Loading...
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
37static struct proc_dir_entry *f2fs_proc_root;
38static struct kmem_cache *f2fs_inode_cachep;
39static struct kset *f2fs_kset;
40
41enum {
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
58static 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 */
76enum {
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
83struct 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
92static 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
105static 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
120static 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
142static 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
152static 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
162static 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) \
170static 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
183F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
184F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
185F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
186F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
187F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
188F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
189F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
190F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
191F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
192F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
193F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
194
195#define ATTR_LIST(name) (&f2fs_attr_##name.attr)
196static 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
211static const struct sysfs_ops f2fs_attr_ops = {
212 .show = f2fs_attr_show,
213 .store = f2fs_attr_store,
214};
215
216static struct kobj_type f2fs_ktype = {
217 .default_attrs = f2fs_attrs,
218 .sysfs_ops = &f2fs_attr_ops,
219 .release = f2fs_sb_release,
220};
221
222void 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
234static 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
241static 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
352static 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
381static 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 */
400static void f2fs_dirty_inode(struct inode *inode, int flags)
401{
402 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
403}
404
405static 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
411static void f2fs_destroy_inode(struct inode *inode)
412{
413 call_rcu(&inode->i_rcu, f2fs_i_callback);
414}
415
416static 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
449int 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
469static 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
480static int f2fs_unfreeze(struct super_block *sb)
481{
482 return 0;
483}
484
485static 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
513static 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 (!(root->d_sb->s_flags & MS_RDONLY) && 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 (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
552static 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
579static 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
584static 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
592static 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
598 sync_filesystem(sb);
599
600 /*
601 * Save the old mount options in case we
602 * need to restore them.
603 */
604 org_mount_opt = sbi->mount_opt;
605 active_logs = sbi->active_logs;
606
607 /* parse mount options */
608 err = parse_options(sb, data);
609 if (err)
610 goto restore_opts;
611
612 /*
613 * Previous and new state of filesystem is RO,
614 * so no point in checking GC conditions.
615 */
616 if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
617 goto skip;
618
619 /*
620 * We stop the GC thread if FS is mounted as RO
621 * or if background_gc = off is passed in mount
622 * option. Also sync the filesystem.
623 */
624 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
625 if (sbi->gc_thread) {
626 stop_gc_thread(sbi);
627 f2fs_sync_fs(sb, 1);
628 }
629 } else if (test_opt(sbi, BG_GC) && !sbi->gc_thread) {
630 err = start_gc_thread(sbi);
631 if (err)
632 goto restore_opts;
633 }
634skip:
635 /* Update the POSIXACL Flag */
636 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
637 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
638 return 0;
639
640restore_opts:
641 sbi->mount_opt = org_mount_opt;
642 sbi->active_logs = active_logs;
643 return err;
644}
645
646static struct super_operations f2fs_sops = {
647 .alloc_inode = f2fs_alloc_inode,
648 .drop_inode = f2fs_drop_inode,
649 .destroy_inode = f2fs_destroy_inode,
650 .write_inode = f2fs_write_inode,
651 .dirty_inode = f2fs_dirty_inode,
652 .show_options = f2fs_show_options,
653 .evict_inode = f2fs_evict_inode,
654 .put_super = f2fs_put_super,
655 .sync_fs = f2fs_sync_fs,
656 .freeze_fs = f2fs_freeze,
657 .unfreeze_fs = f2fs_unfreeze,
658 .statfs = f2fs_statfs,
659 .remount_fs = f2fs_remount,
660};
661
662static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
663 u64 ino, u32 generation)
664{
665 struct f2fs_sb_info *sbi = F2FS_SB(sb);
666 struct inode *inode;
667
668 if (unlikely(ino < F2FS_ROOT_INO(sbi)))
669 return ERR_PTR(-ESTALE);
670 if (unlikely(ino >= NM_I(sbi)->max_nid))
671 return ERR_PTR(-ESTALE);
672
673 /*
674 * f2fs_iget isn't quite right if the inode is currently unallocated!
675 * However f2fs_iget currently does appropriate checks to handle stale
676 * inodes so everything is OK.
677 */
678 inode = f2fs_iget(sb, ino);
679 if (IS_ERR(inode))
680 return ERR_CAST(inode);
681 if (unlikely(generation && inode->i_generation != generation)) {
682 /* we didn't find the right inode.. */
683 iput(inode);
684 return ERR_PTR(-ESTALE);
685 }
686 return inode;
687}
688
689static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
690 int fh_len, int fh_type)
691{
692 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
693 f2fs_nfs_get_inode);
694}
695
696static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
697 int fh_len, int fh_type)
698{
699 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
700 f2fs_nfs_get_inode);
701}
702
703static const struct export_operations f2fs_export_ops = {
704 .fh_to_dentry = f2fs_fh_to_dentry,
705 .fh_to_parent = f2fs_fh_to_parent,
706 .get_parent = f2fs_get_parent,
707};
708
709static loff_t max_file_size(unsigned bits)
710{
711 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
712 loff_t leaf_count = ADDRS_PER_BLOCK;
713
714 /* two direct node blocks */
715 result += (leaf_count * 2);
716
717 /* two indirect node blocks */
718 leaf_count *= NIDS_PER_BLOCK;
719 result += (leaf_count * 2);
720
721 /* one double indirect node block */
722 leaf_count *= NIDS_PER_BLOCK;
723 result += leaf_count;
724
725 result <<= bits;
726 return result;
727}
728
729static int sanity_check_raw_super(struct super_block *sb,
730 struct f2fs_super_block *raw_super)
731{
732 unsigned int blocksize;
733
734 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
735 f2fs_msg(sb, KERN_INFO,
736 "Magic Mismatch, valid(0x%x) - read(0x%x)",
737 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
738 return 1;
739 }
740
741 /* Currently, support only 4KB page cache size */
742 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
743 f2fs_msg(sb, KERN_INFO,
744 "Invalid page_cache_size (%lu), supports only 4KB\n",
745 PAGE_CACHE_SIZE);
746 return 1;
747 }
748
749 /* Currently, support only 4KB block size */
750 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
751 if (blocksize != F2FS_BLKSIZE) {
752 f2fs_msg(sb, KERN_INFO,
753 "Invalid blocksize (%u), supports only 4KB\n",
754 blocksize);
755 return 1;
756 }
757
758 if (le32_to_cpu(raw_super->log_sectorsize) !=
759 F2FS_LOG_SECTOR_SIZE) {
760 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize");
761 return 1;
762 }
763 if (le32_to_cpu(raw_super->log_sectors_per_block) !=
764 F2FS_LOG_SECTORS_PER_BLOCK) {
765 f2fs_msg(sb, KERN_INFO, "Invalid log sectors per block");
766 return 1;
767 }
768 return 0;
769}
770
771static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
772{
773 unsigned int total, fsmeta;
774 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
775 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
776
777 total = le32_to_cpu(raw_super->segment_count);
778 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
779 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
780 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
781 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
782 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
783
784 if (unlikely(fsmeta >= total))
785 return 1;
786
787 if (unlikely(is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
788 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
789 return 1;
790 }
791 return 0;
792}
793
794static void init_sb_info(struct f2fs_sb_info *sbi)
795{
796 struct f2fs_super_block *raw_super = sbi->raw_super;
797 int i;
798
799 sbi->log_sectors_per_block =
800 le32_to_cpu(raw_super->log_sectors_per_block);
801 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
802 sbi->blocksize = 1 << sbi->log_blocksize;
803 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
804 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
805 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
806 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
807 sbi->total_sections = le32_to_cpu(raw_super->section_count);
808 sbi->total_node_count =
809 (le32_to_cpu(raw_super->segment_count_nat) / 2)
810 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
811 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
812 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
813 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
814 sbi->cur_victim_sec = NULL_SECNO;
815 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
816
817 for (i = 0; i < NR_COUNT_TYPE; i++)
818 atomic_set(&sbi->nr_pages[i], 0);
819
820 sbi->dir_level = DEF_DIR_LEVEL;
821}
822
823/*
824 * Read f2fs raw super block.
825 * Because we have two copies of super block, so read the first one at first,
826 * if the first one is invalid, move to read the second one.
