s390/time: add leap seconds to initial system time
[linux/fpc-iii.git] / fs / f2fs / super.c
blob74cc8520b8b1c2f2637ca8d47b5320d0f3a56879
1 /*
2 * fs/f2fs/super.c
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
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.
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>
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "gc.h"
33 #include "trace.h"
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
38 static struct proc_dir_entry *f2fs_proc_root;
39 static struct kmem_cache *f2fs_inode_cachep;
40 static struct kset *f2fs_kset;
42 #ifdef CONFIG_F2FS_FAULT_INJECTION
43 struct f2fs_fault_info f2fs_fault;
45 char *fault_name[FAULT_MAX] = {
46 [FAULT_KMALLOC] = "kmalloc",
47 [FAULT_PAGE_ALLOC] = "page alloc",
48 [FAULT_ALLOC_NID] = "alloc nid",
49 [FAULT_ORPHAN] = "orphan",
50 [FAULT_BLOCK] = "no more block",
51 [FAULT_DIR_DEPTH] = "too big dir depth",
54 static void f2fs_build_fault_attr(unsigned int rate)
56 if (rate) {
57 atomic_set(&f2fs_fault.inject_ops, 0);
58 f2fs_fault.inject_rate = rate;
59 f2fs_fault.inject_type = (1 << FAULT_MAX) - 1;
60 } else {
61 memset(&f2fs_fault, 0, sizeof(struct f2fs_fault_info));
64 #endif
66 /* f2fs-wide shrinker description */
67 static struct shrinker f2fs_shrinker_info = {
68 .scan_objects = f2fs_shrink_scan,
69 .count_objects = f2fs_shrink_count,
70 .seeks = DEFAULT_SEEKS,
73 enum {
74 Opt_gc_background,
75 Opt_disable_roll_forward,
76 Opt_norecovery,
77 Opt_discard,
78 Opt_noheap,
79 Opt_user_xattr,
80 Opt_nouser_xattr,
81 Opt_acl,
82 Opt_noacl,
83 Opt_active_logs,
84 Opt_disable_ext_identify,
85 Opt_inline_xattr,
86 Opt_inline_data,
87 Opt_inline_dentry,
88 Opt_flush_merge,
89 Opt_nobarrier,
90 Opt_fastboot,
91 Opt_extent_cache,
92 Opt_noextent_cache,
93 Opt_noinline_data,
94 Opt_data_flush,
95 Opt_fault_injection,
96 Opt_err,
99 static match_table_t f2fs_tokens = {
100 {Opt_gc_background, "background_gc=%s"},
101 {Opt_disable_roll_forward, "disable_roll_forward"},
102 {Opt_norecovery, "norecovery"},
103 {Opt_discard, "discard"},
104 {Opt_noheap, "no_heap"},
105 {Opt_user_xattr, "user_xattr"},
106 {Opt_nouser_xattr, "nouser_xattr"},
107 {Opt_acl, "acl"},
108 {Opt_noacl, "noacl"},
109 {Opt_active_logs, "active_logs=%u"},
110 {Opt_disable_ext_identify, "disable_ext_identify"},
111 {Opt_inline_xattr, "inline_xattr"},
112 {Opt_inline_data, "inline_data"},
113 {Opt_inline_dentry, "inline_dentry"},
114 {Opt_flush_merge, "flush_merge"},
115 {Opt_nobarrier, "nobarrier"},
116 {Opt_fastboot, "fastboot"},
117 {Opt_extent_cache, "extent_cache"},
118 {Opt_noextent_cache, "noextent_cache"},
119 {Opt_noinline_data, "noinline_data"},
120 {Opt_data_flush, "data_flush"},
121 {Opt_fault_injection, "fault_injection=%u"},
122 {Opt_err, NULL},
125 /* Sysfs support for f2fs */
126 enum {
127 GC_THREAD, /* struct f2fs_gc_thread */
128 SM_INFO, /* struct f2fs_sm_info */
129 NM_INFO, /* struct f2fs_nm_info */
130 F2FS_SBI, /* struct f2fs_sb_info */
131 #ifdef CONFIG_F2FS_FAULT_INJECTION
132 FAULT_INFO_RATE, /* struct f2fs_fault_info */
133 FAULT_INFO_TYPE, /* struct f2fs_fault_info */
134 #endif
137 struct f2fs_attr {
138 struct attribute attr;
139 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
140 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
141 const char *, size_t);
142 int struct_type;
143 int offset;
146 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
148 if (struct_type == GC_THREAD)
149 return (unsigned char *)sbi->gc_thread;
150 else if (struct_type == SM_INFO)
151 return (unsigned char *)SM_I(sbi);
152 else if (struct_type == NM_INFO)
153 return (unsigned char *)NM_I(sbi);
154 else if (struct_type == F2FS_SBI)
155 return (unsigned char *)sbi;
156 #ifdef CONFIG_F2FS_FAULT_INJECTION
157 else if (struct_type == FAULT_INFO_RATE ||
158 struct_type == FAULT_INFO_TYPE)
159 return (unsigned char *)&f2fs_fault;
160 #endif
161 return NULL;
164 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a,
165 struct f2fs_sb_info *sbi, char *buf)
167 struct super_block *sb = sbi->sb;
169 if (!sb->s_bdev->bd_part)
170 return snprintf(buf, PAGE_SIZE, "0\n");
172 return snprintf(buf, PAGE_SIZE, "%llu\n",
173 (unsigned long long)(sbi->kbytes_written +
174 BD_PART_WRITTEN(sbi)));
177 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
178 struct f2fs_sb_info *sbi, char *buf)
180 unsigned char *ptr = NULL;
181 unsigned int *ui;
183 ptr = __struct_ptr(sbi, a->struct_type);
184 if (!ptr)
185 return -EINVAL;
187 ui = (unsigned int *)(ptr + a->offset);
189 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
192 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
193 struct f2fs_sb_info *sbi,
194 const char *buf, size_t count)
196 unsigned char *ptr;
197 unsigned long t;
198 unsigned int *ui;
199 ssize_t ret;
201 ptr = __struct_ptr(sbi, a->struct_type);
202 if (!ptr)
203 return -EINVAL;
205 ui = (unsigned int *)(ptr + a->offset);
207 ret = kstrtoul(skip_spaces(buf), 0, &t);
