_forward, "disable_roll_forward"},
76         {Opt_norecovery, "norecovery"},
77         {Opt_discard, "discard"},
78         {Opt_noheap, "no_heap"},
79         {Opt_user_xattr, "user_xattr"},
80         {Opt_nouser_xattr, "nouser_xattr"},
81         {Opt_acl, "acl"},
82         {Opt_noacl, "noacl"},
83         {Opt_active_logs, "active_logs=%u"},
84         {Opt_disable_ext_identify, "disable_ext_identify"},
85         {Opt_inline_xattr, "inline_xattr"},
86         {Opt_inline_data, "inline_data"},
87         {Opt_inline_dentry, "inline_dentry"},
88         {Opt_flush_merge, "flush_merge"},
89         {Opt_nobarrier, "nobarrier"},
90         {Opt_fastboot, "fastboot"},
91         {Opt_extent_cache, "extent_cache"},
92         {Opt_noextent_cache, "noextent_cache"},
93         {Opt_noinline_data, "noinline_data"},
94         {Opt_err, NULL},
95 };
96
97 /* Sysfs support for f2fs */
98 enum {
99         GC_THREAD,      /* struct f2fs_gc_thread */
100         SM_INFO,        /* struct f2fs_sm_info */
101         NM_INFO,        /* struct f2fs_nm_info */
102         F2FS_SBI,       /* struct f2fs_sb_info */
103 };
104
105 struct f2fs_attr {
106         struct attribute attr;
107         ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
108         ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
109                          const char *, size_t);
110         int struct_type;
111         int offset;
112 };
113
114 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
115 {
116         if (struct_type == GC_THREAD)
117                 return (unsigned char *)sbi->gc_thread;
118         else if (struct_type == SM_INFO)
119                 return (unsigned char *)SM_I(sbi);
120         else if (struct_type == NM_INFO)
121                 return (unsigned char *)NM_I(sbi);
122         else if (struct_type == F2FS_SBI)
123                 return (unsigned char *)sbi;
124         return NULL;
125 }
126
127 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
128                         struct f2fs_sb_info *sbi, char *buf)
129 {
130         unsigned char *ptr = NULL;
131         unsigned int *ui;
132
133         ptr = __struct_ptr(sbi, a->struct_type);
134         if (!ptr)
135                 return -EINVAL;
136
137         ui = (unsigned int *)(ptr + a->offset);
138
139         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
140 }
141
142 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
143                         struct f2fs_sb_info *sbi,
144                         const char *buf, size_t count)
145 {
146         unsigned char *ptr;
147         unsigned long t;
148         unsigned int *ui;
149         ssize_t ret;
150
151         ptr = __struct_ptr(sbi, a->struct_type);
152         if (!ptr)
153                 return -EINVAL;
154
155         ui = (unsigned int *)(ptr + a->offset);
156
157         ret = kstrtoul(skip_spaces(buf), 0, &t);
158         if (ret < 0)
159                 return ret;
160         *ui = t;
161         return count;
162 }
163
164 static ssize_t f2fs_attr_show(struct kobject *kobj,
165                                 struct attribute *attr, char *buf)
166 {
167         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
168                                                                 s_kobj);
169         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
170
171         return a->show ? a->show(a, sbi, buf) : 0;
172 }
173
174 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
175                                                 const char *buf, size_t len)
176 {
177         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
178                                                                         s_kobj);
179         struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
180
181         return a->store ? a->store(a, sbi, buf, len) : 0;
182 }
183
184 static void f2fs_sb_release(struct kobject *kobj)
185 {
186         struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
187                                                                 s_kobj);
188         complete(&sbi->s_kobj_unregister);
189 }
190
191 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
192 static struct f2fs_attr f2fs_attr_##_name = {                   \
193         .attr = {.name = __stringify(_name), .mode = _mode },   \
194         .show   = _show,                                        \
195         .store  = _store,                                       \
196         .struct_type = _struct_type,                            \
197         .offset = _offset                                       \
198 }
199
200 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname)    \
201         F2FS_ATTR_OFFSET(struct_type, name, 0644,               \
202                 f2fs_sbi_show, f2fs_sbi_store,                  \
203                 offsetof(struct struct_name, elname))
204
205 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
206 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
207 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
208 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
209 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
210 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
211 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
212 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
213 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
214 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
215 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
216 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
217 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
218 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
219 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, cp_interval);
220
221 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
222 static struct attribute *f2fs_attrs[] = {
223         ATTR_LIST(gc_min_sleep_time),
224         ATTR_LIST(gc_max_sleep_time),
225         ATTR_LIST(gc_no_gc_sleep_time),
226         ATTR_LIST(gc_idle),
227         ATTR_LIST(reclaim_segments),
228         ATTR_LIST(max_small_discards),
229         ATTR_LIST(batched_trim_sections),
230         ATTR_LIST(ipu_policy),
231         ATTR_LIST(min_ipu_util),
232         ATTR_LIST(min_fsync_blocks),
233         ATTR_LIST(max_victim_search),
234         ATTR_LIST(dir_level),
235         ATTR_LIST(ram_thresh),
236         ATTR_LIST(ra_nid_pages),
237         ATTR_LIST(cp_interval),
238         NULL,
239 };
240
241 static const struct sysfs_ops f2fs_attr_ops = {
242         .show   = f2fs_attr_show,
243         .store  = f2fs_attr_store,
244 };
245
246 static struct kobj_type f2fs_ktype = {
247         .default_attrs  = f2fs_attrs,
248         .sysfs_ops      = &f2fs_attr_ops,
249         .release        = f2fs_sb_release,
250 };
251
252 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
253 {
254         struct va_format vaf;
255         va_list args;
256
257         va_start(args, fmt);
258         vaf.fmt = fmt;
259         vaf.va = &args;
260         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
261         va_end(args);
262 }
263
264 static void init_once(void *foo)
265 {
266         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
267
268         inode_init_once(&fi->vfs_inode);
269 }
270
271 static int parse_options(struct super_block *sb, char *options)
272 {
273         struct f2fs_sb_info *sbi = F2FS_SB(sb);
274         struct request_queue *q;
275         substring_t args[MAX_OPT_ARGS];
276         char *p, *name;
277         int arg = 0;
278
279         if (!options)
280                 return 0;
281
282         while ((p = strsep(&options, ",")) != NULL) {
283                 int token;
284                 if (!*p)
285                         continue;
286                 /*
287                  * Initialize args struct so we know whether arg was
288                  * found; some options take optional arguments.