827 */
828static int read_raw_super_block(struct super_block *sb,
829 struct f2fs_super_block **raw_super,
830 struct buffer_head **raw_super_buf)
831{
832 int block = 0;
833
834retry:
835 *raw_super_buf = sb_bread(sb, block);
836 if (!*raw_super_buf) {
837 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
838 block + 1);
839 if (block == 0) {
840 block++;
841 goto retry;
842 } else {
843 return -EIO;
844 }
845 }
846
847 *raw_super = (struct f2fs_super_block *)
848 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
849
850 /* sanity checking of raw super */
851 if (sanity_check_raw_super(sb, *raw_super)) {
852 brelse(*raw_super_buf);
853 f2fs_msg(sb, KERN_ERR,
854 "Can't find valid F2FS filesystem in %dth superblock",
855 block + 1);
856 if (block == 0) {
857 block++;
858 goto retry;
859 } else {
860 return -EINVAL;
861 }
862 }
863
864 return 0;
865}
866
867static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
868{
869 struct f2fs_sb_info *sbi;
870 struct f2fs_super_block *raw_super;
871 struct buffer_head *raw_super_buf;
872 struct inode *root;
873 long err = -EINVAL;
874 int i;
875
876 /* allocate memory for f2fs-specific super block info */
877 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
878 if (!sbi)
879 return -ENOMEM;
880
881 /* set a block size */
882 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
883 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
884 goto free_sbi;
885 }
886
887 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
888 if (err)
889 goto free_sbi;
890
891 sb->s_fs_info = sbi;
892 /* init some FS parameters */
893 sbi->active_logs = NR_CURSEG_TYPE;
894
895 set_opt(sbi, BG_GC);
896
897#ifdef CONFIG_F2FS_FS_XATTR
898 set_opt(sbi, XATTR_USER);
899#endif
900#ifdef CONFIG_F2FS_FS_POSIX_ACL
901 set_opt(sbi, POSIX_ACL);
902#endif
903 /* parse mount options */
904 err = parse_options(sb, (char *)data);
905 if (err)
906 goto free_sb_buf;
907
908 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
909 sb->s_max_links = F2FS_LINK_MAX;
910 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
911
912 sb->s_op = &f2fs_sops;
913 sb->s_xattr = f2fs_xattr_handlers;
914 sb->s_export_op = &f2fs_export_ops;
915 sb->s_magic = F2FS_SUPER_MAGIC;
916 sb->s_time_gran = 1;
917 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
918 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
919 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
920
921 /* init f2fs-specific super block info */
922 sbi->sb = sb;
923 sbi->raw_super = raw_super;
924 sbi->raw_super_buf = raw_super_buf;
925 mutex_init(&sbi->gc_mutex);
926 mutex_init(&sbi->writepages);
927 mutex_init(&sbi->cp_mutex);
928 mutex_init(&sbi->node_write);
929 sbi->por_doing = false;
930 spin_lock_init(&sbi->stat_lock);
931
932 init_rwsem(&sbi->read_io.io_rwsem);
933 sbi->read_io.sbi = sbi;
934 sbi->read_io.bio = NULL;
935 for (i = 0; i < NR_PAGE_TYPE; i++) {
936 init_rwsem(&sbi->write_io[i].io_rwsem);
937 sbi->write_io[i].sbi = sbi;
938 sbi->write_io[i].bio = NULL;
939 }
940
941 init_rwsem(&sbi->cp_rwsem);
942 init_waitqueue_head(&sbi->cp_wait);
943 init_sb_info(sbi);
944
945 /* get an inode for meta space */
946 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
947 if (IS_ERR(sbi->meta_inode)) {
948 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
949 err = PTR_ERR(sbi->meta_inode);
950 goto free_sb_buf;
951 }
952
953 err = get_valid_checkpoint(sbi);
954 if (err) {
955 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
956 goto free_meta_inode;
957 }
958
959 /* sanity checking of checkpoint */
960 err = -EINVAL;
961 if (sanity_check_ckpt(sbi)) {
962 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
963 goto free_cp;
964 }
965
966 sbi->total_valid_node_count =
967 le32_to_cpu(sbi->ckpt->valid_node_count);
968 sbi->total_valid_inode_count =
969 le32_to_cpu(sbi->ckpt->valid_inode_count);
970 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
971 sbi->total_valid_block_count =
972 le64_to_cpu(sbi->ckpt->valid_block_count);
973 sbi->last_valid_block_count = sbi->total_valid_block_count;
974 sbi->alloc_valid_block_count = 0;
975 INIT_LIST_HEAD(&sbi->dir_inode_list);
976 spin_lock_init(&sbi->dir_inode_lock);
977
978 init_orphan_info(sbi);
979
980 /* setup f2fs internal modules */
981 err = build_segment_manager(sbi);
982 if (err) {
983 f2fs_msg(sb, KERN_ERR,
984 "Failed to initialize F2FS segment manager");
985 goto free_sm;
986 }
987 err = build_node_manager(sbi);
988 if (err) {
989 f2fs_msg(sb, KERN_ERR,
990 "Failed to initialize F2FS node manager");
991 goto free_nm;
992 }
993
994 build_gc_manager(sbi);
995
996 /* get an inode for node space */
997 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
998 if (IS_ERR(sbi->node_inode)) {
999 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1000 err = PTR_ERR(sbi->node_inode);
1001 goto free_nm;
1002 }
1003
1004 /* if there are nt orphan nodes free them */
1005 recover_orphan_inodes(sbi);
1006
1007 /* read root inode and dentry */
1008 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1009 if (IS_ERR(root)) {
1010 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1011 err = PTR_ERR(root);
1012 goto free_node_inode;
1013 }
1014 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1015 err = -EINVAL;
1016 goto free_root_inode;
1017 }
1018
1019 sb->s_root = d_make_root(root); /* allocate root dentry */
1020 if (!sb->s_root) {
1021 err = -ENOMEM;
1022 goto free_root_inode;
1023 }
1024
1025 err = f2fs_build_stats(sbi);
1026 if (err)
1027 goto free_root_inode;
1028
1029 if (f2fs_proc_root)
1030 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1031
1032 if (sbi->s_proc)
1033 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1034 &f2fs_seq_segment_info_fops, sb);
1035
1036 if (test_opt(sbi, DISCARD)) {
1037 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1038 if (!blk_queue_discard(q))
1039 f2fs_msg(sb, KERN_WARNING,
1040 "mounting with \"discard\" option, but "
1041 "the device does not support discard");
1042 }
1043
1044 sbi->s_kobj.kset = f2fs_kset;
1045 init_completion(&sbi->s_kobj_unregister);
1046 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1047 "%s", sb->s_id);
1048 if (err)
1049 goto free_proc;
1050
1051 /* recover fsynced data */
1052 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1053 err = recover_fsync_data(sbi);
1054 if (err)
1055 f2fs_msg(sb, KERN_ERR,
1056 "Cannot recover all fsync data errno=%ld", err);
1057 }
1058
1059 /*
1060 * If filesystem is not mounted as read-only then
1061 * do start the gc_thread.
1062 */
1063 if (!(sb->s_flags & MS_RDONLY)) {
1064 /* After POR, we can run background GC thread.*/
1065 err = start_gc_thread(sbi);
1066 if (err)
1067 goto free_kobj;
1068 }
1069 return 0;
1070
1071free_kobj:
1072 kobject_del(&sbi->s_kobj);
1073free_proc:
1074 if (sbi->s_proc) {
1075 remove_proc_entry("segment_info", sbi->s_proc);
1076 remove_proc_entry(sb->s_id, f2fs_proc_root);
1077 }
1078 f2fs_destroy_stats(sbi);
1079free_root_inode:
1080 dput(sb->s_root);
1081 sb->s_root = NULL;
1082free_node_inode:
1083 iput(sbi->node_inode);
1084free_nm:
1085 destroy_node_manager(sbi);
1086free_sm:
1087 destroy_segment_manager(sbi);
1088free_cp:
1089 kfree(sbi->ckpt);
1090free_meta_inode:
1091 make_bad_inode(sbi->meta_inode);
1092 iput(sbi->meta_inode);
1093free_sb_buf:
1094 brelse(raw_super_buf);
1095free_sbi:
1096 kfree(sbi);
1097 return err;
1098}
1099
1100static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1101 const char *dev_name, void *data)
1102{
1103 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1104}
1105
1106static struct file_system_type f2fs_fs_type = {
1107 .owner = THIS_MODULE,
1108 .name = "f2fs",
1109 .mount = f2fs_mount,
1110 .kill_sb = kill_block_super,
1111 .fs_flags = FS_REQUIRES_DEV,
1112};
1113MODULE_ALIAS_FS("f2fs");
1114
1115static int __init init_inodecache(void)
1116{
1117 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1118 sizeof(struct f2fs_inode_info));
1119 if (!f2fs_inode_cachep)
1120 return -ENOMEM;
1121 return 0;
1122}
1123
1124static void destroy_inodecache(void)
1125{
1126 /*
1127 * Make sure all delayed rcu free inodes are flushed before we
1128 * destroy cache.
1129 */
1130 rcu_barrier();
1131 kmem_cache_destroy(f2fs_inode_cachep);
1132}
1133
1134static int __init init_f2fs_fs(void)
1135{
1136 int err;
1137
1138 err = init_inodecache();
1139 if (err)
1140 goto fail;
1141 err = create_node_manager_caches();
1142 if (err)
1143 goto free_inodecache;
1144 err = create_segment_manager_caches();
1145 if (err)
1146 goto free_node_manager_caches;
1147 err = create_gc_caches();
1148 if (err)
1149 goto free_segment_manager_caches;
1150 err = create_checkpoint_caches();
1151 if (err)
1152 goto free_gc_caches;
1153 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1154 if (!f2fs_kset) {
1155 err = -ENOMEM;
1156 goto free_checkpoint_caches;
1157 }
1158 err = register_filesystem(&f2fs_fs_type);
1159 if (err)
1160 goto free_kset;
1161 f2fs_create_root_stats();
1162 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1163 return 0;
1164
1165free_kset:
1166 kset_unregister(f2fs_kset);
1167free_checkpoint_caches:
1168 destroy_checkpoint_caches();
1169free_gc_caches:
1170 destroy_gc_caches();
1171free_segment_manager_caches:
1172 destroy_segment_manager_caches();
1173free_node_manager_caches:
1174 destroy_node_manager_caches();
1175free_inodecache:
1176 destroy_inodecache();
1177fail:
1178 return err;
1179}
1180
1181static void __exit exit_f2fs_fs(void)
1182{
1183 remove_proc_entry("fs/f2fs", NULL);
1184 f2fs_destroy_root_stats();
1185 unregister_filesystem(&f2fs_fs_type);
1186 destroy_checkpoint_caches();
1187 destroy_gc_caches();
1188 destroy_segment_manager_caches();
1189 destroy_node_manager_caches();
1190 destroy_inodecache();
1191 kset_unregister(f2fs_kset);
1192}
1193
1194module_init(init_f2fs_fs)
1195module_exit(exit_f2fs_fs)
1196
1197MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1198MODULE_DESCRIPTION("Flash Friendly File System");
1199MODULE_LICENSE("GPL");
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