208 if (ret < 0)
209 return ret;
210 #ifdef CONFIG_F2FS_FAULT_INJECTION
211 if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX))
212 return -EINVAL;
213 #endif
214 *ui = t;
215 return count;
218 static ssize_t f2fs_attr_show(struct kobject *kobj,
219 struct attribute *attr, char *buf)
221 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
222 s_kobj);
223 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
225 return a->show ? a->show(a, sbi, buf) : 0;
228 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
229 const char *buf, size_t len)
231 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
232 s_kobj);
233 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
235 return a->store ? a->store(a, sbi, buf, len) : 0;
238 static void f2fs_sb_release(struct kobject *kobj)
240 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
241 s_kobj);
242 complete(&sbi->s_kobj_unregister);
245 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
246 static struct f2fs_attr f2fs_attr_##_name = { \
247 .attr = {.name = __stringify(_name), .mode = _mode }, \
248 .show = _show, \
249 .store = _store, \
250 .struct_type = _struct_type, \
251 .offset = _offset \
254 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
255 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
256 f2fs_sbi_show, f2fs_sbi_store, \
257 offsetof(struct struct_name, elname))
259 #define F2FS_GENERAL_RO_ATTR(name) \
260 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
262 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
263 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
264 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
265 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
266 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
267 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
268 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
269 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
270 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
271 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
272 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
273 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages);
274 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio);
275 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
276 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
277 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]);
278 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]);
279 #ifdef CONFIG_F2FS_FAULT_INJECTION
280 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate);
281 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type);
282 #endif
283 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes);
285 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
286 static struct attribute *f2fs_attrs[] = {
287 ATTR_LIST(gc_min_sleep_time),
288 ATTR_LIST(gc_max_sleep_time),
289 ATTR_LIST(gc_no_gc_sleep_time),
290 ATTR_LIST(gc_idle),
291 ATTR_LIST(reclaim_segments),
292 ATTR_LIST(max_small_discards),
293 ATTR_LIST(batched_trim_sections),
294 ATTR_LIST(ipu_policy),
295 ATTR_LIST(min_ipu_util),
296 ATTR_LIST(min_fsync_blocks),
297 ATTR_LIST(max_victim_search),
298 ATTR_LIST(dir_level),
299 ATTR_LIST(ram_thresh),
300 ATTR_LIST(ra_nid_pages),
301 ATTR_LIST(dirty_nats_ratio),
302 ATTR_LIST(cp_interval),
303 ATTR_LIST(idle_interval),
304 ATTR_LIST(lifetime_write_kbytes),
305 NULL,
308 static const struct sysfs_ops f2fs_attr_ops = {
309 .show = f2fs_attr_show,
310 .store = f2fs_attr_store,
313 static struct kobj_type f2fs_ktype = {
314 .default_attrs = f2fs_attrs,
315 .sysfs_ops = &f2fs_attr_ops,
316 .release = f2fs_sb_release,
319 #ifdef CONFIG_F2FS_FAULT_INJECTION
320 /* sysfs for f2fs fault injection */
321 static struct kobject f2fs_fault_inject;
323 static struct attribute *f2fs_fault_attrs[] = {
324 ATTR_LIST(inject_rate),
325 ATTR_LIST(inject_type),
326 NULL
329 static struct kobj_type f2fs_fault_ktype = {
330 .default_attrs = f2fs_fault_attrs,
331 .sysfs_ops = &f2fs_attr_ops,
333 #endif
335 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
337 struct va_format vaf;
338 va_list args;
340 va_start(args, fmt);
341 vaf.fmt = fmt;
342 vaf.va = &args;
343 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
344 va_end(args);
347 static void init_once(void *foo)
349 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
351 inode_init_once(&fi->vfs_inode);
354 static int parse_options(struct super_block *sb, char *options)
356 struct f2fs_sb_info *sbi = F2FS_SB(sb);
357 struct request_queue *q;
358 substring_t args[MAX_OPT_ARGS];
359 char *p, *name;
360 int arg = 0;
362 #ifdef CONFIG_F2FS_FAULT_INJECTION
363 f2fs_build_fault_attr(0);
364 #endif
366 if (!options)
367 return 0;
369 while ((p = strsep(&options, ",")) != NULL) {
370 int token;
371 if (!*p)
372 continue;
374 * Initialize args struct so we know whether arg was
375 * found; some options take optional arguments.