289                  */
290                 args[0].to = args[0].from = NULL;
291                 token = match_token(p, f2fs_tokens, args);
292
293                 switch (token) {
294                 case Opt_gc_background:
295                         name = match_strdup(&args[0]);
296
297                         if (!name)
298                                 return -ENOMEM;
299                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
300                                 set_opt(sbi, BG_GC);
301                                 clear_opt(sbi, FORCE_FG_GC);
302                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
303                                 clear_opt(sbi, BG_GC);
304                                 clear_opt(sbi, FORCE_FG_GC);
305                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
306                                 set_opt(sbi, BG_GC);
307                                 set_opt(sbi, FORCE_FG_GC);
308                         } else {
309                                 kfree(name);
310                                 return -EINVAL;
311                         }
312                         kfree(name);
313                         break;
314                 case Opt_disable_roll_forward:
315                         set_opt(sbi, DISABLE_ROLL_FORWARD);
316                         break;
317                 case Opt_norecovery:
318                         /* this option mounts f2fs with ro */
319                         set_opt(sbi, DISABLE_ROLL_FORWARD);
320                         if (!f2fs_readonly(sb))
321                                 return -EINVAL;
322                         break;
323                 case Opt_discard:
324                         q = bdev_get_queue(sb->s_bdev);
325                         if (blk_queue_discard(q)) {
326                                 set_opt(sbi, DISCARD);
327                         } else {
328                                 f2fs_msg(sb, KERN_WARNING,
329                                         "mounting with \"discard\" option, but "
330                                         "the device does not support discard");
331                         }
332                         break;
333                 case Opt_noheap:
334                         set_opt(sbi, NOHEAP);
335                         break;
336 #ifdef CONFIG_F2FS_FS_XATTR
337                 case Opt_user_xattr:
338                         set_opt(sbi, XATTR_USER);
339                         break;
340                 case Opt_nouser_xattr:
341                         clear_opt(sbi, XATTR_USER);
342                         break;
343                 case Opt_inline_xattr:
344                         set_opt(sbi, INLINE_XATTR);
345                         break;
346 #else
347                 case Opt_user_xattr:
348                         f2fs_msg(sb, KERN_INFO,
349                                 "user_xattr options not supported");
350                         break;
351                 case Opt_nouser_xattr:
352                         f2fs_msg(sb, KERN_INFO,
353                                 "nouser_xattr options not supported");
354                         break;
355                 case Opt_inline_xattr:
356                         f2fs_msg(sb, KERN_INFO,
357                                 "inline_xattr options not supported");
358                         break;
359 #endif
360 #ifdef CONFIG_F2FS_FS_POSIX_ACL
361                 case Opt_acl:
362                         set_opt(sbi, POSIX_ACL);
363                         break;
364                 case Opt_noacl:
365                         clear_opt(sbi, POSIX_ACL);
366                         break;
367 #else
368                 case Opt_acl:
369                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
370                         break;
371                 case Opt_noacl:
372                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
373                         break;
374 #endif
375                 case Opt_active_logs:
376                         if (args->from && match_int(args, &arg))
377                                 return -EINVAL;
378                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
379                                 return -EINVAL;
380                         sbi->active_logs = arg;
381                         break;
382                 case Opt_disable_ext_identify:
383                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
384                         break;
385                 case Opt_inline_data:
386                         set_opt(sbi, INLINE_DATA);
387                         break;
388                 case Opt_inline_dentry:
389                         set_opt(sbi, INLINE_DENTRY);
390                         break;
391                 case Opt_flush_merge:
392                         set_opt(sbi, FLUSH_MERGE);
393                         break;
394                 case Opt_nobarrier:
395                         set_opt(sbi, NOBARRIER);
396                         break;
397                 case Opt_fastboot:
398                         set_opt(sbi, FASTBOOT);
399                         break;
400                 case Opt_extent_cache:
401                         set_opt(sbi, EXTENT_CACHE);
402                         break;
403                 case Opt_noextent_cache:
404                         clear_opt(sbi, EXTENT_CACHE);
405                         break;
406                 case Opt_noinline_data:
407                         clear_opt(sbi, INLINE_DATA);
408                         break;
409                 default:
410                         f2fs_msg(sb, KERN_ERR,
411                                 "Unrecognized mount option \"%s\" or missing value",
412                                 p);
413                         return -EINVAL;
414                 }
415         }
416         return 0;
417 }
418
419 static struct inode *f2fs_alloc_inode(struct super_block *sb)
420 {
421         struct f2fs_inode_info *fi;
422
423         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
424         if (!fi)
425                 return NULL;
426
427         init_once((void *) fi);
428
429         /* Initialize f2fs-specific inode info */
430         fi->vfs_inode.i_version = 1;
431         atomic_set(&fi->dirty_pages, 0);
432         fi->i_current_depth = 1;
433         fi->i_advise = 0;
434         init_rwsem(&fi->i_sem);
435         INIT_LIST_HEAD(&fi->inmem_pages);
436         mutex_init(&fi->inmem_lock);
437
438         set_inode_flag(fi, FI_NEW_INODE);
439
440         if (test_opt(F2FS_SB(sb), INLINE_XATTR))
441                 set_inode_flag(fi, FI_INLINE_XATTR);
442
443         /* Will be used by directory only */
444         fi->i_dir_level = F2FS_SB(sb)->dir_level;
445
446 #ifdef CONFIG_F2FS_FS_ENCRYPTION
447         fi->i_crypt_info = NULL;
448 #endif
449         return &fi->vfs_inode;
450 }
451
452 static int f2fs_drop_inode(struct inode *inode)
453 {
454         /*
455          * This is to avoid a deadlock condition like below.
456          * writeback_single_inode(inode)
457          *  - f2fs_write_data_page
458          *    - f2fs_gc -> iput -> evict
459          *       - inode_wait_for_writeback(inode)
460          */
461         if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
462                 if (!inode->i_nlink && !is_bad_inode(inode)) {
463                         /* to avoid evict_inode call simultaneously */
464                         atomic_inc(&inode->i_count);
465                         spin_unlock(&inode->i_lock);
466
467                         /* some remained atomic pages should discarded */
468                         if (f2fs_is_atomic_file(inode))
469                                 commit_inmem_pages(inode, true);
470
471                         /* should remain fi->extent_tree for writepage */
472                         f2fs_destroy_extent_node(inode);
473
474                         sb_start_intwrite(inode->i_sb);
475                         i_size_write(inode, 0);
476
477                         if (F2FS_HAS_BLOCKS(inode))
478                                 f2fs_truncate(inode, true);
479
480                         sb_end_intwrite(inode->i_sb);
481
482 #ifdef CONFIG_F2FS_FS_ENCRYPTION
483                         if (F2FS_I(inode)->i_crypt_info)
484                                 f2fs_free_encryption_info(inode,
485                                         F2FS_I(inode)->i_crypt_info);
486 #endif
487                         spin_lock(&inode->i_lock);
488                         atomic_dec(&inode->i_count);
489                 }
490                 return 0;
491         }
492         return generic_drop_inode(inode);
493 }
494
495 /*
496  * f2fs_dirty_inode() is called from __mark_inode_dirty()
497  *
498  * We should call set_dirty_inode to write the dirty inode through write_inode.