38static struct proc_dir_entry *f2fs_proc_root;
39static struct kmem_cache *f2fs_inode_cachep;
40static struct kset *f2fs_kset;
41
42#ifdef CONFIG_F2FS_FAULT_INJECTION
43
44char *fault_name[FAULT_MAX] = {
45 [FAULT_KMALLOC] = "kmalloc",
46 [FAULT_PAGE_ALLOC] = "page alloc",
47 [FAULT_ALLOC_NID] = "alloc nid",
48 [FAULT_ORPHAN] = "orphan",
49 [FAULT_BLOCK] = "no more block",
50 [FAULT_DIR_DEPTH] = "too big dir depth",
51 [FAULT_EVICT_INODE] = "evict_inode fail",
52 [FAULT_IO] = "IO error",
53 [FAULT_CHECKPOINT] = "checkpoint error",
54};
55
56static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
57 unsigned int rate)
58{
59 struct f2fs_fault_info *ffi = &sbi->fault_info;
60
61 if (rate) {
62 atomic_set(&ffi->inject_ops, 0);
63 ffi->inject_rate = rate;
64 ffi->inject_type = (1 << FAULT_MAX) - 1;
65 } else {
66 memset(ffi, 0, sizeof(struct f2fs_fault_info));
67 }
68}
69#endif
70
71/* f2fs-wide shrinker description */
72static struct shrinker f2fs_shrinker_info = {
73 .scan_objects = f2fs_shrink_scan,
74 .count_objects = f2fs_shrink_count,
75 .seeks = DEFAULT_SEEKS,
76};
77
78enum {
79 Opt_gc_background,
80 Opt_disable_roll_forward,
81 Opt_norecovery,
82 Opt_discard,
83 Opt_nodiscard,
84 Opt_noheap,
85 Opt_user_xattr,
86 Opt_nouser_xattr,
87 Opt_acl,
88 Opt_noacl,
89 Opt_active_logs,
90 Opt_disable_ext_identify,
91 Opt_inline_xattr,
92 Opt_inline_data,
93 Opt_inline_dentry,
94 Opt_noinline_dentry,
95 Opt_flush_merge,
96 Opt_noflush_merge,
97 Opt_nobarrier,
98 Opt_fastboot,
99 Opt_extent_cache,
100 Opt_noextent_cache,
101 Opt_noinline_data,
102 Opt_data_flush,
103 Opt_mode,
104 Opt_fault_injection,
105 Opt_lazytime,
106 Opt_nolazytime,
107 Opt_err,
108};
109
110static match_table_t f2fs_tokens = {
111 {Opt_gc_background, "background_gc=%s"},
112 {Opt_disable_roll_forward, "disable_roll_forward"},
113 {Opt_norecovery, "norecovery"},
114 {Opt_discard, "discard"},
115 {Opt_nodiscard, "nodiscard"},
116 {Opt_noheap, "no_heap"},
117 {Opt_user_xattr, "user_xattr"},
118 {Opt_nouser_xattr, "nouser_xattr"},
119 {Opt_acl, "acl"},
120 {Opt_noacl, "noacl"},
121 {Opt_active_logs, "active_logs=%u"},
122 {Opt_disable_ext_identify, "disable_ext_identify"},
123 {Opt_inline_xattr, "inline_xattr"},
124 {Opt_inline_data, "inline_data"},
125 {Opt_inline_dentry, "inline_dentry"},
126 {Opt_noinline_dentry, "noinline_dentry"},
127 {Opt_flush_merge, "flush_merge"},
128 {Opt_noflush_merge, "noflush_merge"},
129 {Opt_nobarrier, "nobarrier"},
130 {Opt_fastboot, "fastboot"},
131 {Opt_extent_cache, "extent_cache"},
132 {Opt_noextent_cache, "noextent_cache"},
133 {Opt_noinline_data, "noinline_data"},
134 {Opt_data_flush, "data_flush"},
135 {Opt_mode, "mode=%s"},
136 {Opt_fault_injection, "fault_injection=%u"},
137 {Opt_lazytime, "lazytime"},
138 {Opt_nolazytime, "nolazytime"},
139 {Opt_err, NULL},
140};
141
142/* Sysfs support for f2fs */
143enum {
144 GC_THREAD, /* struct f2fs_gc_thread */
145 SM_INFO, /* struct f2fs_sm_info */
146 NM_INFO, /* struct f2fs_nm_info */
147 F2FS_SBI, /* struct f2fs_sb_info */
148#ifdef CONFIG_F2FS_FAULT_INJECTION
149 FAULT_INFO_RATE, /* struct f2fs_fault_info */
150 FAULT_INFO_TYPE, /* struct f2fs_fault_info */
151#endif
152};
153
154struct f2fs_attr {
155 struct attribute attr;
156 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
157 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
158 const char *, size_t);
159 int struct_type;
160 int offset;
161};
162
163static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
164{
165 if (struct_type == GC_THREAD)
166 return (unsigned char *)sbi->gc_thread;
167 else if (struct_type == SM_INFO)
168 return (unsigned char *)SM_I(sbi);
169 else if (struct_type == NM_INFO)
170 return (unsigned char *)NM_I(sbi);
171 else if (struct_type == F2FS_SBI)
172 return (unsigned char *)sbi;
173#ifdef CONFIG_F2FS_FAULT_INJECTION
174 else if (struct_type == FAULT_INFO_RATE ||
175 struct_type == FAULT_INFO_TYPE)
176 return (unsigned char *)&sbi->fault_info;
177#endif
178 return NULL;
179}
180
181static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
182 struct f2fs_sb_info *sbi, char *buf)
183{
184 struct super_block *sb = sbi->sb;
185
186 if (!sb->s_bdev->bd_part)
187 return snprintf(buf, PAGE_SIZE, "0\n");
188
189 return snprintf(buf, PAGE_SIZE, "%llu\n",
190 (unsigned long long)(sbi->kbytes_written +
191 BD_PART_WRITTEN(sbi)));
192}
193
194static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
195 struct f2fs_sb_info *sbi, char *buf)
196{
197 unsigned char *ptr = NULL;
198 unsigned int *ui;
199
200 ptr = __struct_ptr(sbi, a->struct_type);
201 if (!ptr)
202 return -EINVAL;
203
204 ui = (unsigned int *)(ptr + a->offset);
205
206 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
207}
208
209static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
210 struct f2fs_sb_info *sbi,
211 const char *buf, size_t count)
212{
213 unsigned char *ptr;
214 unsigned long t;
215 unsigned int *ui;
216 ssize_t ret;
217
218 ptr = __struct_ptr(sbi, a->struct_type);
219 if (!ptr)
220 return -EINVAL;
221
222 ui = (unsigned int *)(ptr + a->offset);
223
224 ret = kstrtoul(skip_spaces(buf), 0, &t);
225 if (ret < 0)
226 return ret;
227#ifdef CONFIG_F2FS_FAULT_INJECTION
228 if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
229 return -EINVAL;
230#endif
231 *ui = t;
232 return count;
233}
234
235static ssize_t f2fs_attr_show(struct kobject *kobj,
236 struct attribute *attr, char *buf)
237{
238 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
239 s_kobj);
240 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
241
242 return a->show ? a->show(a, sbi, buf) : 0;
243}
244
245static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
246 const char *buf, size_t len)
247{
248 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
249 s_kobj);
250 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
251
252 return a->store ? a->store(a, sbi, buf, len) : 0;
253}
254
255static void f2fs_sb_release(struct kobject *kobj)
256{
257 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
258 s_kobj);
259 complete(&sbi->s_kobj_unregister);
260}
261
262#define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
263static struct f2fs_attr f2fs_attr_##_name = { \
264 .attr = {.name = __stringify(_name), .mode = _mode }, \
265 .show = _show, \
266 .store = _store, \
267 .struct_type = _struct_type, \
268 .offset = _offset \
269}
270
271#define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
272 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
273 f2fs_sbi_show, f2fs_sbi_store, \
274 offsetof(struct struct_name, elname))
275
276#define F2FS_GENERAL_RO_ATTR(name) \
277static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
278
279F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
280F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
281F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
282F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
283F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
284F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
285F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
286F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
287F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
288F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
289F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
290F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
291F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
292F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
293F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
294F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
295F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
296#ifdef CONFIG_F2FS_FAULT_INJECTION
297F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
298F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
299#endif
300F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
301
302#define ATTR_LIST(name) (&f2fs_attr_##name.attr)
303static struct attribute *f2fs_attrs[] = {
304 ATTR_LIST(gc_min_sleep_time),
305 ATTR_LIST(gc_max_sleep_time),
306 ATTR_LIST(gc_no_gc_sleep_time),
307 ATTR_LIST(gc_idle),
308 ATTR_LIST(reclaim_segments),
309 ATTR_LIST(max_small_discards),
310 ATTR_LIST(batched_trim_sections),
311 ATTR_LIST(ipu_policy),
312 ATTR_LIST(min_ipu_util),
313 ATTR_LIST(min_fsync_blocks),
314 ATTR_LIST(max_victim_search),
315 ATTR_LIST(dir_level),
316 ATTR_LIST(ram_thresh),
317 ATTR_LIST(ra_nid_pages),
318 ATTR_LIST(dirty_nats_ratio),
319 ATTR_LIST(cp_interval),
320 ATTR_LIST(idle_interval),
321#ifdef CONFIG_F2FS_FAULT_INJECTION
322 ATTR_LIST(inject_rate),
323 ATTR_LIST(inject_type),
324#endif
325 ATTR_LIST(lifetime_write_kbytes),
326 NULL,
327};
328
329static const struct sysfs_ops f2fs_attr_ops = {
330 .show = f2fs_attr_show,
331 .store = f2fs_attr_store,
332};
333
334static struct kobj_type f2fs_ktype = {
335 .default_attrs = f2fs_attrs,
336 .sysfs_ops = &f2fs_attr_ops,
337 .release = f2fs_sb_release,
338};
339
340void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
341{
342 struct va_format vaf;
343 va_list args;
344
345 va_start(args, fmt);
346 vaf.fmt = fmt;
347 vaf.va = &args;
348 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
349 va_end(args);
350}
351
352static void init_once(void *foo)
353{
354 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
355
356 inode_init_once(&fi->vfs_inode);
357}
358
359static int parse_options(struct super_block *sb, char *options)
360{
361 struct f2fs_sb_info *sbi = F2FS_SB(sb);
362 struct request_queue *q;
363 substring_t args[MAX_OPT_ARGS];
364 char *p, *name;
365 int arg = 0;
366
367 if (!options)
368 return 0;
369
370 while ((p = strsep(&options, ",")) != NULL) {
371 int token;
372 if (!*p)
373 continue;
374 /*
375 * Initialize args struct so we know whether arg was
376 * found; some options take optional arguments.
377 */
378 args[0].to = args[0].from = NULL;
379 token = match_token(p, f2fs_tokens, args);
380
381 switch (token) {
382 case Opt_gc_background:
383 name = match_strdup(&args[0]);