377 args[0].to = args[0].from = NULL;
378 token = match_token(p, f2fs_tokens, args);
380 switch (token) {
381 case Opt_gc_background:
382 name = match_strdup(&args[0]);
384 if (!name)
385 return -ENOMEM;
386 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
387 set_opt(sbi, BG_GC);
388 clear_opt(sbi, FORCE_FG_GC);
389 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
390 clear_opt(sbi, BG_GC);
391 clear_opt(sbi, FORCE_FG_GC);
392 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
393 set_opt(sbi, BG_GC);
394 set_opt(sbi, FORCE_FG_GC);
395 } else {
396 kfree(name);
397 return -EINVAL;
399 kfree(name);
400 break;
401 case Opt_disable_roll_forward:
402 set_opt(sbi, DISABLE_ROLL_FORWARD);
403 break;
404 case Opt_norecovery:
405 /* this option mounts f2fs with ro */
406 set_opt(sbi, DISABLE_ROLL_FORWARD);
407 if (!f2fs_readonly(sb))
408 return -EINVAL;
409 break;
410 case Opt_discard:
411 q = bdev_get_queue(sb->s_bdev);
412 if (blk_queue_discard(q)) {
413 set_opt(sbi, DISCARD);
414 } else {
415 f2fs_msg(sb, KERN_WARNING,
416 "mounting with \"discard\" option, but "
417 "the device does not support discard");
419 break;
420 case Opt_noheap:
421 set_opt(sbi, NOHEAP);
422 break;
423 #ifdef CONFIG_F2FS_FS_XATTR
424 case Opt_user_xattr:
425 set_opt(sbi, XATTR_USER);
426 break;
427 case Opt_nouser_xattr:
428 clear_opt(sbi, XATTR_USER);
429 break;
430 case Opt_inline_xattr:
431 set_opt(sbi, INLINE_XATTR);
432 break;
433 #else
434 case Opt_user_xattr:
435 f2fs_msg(sb, KERN_INFO,
436 "user_xattr options not supported");
437 break;
438 case Opt_nouser_xattr:
439 f2fs_msg(sb, KERN_INFO,
440 "nouser_xattr options not supported");
441 break;
442 case Opt_inline_xattr:
443 f2fs_msg(sb, KERN_INFO,
444 "inline_xattr options not supported");
445 break;
446 #endif
447 #ifdef CONFIG_F2FS_FS_POSIX_ACL
448 case Opt_acl:
449 set_opt(sbi, POSIX_ACL);
450 break;
451 case Opt_noacl:
452 clear_opt(sbi, POSIX_ACL);
453 break;
454 #else
455 case Opt_acl:
456 f2fs_msg(sb, KERN_INFO, "acl options not supported");
457 break;
458 case Opt_noacl:
459 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
460 break;
461 #endif
462 case Opt_active_logs:
463 if (args->from && match_int(args, &arg))
464 return -EINVAL;
465 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
466 return -EINVAL;
467 sbi->active_logs = arg;
468 break;
469 case Opt_disable_ext_identify:
470 set_opt(sbi, DISABLE_EXT_IDENTIFY);
471 break;
472 case Opt_inline_data:
473 set_opt(sbi, INLINE_DATA);
474 break;
475 case Opt_inline_dentry:
476 set_opt(sbi, INLINE_DENTRY);
477 break;
478 case Opt_flush_merge:
479 set_opt(sbi, FLUSH_MERGE);
480 break;
481 case Opt_nobarrier:
482 set_opt(sbi, NOBARRIER);
483 break;
484 case Opt_fastboot:
485 set_opt(sbi, FASTBOOT);
486 break;
487 case Opt_extent_cache:
488 set_opt(sbi, EXTENT_CACHE);
489 break;
490 case Opt_noextent_cache:
491 clear_opt(sbi, EXTENT_CACHE);
492 break;
493 case Opt_noinline_data:
494 clear_opt(sbi, INLINE_DATA);
495 break;
496 case Opt_data_flush:
497 set_opt(sbi, DATA_FLUSH);
498 break;
499 case Opt_fault_injection:
500 if (args->from && match_int(args, &arg))
501 return -EINVAL;
502 #ifdef CONFIG_F2FS_FAULT_INJECTION
503 f2fs_build_fault_attr(arg);
504 #else
505 f2fs_msg(sb, KERN_INFO,
506 "FAULT_INJECTION was not selected");
507 #endif
508 break;
509 default:
510 f2fs_msg(sb, KERN_ERR,
511 "Unrecognized mount option \"%s\" or missing value",
513 return -EINVAL;
516 return 0;
519 static struct inode *f2fs_alloc_inode(struct super_block *sb)
521 struct f2fs_inode_info *fi;
523 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
524 if (!fi)
525 return NULL;
527 init_once((void *) fi);
529 if (percpu_counter_init(&fi->dirty_pages, 0, GFP_NOFS)) {
530 kmem_cache_free(f2fs_inode_cachep, fi);
531 return NULL;
534 /* Initialize f2fs-specific inode info */
535 fi->vfs_inode.i_version = 1;
536 fi->i_current_depth = 1;
537 fi->i_advise = 0;
538 init_rwsem(&fi->i_sem);
539 INIT_LIST_HEAD(&fi->dirty_list);
540 INIT_LIST_HEAD(&fi->inmem_pages);
541 mutex_init(&fi->inmem_lock);
543 set_inode_flag(fi, FI_NEW_INODE);
545 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
546 set_inode_flag(fi, FI_INLINE_XATTR);
548 /* Will be used by directory only */
549 fi->i_dir_level = F2FS_SB(sb)->dir_level;
550 return &fi->vfs_inode;
553 static int f2fs_drop_inode(struct inode *inode)
556 * This is to avoid a deadlock condition like below.
557 * writeback_single_inode(inode)
558 * - f2fs_write_data_page
559 * - f2fs_gc -> iput -> evict
560 * - inode_wait_for_writeback(inode)
562 if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
563 if (!inode->i_nlink && !is_bad_inode(inode)) {
564 /* to avoid evict_inode call simultaneously */
565 atomic_inc(&inode->i_count);
566 spin_unlock(&inode->i_lock);
568 /* some remained atomic pages should discarded */
569 if (f2fs_is_atomic_file(inode))
570 drop_inmem_pages(inode);
572 /* should remain fi->extent_tree for writepage */
573 f2fs_destroy_extent_node(inode);
575 sb_start_intwrite(inode->i_sb);
576 i_size_write(inode, 0);
578 if (F2FS_HAS_BLOCKS(inode))
579 f2fs_truncate(inode, true);
581 sb_end_intwrite(inode->i_sb);
583 fscrypt_put_encryption_info(inode, NULL);
584 spin_lock(&inode->i_lock);
585 atomic_dec(&inode->i_count);
587 return 0;
589 return generic_drop_inode(inode);
593 * f2fs_dirty_inode() is called from __mark_inode_dirty()
595 * We should call set_dirty_inode to write the dirty inode through write_inode.