499  */
500 static void f2fs_dirty_inode(struct inode *inode, int flags)
501 {
502         set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
503 }
504
505 static void f2fs_i_callback(struct rcu_head *head)
506 {
507         struct inode *inode = container_of(head, struct inode, i_rcu);
508         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
509 }
510
511 static void f2fs_destroy_inode(struct inode *inode)
512 {
513         call_rcu(&inode->i_rcu, f2fs_i_callback);
514 }
515
516 static void f2fs_put_super(struct super_block *sb)
517 {
518         struct f2fs_sb_info *sbi = F2FS_SB(sb);
519
520         if (sbi->s_proc) {
521                 remove_proc_entry("segment_info", sbi->s_proc);
522                 remove_proc_entry(sb->s_id, f2fs_proc_root);
523         }
524         kobject_del(&sbi->s_kobj);
525
526         stop_gc_thread(sbi);
527
528         /* prevent remaining shrinker jobs */
529         mutex_lock(&sbi->umount_mutex);
530
531         /*
532          * We don't need to do checkpoint when superblock is clean.
533          * But, the previous checkpoint was not done by umount, it needs to do
534          * clean checkpoint again.
535          */
536         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
537                         !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
538                 struct cp_control cpc = {
539                         .reason = CP_UMOUNT,
540                 };
541                 write_checkpoint(sbi, &cpc);
542         }
543
544         /* write_checkpoint can update stat informaion */
545         f2fs_destroy_stats(sbi);
546
547         /*
548          * normally superblock is clean, so we need to release this.
549          * In addition, EIO will skip do checkpoint, we need this as well.
550          */
551         release_dirty_inode(sbi);
552         release_discard_addrs(sbi);
553
554         f2fs_leave_shrinker(sbi);
555         mutex_unlock(&sbi->umount_mutex);
556
557         iput(sbi->node_inode);
558         iput(sbi->meta_inode);
559
560         /* destroy f2fs internal modules */
561         destroy_node_manager(sbi);
562         destroy_segment_manager(sbi);
563
564         kfree(sbi->ckpt);
565         kobject_put(&sbi->s_kobj);
566         wait_for_completion(&sbi->s_kobj_unregister);
567
568         sb->s_fs_info = NULL;
569         brelse(sbi->raw_super_buf);
570         kfree(sbi);
571 }
572
573 int f2fs_sync_fs(struct super_block *sb, int sync)
574 {
575         struct f2fs_sb_info *sbi = F2FS_SB(sb);
576
577         trace_f2fs_sync_fs(sb, sync);
578
579         if (sync) {
580                 struct cp_control cpc;
581
582                 cpc.reason = __get_cp_reason(sbi);
583
584                 mutex_lock(&sbi->gc_mutex);
585                 write_checkpoint(sbi, &cpc);
586                 mutex_unlock(&sbi->gc_mutex);
587         } else {
588                 f2fs_balance_fs(sbi);
589         }
590         f2fs_trace_ios(NULL, 1);
591
592         return 0;
593 }
594
595 static int f2fs_freeze(struct super_block *sb)
596 {
597         int err;
598
599         if (f2fs_readonly(sb))
600                 return 0;
601
602         err = f2fs_sync_fs(sb, 1);
603         return err;
604 }
605
606 static int f2fs_unfreeze(struct super_block *sb)
607 {
608         return 0;
609 }
610
611 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
612 {
613         struct super_block *sb = dentry->d_sb;
614         struct f2fs_sb_info *sbi = F2FS_SB(sb);
615         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
616         block_t total_count, user_block_count, start_count, ovp_count;
617
618         total_count = le64_to_cpu(sbi->raw_super->block_count);
619         user_block_count = sbi->user_block_count;
620         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
621         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
622         buf->f_type = F2FS_SUPER_MAGIC;
623         buf->f_bsize = sbi->blocksize;
624
625         buf->f_blocks = total_count - start_count;
626         buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
627         buf->f_bavail = user_block_count - valid_user_blocks(sbi);
628
629         buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
630         buf->f_ffree = buf->f_files - valid_inode_count(sbi);
631
632         buf->f_namelen = F2FS_NAME_LEN;
633         buf->f_fsid.val[0] = (u32)id;
634         buf->f_fsid.val[1] = (u32)(id >> 32);
635
636         return 0;
637 }
638
639 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
640 {
641         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
642
643         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
644                 if (test_opt(sbi, FORCE_FG_GC))
645                         seq_printf(seq, ",background_gc=%s", "sync");
646                 else
647                         seq_printf(seq, ",background_gc=%s", "on");
648         } else {
649                 seq_printf(seq, ",background_gc=%s", "off");
650         }
651         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
652                 seq_puts(seq, ",disable_roll_forward");
653         if (test_opt(sbi, DISCARD))
654                 seq_puts(seq, ",discard");
655         if (test_opt(sbi, NOHEAP))
656                 seq_puts(seq, ",no_heap_alloc");
657 #ifdef CONFIG_F2FS_FS_XATTR
658         if (test_opt(sbi, XATTR_USER))
659                 seq_puts(seq, ",user_xattr");
660         else
661                 seq_puts(seq, ",nouser_xattr");
662         if (test_opt(sbi, INLINE_XATTR))
663                 seq_puts(seq, ",inline_xattr");
664 #endif
665 #ifdef CONFIG_F2FS_FS_POSIX_ACL
666         if (test_opt(sbi, POSIX_ACL))
667                 seq_puts(seq, ",acl");
668         else
669                 seq_puts(seq, ",noacl");
670 #endif
671         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
672                 seq_puts(seq, ",disable_ext_identify");
673         if (test_opt(sbi, INLINE_DATA))
674                 seq_puts(seq, ",inline_data");
675         else
676                 seq_puts(seq, ",noinline_data");
677         if (test_opt(sbi, INLINE_DENTRY))
678                 seq_puts(seq, ",inline_dentry");
679         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
680                 seq_puts(seq, ",flush_merge");
681         if (test_opt(sbi, NOBARRIER))
682                 seq_puts(seq, ",nobarrier");
683         if (test_opt(sbi, FASTBOOT))
684                 seq_puts(seq, ",fastboot");
685         if (test_opt(sbi, EXTENT_CACHE))
686                 seq_puts(seq, ",extent_cache");
687         else
688                 seq_puts(seq, ",noextent_cache");
689         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
690
691         return 0;
692 }
693
694 static int segment_info_seq_show(struct seq_file *seq, void *offset)
695 {
696         struct super_block *sb = seq->private;
697         struct f2fs_sb_info *sbi = F2FS_SB(sb);
698         unsigned int total_segs =
699                         le32_to_cpu(sbi->raw_super->segment_count_main);
700         int i;
701
702         seq_puts(seq, "format: segment_type|valid_blocks\n"
703                 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
704
705         for (i = 0; i < total_segs; i++) {
706                 struct seg_entry *se = get_seg_entry(sbi, i);
707
708                 if ((i % 10) == 0)
709                         seq_printf(seq, "%-10d", i);
710                 seq_printf(seq, "%d|%-3u", se->type,
711                                         get_valid_blocks(sbi, i, 1));
712                 if ((i % 10) == 9 || i == (total_segs - 1))
713                         seq_putc(seq, '\n');
714                 else
715                         seq_putc(seq, ' ');
716         }
717
718         return 0;
719 }
720
721 static int segment_info_open_fs(struct inode *inode, struct file *file)
722 {
723         return single_open(file, segment_info_seq_show, PDE_DATA(inode));
724 }
725
726 static const struct file_operations f2fs_seq_segment_info_fops = {
727         .owner = THIS_MODULE,
728         .open = segment_info_open_fs,
729         .read = seq_read,
730         .llseek = seq_lseek,
731         .release = single_release,
732 };
733
734 static void default_options(struct f2fs_sb_info *sbi)
735 {
736         /* init some FS parameters */
737         sbi->active_logs = NR_CURSEG_TYPE;
738
739         set_opt(sbi, BG_GC);
740         set_opt(sbi, INLINE_DATA);
741         set_opt(sbi, EXTENT_CACHE);
742
743 #ifdef CONFIG_F2FS_FS_XATTR
744         set_opt(sbi, XATTR_USER);
745 #endif
746 #ifdef CONFIG_F2FS_FS_POSIX_ACL
747         set_opt(sbi, POSIX_ACL);
748 #endif
749 }
750
751 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
752 {
753         struct f2fs_sb_info *sbi = F2FS_SB(sb);
754         struct f2fs_mount_info org_mount_opt;
755         int err, active_logs;
756         bool need_restart_gc = false;
757         bool need_stop_gc = false;
758         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
759
760         sync_filesystem(sb);
761
762         /*
763          * Save the old mount options in case we
764          * need to restore them.