384
385 if (!name)
386 return -ENOMEM;
387 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
388 set_opt(sbi, BG_GC);
389 clear_opt(sbi, FORCE_FG_GC);
390 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
391 clear_opt(sbi, BG_GC);
392 clear_opt(sbi, FORCE_FG_GC);
393 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
394 set_opt(sbi, BG_GC);
395 set_opt(sbi, FORCE_FG_GC);
396 } else {
397 kfree(name);
398 return -EINVAL;
399 }
400 kfree(name);
401 break;
402 case Opt_disable_roll_forward:
403 set_opt(sbi, DISABLE_ROLL_FORWARD);
404 break;
405 case Opt_norecovery:
406 /* this option mounts f2fs with ro */
407 set_opt(sbi, DISABLE_ROLL_FORWARD);
408 if (!f2fs_readonly(sb))
409 return -EINVAL;
410 break;
411 case Opt_discard:
412 q = bdev_get_queue(sb->s_bdev);
413 if (blk_queue_discard(q)) {
414 set_opt(sbi, DISCARD);
415 } else if (!f2fs_sb_mounted_blkzoned(sb)) {
416 f2fs_msg(sb, KERN_WARNING,
417 "mounting with \"discard\" option, but "
418 "the device does not support discard");
419 }
420 break;
421 case Opt_nodiscard:
422 if (f2fs_sb_mounted_blkzoned(sb)) {
423 f2fs_msg(sb, KERN_WARNING,
424 "discard is required for zoned block devices");
425 return -EINVAL;
426 }
427 clear_opt(sbi, DISCARD);
428 break;
429 case Opt_noheap:
430 set_opt(sbi, NOHEAP);
431 break;
432#ifdef CONFIG_F2FS_FS_XATTR
433 case Opt_user_xattr:
434 set_opt(sbi, XATTR_USER);
435 break;
436 case Opt_nouser_xattr:
437 clear_opt(sbi, XATTR_USER);
438 break;
439 case Opt_inline_xattr:
440 set_opt(sbi, INLINE_XATTR);
441 break;
442#else
443 case Opt_user_xattr:
444 f2fs_msg(sb, KERN_INFO,
445 "user_xattr options not supported");
446 break;
447 case Opt_nouser_xattr:
448 f2fs_msg(sb, KERN_INFO,
449 "nouser_xattr options not supported");
450 break;
451 case Opt_inline_xattr:
452 f2fs_msg(sb, KERN_INFO,
453 "inline_xattr options not supported");
454 break;
455#endif
456#ifdef CONFIG_F2FS_FS_POSIX_ACL
457 case Opt_acl:
458 set_opt(sbi, POSIX_ACL);
459 break;
460 case Opt_noacl:
461 clear_opt(sbi, POSIX_ACL);
462 break;
463#else
464 case Opt_acl:
465 f2fs_msg(sb, KERN_INFO, "acl options not supported");
466 break;
467 case Opt_noacl:
468 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
469 break;
470#endif
471 case Opt_active_logs:
472 if (args->from && match_int(args, &arg))
473 return -EINVAL;
474 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
475 return -EINVAL;
476 sbi->active_logs = arg;
477 break;
478 case Opt_disable_ext_identify:
479 set_opt(sbi, DISABLE_EXT_IDENTIFY);
480 break;
481 case Opt_inline_data:
482 set_opt(sbi, INLINE_DATA);
483 break;
484 case Opt_inline_dentry:
485 set_opt(sbi, INLINE_DENTRY);
486 break;
487 case Opt_noinline_dentry:
488 clear_opt(sbi, INLINE_DENTRY);
489 break;
490 case Opt_flush_merge:
491 set_opt(sbi, FLUSH_MERGE);
492 break;
493 case Opt_noflush_merge:
494 clear_opt(sbi, FLUSH_MERGE);
495 break;
496 case Opt_nobarrier:
497 set_opt(sbi, NOBARRIER);
498 break;
499 case Opt_fastboot:
500 set_opt(sbi, FASTBOOT);
501 break;
502 case Opt_extent_cache:
503 set_opt(sbi, EXTENT_CACHE);
504 break;
505 case Opt_noextent_cache:
506 clear_opt(sbi, EXTENT_CACHE);
507 break;
508 case Opt_noinline_data:
509 clear_opt(sbi, INLINE_DATA);
510 break;
511 case Opt_data_flush:
512 set_opt(sbi, DATA_FLUSH);
513 break;
514 case Opt_mode:
515 name = match_strdup(&args[0]);
516
517 if (!name)
518 return -ENOMEM;
519 if (strlen(name) == 8 &&
520 !strncmp(name, "adaptive", 8)) {
521 if (f2fs_sb_mounted_blkzoned(sb)) {
522 f2fs_msg(sb, KERN_WARNING,
523 "adaptive mode is not allowed with "
524 "zoned block device feature");
525 kfree(name);
526 return -EINVAL;
527 }
528 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
529 } else if (strlen(name) == 3 &&
530 !strncmp(name, "lfs", 3)) {
531 set_opt_mode(sbi, F2FS_MOUNT_LFS);
532 } else {
533 kfree(name);
534 return -EINVAL;
535 }
536 kfree(name);
537 break;
538 case Opt_fault_injection:
539 if (args->from && match_int(args, &arg))
540 return -EINVAL;
541#ifdef CONFIG_F2FS_FAULT_INJECTION
542 f2fs_build_fault_attr(sbi, arg);
543#else
544 f2fs_msg(sb, KERN_INFO,
545 "FAULT_INJECTION was not selected");
546#endif
547 break;
548 case Opt_lazytime:
549 sb->s_flags |= MS_LAZYTIME;
550 break;
551 case Opt_nolazytime:
552 sb->s_flags &= ~MS_LAZYTIME;
553 break;
554 default:
555 f2fs_msg(sb, KERN_ERR,
556 "Unrecognized mount option \"%s\" or missing value",
557 p);
558 return -EINVAL;
559 }
560 }
561 return 0;
562}
563
564static struct inode *f2fs_alloc_inode(struct super_block *sb)
565{
566 struct f2fs_inode_info *fi;
567
568 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
569 if (!fi)
570 return NULL;
571
572 init_once((void *) fi);
573
574 /* Initialize f2fs-specific inode info */
575 fi->vfs_inode.i_version = 1;
576 atomic_set(&fi->dirty_pages, 0);
577 fi->i_current_depth = 1;
578 fi->i_advise = 0;
579 init_rwsem(&fi->i_sem);
580 INIT_LIST_HEAD(&fi->dirty_list);
581 INIT_LIST_HEAD(&fi->gdirty_list);
582 INIT_LIST_HEAD(&fi->inmem_pages);
583 mutex_init(&fi->inmem_lock);
584 init_rwsem(&fi->dio_rwsem[READ]);
585 init_rwsem(&fi->dio_rwsem[WRITE]);
586
587 /* Will be used by directory only */
588 fi->i_dir_level = F2FS_SB(sb)->dir_level;
589 return &fi->vfs_inode;
590}
591
592static int f2fs_drop_inode(struct inode *inode)
593{
594 /*
595 * This is to avoid a deadlock condition like below.
596 * writeback_single_inode(inode)
597 * - f2fs_write_data_page
598 * - f2fs_gc -> iput -> evict
599 * - inode_wait_for_writeback(inode)
600 */
601 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
602 if (!inode->i_nlink && !is_bad_inode(inode)) {
603 /* to avoid evict_inode call simultaneously */
604 atomic_inc(&inode->i_count);
605 spin_unlock(&inode->i_lock);
606
607 /* some remained atomic pages should discarded */
608 if (f2fs_is_atomic_file(inode))
609 drop_inmem_pages(inode);
610
611 /* should remain fi->extent_tree for writepage */
612 f2fs_destroy_extent_node(inode);
613
614 sb_start_intwrite(inode->i_sb);
615 f2fs_i_size_write(inode, 0);
616
617 if (F2FS_HAS_BLOCKS(inode))
618 f2fs_truncate(inode);
619
620 sb_end_intwrite(inode->i_sb);
621
622 fscrypt_put_encryption_info(inode, NULL);
623 spin_lock(&inode->i_lock);
624 atomic_dec(&inode->i_count);
625 }
626 return 0;
627 }
628
629 return generic_drop_inode(inode);
630}
631
632int f2fs_inode_dirtied(struct inode *inode, bool sync)
633{
634 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
635 int ret = 0;
636
637 spin_lock(&sbi->inode_lock[DIRTY_META]);
638 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
639 ret = 1;
640 } else {
641 set_inode_flag(inode, FI_DIRTY_INODE);
642 stat_inc_dirty_inode(sbi, DIRTY_META);
643 }
644 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
645 list_add_tail(&F2FS_I(inode)->gdirty_list,
646 &sbi->inode_list[DIRTY_META]);
647 inc_page_count(sbi, F2FS_DIRTY_IMETA);
648 }
649 spin_unlock(&sbi->inode_lock[DIRTY_META]);
650 return ret;
651}
652
653void f2fs_inode_synced(struct inode *inode)
654{
655 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
656
657 spin_lock(&sbi->inode_lock[DIRTY_META]);
658 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
659 spin_unlock(&sbi->inode_lock[DIRTY_META]);
660 return;
661 }
662 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
663 list_del_init(&F2FS_I(inode)->gdirty_list);
664 dec_page_count(sbi, F2FS_DIRTY_IMETA);
665 }
666 clear_inode_flag(inode, FI_DIRTY_INODE);
667 clear_inode_flag(inode, FI_AUTO_RECOVER);
668 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
669 spin_unlock(&sbi->inode_lock[DIRTY_META]);
670}
671
672/*
673 * f2fs_dirty_inode() is called from __mark_inode_dirty()
674 *
675 * We should call set_dirty_inode to write the dirty inode through write_inode.
676 */
677static void f2fs_dirty_inode(struct inode *inode, int flags)
678{
679 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
680
681 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
682 inode->i_ino == F2FS_META_INO(sbi))
683 return;
684
685 if (flags == I_DIRTY_TIME)
686 return;
687
688 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
689 clear_inode_flag(inode, FI_AUTO_RECOVER);
690
691 f2fs_inode_dirtied(inode, false);
692}
693
694static void f2fs_i_callback(struct rcu_head *head)
695{
696 struct inode *inode = container_of(head, struct inode, i_rcu);
697 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
698}
699
700static void f2fs_destroy_inode(struct inode *inode)
701{
702 call_rcu(&inode->i_rcu, f2fs_i_callback);
703}
704
705static void destroy_percpu_info(struct f2fs_sb_info *sbi)
706{
707 percpu_counter_destroy(&sbi->alloc_valid_block_count);
708 percpu_counter_destroy(&sbi->total_valid_inode_count);
709}
710
711static void destroy_device_list(struct f2fs_sb_info *sbi)
712{
713 int i;
714
715 for (i = 0; i < sbi->s_ndevs; i++) {
716 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
717#ifdef CONFIG_BLK_DEV_ZONED
718 kfree(FDEV(i).blkz_type);
719#endif
720 }
721 kfree(sbi->devs);
722}
723
724static void f2fs_put_super(struct super_block *sb)
725{
726 struct f2fs_sb_info *sbi = F2FS_SB(sb);
727
728 if (sbi->s_proc) {
729 remove_proc_entry("segment_info", sbi->s_proc);
730 remove_proc_entry("segment_bits", sbi->s_proc);
731 remove_proc_entry(sb->s_id, f2fs_proc_root);
732 }
733 kobject_del(&sbi->s_kobj);
734
735 stop_gc_thread(sbi);
736
737 /* prevent remaining shrinker jobs */
738 mutex_lock(&sbi->umount_mutex);
739
740 /*
741 * We don't need to do checkpoint when superblock is clean.
742 * But, the previous checkpoint was not done by umount, it needs to do
743 * clean checkpoint again.
744 */
745 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
746 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
747 struct cp_control cpc = {
748 .reason = CP_UMOUNT,
749 };
750 write_checkpoint(sbi, &cpc);
751 }
752
753 /* write_checkpoint can update stat informaion */
754 f2fs_destroy_stats(sbi);
755
756 /*
757 * normally superblock is clean, so we need to release this.
758 * In addition, EIO will skip do checkpoint, we need this as well.