597 static void f2fs_dirty_inode(struct inode *inode, int flags)
599 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
602 static void f2fs_i_callback(struct rcu_head *head)
604 struct inode *inode = container_of(head, struct inode, i_rcu);
605 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
608 static void f2fs_destroy_inode(struct inode *inode)
610 percpu_counter_destroy(&F2FS_I(inode)->dirty_pages);
611 call_rcu(&inode->i_rcu, f2fs_i_callback);
614 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
616 int i;
618 for (i = 0; i < NR_COUNT_TYPE; i++)
619 percpu_counter_destroy(&sbi->nr_pages[i]);
620 percpu_counter_destroy(&sbi->alloc_valid_block_count);
621 percpu_counter_destroy(&sbi->total_valid_inode_count);
624 static void f2fs_put_super(struct super_block *sb)
626 struct f2fs_sb_info *sbi = F2FS_SB(sb);
628 if (sbi->s_proc) {
629 remove_proc_entry("segment_info", sbi->s_proc);
630 remove_proc_entry("segment_bits", sbi->s_proc);
631 remove_proc_entry(sb->s_id, f2fs_proc_root);
633 kobject_del(&sbi->s_kobj);
635 stop_gc_thread(sbi);
637 /* prevent remaining shrinker jobs */
638 mutex_lock(&sbi->umount_mutex);
641 * We don't need to do checkpoint when superblock is clean.
642 * But, the previous checkpoint was not done by umount, it needs to do
643 * clean checkpoint again.
645 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
646 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
647 struct cp_control cpc = {
648 .reason = CP_UMOUNT,
650 write_checkpoint(sbi, &cpc);
653 /* write_checkpoint can update stat informaion */
654 f2fs_destroy_stats(sbi);
657 * normally superblock is clean, so we need to release this.
658 * In addition, EIO will skip do checkpoint, we need this as well.
660 release_ino_entry(sbi, true);
661 release_discard_addrs(sbi);
663 f2fs_leave_shrinker(sbi);
664 mutex_unlock(&sbi->umount_mutex);
666 /* our cp_error case, we can wait for any writeback page */
667 f2fs_flush_merged_bios(sbi);
669 iput(sbi->node_inode);
670 iput(sbi->meta_inode);
672 /* destroy f2fs internal modules */
673 destroy_node_manager(sbi);
674 destroy_segment_manager(sbi);
676 kfree(sbi->ckpt);
677 kobject_put(&sbi->s_kobj);
678 wait_for_completion(&sbi->s_kobj_unregister);
680 sb->s_fs_info = NULL;
681 if (sbi->s_chksum_driver)
682 crypto_free_shash(sbi->s_chksum_driver);
683 kfree(sbi->raw_super);
685 destroy_percpu_info(sbi);
686 kfree(sbi);
689 int f2fs_sync_fs(struct super_block *sb, int sync)
691 struct f2fs_sb_info *sbi = F2FS_SB(sb);
692 int err = 0;
694 trace_f2fs_sync_fs(sb, sync);
696 if (sync) {
697 struct cp_control cpc;
699 cpc.reason = __get_cp_reason(sbi);
701 mutex_lock(&sbi->gc_mutex);
702 err = write_checkpoint(sbi, &cpc);
703 mutex_unlock(&sbi->gc_mutex);
705 f2fs_trace_ios(NULL, 1);
707 return err;
710 static int f2fs_freeze(struct super_block *sb)
712 int err;
714 if (f2fs_readonly(sb))
715 return 0;
717 err = f2fs_sync_fs(sb, 1);
718 return err;
721 static int f2fs_unfreeze(struct super_block *sb)
723 return 0;
726 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
728 struct super_block *sb = dentry->d_sb;
729 struct f2fs_sb_info *sbi = F2FS_SB(sb);
730 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
731 block_t total_count, user_block_count, start_count, ovp_count;
733 total_count = le64_to_cpu(sbi->raw_super->block_count);
734 user_block_count = sbi->user_block_count;
735 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
736 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
737 buf->f_type = F2FS_SUPER_MAGIC;
738 buf->f_bsize = sbi->blocksize;
740 buf->f_blocks = total_count - start_count;
741 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
742 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
744 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
745 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
747 buf->f_namelen = F2FS_NAME_LEN;
748 buf->f_fsid.val[0] = (u32)id;
749 buf->f_fsid.val[1] = (u32)(id >> 32);
751 return 0;
754 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
756 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
758 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
759 if (test_opt(sbi, FORCE_FG_GC))
760 seq_printf(seq, ",background_gc=%s", "sync");
761 else
762 seq_printf(seq, ",background_gc=%s", "on");
763 } else {
764 seq_printf(seq, ",background_gc=%s", "off");
766 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
767 seq_puts(seq, ",disable_roll_forward");
768 if (test_opt(sbi, DISCARD))
769 seq_puts(seq, ",discard");
770 if (test_opt(sbi, NOHEAP))
771 seq_puts(seq, ",no_heap_alloc");
772 #ifdef CONFIG_F2FS_FS_XATTR
773 if (test_opt(sbi, XATTR_USER))
774 seq_puts(seq, ",user_xattr");
775 else
776 seq_puts(seq, ",nouser_xattr");
777 if (test_opt(sbi, INLINE_XATTR))
778 seq_puts(seq, ",inline_xattr");
779 #endif
780 #ifdef CONFIG_F2FS_FS_POSIX_ACL
781 if (test_opt(sbi, POSIX_ACL))
782 seq_puts(seq, ",acl");
783 else
784 seq_puts(seq, ",noacl");
785 #endif
786 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
787 seq_puts(seq, ",disable_ext_identify");
788 if (test_opt(sbi, INLINE_DATA))
789 seq_puts(seq, ",inline_data");
790 else
791 seq_puts(seq, ",noinline_data");
792 if (test_opt(sbi, INLINE_DENTRY))
793 seq_puts(seq, ",inline_dentry");
794 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
795 seq_puts(seq, ",flush_merge");
796 if (test_opt(sbi, NOBARRIER))
797 seq_puts(seq, ",nobarrier");
798 if (test_opt(sbi, FASTBOOT))
799 seq_puts(seq, ",fastboot");
800 if (test_opt(sbi, EXTENT_CACHE))
801 seq_puts(seq, ",extent_cache");
802 else
803 seq_puts(seq, ",noextent_cache");
804 if (test_opt(sbi, DATA_FLUSH))
805 seq_puts(seq, ",data_flush");
806 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
808 return 0;
811 static int segment_info_seq_show(struct seq_file *seq, void *offset)
813 struct super_block *sb = seq->private;
814 struct f2fs_sb_info *sbi = F2FS_SB(sb);
815 unsigned int total_segs =
816 le32_to_cpu(sbi->raw_super->segment_count_main);
817 int i;
819 seq_puts(seq, "format: segment_type|valid_blocks\n"