765          */
766         org_mount_opt = sbi->mount_opt;
767         active_logs = sbi->active_logs;
768
769         sbi->mount_opt.opt = 0;
770         default_options(sbi);
771
772         /* parse mount options */
773         err = parse_options(sb, data);
774         if (err)
775                 goto restore_opts;
776
777         /*
778          * Previous and new state of filesystem is RO,
779          * so skip checking GC and FLUSH_MERGE conditions.
780          */
781         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
782                 goto skip;
783
784         /* disallow enable/disable extent_cache dynamically */
785         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
786                 err = -EINVAL;
787                 f2fs_msg(sbi->sb, KERN_WARNING,
788                                 "switch extent_cache option is not allowed");
789                 goto restore_opts;
790         }
791
792         /*
793          * We stop the GC thread if FS is mounted as RO
794          * or if background_gc = off is passed in mount
795          * option. Also sync the filesystem.
796          */
797         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
798                 if (sbi->gc_thread) {
799                         stop_gc_thread(sbi);
800                         f2fs_sync_fs(sb, 1);
801                         need_restart_gc = true;
802                 }
803         } else if (!sbi->gc_thread) {
804                 err = start_gc_thread(sbi);
805                 if (err)
806                         goto restore_opts;
807                 need_stop_gc = true;
808         }
809
810         /*
811          * We stop issue flush thread if FS is mounted as RO
812          * or if flush_merge is not passed in mount option.
813          */
814         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
815                 destroy_flush_cmd_control(sbi);
816         } else if (!SM_I(sbi)->cmd_control_info) {
817                 err = create_flush_cmd_control(sbi);
818                 if (err)
819                         goto restore_gc;
820         }
821 skip:
822         /* Update the POSIXACL Flag */
823          sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
824                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
825         return 0;
826 restore_gc:
827         if (need_restart_gc) {
828                 if (start_gc_thread(sbi))
829                         f2fs_msg(sbi->sb, KERN_WARNING,
830                                 "background gc thread has stopped");
831         } else if (need_stop_gc) {
832                 stop_gc_thread(sbi);
833         }
834 restore_opts:
835         sbi->mount_opt = org_mount_opt;
836         sbi->active_logs = active_logs;
837         return err;
838 }
839
840 static struct super_operations f2fs_sops = {
841         .alloc_inode    = f2fs_alloc_inode,
842         .drop_inode     = f2fs_drop_inode,
843         .destroy_inode  = f2fs_destroy_inode,
844         .write_inode    = f2fs_write_inode,
845         .dirty_inode    = f2fs_dirty_inode,
846         .show_options   = f2fs_show_options,
847         .evict_inode    = f2fs_evict_inode,
848         .put_super      = f2fs_put_super,
849         .sync_fs        = f2fs_sync_fs,
850         .freeze_fs      = f2fs_freeze,
851         .unfreeze_fs    = f2fs_unfreeze,
852         .statfs         = f2fs_statfs,
853         .remount_fs     = f2fs_remount,
854 };
855
856 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
857                 u64 ino, u32 generation)
858 {
859         struct f2fs_sb_info *sbi = F2FS_SB(sb);
860         struct inode *inode;
861
862         if (check_nid_range(sbi, ino))
863                 return ERR_PTR(-ESTALE);
864
865         /*
866          * f2fs_iget isn't quite right if the inode is currently unallocated!
867          * However f2fs_iget currently does appropriate checks to handle stale
868          * inodes so everything is OK.