759 */
760 release_ino_entry(sbi, true);
761
762 f2fs_leave_shrinker(sbi);
763 mutex_unlock(&sbi->umount_mutex);
764
765 /* our cp_error case, we can wait for any writeback page */
766 f2fs_flush_merged_bios(sbi);
767
768 iput(sbi->node_inode);
769 iput(sbi->meta_inode);
770
771 /* destroy f2fs internal modules */
772 destroy_node_manager(sbi);
773 destroy_segment_manager(sbi);
774
775 kfree(sbi->ckpt);
776 kobject_put(&sbi->s_kobj);
777 wait_for_completion(&sbi->s_kobj_unregister);
778
779 sb->s_fs_info = NULL;
780 if (sbi->s_chksum_driver)
781 crypto_free_shash(sbi->s_chksum_driver);
782 kfree(sbi->raw_super);
783
784 destroy_device_list(sbi);
785
786 destroy_percpu_info(sbi);
787 kfree(sbi);
788}
789
790int f2fs_sync_fs(struct super_block *sb, int sync)
791{
792 struct f2fs_sb_info *sbi = F2FS_SB(sb);
793 int err = 0;
794
795 trace_f2fs_sync_fs(sb, sync);
796
797 if (sync) {
798 struct cp_control cpc;
799
800 cpc.reason = __get_cp_reason(sbi);
801
802 mutex_lock(&sbi->gc_mutex);
803 err = write_checkpoint(sbi, &cpc);
804 mutex_unlock(&sbi->gc_mutex);
805 }
806 f2fs_trace_ios(NULL, 1);
807
808 return err;
809}
810
811static int f2fs_freeze(struct super_block *sb)
812{
813 if (f2fs_readonly(sb))
814 return 0;
815
816 /* IO error happened before */
817 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
818 return -EIO;
819
820 /* must be clean, since sync_filesystem() was already called */
821 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
822 return -EINVAL;
823 return 0;
824}
825
826static int f2fs_unfreeze(struct super_block *sb)
827{
828 return 0;
829}
830
831static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
832{
833 struct super_block *sb = dentry->d_sb;
834 struct f2fs_sb_info *sbi = F2FS_SB(sb);
835 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
836 block_t total_count, user_block_count, start_count, ovp_count;
837
838 total_count = le64_to_cpu(sbi->raw_super->block_count);
839 user_block_count = sbi->user_block_count;
840 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
841 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
842 buf->f_type = F2FS_SUPER_MAGIC;
843 buf->f_bsize = sbi->blocksize;
844
845 buf->f_blocks = total_count - start_count;
846 buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
847 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
848
849 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
850 buf->f_ffree = min(buf->f_files - valid_node_count(sbi),
851 buf->f_bavail);
852
853 buf->f_namelen = F2FS_NAME_LEN;
854 buf->f_fsid.val[0] = (u32)id;
855 buf->f_fsid.val[1] = (u32)(id >> 32);
856
857 return 0;
858}
859
860static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
861{
862 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
863
864 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
865 if (test_opt(sbi, FORCE_FG_GC))
866 seq_printf(seq, ",background_gc=%s", "sync");
867 else
868 seq_printf(seq, ",background_gc=%s", "on");
869 } else {
870 seq_printf(seq, ",background_gc=%s", "off");
871 }
872 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
873 seq_puts(seq, ",disable_roll_forward");
874 if (test_opt(sbi, DISCARD))
875 seq_puts(seq, ",discard");
876 if (test_opt(sbi, NOHEAP))
877 seq_puts(seq, ",no_heap_alloc");
878#ifdef CONFIG_F2FS_FS_XATTR
879 if (test_opt(sbi, XATTR_USER))
880 seq_puts(seq, ",user_xattr");
881 else
882 seq_puts(seq, ",nouser_xattr");
883 if (test_opt(sbi, INLINE_XATTR))
884 seq_puts(seq, ",inline_xattr");
885#endif
886#ifdef CONFIG_F2FS_FS_POSIX_ACL
887 if (test_opt(sbi, POSIX_ACL))
888 seq_puts(seq, ",acl");
889 else
890 seq_puts(seq, ",noacl");
891#endif
892 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
893 seq_puts(seq, ",disable_ext_identify");
894 if (test_opt(sbi, INLINE_DATA))
895 seq_puts(seq, ",inline_data");
896 else
897 seq_puts(seq, ",noinline_data");
898 if (test_opt(sbi, INLINE_DENTRY))
899 seq_puts(seq, ",inline_dentry");
900 else
901 seq_puts(seq, ",noinline_dentry");
902 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
903 seq_puts(seq, ",flush_merge");
904 if (test_opt(sbi, NOBARRIER))
905 seq_puts(seq, ",nobarrier");
906 if (test_opt(sbi, FASTBOOT))
907 seq_puts(seq, ",fastboot");
908 if (test_opt(sbi, EXTENT_CACHE))
909 seq_puts(seq, ",extent_cache");
910 else
911 seq_puts(seq, ",noextent_cache");
912 if (test_opt(sbi, DATA_FLUSH))
913 seq_puts(seq, ",data_flush");
914
915 seq_puts(seq, ",mode=");
916 if (test_opt(sbi, ADAPTIVE))
917 seq_puts(seq, "adaptive");
918 else if (test_opt(sbi, LFS))
919 seq_puts(seq, "lfs");
920 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
921
922 return 0;
923}
924
925static int segment_info_seq_show(struct seq_file *seq, void *offset)
926{
927 struct super_block *sb = seq->private;
928 struct f2fs_sb_info *sbi = F2FS_SB(sb);
929 unsigned int total_segs =
930 le32_to_cpu(sbi->raw_super->segment_count_main);
931 int i;
932
933 seq_puts(seq, "format: segment_type|valid_blocks\n"
934 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
935
936 for (i = 0; i < total_segs; i++) {
937 struct seg_entry *se = get_seg_entry(sbi, i);
938
939 if ((i % 10) == 0)
940 seq_printf(seq, "%-10d", i);
941 seq_printf(seq, "%d|%-3u", se->type,
942 get_valid_blocks(sbi, i, 1));
943 if ((i % 10) == 9 || i == (total_segs - 1))
944 seq_putc(seq, '\n');
945 else
946 seq_putc(seq, ' ');
947 }
948
949 return 0;
950}
951
952static int segment_bits_seq_show(struct seq_file *seq, void *offset)
953{
954 struct super_block *sb = seq->private;
955 struct f2fs_sb_info *sbi = F2FS_SB(sb);
956 unsigned int total_segs =
957 le32_to_cpu(sbi->raw_super->segment_count_main);
958 int i, j;
959
960 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
961 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
962
963 for (i = 0; i < total_segs; i++) {
964 struct seg_entry *se = get_seg_entry(sbi, i);
965
966 seq_printf(seq, "%-10d", i);
967 seq_printf(seq, "%d|%-3u|", se->type,
968 get_valid_blocks(sbi, i, 1));
969 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
970 seq_printf(seq, " %.2x", se->cur_valid_map[j]);
971 seq_putc(seq, '\n');
972 }
973 return 0;
974}
975
976#define F2FS_PROC_FILE_DEF(_name) \
977static int _name##_open_fs(struct inode *inode, struct file *file) \
978{ \
979 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
980} \
981 \
982static const struct file_operations f2fs_seq_##_name##_fops = { \
983 .open = _name##_open_fs, \
984 .read = seq_read, \
985 .llseek = seq_lseek, \
986 .release = single_release, \
987};
988
989F2FS_PROC_FILE_DEF(segment_info);
990F2FS_PROC_FILE_DEF(segment_bits);
991
992static void default_options(struct f2fs_sb_info *sbi)
993{
994 /* init some FS parameters */
995 sbi->active_logs = NR_CURSEG_TYPE;
996
997 set_opt(sbi, BG_GC);
998 set_opt(sbi, INLINE_DATA);
999 set_opt(sbi, INLINE_DENTRY);
1000 set_opt(sbi, EXTENT_CACHE);
1001 sbi->sb->s_flags |= MS_LAZYTIME;
1002 set_opt(sbi, FLUSH_MERGE);
1003 if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1004 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1005 set_opt(sbi, DISCARD);
1006 } else {
1007 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1008 }
1009
1010#ifdef CONFIG_F2FS_FS_XATTR
1011 set_opt(sbi, XATTR_USER);
1012#endif
1013#ifdef CONFIG_F2FS_FS_POSIX_ACL
1014 set_opt(sbi, POSIX_ACL);
1015#endif
1016
1017#ifdef CONFIG_F2FS_FAULT_INJECTION
1018 f2fs_build_fault_attr(sbi, 0);
1019#endif
1020}
1021
1022static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1023{
1024 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1025 struct f2fs_mount_info org_mount_opt;
1026 int err, active_logs;
1027 bool need_restart_gc = false;
1028 bool need_stop_gc = false;
1029 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1030#ifdef CONFIG_F2FS_FAULT_INJECTION
1031 struct f2fs_fault_info ffi = sbi->fault_info;
1032#endif
1033
1034 /*
1035 * Save the old mount options in case we
1036 * need to restore them.
1037 */
1038 org_mount_opt = sbi->mount_opt;
1039 active_logs = sbi->active_logs;
1040
1041 /* recover superblocks we couldn't write due to previous RO mount */
1042 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1043 err = f2fs_commit_super(sbi, false);
1044 f2fs_msg(sb, KERN_INFO,
1045 "Try to recover all the superblocks, ret: %d", err);
1046 if (!err)
1047 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1048 }
1049
1050 sbi->mount_opt.opt = 0;
1051 default_options(sbi);
1052
1053 /* parse mount options */
1054 err = parse_options(sb, data);
1055 if (err)
1056 goto restore_opts;
1057
1058 /*
1059 * Previous and new state of filesystem is RO,
1060 * so skip checking GC and FLUSH_MERGE conditions.
1061 */
1062 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1063 goto skip;
1064
1065 /* disallow enable/disable extent_cache dynamically */
1066 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1067 err = -EINVAL;
1068 f2fs_msg(sbi->sb, KERN_WARNING,
1069 "switch extent_cache option is not allowed");
1070 goto restore_opts;
1071 }
1072
1073 /*
1074 * We stop the GC thread if FS is mounted as RO
1075 * or if background_gc = off is passed in mount
1076 * option. Also sync the filesystem.
1077 */
1078 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1079 if (sbi->gc_thread) {
1080 stop_gc_thread(sbi);
1081 need_restart_gc = true;
1082 }
1083 } else if (!sbi->gc_thread) {
1084 err = start_gc_thread(sbi);
1085 if (err)
1086 goto restore_opts;
1087 need_stop_gc = true;
1088 }
1089
1090 if (*flags & MS_RDONLY) {
1091 writeback_inodes_sb(sb, WB_REASON_SYNC);
1092 sync_inodes_sb(sb);
1093
1094 set_sbi_flag(sbi, SBI_IS_DIRTY);
1095 set_sbi_flag(sbi, SBI_IS_CLOSE);
1096 f2fs_sync_fs(sb, 1);
1097 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1098 }
1099
1100 /*
1101 * We stop issue flush thread if FS is mounted as RO
1102 * or if flush_merge is not passed in mount option.
1103 */
1104 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1105 clear_opt(sbi, FLUSH_MERGE);
1106 destroy_flush_cmd_control(sbi, false);
1107 } else {
1108 err = create_flush_cmd_control(sbi);
1109 if (err)
1110 goto restore_gc;
1111 }
1112skip:
1113 /* Update the POSIXACL Flag */
1114 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1115 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1116
1117 return 0;
1118restore_gc:
1119 if (need_restart_gc) {
1120 if (start_gc_thread(sbi))
1121 f2fs_msg(sbi->sb, KERN_WARNING,
1122 "background gc thread has stopped");
1123 } else if (need_stop_gc) {
1124 stop_gc_thread(sbi);
1125 }
1126restore_opts:
1127 sbi->mount_opt = org_mount_opt;
1128 sbi->active_logs = active_logs;
1129#ifdef CONFIG_F2FS_FAULT_INJECTION
1130 sbi->fault_info = ffi;
1131#endif
1132 return err;
1133}
1134
1135static struct super_operations f2fs_sops = {
1136 .alloc_inode = f2fs_alloc_inode,
1137 .drop_inode = f2fs_drop_inode,
1138 .destroy_inode = f2fs_destroy_inode,
1139 .write_inode = f2fs_write_inode,
1140 .dirty_inode = f2fs_dirty_inode,
1141 .show_options = f2fs_show_options,
1142 .evict_inode = f2fs_evict_inode,
1143 .put_super = f2fs_put_super,
1144 .sync_fs = f2fs_sync_fs,
1145 .freeze_fs = f2fs_freeze,
1146 .unfreeze_fs = f2fs_unfreeze,
1147 .statfs = f2fs_statfs,
1148 .remount_fs = f2fs_remount,
1149};
1150
1151#ifdef CONFIG_F2FS_FS_ENCRYPTION
1152static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1153{
1154 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1155 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1156 ctx, len, NULL);
1157}
1158
1159static int f2fs_key_prefix(struct inode *inode, u8 **key)
1160{
1161 *key = F2FS_I_SB(inode)->key_prefix;
1162 return F2FS_I_SB(inode)->key_prefix_size;
1163}
1164
1165static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1166 void *fs_data)
1167{
1168 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1169 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1170 ctx, len, fs_data, XATTR_CREATE);
1171}
1172
1173static unsigned f2fs_max_namelen(struct inode *inode)
1174{
1175 return S_ISLNK(inode->i_mode) ?