820 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
822 for (i = 0; i < total_segs; i++) {
823 struct seg_entry *se = get_seg_entry(sbi, i);
825 if ((i % 10) == 0)
826 seq_printf(seq, "%-10d", i);
827 seq_printf(seq, "%d|%-3u", se->type,
828 get_valid_blocks(sbi, i, 1));
829 if ((i % 10) == 9 || i == (total_segs - 1))
830 seq_putc(seq, '\n');
831 else
832 seq_putc(seq, ' ');
835 return 0;
838 static int segment_bits_seq_show(struct seq_file *seq, void *offset)
840 struct super_block *sb = seq->private;
841 struct f2fs_sb_info *sbi = F2FS_SB(sb);
842 unsigned int total_segs =
843 le32_to_cpu(sbi->raw_super->segment_count_main);
844 int i, j;
846 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n"
847 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
849 for (i = 0; i < total_segs; i++) {
850 struct seg_entry *se = get_seg_entry(sbi, i);
852 seq_printf(seq, "%-10d", i);
853 seq_printf(seq, "%d|%-3u|", se->type,
854 get_valid_blocks(sbi, i, 1));
855 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++)
856 seq_printf(seq, "%x ", se->cur_valid_map[j]);
857 seq_putc(seq, '\n');
859 return 0;
862 #define F2FS_PROC_FILE_DEF(_name) \
863 static int _name##_open_fs(struct inode *inode, struct file *file) \
865 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
868 static const struct file_operations f2fs_seq_##_name##_fops = { \
869 .owner = THIS_MODULE, \
870 .open = _name##_open_fs, \
871 .read = seq_read, \
872 .llseek = seq_lseek, \
873 .release = single_release, \
876 F2FS_PROC_FILE_DEF(segment_info);
877 F2FS_PROC_FILE_DEF(segment_bits);
879 static void default_options(struct f2fs_sb_info *sbi)
881 /* init some FS parameters */
882 sbi->active_logs = NR_CURSEG_TYPE;
884 set_opt(sbi, BG_GC);
885 set_opt(sbi, INLINE_DATA);
886 set_opt(sbi, EXTENT_CACHE);
888 #ifdef CONFIG_F2FS_FS_XATTR
889 set_opt(sbi, XATTR_USER);
890 #endif
891 #ifdef CONFIG_F2FS_FS_POSIX_ACL
892 set_opt(sbi, POSIX_ACL);
893 #endif
896 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
898 struct f2fs_sb_info *sbi = F2FS_SB(sb);
899 struct f2fs_mount_info org_mount_opt;
900 int err, active_logs;
901 bool need_restart_gc = false;
902 bool need_stop_gc = false;
903 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
906 * Save the old mount options in case we
907 * need to restore them.
909 org_mount_opt = sbi->mount_opt;
910 active_logs = sbi->active_logs;
912 /* recover superblocks we couldn't write due to previous RO mount */
913 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
914 err = f2fs_commit_super(sbi, false);
915 f2fs_msg(sb, KERN_INFO,
916 "Try to recover all the superblocks, ret: %d", err);
917 if (!err)
918 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
921 sbi->mount_opt.opt = 0;
922 default_options(sbi);
924 /* parse mount options */
925 err = parse_options(sb, data);
926 if (err)
927 goto restore_opts;
930 * Previous and new state of filesystem is RO,
931 * so skip checking GC and FLUSH_MERGE conditions.
933 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
934 goto skip;
936 /* disallow enable/disable extent_cache dynamically */
937 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
938 err = -EINVAL;
939 f2fs_msg(sbi->sb, KERN_WARNING,
940 "switch extent_cache option is not allowed");
941 goto restore_opts;
945 * We stop the GC thread if FS is mounted as RO
946 * or if background_gc = off is passed in mount
947 * option. Also sync the filesystem.
949 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
950 if (sbi->gc_thread) {
951 stop_gc_thread(sbi);
952 need_restart_gc = true;
954 } else if (!sbi->gc_thread) {
955 err = start_gc_thread(sbi);
956 if (err)
957 goto restore_opts;
958 need_stop_gc = true;
961 if (*flags & MS_RDONLY) {
962 writeback_inodes_sb(sb, WB_REASON_SYNC);
963 sync_inodes_sb(sb);
965 set_sbi_flag(sbi, SBI_IS_DIRTY);
966 set_sbi_flag(sbi, SBI_IS_CLOSE);
967 f2fs_sync_fs(sb, 1);
968 clear_sbi_flag(sbi, SBI_IS_CLOSE);
972 * We stop issue flush thread if FS is mounted as RO
973 * or if flush_merge is not passed in mount option.
975 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
976 destroy_flush_cmd_control(sbi);
977 } else if (!SM_I(sbi)->cmd_control_info) {
978 err = create_flush_cmd_control(sbi);
979 if (err)
980 goto restore_gc;
982 skip:
983 /* Update the POSIXACL Flag */
984 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
985 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
987 return 0;
988 restore_gc:
989 if (need_restart_gc) {
990 if (start_gc_thread(sbi))
991 f2fs_msg(sbi->sb, KERN_WARNING,
992 "background gc thread has stopped");
993 } else if (need_stop_gc) {
994 stop_gc_thread(sbi);
996 restore_opts:
997 sbi->mount_opt = org_mount_opt;
998 sbi->active_logs = active_logs;
999 return err;
1002 static struct super_operations f2fs_sops = {
1003 .alloc_inode = f2fs_alloc_inode,
1004 .drop_inode = f2fs_drop_inode,
1005 .destroy_inode = f2fs_destroy_inode,
1006 .write_inode = f2fs_write_inode,
1007 .dirty_inode = f2fs_dirty_inode,
1008 .show_options = f2fs_show_options,
1009 .evict_inode = f2fs_evict_inode,
1010 .put_super = f2fs_put_super,
1011 .sync_fs = f2fs_sync_fs,
1012 .freeze_fs = f2fs_freeze,
1013 .unfreeze_fs = f2fs_unfreeze,
1014 .statfs = f2fs_statfs,
1015 .remount_fs = f2fs_remount,
1018 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1019 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1021 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1022 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1023 ctx, len, NULL);
1026 static int f2fs_key_prefix(struct inode *inode, u8 **key)
1028 *key = F2FS_I_SB(inode)->key_prefix;
1029 return F2FS_I_SB(inode)->key_prefix_size;
1032 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1033 void *fs_data)
1035 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1036 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1037 ctx, len, fs_data, XATTR_CREATE);
1040 static unsigned f2fs_max_namelen(struct inode *inode)
1042 return S_ISLNK(inode->i_mode) ?