869          */
870         inode = f2fs_iget(sb, ino);
871         if (IS_ERR(inode))
872                 return ERR_CAST(inode);
873         if (unlikely(generation && inode->i_generation != generation)) {
874                 /* we didn't find the right inode.. */
875                 iput(inode);
876                 return ERR_PTR(-ESTALE);
877         }
878         return inode;
879 }
880
881 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
882                 int fh_len, int fh_type)
883 {
884         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
885                                     f2fs_nfs_get_inode);
886 }
887
888 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
889                 int fh_len, int fh_type)
890 {
891         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
892                                     f2fs_nfs_get_inode);
893 }
894
895 static const struct export_operations f2fs_export_ops = {
896         .fh_to_dentry = f2fs_fh_to_dentry,
897         .fh_to_parent = f2fs_fh_to_parent,
898         .get_parent = f2fs_get_parent,
899 };
900
901 static loff_t max_file_size(unsigned bits)
902 {
903         loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
904         loff_t leaf_count = ADDRS_PER_BLOCK;
905
906         /* two direct node blocks */
907         result += (leaf_count * 2);
908
909         /* two indirect node blocks */
910         leaf_count *= NIDS_PER_BLOCK;
911         result += (leaf_count * 2);
912
913         /* one double indirect node block */
914         leaf_count *= NIDS_PER_BLOCK;
915         result += leaf_count;
916
917         result <<= bits;
918         return result;
919 }
920
921 static inline bool sanity_check_area_boundary(struct super_block *sb,
922                                         struct f2fs_super_block *raw_super)
923 {
924         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
925         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
926         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
927         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
928         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
929         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
930         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
931         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
932         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
933         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
934         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
935         u32 segment_count = le32_to_cpu(raw_super->segment_count);
936         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
937
938         if (segment0_blkaddr != cp_blkaddr) {
939                 f2fs_msg(sb, KERN_INFO,
940                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
941                         segment0_blkaddr, cp_blkaddr);
942                 return true;
943         }
944
945         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
946                                                         sit_blkaddr) {
947                 f2fs_msg(sb, KERN_INFO,
948                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
949                         cp_blkaddr, sit_blkaddr,
950                         segment_count_ckpt << log_blocks_per_seg);
951                 return true;
952         }
953
954         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
955                                                         nat_blkaddr) {
956                 f2fs_msg(sb, KERN_INFO,
957                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
958                         sit_blkaddr, nat_blkaddr,
959                         segment_count_sit << log_blocks_per_seg);
960                 return true;
961         }
962
963         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
964                                                         ssa_blkaddr) {
965                 f2fs_msg(sb, KERN_INFO,
966                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
967                         nat_blkaddr, ssa_blkaddr,
968                         segment_count_nat << log_blocks_per_seg);
969                 return true;
970         }
971
972         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
973                                                         main_blkaddr) {
974                 f2fs_msg(sb, KERN_INFO,
975                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
976                         ssa_blkaddr, main_blkaddr,
977                         segment_count_ssa << log_blocks_per_seg);
978                 return true;
979         }
980
981         if (main_blkaddr + (segment_count_main << log_blocks_per_seg) !=
982                 segment0_blkaddr + (segment_count << log_blocks_per_seg)) {
983                 f2fs_msg(sb, KERN_INFO,
984                         "Wrong MAIN_AREA boundary, start(%u) end(%u) blocks(%u)",
985                         main_blkaddr,
986                         segment0_blkaddr + (segment_count << log_blocks_per_seg),
987                         segment_count_main << log_blocks_per_seg);
988                 return true;
989         }
990
991         return false;
992 }
993
994 static int sanity_check_raw_super(struct super_block *sb,
995                         struct f2fs_super_block *raw_super)
996 {
997         block_t segment_count, segs_per_sec, secs_per_zone;
998         block_t total_sections, blocks_per_seg;
999         unsigned int blocksize;
1000
1001         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1002                 f2fs_msg(sb, KERN_INFO,
1003                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1004                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1005                 return 1;
1006         }
1007
1008         /* Currently, support only 4KB page cache size */
1009         if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
1010                 f2fs_msg(sb, KERN_INFO,
1011                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1012                         PAGE_CACHE_SIZE);
1013                 return 1;
1014         }
1015
1016         /* Currently, support only 4KB block size */
1017         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1018         if (blocksize != F2FS_BLKSIZE) {
1019                 f2fs_msg(sb, KERN_INFO,
1020                         "Invalid blocksize (%u), supports only 4KB\n",
1021                         blocksize);
1022                 return 1;
1023         }
1024
1025         /* check log blocks per segment */
1026         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1027                 f2fs_msg(sb, KERN_INFO,
1028                         "Invalid log blocks per segment (%u)\n",
1029                         le32_to_cpu(raw_super->log_blocks_per_seg));
1030                 return 1;
1031         }
1032
1033         /* Currently, support 512/1024/2048/4096 bytes sector size */
1034         if (le32_to_cpu(raw_super->log_sectorsize) >
1035                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1036                 le32_to_cpu(raw_super->log_sectorsize) <
1037                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1038                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1039                         le32_to_cpu(raw_super->log_sectorsize));
1040                 return 1;
1041         }
1042         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1043                 le32_to_cpu(raw_super->log_sectorsize) !=
1044                         F2FS_MAX_LOG_SECTOR_SIZE) {
1045                 f2fs_msg(sb, KERN_INFO,
1046                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1047                         le32_to_cpu(raw_super->log_sectors_per_block),
1048                         le32_to_cpu(raw_super->log_sectorsize));
1049                 return 1;
1050         }
1051
1052         segment_count = le32_to_cpu(raw_super->segment_count);
1053         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1054         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1055         total_sections = le32_to_cpu(raw_super->section_count);
1056
1057         /* blocks_per_seg should be 512, given the above check */
1058         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
1059
1060         if (segment_count > F2FS_MAX_SEGMENT ||
1061                                 segment_count < F2FS_MIN_SEGMENTS) {
1062                 f2fs_msg(sb, KERN_INFO,
1063                         "Invalid segment count (%u)",
1064                         segment_count);
1065                 return 1;
1066         }
1067
1068         if (total_sections > segment_count ||
1069                         total_sections < F2FS_MIN_SEGMENTS ||
1070                         segs_per_sec > segment_count || !segs_per_sec) {
1071                 f2fs_msg(sb, KERN_INFO,
1072                         "Invalid segment/section count (%u, %u x %u)",
1073                         segment_count, total_sections, segs_per_sec);
1074                 return 1;
1075         }
1076
1077         if ((segment_count / segs_per_sec) < total_sections) {
1078                 f2fs_msg(sb, KERN_INFO,
1079                         "Small segment_count (%u < %u * %u)",
1080                         segment_count, segs_per_sec, total_sections);
1081                 return 1;
1082         }
1083
1084         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
1085                 f2fs_msg(sb, KERN_INFO,
1086                         "Wrong segment_count / block_count (%u > %llu)",
1087                         segment_count, le64_to_cpu(raw_super->block_count));
1088                 return 1;
1089         }
1090
1091         if (secs_per_zone > total_sections || !secs_per_zone) {
1092                 f2fs_msg(sb, KERN_INFO,