1176 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1177}
1178
1179static struct fscrypt_operations f2fs_cryptops = {
1180 .get_context = f2fs_get_context,
1181 .key_prefix = f2fs_key_prefix,
1182 .set_context = f2fs_set_context,
1183 .is_encrypted = f2fs_encrypted_inode,
1184 .empty_dir = f2fs_empty_dir,
1185 .max_namelen = f2fs_max_namelen,
1186};
1187#else
1188static struct fscrypt_operations f2fs_cryptops = {
1189 .is_encrypted = f2fs_encrypted_inode,
1190};
1191#endif
1192
1193static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1194 u64 ino, u32 generation)
1195{
1196 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1197 struct inode *inode;
1198
1199 if (check_nid_range(sbi, ino))
1200 return ERR_PTR(-ESTALE);
1201
1202 /*
1203 * f2fs_iget isn't quite right if the inode is currently unallocated!
1204 * However f2fs_iget currently does appropriate checks to handle stale
1205 * inodes so everything is OK.
1206 */
1207 inode = f2fs_iget(sb, ino);
1208 if (IS_ERR(inode))
1209 return ERR_CAST(inode);
1210 if (unlikely(generation && inode->i_generation != generation)) {
1211 /* we didn't find the right inode.. */
1212 iput(inode);
1213 return ERR_PTR(-ESTALE);
1214 }
1215 return inode;
1216}
1217
1218static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1219 int fh_len, int fh_type)
1220{
1221 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1222 f2fs_nfs_get_inode);
1223}
1224
1225static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1226 int fh_len, int fh_type)
1227{
1228 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1229 f2fs_nfs_get_inode);
1230}
1231
1232static const struct export_operations f2fs_export_ops = {
1233 .fh_to_dentry = f2fs_fh_to_dentry,
1234 .fh_to_parent = f2fs_fh_to_parent,
1235 .get_parent = f2fs_get_parent,
1236};
1237
1238static loff_t max_file_blocks(void)
1239{
1240 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1241 loff_t leaf_count = ADDRS_PER_BLOCK;
1242
1243 /* two direct node blocks */
1244 result += (leaf_count * 2);
1245
1246 /* two indirect node blocks */
1247 leaf_count *= NIDS_PER_BLOCK;
1248 result += (leaf_count * 2);
1249
1250 /* one double indirect node block */
1251 leaf_count *= NIDS_PER_BLOCK;
1252 result += leaf_count;
1253
1254 return result;
1255}
1256
1257static int __f2fs_commit_super(struct buffer_head *bh,
1258 struct f2fs_super_block *super)
1259{
1260 lock_buffer(bh);
1261 if (super)
1262 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1263 set_buffer_uptodate(bh);
1264 set_buffer_dirty(bh);
1265 unlock_buffer(bh);
1266
1267 /* it's rare case, we can do fua all the time */
1268 return __sync_dirty_buffer(bh, REQ_PREFLUSH | REQ_FUA);
1269}
1270
1271static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1272 struct buffer_head *bh)
1273{
1274 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1275 (bh->b_data + F2FS_SUPER_OFFSET);
1276 struct super_block *sb = sbi->sb;
1277 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1278 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1279 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1280 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1281 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1282 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1283 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1284 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1285 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1286 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1287 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1288 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1289 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1290 u64 main_end_blkaddr = main_blkaddr +
1291 (segment_count_main << log_blocks_per_seg);
1292 u64 seg_end_blkaddr = segment0_blkaddr +
1293 (segment_count << log_blocks_per_seg);
1294
1295 if (segment0_blkaddr != cp_blkaddr) {
1296 f2fs_msg(sb, KERN_INFO,
1297 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1298 segment0_blkaddr, cp_blkaddr);
1299 return true;
1300 }
1301
1302 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1303 sit_blkaddr) {
1304 f2fs_msg(sb, KERN_INFO,
1305 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1306 cp_blkaddr, sit_blkaddr,
1307 segment_count_ckpt << log_blocks_per_seg);
1308 return true;
1309 }
1310
1311 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1312 nat_blkaddr) {
1313 f2fs_msg(sb, KERN_INFO,
1314 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1315 sit_blkaddr, nat_blkaddr,
1316 segment_count_sit << log_blocks_per_seg);
1317 return true;
1318 }
1319
1320 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1321 ssa_blkaddr) {
1322 f2fs_msg(sb, KERN_INFO,
1323 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1324 nat_blkaddr, ssa_blkaddr,
1325 segment_count_nat << log_blocks_per_seg);
1326 return true;
1327 }
1328
1329 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1330 main_blkaddr) {
1331 f2fs_msg(sb, KERN_INFO,
1332 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1333 ssa_blkaddr, main_blkaddr,
1334 segment_count_ssa << log_blocks_per_seg);
1335 return true;
1336 }
1337
1338 if (main_end_blkaddr > seg_end_blkaddr) {
1339 f2fs_msg(sb, KERN_INFO,
1340 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1341 main_blkaddr,
1342 segment0_blkaddr +
1343 (segment_count << log_blocks_per_seg),
1344 segment_count_main << log_blocks_per_seg);
1345 return true;
1346 } else if (main_end_blkaddr < seg_end_blkaddr) {
1347 int err = 0;
1348 char *res;
1349
1350 /* fix in-memory information all the time */
1351 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1352 segment0_blkaddr) >> log_blocks_per_seg);
1353
1354 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1355 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1356 res = "internally";
1357 } else {
1358 err = __f2fs_commit_super(bh, NULL);
1359 res = err ? "failed" : "done";
1360 }
1361 f2fs_msg(sb, KERN_INFO,
1362 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1363 res, main_blkaddr,
1364 segment0_blkaddr +
1365 (segment_count << log_blocks_per_seg),
1366 segment_count_main << log_blocks_per_seg);
1367 if (err)
1368 return true;
1369 }
1370 return false;
1371}
1372
1373static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1374 struct buffer_head *bh)
1375{
1376 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1377 (bh->b_data + F2FS_SUPER_OFFSET);
1378 struct super_block *sb = sbi->sb;
1379 unsigned int blocksize;
1380
1381 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1382 f2fs_msg(sb, KERN_INFO,
1383 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1384 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1385 return 1;
1386 }
1387
1388 /* Currently, support only 4KB page cache size */
1389 if (F2FS_BLKSIZE != PAGE_SIZE) {
1390 f2fs_msg(sb, KERN_INFO,
1391 "Invalid page_cache_size (%lu), supports only 4KB\n",
1392 PAGE_SIZE);
1393 return 1;
1394 }
1395
1396 /* Currently, support only 4KB block size */
1397 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1398 if (blocksize != F2FS_BLKSIZE) {
1399 f2fs_msg(sb, KERN_INFO,
1400 "Invalid blocksize (%u), supports only 4KB\n",
1401 blocksize);
1402 return 1;
1403 }
1404
1405 /* check log blocks per segment */
1406 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1407 f2fs_msg(sb, KERN_INFO,
1408 "Invalid log blocks per segment (%u)\n",
1409 le32_to_cpu(raw_super->log_blocks_per_seg));
1410 return 1;
1411 }
1412
1413 /* Currently, support 512/1024/2048/4096 bytes sector size */
1414 if (le32_to_cpu(raw_super->log_sectorsize) >
1415 F2FS_MAX_LOG_SECTOR_SIZE ||
1416 le32_to_cpu(raw_super->log_sectorsize) <
1417 F2FS_MIN_LOG_SECTOR_SIZE) {
1418 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1419 le32_to_cpu(raw_super->log_sectorsize));
1420 return 1;
1421 }
1422 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1423 le32_to_cpu(raw_super->log_sectorsize) !=
1424 F2FS_MAX_LOG_SECTOR_SIZE) {
1425 f2fs_msg(sb, KERN_INFO,
1426 "Invalid log sectors per block(%u) log sectorsize(%u)",
1427 le32_to_cpu(raw_super->log_sectors_per_block),
1428 le32_to_cpu(raw_super->log_sectorsize));
1429 return 1;
1430 }
1431
1432 /* check reserved ino info */
1433 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1434 le32_to_cpu(raw_super->meta_ino) != 2 ||
1435 le32_to_cpu(raw_super->root_ino) != 3) {
1436 f2fs_msg(sb, KERN_INFO,
1437 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1438 le32_to_cpu(raw_super->node_ino),
1439 le32_to_cpu(raw_super->meta_ino),
1440 le32_to_cpu(raw_super->root_ino));
1441 return 1;
1442 }
1443
1444 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1445 if (sanity_check_area_boundary(sbi, bh))
1446 return 1;
1447
1448 return 0;
1449}
1450
1451int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1452{
1453 unsigned int total, fsmeta;
1454 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1455 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1456 unsigned int ovp_segments, reserved_segments;
1457
1458 total = le32_to_cpu(raw_super->segment_count);
1459 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1460 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1461 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1462 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1463 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1464
1465 if (unlikely(fsmeta >= total))
1466 return 1;
1467
1468 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1469 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1470
1471 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1472 ovp_segments == 0 || reserved_segments == 0)) {
1473 f2fs_msg(sbi->sb, KERN_ERR,
1474 "Wrong layout: check mkfs.f2fs version");
1475 return 1;
1476 }
1477
1478 if (unlikely(f2fs_cp_error(sbi))) {
1479 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1480 return 1;
1481 }
1482 return 0;
1483}
1484
1485static void init_sb_info(struct f2fs_sb_info *sbi)
1486{
1487 struct f2fs_super_block *raw_super = sbi->raw_super;
1488 int i;
1489
1490 sbi->log_sectors_per_block =
1491 le32_to_cpu(raw_super->log_sectors_per_block);
1492 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1493 sbi->blocksize = 1 << sbi->log_blocksize;
1494 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1495 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1496 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1497 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1498 sbi->total_sections = le32_to_cpu(raw_super->section_count);
1499 sbi->total_node_count =
1500 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1501 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1502 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1503 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1504 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1505 sbi->cur_victim_sec = NULL_SECNO;
1506 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1507
1508 sbi->dir_level = DEF_DIR_LEVEL;
1509 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1510 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1511 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1512
1513 for (i = 0; i < NR_COUNT_TYPE; i++)
1514 atomic_set(&sbi->nr_pages[i], 0);
1515
1516 INIT_LIST_HEAD(&sbi->s_list);
1517 mutex_init(&sbi->umount_mutex);
1518 mutex_init(&sbi->wio_mutex[NODE]);
1519 mutex_init(&sbi->wio_mutex[DATA]);
1520 spin_lock_init(&sbi->cp_lock);
1521
1522#ifdef CONFIG_F2FS_FS_ENCRYPTION
1523 memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1524 F2FS_KEY_DESC_PREFIX_SIZE);
1525 sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1526#endif
1527}
1528
1529static int init_percpu_info(struct f2fs_sb_info *sbi)
1530{
1531 int err;
1532
1533 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1534 if (err)
1535 return err;
1536
1537 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1538 GFP_KERNEL);
1539}
1540
1541#ifdef CONFIG_BLK_DEV_ZONED
1542static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
1543{
1544 struct block_device *bdev = FDEV(devi).bdev;
1545 sector_t nr_sectors = bdev->bd_part->nr_sects;
1546 sector_t sector = 0;
1547 struct blk_zone *zones;
1548 unsigned int i, nr_zones;
1549 unsigned int n = 0;
1550 int err = -EIO;
1551
1552 if (!f2fs_sb_mounted_blkzoned(sbi->sb))
1553 return 0;
1554
1555 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
1556 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
1557 return -EINVAL;
1558 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
1559 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
1560 __ilog2_u32(sbi->blocks_per_blkz))
1561 return -EINVAL;
1562 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
1563 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
1564 sbi->log_blocks_per_blkz;
1565 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
1566 FDEV(devi).nr_blkz++;
1567
1568 FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
1569 if (!FDEV(devi).blkz_type)
1570 return -ENOMEM;
1571
1572#define F2FS_REPORT_NR_ZONES 4096
1573
1574 zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
1575 GFP_KERNEL);
1576 if (!zones)
1577 return -ENOMEM;
1578
1579 /* Get block zones type */
1580 while (zones && sector < nr_sectors) {
1581
1582 nr_zones = F2FS_REPORT_NR_ZONES;
1583 err = blkdev_report_zones(bdev, sector,
1584 zones, &nr_zones,
1585 GFP_KERNEL);
1586 if (err)
1587 break;
1588 if (!nr_zones) {
1589 err = -EIO;
1590 break;
1591 }
1592
1593 for (i = 0; i < nr_zones; i++) {
1594 FDEV(devi).blkz_type[n] = zones[i].type;
1595 sector += zones[i].len;
1596 n++;
1597 }
1598 }
1599
1600 kfree(zones);
1601
1602 return err;
1603}
1604#endif
1605
1606/*
1607 * Read f2fs raw super block.