1043 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1046 static struct fscrypt_operations f2fs_cryptops = {
1047 .get_context = f2fs_get_context,
1048 .key_prefix = f2fs_key_prefix,
1049 .set_context = f2fs_set_context,
1050 .is_encrypted = f2fs_encrypted_inode,
1051 .empty_dir = f2fs_empty_dir,
1052 .max_namelen = f2fs_max_namelen,
1054 #else
1055 static struct fscrypt_operations f2fs_cryptops = {
1056 .is_encrypted = f2fs_encrypted_inode,
1058 #endif
1060 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1061 u64 ino, u32 generation)
1063 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1064 struct inode *inode;
1066 if (check_nid_range(sbi, ino))
1067 return ERR_PTR(-ESTALE);
1070 * f2fs_iget isn't quite right if the inode is currently unallocated!
1071 * However f2fs_iget currently does appropriate checks to handle stale
1072 * inodes so everything is OK.
1074 inode = f2fs_iget(sb, ino);
1075 if (IS_ERR(inode))
1076 return ERR_CAST(inode);
1077 if (unlikely(generation && inode->i_generation != generation)) {
1078 /* we didn't find the right inode.. */
1079 iput(inode);
1080 return ERR_PTR(-ESTALE);
1082 return inode;
1085 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1086 int fh_len, int fh_type)
1088 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1089 f2fs_nfs_get_inode);
1092 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1093 int fh_len, int fh_type)
1095 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1096 f2fs_nfs_get_inode);
1099 static const struct export_operations f2fs_export_ops = {
1100 .fh_to_dentry = f2fs_fh_to_dentry,
1101 .fh_to_parent = f2fs_fh_to_parent,
1102 .get_parent = f2fs_get_parent,
1105 static loff_t max_file_blocks(void)
1107 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
1108 loff_t leaf_count = ADDRS_PER_BLOCK;
1110 /* two direct node blocks */
1111 result += (leaf_count * 2);
1113 /* two indirect node blocks */
1114 leaf_count *= NIDS_PER_BLOCK;
1115 result += (leaf_count * 2);
1117 /* one double indirect node block */
1118 leaf_count *= NIDS_PER_BLOCK;
1119 result += leaf_count;
1121 return result;
1124 static int __f2fs_commit_super(struct buffer_head *bh,
1125 struct f2fs_super_block *super)
1127 lock_buffer(bh);
1128 if (super)
1129 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1130 set_buffer_uptodate(bh);
1131 set_buffer_dirty(bh);
1132 unlock_buffer(bh);
1134 /* it's rare case, we can do fua all the time */
1135 return __sync_dirty_buffer(bh, WRITE_FLUSH_FUA);
1138 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1139 struct buffer_head *bh)
1141 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1142 (bh->b_data + F2FS_SUPER_OFFSET);
1143 struct super_block *sb = sbi->sb;
1144 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1145 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1146 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1147 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1148 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1149 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1150 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1151 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1152 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1153 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1154 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1155 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1156 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1157 u64 main_end_blkaddr = main_blkaddr +
1158 (segment_count_main << log_blocks_per_seg);
1159 u64 seg_end_blkaddr = segment0_blkaddr +
1160 (segment_count << log_blocks_per_seg);
1162 if (segment0_blkaddr != cp_blkaddr) {
1163 f2fs_msg(sb, KERN_INFO,
1164 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1165 segment0_blkaddr, cp_blkaddr);
1166 return true;
1169 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1170 sit_blkaddr) {
1171 f2fs_msg(sb, KERN_INFO,
1172 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1173 cp_blkaddr, sit_blkaddr,
1174 segment_count_ckpt << log_blocks_per_seg);
1175 return true;
1178 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1179 nat_blkaddr) {
1180 f2fs_msg(sb, KERN_INFO,
1181 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1182 sit_blkaddr, nat_blkaddr,
1183 segment_count_sit << log_blocks_per_seg);
1184 return true;
1187 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1188 ssa_blkaddr) {
1189 f2fs_msg(sb, KERN_INFO,
1190 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1191 nat_blkaddr, ssa_blkaddr,
1192 segment_count_nat << log_blocks_per_seg);
1193 return true;
1196 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1197 main_blkaddr) {
1198 f2fs_msg(sb, KERN_INFO,
1199 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1200 ssa_blkaddr, main_blkaddr,
1201 segment_count_ssa << log_blocks_per_seg);
1202 return true;
1205 if (main_end_blkaddr > seg_end_blkaddr) {
1206 f2fs_msg(sb, KERN_INFO,
1207 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1208 main_blkaddr,
1209 segment0_blkaddr +
1210 (segment_count << log_blocks_per_seg),
1211 segment_count_main << log_blocks_per_seg);
1212 return true;
1213 } else if (main_end_blkaddr < seg_end_blkaddr) {
1214 int err = 0;
1215 char *res;
1217 /* fix in-memory information all the time */
1218 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1219 segment0_blkaddr) >> log_blocks_per_seg);
1221 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1222 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1223 res = "internally";
1224 } else {