1093                         "Wrong secs_per_zone / total_sections (%u, %u)",
1094                         secs_per_zone, total_sections);
1095                 return 1;
1096         }
1097         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION) {
1098                 f2fs_msg(sb, KERN_INFO,
1099                         "Corrupted extension count (%u > %u)",
1100                         le32_to_cpu(raw_super->extension_count),
1101                         F2FS_MAX_EXTENSION);
1102                 return 1;
1103         }
1104
1105         if (le32_to_cpu(raw_super->cp_payload) >
1106                                 (blocks_per_seg - F2FS_CP_PACKS)) {
1107                 f2fs_msg(sb, KERN_INFO,
1108                         "Insane cp_payload (%u > %u)",
1109                         le32_to_cpu(raw_super->cp_payload),
1110                         blocks_per_seg - F2FS_CP_PACKS);
1111                 return 1;
1112         }
1113
1114         /* check reserved ino info */
1115         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1116                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1117                 le32_to_cpu(raw_super->root_ino) != 3) {
1118                 f2fs_msg(sb, KERN_INFO,
1119                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1120                         le32_to_cpu(raw_super->node_ino),
1121                         le32_to_cpu(raw_super->meta_ino),
1122                         le32_to_cpu(raw_super->root_ino));
1123                 return 1;
1124         }
1125
1126         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1127         if (sanity_check_area_boundary(sb, raw_super))
1128                 return 1;
1129
1130         return 0;
1131 }
1132
1133 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1134 {
1135         unsigned int total, fsmeta;
1136         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1137         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1138         unsigned int ovp_segments, reserved_segments;
1139         unsigned int main_segs, blocks_per_seg;
1140         unsigned int sit_segs, nat_segs;
1141         unsigned int sit_bitmap_size, nat_bitmap_size;
1142         unsigned int log_blocks_per_seg;
1143         unsigned int segment_count_main;
1144         unsigned int cp_pack_start_sum, cp_payload;
1145         block_t user_block_count;
1146         int i, j;
1147
1148         total = le32_to_cpu(raw_super->segment_count);
1149         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1150         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
1151         fsmeta += sit_segs;
1152         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
1153         fsmeta += nat_segs;
1154         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1155         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1156
1157         if (unlikely(fsmeta >= total))
1158                 return 1;
1159
1160         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1161         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1162
1163         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1164                         ovp_segments == 0 || reserved_segments == 0)) {
1165                 f2fs_msg(sbi->sb, KERN_ERR,
1166                         "Wrong layout: check mkfs.f2fs version");
1167                 return 1;
1168         }
1169
1170         user_block_count = le64_to_cpu(ckpt->user_block_count);
1171         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1172         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1173         if (!user_block_count || user_block_count >=
1174                         segment_count_main << log_blocks_per_seg) {
1175                 f2fs_msg(sbi->sb, KERN_ERR,
1176                         "Wrong user_block_count: %u", user_block_count);
1177                 return 1;
1178         }
1179
1180         main_segs = le32_to_cpu(raw_super->segment_count_main);
1181         blocks_per_seg = sbi->blocks_per_seg;
1182
1183         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
1184                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
1185                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
1186                         return 1;
1187                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
1188                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
1189                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
1190                                 f2fs_msg(sbi->sb, KERN_ERR,
1191                                         "Node segment (%u, %u) has the same "
1192                                         "segno: %u", i, j,
1193                                         le32_to_cpu(ckpt->cur_node_segno[i]));
1194                                 return 1;
1195                         }
1196                 }
1197         }
1198         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
1199                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
1200                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
1201                         return 1;
1202                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
1203                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
1204                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
1205                                 f2fs_msg(sbi->sb, KERN_ERR,
1206                                         "Data segment (%u, %u) has the same "
1207                                         "segno: %u", i, j,
1208                                         le32_to_cpu(ckpt->cur_data_segno[i]));
1209                                 return 1;
1210                         }
1211                 }
1212         }
1213         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
1214                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
1215                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
1216                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
1217                                 f2fs_msg(sbi->sb, KERN_ERR,
1218                                         "Node segment (%u) and Data segment (%u)"
1219                                         " has the same segno: %u", i, j,
1220                                         le32_to_cpu(ckpt->cur_node_segno[i]));
1221                                 return 1;
1222                         }
1223                 }
1224         }
1225
1226         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
1227         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
1228
1229         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
1230                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
1231                 f2fs_msg(sbi->sb, KERN_ERR,
1232                         "Wrong bitmap size: sit: %u, nat:%u",
1233                         sit_bitmap_size, nat_bitmap_size);
1234                 return 1;
1235         }
1236
1237         cp_pack_start_sum = __start_sum_addr(sbi);
1238         cp_payload = __cp_payload(sbi);
1239         if (cp_pack_start_sum < cp_payload + 1 ||
1240                 cp_pack_start_sum > blocks_per_seg - 1 -
1241                         NR_CURSEG_TYPE) {
1242                 f2fs_msg(sbi->sb, KERN_ERR,
1243                         "Wrong cp_pack_start_sum: %u",
1244                         cp_pack_start_sum);
1245                 return 1;
1246         }
1247
1248         if (unlikely(f2fs_cp_error(sbi))) {
1249                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1250                 return 1;
1251         }
1252         return 0;
1253 }
1254
1255 static void init_sb_info(struct f2fs_sb_info *sbi)
1256 {
1257         struct f2fs_super_block *raw_super = sbi->raw_super;
1258         int i;
1259
1260         sbi->log_sectors_per_block =
1261                 le32_to_cpu(raw_super->log_sectors_per_block);
1262         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1263         sbi->blocksize = 1 << sbi->log_blocksize;
1264         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1265         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1266         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1267         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1268         sbi->total_sections = le32_to_cpu(raw_super->section_count);
1269         sbi->total_node_count =
1270                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1271                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1272         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1273         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1274         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1275         sbi->cur_victim_sec = NULL_SECNO;
1276         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1277
1278         for (i = 0; i < NR_COUNT_TYPE; i++)
1279                 atomic_set(&sbi->nr_pages[i], 0);
1280
1281         sbi->dir_level = DEF_DIR_LEVEL;
1282         sbi->cp_interval = DEF_CP_INTERVAL;
1283         clear_sbi_flag(sbi, SBI_NEED_FSCK);
1284
1285         INIT_LIST_HEAD(&sbi->s_list);
1286         mutex_init(&sbi->umount_mutex);
1287 }
1288
1289 /*
1290  * Read f2fs raw super block.
1291  * Because we have two copies of super block, so read the first one at first,
1292  * if the first one is invalid, move to read the second one.
1293  */
1294 static int read_raw_super_block(struct super_block *sb,
1295                         struct f2fs_super_block **raw_super,
1296                         struct buffer_head **raw_super_buf,
1297                         int *recovery)
1298 {
1299         int block = 0;
1300         struct buffer_head *buffer;
1301         struct f2fs_super_block *super;
1302         int err = 0;
1303
1304 retry:
1305         buffer = sb_bread(sb, block);