1608 * Because we have two copies of super block, so read both of them
1609 * to get the first valid one. If any one of them is broken, we pass
1610 * them recovery flag back to the caller.
1611 */
1612static int read_raw_super_block(struct f2fs_sb_info *sbi,
1613 struct f2fs_super_block **raw_super,
1614 int *valid_super_block, int *recovery)
1615{
1616 struct super_block *sb = sbi->sb;
1617 int block;
1618 struct buffer_head *bh;
1619 struct f2fs_super_block *super;
1620 int err = 0;
1621
1622 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1623 if (!super)
1624 return -ENOMEM;
1625
1626 for (block = 0; block < 2; block++) {
1627 bh = sb_bread(sb, block);
1628 if (!bh) {
1629 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1630 block + 1);
1631 err = -EIO;
1632 continue;
1633 }
1634
1635 /* sanity checking of raw super */
1636 if (sanity_check_raw_super(sbi, bh)) {
1637 f2fs_msg(sb, KERN_ERR,
1638 "Can't find valid F2FS filesystem in %dth superblock",
1639 block + 1);
1640 err = -EINVAL;
1641 brelse(bh);
1642 continue;
1643 }
1644
1645 if (!*raw_super) {
1646 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1647 sizeof(*super));
1648 *valid_super_block = block;
1649 *raw_super = super;
1650 }
1651 brelse(bh);
1652 }
1653
1654 /* Fail to read any one of the superblocks*/
1655 if (err < 0)
1656 *recovery = 1;
1657
1658 /* No valid superblock */
1659 if (!*raw_super)
1660 kfree(super);
1661 else
1662 err = 0;
1663
1664 return err;
1665}
1666
1667int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1668{
1669 struct buffer_head *bh;
1670 int err;
1671
1672 if ((recover && f2fs_readonly(sbi->sb)) ||
1673 bdev_read_only(sbi->sb->s_bdev)) {
1674 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1675 return -EROFS;
1676 }
1677
1678 /* write back-up superblock first */
1679 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1680 if (!bh)
1681 return -EIO;
1682 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1683 brelse(bh);
1684
1685 /* if we are in recovery path, skip writing valid superblock */
1686 if (recover || err)
1687 return err;
1688
1689 /* write current valid superblock */
1690 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1691 if (!bh)
1692 return -EIO;
1693 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1694 brelse(bh);
1695 return err;
1696}
1697
1698static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
1699{
1700 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1701 unsigned int max_devices = MAX_DEVICES;
1702 int i;
1703
1704 /* Initialize single device information */
1705 if (!RDEV(0).path[0]) {
1706 if (!bdev_is_zoned(sbi->sb->s_bdev))
1707 return 0;
1708 max_devices = 1;
1709 }
1710
1711 /*
1712 * Initialize multiple devices information, or single
1713 * zoned block device information.
1714 */
1715 sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info),
1716 GFP_KERNEL);
1717 if (!sbi->devs)
1718 return -ENOMEM;
1719
1720 for (i = 0; i < max_devices; i++) {
1721
1722 if (i > 0 && !RDEV(i).path[0])
1723 break;
1724
1725 if (max_devices == 1) {
1726 /* Single zoned block device mount */
1727 FDEV(0).bdev =
1728 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
1729 sbi->sb->s_mode, sbi->sb->s_type);
1730 } else {
1731 /* Multi-device mount */
1732 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
1733 FDEV(i).total_segments =
1734 le32_to_cpu(RDEV(i).total_segments);
1735 if (i == 0) {
1736 FDEV(i).start_blk = 0;
1737 FDEV(i).end_blk = FDEV(i).start_blk +
1738 (FDEV(i).total_segments <<
1739 sbi->log_blocks_per_seg) - 1 +
1740 le32_to_cpu(raw_super->segment0_blkaddr);
1741 } else {
1742 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
1743 FDEV(i).end_blk = FDEV(i).start_blk +
1744 (FDEV(i).total_segments <<
1745 sbi->log_blocks_per_seg) - 1;
1746 }
1747 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
1748 sbi->sb->s_mode, sbi->sb->s_type);
1749 }
1750 if (IS_ERR(FDEV(i).bdev))
1751 return PTR_ERR(FDEV(i).bdev);
1752
1753 /* to release errored devices */
1754 sbi->s_ndevs = i + 1;
1755
1756#ifdef CONFIG_BLK_DEV_ZONED
1757 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
1758 !f2fs_sb_mounted_blkzoned(sbi->sb)) {
1759 f2fs_msg(sbi->sb, KERN_ERR,
1760 "Zoned block device feature not enabled\n");
1761 return -EINVAL;
1762 }
1763 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
1764 if (init_blkz_info(sbi, i)) {
1765 f2fs_msg(sbi->sb, KERN_ERR,
1766 "Failed to initialize F2FS blkzone information");
1767 return -EINVAL;
1768 }
1769 if (max_devices == 1)
1770 break;
1771 f2fs_msg(sbi->sb, KERN_INFO,
1772 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
1773 i, FDEV(i).path,
1774 FDEV(i).total_segments,
1775 FDEV(i).start_blk, FDEV(i).end_blk,
1776 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
1777 "Host-aware" : "Host-managed");
1778 continue;
1779 }
1780#endif
1781 f2fs_msg(sbi->sb, KERN_INFO,
1782 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
1783 i, FDEV(i).path,
1784 FDEV(i).total_segments,
1785 FDEV(i).start_blk, FDEV(i).end_blk);
1786 }
1787 return 0;
1788}
1789
1790static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1791{
1792 struct f2fs_sb_info *sbi;
1793 struct f2fs_super_block *raw_super;
1794 struct inode *root;
1795 int err;
1796 bool retry = true, need_fsck = false;
1797 char *options = NULL;
1798 int recovery, i, valid_super_block;
1799 struct curseg_info *seg_i;
1800
1801try_onemore:
1802 err = -EINVAL;
1803 raw_super = NULL;
1804 valid_super_block = -1;
1805 recovery = 0;
1806
1807 /* allocate memory for f2fs-specific super block info */
1808 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1809 if (!sbi)
1810 return -ENOMEM;
1811
1812 sbi->sb = sb;
1813
1814 /* Load the checksum driver */
1815 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1816 if (IS_ERR(sbi->s_chksum_driver)) {
1817 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1818 err = PTR_ERR(sbi->s_chksum_driver);
1819 sbi->s_chksum_driver = NULL;
1820 goto free_sbi;
1821 }
1822
1823 /* set a block size */
1824 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1825 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1826 goto free_sbi;
1827 }
1828
1829 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1830 &recovery);
1831 if (err)
1832 goto free_sbi;
1833
1834 sb->s_fs_info = sbi;
1835 sbi->raw_super = raw_super;
1836
1837 /*
1838 * The BLKZONED feature indicates that the drive was formatted with
1839 * zone alignment optimization. This is optional for host-aware
1840 * devices, but mandatory for host-managed zoned block devices.