1225 err = __f2fs_commit_super(bh, NULL);
1226 res = err ? "failed" : "done";
1228 f2fs_msg(sb, KERN_INFO,
1229 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1230 res, main_blkaddr,
1231 segment0_blkaddr +
1232 (segment_count << log_blocks_per_seg),
1233 segment_count_main << log_blocks_per_seg);
1234 if (err)
1235 return true;
1237 return false;
1240 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1241 struct buffer_head *bh)
1243 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1244 (bh->b_data + F2FS_SUPER_OFFSET);
1245 struct super_block *sb = sbi->sb;
1246 unsigned int blocksize;
1248 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1249 f2fs_msg(sb, KERN_INFO,
1250 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1251 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1252 return 1;
1255 /* Currently, support only 4KB page cache size */
1256 if (F2FS_BLKSIZE != PAGE_SIZE) {
1257 f2fs_msg(sb, KERN_INFO,
1258 "Invalid page_cache_size (%lu), supports only 4KB\n",
1259 PAGE_SIZE);
1260 return 1;
1263 /* Currently, support only 4KB block size */
1264 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1265 if (blocksize != F2FS_BLKSIZE) {
1266 f2fs_msg(sb, KERN_INFO,
1267 "Invalid blocksize (%u), supports only 4KB\n",
1268 blocksize);
1269 return 1;
1272 /* check log blocks per segment */
1273 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1274 f2fs_msg(sb, KERN_INFO,
1275 "Invalid log blocks per segment (%u)\n",
1276 le32_to_cpu(raw_super->log_blocks_per_seg));
1277 return 1;
1280 /* Currently, support 512/1024/2048/4096 bytes sector size */
1281 if (le32_to_cpu(raw_super->log_sectorsize) >
1282 F2FS_MAX_LOG_SECTOR_SIZE ||
1283 le32_to_cpu(raw_super->log_sectorsize) <
1284 F2FS_MIN_LOG_SECTOR_SIZE) {
1285 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1286 le32_to_cpu(raw_super->log_sectorsize));
1287 return 1;
1289 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1290 le32_to_cpu(raw_super->log_sectorsize) !=
1291 F2FS_MAX_LOG_SECTOR_SIZE) {
1292 f2fs_msg(sb, KERN_INFO,
1293 "Invalid log sectors per block(%u) log sectorsize(%u)",
1294 le32_to_cpu(raw_super->log_sectors_per_block),
1295 le32_to_cpu(raw_super->log_sectorsize));
1296 return 1;
1299 /* check reserved ino info */
1300 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1301 le32_to_cpu(raw_super->meta_ino) != 2 ||
1302 le32_to_cpu(raw_super->root_ino) != 3) {
1303 f2fs_msg(sb, KERN_INFO,
1304 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1305 le32_to_cpu(raw_super->node_ino),
1306 le32_to_cpu(raw_super->meta_ino),
1307 le32_to_cpu(raw_super->root_ino));
1308 return 1;
1311 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1312 if (sanity_check_area_boundary(sbi, bh))
1313 return 1;
1315 return 0;
1318 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1320 unsigned int total, fsmeta;
1321 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1322 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1324 total = le32_to_cpu(raw_super->segment_count);
1325 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1326 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
1327 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
1328 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1329 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1331 if (unlikely(fsmeta >= total))
1332 return 1;
1334 if (unlikely(f2fs_cp_error(sbi))) {
1335 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
1336 return 1;
1338 return 0;
1341 static void init_sb_info(struct f2fs_sb_info *sbi)
1343 struct f2fs_super_block *raw_super = sbi->raw_super;
1345 sbi->log_sectors_per_block =
1346 le32_to_cpu(raw_super->log_sectors_per_block);
1347 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
1348 sbi->blocksize = 1 << sbi->log_blocksize;
1349 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1350 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1351 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1352 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1353 sbi->total_sections = le32_to_cpu(raw_super->section_count);
1354 sbi->total_node_count =
1355 (le32_to_cpu(raw_super->segment_count_nat) / 2)
1356 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1357 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
1358 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
1359 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
1360 sbi->cur_victim_sec = NULL_SECNO;
1361 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
1363 sbi->dir_level = DEF_DIR_LEVEL;
1364 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
1365 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
1366 clear_sbi_flag(sbi, SBI_NEED_FSCK);
1368 INIT_LIST_HEAD(&sbi->s_list);
1369 mutex_init(&sbi->umount_mutex);
1371 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1372 memcpy(sbi->key_prefix, F2FS_KEY_DESC_PREFIX,
1373 F2FS_KEY_DESC_PREFIX_SIZE);
1374 sbi->key_prefix_size = F2FS_KEY_DESC_PREFIX_SIZE;
1375 #endif
1378 static int init_percpu_info(struct f2fs_sb_info *sbi)
1380 int i, err;
1382 for (i = 0; i < NR_COUNT_TYPE; i++) {
1383 err = percpu_counter_init(&sbi->nr_pages[i], 0, GFP_KERNEL);
1384 if (err)
1385 return err;
1388 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
1389 if (err)
1390 return err;
1392 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
1393 GFP_KERNEL);
1397 * Read f2fs raw super block.
1398 * Because we have two copies of super block, so read both of them
1399 * to get the first valid one. If any one of them is broken, we pass
1400 * them recovery flag back to the caller.