1306         if (!buffer) {
1307                 *recovery = 1;
1308                 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1309                                 block + 1);
1310                 if (block == 0) {
1311                         block++;
1312                         goto retry;
1313                 } else {
1314                         err = -EIO;
1315                         goto out;
1316                 }
1317         }
1318
1319         super = (struct f2fs_super_block *)
1320                 ((char *)(buffer)->b_data + F2FS_SUPER_OFFSET);
1321
1322         /* sanity checking of raw super */
1323         if (sanity_check_raw_super(sb, super)) {
1324                 brelse(buffer);
1325                 *recovery = 1;
1326                 f2fs_msg(sb, KERN_ERR,
1327                         "Can't find valid F2FS filesystem in %dth superblock",
1328                                                                 block + 1);
1329                 if (block == 0) {
1330                         block++;
1331                         goto retry;
1332                 } else {
1333                         err = -EINVAL;
1334                         goto out;
1335                 }
1336         }
1337
1338         if (!*raw_super) {
1339                 *raw_super_buf = buffer;
1340                 *raw_super = super;
1341         } else {
1342                 /* already have a valid superblock */
1343                 brelse(buffer);
1344         }
1345
1346         /* check the validity of the second superblock */
1347         if (block == 0) {
1348                 block++;
1349                 goto retry;
1350         }
1351
1352 out:
1353         /* No valid superblock */
1354         if (!*raw_super)
1355                 return err;
1356
1357         return 0;
1358 }
1359
1360 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1361 {
1362         struct buffer_head *sbh = sbi->raw_super_buf;
1363         sector_t block = sbh->b_blocknr;
1364         int err;
1365
1366         /* write back-up superblock first */
1367         sbh->b_blocknr = block ? 0 : 1;
1368         mark_buffer_dirty(sbh);
1369         err = sync_dirty_buffer(sbh);
1370
1371         sbh->b_blocknr = block;
1372
1373         /* if we are in recovery path, skip writing valid superblock */
1374         if (recover || err)
1375                 goto out;
1376
1377         /* write current valid superblock */
1378         mark_buffer_dirty(sbh);
1379         err = sync_dirty_buffer(sbh);
1380 out:
1381         clear_buffer_write_io_error(sbh);
1382         set_buffer_uptodate(sbh);
1383         return err;
1384 }
1385
1386 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1387 {
1388         struct f2fs_sb_info *sbi;
1389         struct f2fs_super_block *raw_super;
1390         struct buffer_head *raw_super_buf;
1391         struct inode *root;
1392         long err;
1393         bool retry = true, need_fsck = false;
1394         char *options = NULL;
1395         int recovery, i;
1396
1397 try_onemore:
1398         err = -EINVAL;
1399         raw_super = NULL;
1400         raw_super_buf = NULL;
1401         recovery = 0;
1402
1403         /* allocate memory for f2fs-specific super block info */
1404         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1405         if (!sbi)
1406                 return -ENOMEM;
1407
1408         /* set a block size */
1409         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1410                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1411                 goto free_sbi;
1412         }
1413
1414         err = read_raw_super_block(sb, &raw_super, &raw_super_buf, &recovery);
1415         if (err)
1416                 goto free_sbi;
1417
1418         sb->s_fs_info = sbi;
1419         default_options(sbi);
1420         /* parse mount options */
1421         options = kstrdup((const char *)data, GFP_KERNEL);
1422         if (data && !options) {
1423                 err = -ENOMEM;
1424                 goto free_sb_buf;
1425         }
1426
1427         err = parse_options(sb, options);
1428         if (err)
1429                 goto free_options;
1430
1431         sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
1432         sb->s_max_links = F2FS_LINK_MAX;
1433         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1434
1435         sb->s_op = &f2fs_sops;
1436         sb->s_xattr = f2fs_xattr_handlers;
1437         sb->s_export_op = &f2fs_export_ops;
1438         sb->s_magic = F2FS_SUPER_MAGIC;
1439         sb->s_time_gran = 1;
1440         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1441                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1442         memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1443
1444         /* init f2fs-specific super block info */
1445         sbi->sb = sb;
1446         sbi->raw_super = raw_super;
1447         sbi->raw_super_buf = raw_super_buf;
1448         mutex_init(&sbi->gc_mutex);
1449         mutex_init(&sbi->writepages);
1450         mutex_init(&sbi->cp_mutex);
1451         init_rwsem(&sbi->node_write);
1452
1453         /* disallow all the data/node/meta page writes */
1454         set_sbi_flag(sbi, SBI_POR_DOING);
1455         spin_lock_init(&sbi->stat_lock);
1456
1457         init_rwsem(&sbi->read_io.io_rwsem);
1458         sbi->read_io.sbi = sbi;
1459         sbi->read_io.bio = NULL;
1460         for (i = 0; i < NR_PAGE_TYPE; i++) {
1461                 init_rwsem(&sbi->write_io[i].io_rwsem);
1462                 sbi->write_io[i].sbi = sbi;
1463                 sbi->write_io[i].bio = NULL;
1464         }
1465
1466         init_rwsem(&sbi->cp_rwsem);
1467         init_waitqueue_head(&sbi->cp_wait);
1468         init_sb_info(sbi);
1469
1470         /* get an inode for meta space */
1471         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1472         if (IS_ERR(sbi->meta_inode)) {
1473                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1474                 err = PTR_ERR(sbi->meta_inode);
1475                 goto free_options;
1476         }
1477
1478         err = get_valid_checkpoint(sbi);
1479         if (err) {
1480                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1481                 goto free_meta_inode;
1482         }
1483
1484         sbi->total_valid_node_count =
1485                                 le32_to_cpu(sbi->ckpt->valid_node_count);
1486         sbi->total_valid_inode_count =
1487                                 le32_to_cpu(sbi->ckpt->valid_inode_count);
1488         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1489         sbi->total_valid_block_count =
1490                                 le64_to_cpu(sbi->ckpt->valid_block_count);
1491         sbi->last_valid_block_count = sbi->total_valid_block_count;
1492         sbi->alloc_valid_block_count = 0;
1493         INIT_LIST_HEAD(&sbi->dir_inode_list);
1494         spin_lock_init(&sbi->dir_inode_lock);
1495
1496         init_extent_cache_info(sbi);
1497
1498         init_ino_entry_info(sbi);
1499
1500         /* setup f2fs internal modules */
1501         err = build_segment_manager(sbi);
1502         if (err) {
1503                 f2fs_msg(sb, KERN_ERR,
1504                         "Failed to initialize F2FS segment manager");
1505                 goto free_sm;
1506         }
1507         err = build_node_manager(sbi);
1508         if (err) {
1509                 f2fs_msg(sb, KERN_ERR,
1510                         "Failed to initialize F2FS node manager");
1511                 goto free_nm;
1512         }
1513
1514         build_gc_manager(sbi);
1515
1516         /* get an inode for node space */
1517         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1518         if (IS_ERR(sbi->node_inode)) {
1519                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1520                 err = PTR_ERR(sbi->node_inode);
1521                 goto free_nm;
1522         }
1523
1524         f2fs_join_shrinker(sbi);
1525
1526         /* if there are nt orphan nodes free them */
1527         err = recover_orphan_inodes(sbi);
1528         if (err)
1529                 goto free_node_inode;
1530
1531         /* read root inode and dentry */
1532         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1533         if (IS_ERR(root)) {
1534                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1535                 err = PTR_ERR(root);
1536                 goto free_node_inode;
1537         }
1538         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1539                 iput(root);
1540                 err = -EINVAL;
1541                 goto free_node_inode;
1542         }
1543
1544         sb->s_root = d_make_root(root); /* allocate root dentry */
1545         if (!sb->s_root) {
1546                 err = -ENOMEM;
1547                 goto free_root_inode;
1548         }
1549
1550         err = f2fs_build_stats(sbi);
1551         if (err)
1552                 goto free_root_inode;
1553
1554         if (f2fs_proc_root)
1555                 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1556
1557         if (sbi->s_proc)
1558                 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1559                                  &f2fs_seq_segment_info_fops, sb);
1560
1561         sbi->s_kobj.kset = f2fs_kset;
1562         init_completion(&sbi->s_kobj_unregister);
1563         err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1564                                                         "%s", sb->s_id);
1565         if (err)
1566                 goto free_proc;
1567
1568         /* recover fsynced data */
1569         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1570                 /*
1571                  * mount should be failed, when device has readonly mode, and
1572                  * previous checkpoint was not done by clean system shutdown.
