Loading...
Note: File does not exist in v3.1.
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");