1841 */
1842#ifndef CONFIG_BLK_DEV_ZONED
1843 if (f2fs_sb_mounted_blkzoned(sb)) {
1844 f2fs_msg(sb, KERN_ERR,
1845 "Zoned block device support is not enabled\n");
1846 goto free_sb_buf;
1847 }
1848#endif
1849 default_options(sbi);
1850 /* parse mount options */
1851 options = kstrdup((const char *)data, GFP_KERNEL);
1852 if (data && !options) {
1853 err = -ENOMEM;
1854 goto free_sb_buf;
1855 }
1856
1857 err = parse_options(sb, options);
1858 if (err)
1859 goto free_options;
1860
1861 sbi->max_file_blocks = max_file_blocks();
1862 sb->s_maxbytes = sbi->max_file_blocks <<
1863 le32_to_cpu(raw_super->log_blocksize);
1864 sb->s_max_links = F2FS_LINK_MAX;
1865 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1866
1867 sb->s_op = &f2fs_sops;
1868 sb->s_cop = &f2fs_cryptops;
1869 sb->s_xattr = f2fs_xattr_handlers;
1870 sb->s_export_op = &f2fs_export_ops;
1871 sb->s_magic = F2FS_SUPER_MAGIC;
1872 sb->s_time_gran = 1;
1873 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1874 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1875 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1876
1877 /* init f2fs-specific super block info */
1878 sbi->valid_super_block = valid_super_block;
1879 mutex_init(&sbi->gc_mutex);
1880 mutex_init(&sbi->cp_mutex);
1881 init_rwsem(&sbi->node_write);
1882
1883 /* disallow all the data/node/meta page writes */
1884 set_sbi_flag(sbi, SBI_POR_DOING);
1885 spin_lock_init(&sbi->stat_lock);
1886
1887 init_rwsem(&sbi->read_io.io_rwsem);
1888 sbi->read_io.sbi = sbi;
1889 sbi->read_io.bio = NULL;
1890 for (i = 0; i < NR_PAGE_TYPE; i++) {
1891 init_rwsem(&sbi->write_io[i].io_rwsem);
1892 sbi->write_io[i].sbi = sbi;
1893 sbi->write_io[i].bio = NULL;
1894 }
1895
1896 init_rwsem(&sbi->cp_rwsem);
1897 init_waitqueue_head(&sbi->cp_wait);
1898 init_sb_info(sbi);
1899
1900 err = init_percpu_info(sbi);
1901 if (err)
1902 goto free_options;
1903
1904 /* get an inode for meta space */
1905 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1906 if (IS_ERR(sbi->meta_inode)) {
1907 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1908 err = PTR_ERR(sbi->meta_inode);
1909 goto free_options;
1910 }
1911
1912 err = get_valid_checkpoint(sbi);
1913 if (err) {
1914 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1915 goto free_meta_inode;
1916 }
1917
1918 /* Initialize device list */
1919 err = f2fs_scan_devices(sbi);
1920 if (err) {
1921 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
1922 goto free_devices;
1923 }
1924
1925 sbi->total_valid_node_count =
1926 le32_to_cpu(sbi->ckpt->valid_node_count);
1927 percpu_counter_set(&sbi->total_valid_inode_count,
1928 le32_to_cpu(sbi->ckpt->valid_inode_count));
1929 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1930 sbi->total_valid_block_count =
1931 le64_to_cpu(sbi->ckpt->valid_block_count);
1932 sbi->last_valid_block_count = sbi->total_valid_block_count;
1933
1934 for (i = 0; i < NR_INODE_TYPE; i++) {
1935 INIT_LIST_HEAD(&sbi->inode_list[i]);
1936 spin_lock_init(&sbi->inode_lock[i]);
1937 }
1938
1939 init_extent_cache_info(sbi);
1940
1941 init_ino_entry_info(sbi);
1942
1943 /* setup f2fs internal modules */
1944 err = build_segment_manager(sbi);
1945 if (err) {
1946 f2fs_msg(sb, KERN_ERR,
1947 "Failed to initialize F2FS segment manager");
1948 goto free_sm;
1949 }
1950 err = build_node_manager(sbi);
1951 if (err) {
1952 f2fs_msg(sb, KERN_ERR,
1953 "Failed to initialize F2FS node manager");
1954 goto free_nm;
1955 }
1956
1957 /* For write statistics */
1958 if (sb->s_bdev->bd_part)
1959 sbi->sectors_written_start =
1960 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1961
1962 /* Read accumulated write IO statistics if exists */
1963 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1964 if (__exist_node_summaries(sbi))
1965 sbi->kbytes_written =
1966 le64_to_cpu(seg_i->journal->info.kbytes_written);
1967
1968 build_gc_manager(sbi);
1969
1970 /* get an inode for node space */
1971 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1972 if (IS_ERR(sbi->node_inode)) {
1973 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1974 err = PTR_ERR(sbi->node_inode);
1975 goto free_nm;
1976 }
1977
1978 f2fs_join_shrinker(sbi);
1979
1980 /* if there are nt orphan nodes free them */
1981 err = recover_orphan_inodes(sbi);
1982 if (err)
1983 goto free_node_inode;
1984
1985 /* read root inode and dentry */
1986 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1987 if (IS_ERR(root)) {
1988 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1989 err = PTR_ERR(root);
1990 goto free_node_inode;
1991 }
1992 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1993 iput(root);
1994 err = -EINVAL;
1995 goto free_node_inode;
1996 }
1997
1998 sb->s_root = d_make_root(root); /* allocate root dentry */
1999 if (!sb->s_root) {
2000 err = -ENOMEM;
2001 goto free_root_inode;
2002 }
2003
2004 err = f2fs_build_stats(sbi);
2005 if (err)
2006 goto free_root_inode;
2007
2008 if (f2fs_proc_root)
2009 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
2010
2011 if (sbi->s_proc) {
2012 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
2013 &f2fs_seq_segment_info_fops, sb);
2014 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
2015 &f2fs_seq_segment_bits_fops, sb);
2016 }
2017
2018 sbi->s_kobj.kset = f2fs_kset;
2019 init_completion(&sbi->s_kobj_unregister);
2020 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
2021 "%s", sb->s_id);
2022 if (err)
2023 goto free_proc;
2024
2025 /* recover fsynced data */
2026 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2027 /*
2028 * mount should be failed, when device has readonly mode, and
2029 * previous checkpoint was not done by clean system shutdown.
2030 */
2031 if (bdev_read_only(sb->s_bdev) &&
2032 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2033 err = -EROFS;
2034 goto free_kobj;
2035 }
2036
2037 if (need_fsck)
2038 set_sbi_flag(sbi, SBI_NEED_FSCK);
2039
2040 if (!retry)
2041 goto skip_recovery;
2042
2043 err = recover_fsync_data(sbi, false);
2044 if (err < 0) {
2045 need_fsck = true;
2046 f2fs_msg(sb, KERN_ERR,
2047 "Cannot recover all fsync data errno=%d", err);
2048 goto free_kobj;
2049 }
2050 } else {
2051 err = recover_fsync_data(sbi, true);
2052
2053 if (!f2fs_readonly(sb) && err > 0) {
2054 err = -EINVAL;
2055 f2fs_msg(sb, KERN_ERR,
2056 "Need to recover fsync data");
2057 goto free_kobj;
2058 }
2059 }
2060skip_recovery:
2061 /* recover_fsync_data() cleared this already */
2062 clear_sbi_flag(sbi, SBI_POR_DOING);
2063
2064 /*
2065 * If filesystem is not mounted as read-only then
2066 * do start the gc_thread.
2067 */
2068 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2069 /* After POR, we can run background GC thread.*/
2070 err = start_gc_thread(sbi);
2071 if (err)
2072 goto free_kobj;
2073 }
2074 kfree(options);
2075
2076 /* recover broken superblock */
2077 if (recovery) {
2078 err = f2fs_commit_super(sbi, true);
2079 f2fs_msg(sb, KERN_INFO,
2080 "Try to recover %dth superblock, ret: %d",
2081 sbi->valid_super_block ? 1 : 2, err);
2082 }
2083
2084 f2fs_update_time(sbi, CP_TIME);
2085 f2fs_update_time(sbi, REQ_TIME);
2086 return 0;
2087
2088free_kobj:
2089 f2fs_sync_inode_meta(sbi);
2090 kobject_del(&sbi->s_kobj);
2091 kobject_put(&sbi->s_kobj);
2092 wait_for_completion(&sbi->s_kobj_unregister);
2093free_proc:
2094 if (sbi->s_proc) {
2095 remove_proc_entry("segment_info", sbi->s_proc);
2096 remove_proc_entry("segment_bits", sbi->s_proc);
2097 remove_proc_entry(sb->s_id, f2fs_proc_root);
2098 }
2099 f2fs_destroy_stats(sbi);
2100free_root_inode:
2101 dput(sb->s_root);
2102 sb->s_root = NULL;
2103free_node_inode:
2104 truncate_inode_pages_final(NODE_MAPPING(sbi));
2105 mutex_lock(&sbi->umount_mutex);
2106 release_ino_entry(sbi, true);
2107 f2fs_leave_shrinker(sbi);
2108 /*
2109 * Some dirty meta pages can be produced by recover_orphan_inodes()
2110 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2111 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2112 * falls into an infinite loop in sync_meta_pages().
2113 */
2114 truncate_inode_pages_final(META_MAPPING(sbi));
2115 iput(sbi->node_inode);
2116 mutex_unlock(&sbi->umount_mutex);
2117free_nm:
2118 destroy_node_manager(sbi);
2119free_sm:
2120 destroy_segment_manager(sbi);
2121free_devices:
2122 destroy_device_list(sbi);
2123 kfree(sbi->ckpt);
2124free_meta_inode:
2125 make_bad_inode(sbi->meta_inode);
2126 iput(sbi->meta_inode);
2127free_options:
2128 destroy_percpu_info(sbi);
2129 kfree(options);
2130free_sb_buf:
2131 kfree(raw_super);
2132free_sbi:
2133 if (sbi->s_chksum_driver)
2134 crypto_free_shash(sbi->s_chksum_driver);
2135 kfree(sbi);
2136
2137 /* give only one another chance */
2138 if (retry) {
2139 retry = false;
2140 shrink_dcache_sb(sb);
2141 goto try_onemore;
2142 }
2143 return err;
2144}
2145
2146static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2147 const char *dev_name, void *data)
2148{
2149 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2150}
2151
2152static void kill_f2fs_super(struct super_block *sb)
2153{
2154 if (sb->s_root)
2155 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2156 kill_block_super(sb);
2157}
2158
2159static struct file_system_type f2fs_fs_type = {
2160 .owner = THIS_MODULE,
2161 .name = "f2fs",
2162 .mount = f2fs_mount,
2163 .kill_sb = kill_f2fs_super,
2164 .fs_flags = FS_REQUIRES_DEV,
2165};
2166MODULE_ALIAS_FS("f2fs");
2167
2168static int __init init_inodecache(void)
2169{
2170 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2171 sizeof(struct f2fs_inode_info), 0,
2172 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2173 if (!f2fs_inode_cachep)
2174 return -ENOMEM;
2175 return 0;
2176}
2177
2178static void destroy_inodecache(void)
2179{
2180 /*
2181 * Make sure all delayed rcu free inodes are flushed before we
2182 * destroy cache.
2183 */
2184 rcu_barrier();
2185 kmem_cache_destroy(f2fs_inode_cachep);
2186}
2187
2188static int __init init_f2fs_fs(void)
2189{
2190 int err;
2191
2192 f2fs_build_trace_ios();
2193
2194 err = init_inodecache();
2195 if (err)
2196 goto fail;
2197 err = create_node_manager_caches();
2198 if (err)
2199 goto free_inodecache;
2200 err = create_segment_manager_caches();
2201 if (err)
2202 goto free_node_manager_caches;
2203 err = create_checkpoint_caches();
2204 if (err)
2205 goto free_segment_manager_caches;
2206 err = create_extent_cache();
2207 if (err)
2208 goto free_checkpoint_caches;
2209 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
2210 if (!f2fs_kset) {
2211 err = -ENOMEM;
2212 goto free_extent_cache;
2213 }
2214 err = register_shrinker(&f2fs_shrinker_info);
2215 if (err)
2216 goto free_kset;
2217
2218 err = register_filesystem(&f2fs_fs_type);
2219 if (err)
2220 goto free_shrinker;
2221 err = f2fs_create_root_stats();
2222 if (err)
2223 goto free_filesystem;
2224 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
2225 return 0;
2226
2227free_filesystem:
2228 unregister_filesystem(&f2fs_fs_type);
2229free_shrinker:
2230 unregister_shrinker(&f2fs_shrinker_info);
2231free_kset:
2232 kset_unregister(f2fs_kset);
2233free_extent_cache:
2234 destroy_extent_cache();
2235free_checkpoint_caches:
2236 destroy_checkpoint_caches();
2237free_segment_manager_caches:
2238 destroy_segment_manager_caches();
2239free_node_manager_caches:
2240 destroy_node_manager_caches();
2241free_inodecache:
2242 destroy_inodecache();
2243fail:
2244 return err;
2245}
2246
2247static void __exit exit_f2fs_fs(void)
2248{
2249 remove_proc_entry("fs/f2fs", NULL);
2250 f2fs_destroy_root_stats();
2251 unregister_filesystem(&f2fs_fs_type);
2252 unregister_shrinker(&f2fs_shrinker_info);
2253 kset_unregister(f2fs_kset);
2254 destroy_extent_cache();
2255 destroy_checkpoint_caches();
2256 destroy_segment_manager_caches();
2257 destroy_node_manager_caches();
2258 destroy_inodecache();
2259 f2fs_destroy_trace_ios();
2260}
2261
2262module_init(init_f2fs_fs)
2263module_exit(exit_f2fs_fs)
2264
2265MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2266MODULE_DESCRIPTION("Flash Friendly File System");
2267MODULE_LICENSE("GPL");
2268