1573                  */
1574                 if (bdev_read_only(sb->s_bdev) &&
1575                                 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1576                         err = -EROFS;
1577                         goto free_kobj;
1578                 }
1579
1580                 if (need_fsck)
1581                         set_sbi_flag(sbi, SBI_NEED_FSCK);
1582
1583                 if (!retry)
1584                         goto skip_recovery;
1585
1586                 err = recover_fsync_data(sbi, false);
1587                 if (err < 0) {
1588                         need_fsck = true;
1589                         f2fs_msg(sb, KERN_ERR,
1590                                 "Cannot recover all fsync data errno=%ld", err);
1591                         goto free_kobj;
1592                 }
1593         } else {
1594                 err = recover_fsync_data(sbi, true);
1595
1596                 if (!f2fs_readonly(sb) && err > 0) {
1597                         err = -EINVAL;
1598                         f2fs_msg(sb, KERN_ERR,
1599                                 "Need to recover fsync data");
1600                         goto free_kobj;
1601                 }
1602         }
1603 skip_recovery:
1604         /* recover_fsync_data() cleared this already */
1605         clear_sbi_flag(sbi, SBI_POR_DOING);
1606
1607         /*
1608          * If filesystem is not mounted as read-only then
1609          * do start the gc_thread.
1610          */
1611         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1612                 /* After POR, we can run background GC thread.*/
1613                 err = start_gc_thread(sbi);
1614                 if (err)
1615                         goto free_kobj;
1616         }
1617         kfree(options);
1618
1619         /* recover broken superblock */
1620         if (recovery && !f2fs_readonly(sb) && !bdev_read_only(sb->s_bdev)) {
1621                 f2fs_msg(sb, KERN_INFO, "Recover invalid superblock");
1622                 f2fs_commit_super(sbi, true);
1623         }
1624
1625         sbi->cp_expires = round_jiffies_up(jiffies);
1626
1627         return 0;
1628
1629 free_kobj:
1630         kobject_del(&sbi->s_kobj);
1631 free_proc:
1632         if (sbi->s_proc) {
1633                 remove_proc_entry("segment_info", sbi->s_proc);
1634                 remove_proc_entry(sb->s_id, f2fs_proc_root);
1635         }
1636         f2fs_destroy_stats(sbi);
1637 free_root_inode:
1638         dput(sb->s_root);
1639         sb->s_root = NULL;
1640 free_node_inode:
1641         mutex_lock(&sbi->umount_mutex);
1642         f2fs_leave_shrinker(sbi);
1643         iput(sbi->node_inode);
1644         mutex_unlock(&sbi->umount_mutex);
1645 free_nm:
1646         destroy_node_manager(sbi);
1647 free_sm:
1648         destroy_segment_manager(sbi);
1649         kfree(sbi->ckpt);
1650 free_meta_inode:
1651         make_bad_inode(sbi->meta_inode);
1652         iput(sbi->meta_inode);
1653 free_options:
1654         kfree(options);
1655 free_sb_buf:
1656         brelse(raw_super_buf);
1657 free_sbi:
1658         kfree(sbi);
1659
1660         /* give only one another chance */
1661         if (retry) {
1662                 retry = false;
1663                 shrink_dcache_sb(sb);
1664                 goto try_onemore;
1665         }
1666         return err;
1667 }
1668
1669 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1670                         const char *dev_name, void *data)
1671 {
1672         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1673 }
1674
1675 static void kill_f2fs_super(struct super_block *sb)
1676 {
1677         if (sb->s_root)
1678                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1679         kill_block_super(sb);
1680 }
1681
1682 static struct file_system_type f2fs_fs_type = {
1683         .owner          = THIS_MODULE,
1684         .name           = "f2fs",
1685         .mount          = f2fs_mount,
1686         .kill_sb        = kill_f2fs_super,
1687         .fs_flags       = FS_REQUIRES_DEV,
1688 };
1689 MODULE_ALIAS_FS("f2fs");
1690
1691 static int __init init_inodecache(void)
1692 {
1693         f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1694                         sizeof(struct f2fs_inode_info));
1695         if (!f2fs_inode_cachep)
1696                 return -ENOMEM;
1697         return 0;
1698 }
1699
1700 static void destroy_inodecache(void)
1701 {
1702         /*
1703          * Make sure all delayed rcu free inodes are flushed before we
1704          * destroy cache.
1705          */
1706         rcu_barrier();
1707         kmem_cache_destroy(f2fs_inode_cachep);
1708 }
1709
1710 static int __init init_f2fs_fs(void)
1711 {
1712         int err;
1713
1714         if (PAGE_SIZE != F2FS_BLKSIZE) {
1715                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
1716                                 PAGE_SIZE, F2FS_BLKSIZE);
1717                 return -EINVAL;
1718         }
1719
1720         f2fs_build_trace_ios();
1721
1722         err = init_inodecache();
1723         if (err)
1724                 goto fail;
1725         err = create_node_manager_caches();
1726         if (err)
1727                 goto free_inodecache;
1728         err = create_segment_manager_caches();
1729         if (err)
1730                 goto free_node_manager_caches;
1731         err = create_checkpoint_caches();
1732         if (err)
1733                 goto free_segment_manager_caches;
1734         err = create_extent_cache();
1735         if (err)
1736                 goto free_checkpoint_caches;
1737         f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1738         if (!f2fs_kset) {
1739                 err = -ENOMEM;
1740                 goto free_extent_cache;
1741         }
1742         err = f2fs_init_crypto();
1743         if (err)
1744                 goto free_kset;
1745
1746         err = register_shrinker(&f2fs_shrinker_info);
1747         if (err)
1748                 goto free_crypto;
1749
1750         err = register_filesystem(&f2fs_fs_type);
1751         if (err)
1752                 goto free_shrinker;
1753         f2fs_create_root_stats();
1754         f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1755         return 0;
1756
1757 free_shrinker:
1758         unregister_shrinker(&f2fs_shrinker_info);
1759 free_crypto:
1760         f2fs_exit_crypto();
1761 free_kset:
1762         kset_unregister(f2fs_kset);
1763 free_extent_cache:
1764         destroy_extent_cache();
1765 free_checkpoint_caches:
1766         destroy_checkpoint_caches();
1767 free_segment_manager_caches:
1768         destroy_segment_manager_caches();
1769 free_node_manager_caches:
1770         destroy_node_manager_caches();
1771 free_inodecache:
1772         destroy_inodecache();
1773 fail:
1774         return err;
1775 }
1776
1777 static void __exit exit_f2fs_fs(void)
1778 {
1779         remove_proc_entry("fs/f2fs", NULL);
1780         f2fs_destroy_root_stats();
1781         unregister_shrinker(&f2fs_shrinker_info);
1782         unregister_filesystem(&f2fs_fs_type);
1783         f2fs_exit_crypto();
1784         destroy_extent_cache();
1785         destroy_checkpoint_caches();
1786         destroy_segment_manager_caches();
1787         destroy_node_manager_caches();
1788         destroy_inodecache();
1789         kset_unregister(f2fs_kset);
1790         f2fs_destroy_trace_ios();
1791 }
1792
1793 module_init(init_f2fs_fs)
1794 module_exit(exit_f2fs_fs)
1795
1796 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1797 MODULE_DESCRIPTION("Flash Friendly File System");
1798 MODULE_LICENSE("GPL");