1402 static int read_raw_super_block(struct f2fs_sb_info *sbi,
1403 struct f2fs_super_block **raw_super,
1404 int *valid_super_block, int *recovery)
1406 struct super_block *sb = sbi->sb;
1407 int block;
1408 struct buffer_head *bh;
1409 struct f2fs_super_block *super;
1410 int err = 0;
1412 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
1413 if (!super)
1414 return -ENOMEM;
1416 for (block = 0; block < 2; block++) {
1417 bh = sb_bread(sb, block);
1418 if (!bh) {
1419 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1420 block + 1);
1421 err = -EIO;
1422 continue;
1425 /* sanity checking of raw super */
1426 if (sanity_check_raw_super(sbi, bh)) {
1427 f2fs_msg(sb, KERN_ERR,
1428 "Can't find valid F2FS filesystem in %dth superblock",
1429 block + 1);
1430 err = -EINVAL;
1431 brelse(bh);
1432 continue;
1435 if (!*raw_super) {
1436 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
1437 sizeof(*super));
1438 *valid_super_block = block;
1439 *raw_super = super;
1441 brelse(bh);
1444 /* Fail to read any one of the superblocks*/
1445 if (err < 0)
1446 *recovery = 1;
1448 /* No valid superblock */
1449 if (!*raw_super)
1450 kfree(super);
1451 else
1452 err = 0;
1454 return err;
1457 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
1459 struct buffer_head *bh;
1460 int err;
1462 if ((recover && f2fs_readonly(sbi->sb)) ||
1463 bdev_read_only(sbi->sb->s_bdev)) {
1464 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1465 return -EROFS;
1468 /* write back-up superblock first */
1469 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
1470 if (!bh)
1471 return -EIO;
1472 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1473 brelse(bh);
1475 /* if we are in recovery path, skip writing valid superblock */
1476 if (recover || err)
1477 return err;
1479 /* write current valid superblock */
1480 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
1481 if (!bh)
1482 return -EIO;
1483 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
1484 brelse(bh);
1485 return err;
1488 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1490 struct f2fs_sb_info *sbi;
1491 struct f2fs_super_block *raw_super;
1492 struct inode *root;
1493 int err;
1494 bool retry = true, need_fsck = false;
1495 char *options = NULL;
1496 int recovery, i, valid_super_block;
1497 struct curseg_info *seg_i;
1499 try_onemore:
1500 err = -EINVAL;
1501 raw_super = NULL;
1502 valid_super_block = -1;
1503 recovery = 0;
1505 /* allocate memory for f2fs-specific super block info */
1506 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1507 if (!sbi)
1508 return -ENOMEM;
1510 sbi->sb = sb;
1512 /* Load the checksum driver */
1513 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
1514 if (IS_ERR(sbi->s_chksum_driver)) {
1515 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
1516 err = PTR_ERR(sbi->s_chksum_driver);
1517 sbi->s_chksum_driver = NULL;
1518 goto free_sbi;
1521 /* set a block size */
1522 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
1523 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
1524 goto free_sbi;
1527 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
1528 &recovery);
1529 if (err)
1530 goto free_sbi;
1532 sb->s_fs_info = sbi;
1533 default_options(sbi);
1534 /* parse mount options */
1535 options = kstrdup((const char *)data, GFP_KERNEL);
1536 if (data && !options) {
1537 err = -ENOMEM;
1538 goto free_sb_buf;
1541 err = parse_options(sb, options);
1542 if (err)
1543 goto free_options;
1545 sbi->max_file_blocks = max_file_blocks();
1546 sb->s_maxbytes = sbi->max_file_blocks <<
1547 le32_to_cpu(raw_super->log_blocksize);
1548 sb->s_max_links = F2FS_LINK_MAX;
1549 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1551 sb->s_op = &f2fs_sops;
1552 sb->s_cop = &f2fs_cryptops;
1553 sb->s_xattr = f2fs_xattr_handlers;
1554 sb->s_export_op = &f2fs_export_ops;
1555 sb->s_magic = F2FS_SUPER_MAGIC;
1556 sb->s_time_gran = 1;
1557 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1558 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1559 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1561 /* init f2fs-specific super block info */
1562 sbi->raw_super = raw_super;
1563 sbi->valid_super_block = valid_super_block;
1564 mutex_init(&sbi->gc_mutex);
1565 mutex_init(&sbi->writepages);
1566 mutex_init(&sbi->cp_mutex);
1567 init_rwsem(&sbi->node_write);
1569 /* disallow all the data/node/meta page writes */
1570 set_sbi_flag(sbi, SBI_POR_DOING);
1571 spin_lock_init(&sbi->stat_lock);
1573 init_rwsem(&sbi->read_io.io_rwsem);
1574 sbi->read_io.sbi = sbi;
1575 sbi->read_io.bio = NULL;
1576 for (i = 0; i < NR_PAGE_TYPE; i++) {
1577 init_rwsem(&sbi->write_io[i].io_rwsem);
1578 sbi->write_io[i].sbi = sbi;
1579 sbi->write_io[i].bio = NULL;
1582 init_rwsem(&sbi->cp_rwsem);
1583 init_waitqueue_head(&sbi->cp_wait);
1584 init_sb_info(sbi);
1586 err = init_percpu_info(sbi);
1587 if (err)
1588 goto free_options;
1590 /* get an inode for meta space */
1591 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1592 if (IS_ERR(sbi->meta_inode)) {
1593 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1594 err = PTR_ERR(sbi->meta_inode);
1595 goto free_options;
1598 err = get_valid_checkpoint(sbi);
1599 if (err) {
1600 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1601 goto free_meta_inode;
1604 sbi->total_valid_node_count =
1605 le32_to_cpu(sbi->ckpt->valid_node_count);
1606 percpu_counter_set(&sbi->total_valid_inode_count,
1607 le32_to_cpu(sbi->ckpt->valid_inode_count));
1608 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1609 sbi->total_valid_block_count =
1610 le64_to_cpu(sbi->ckpt->valid_block_count);
1611 sbi->last_valid_block_count = sbi->total_valid_block_count;
1613 for (i = 0; i < NR_INODE_TYPE; i++) {
1614 INIT_LIST_HEAD(&sbi->inode_list[i]);
1615 spin_lock_init(&sbi->inode_lock[i]);
1618 init_extent_cache_info(sbi);
1620 init_ino_entry_info(sbi);
1622 /* setup f2fs internal modules */
1623 err = build_segment_manager(sbi);
1624 if (err) {
1625 f2fs_msg(sb, KERN_ERR,
1626 "Failed to initialize F2FS segment manager");
1627 goto free_sm;
1629 err = build_node_manager(sbi);
1630 if (err) {
1631 f2fs_msg(sb, KERN_ERR,
1632 "Failed to initialize F2FS node manager");
1633 goto free_nm;
1636 /* For write statistics */
1637 if (sb->s_bdev->bd_part)
1638 sbi->sectors_written_start =
1639 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
1641 /* Read accumulated write IO statistics if exists */
1642 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1643 if (__exist_node_summaries(sbi))
1644 sbi->kbytes_written =
1645 le64_to_cpu(seg_i->journal->info.kbytes_written);
1647 build_gc_manager(sbi);
1649 /* get an inode for node space */
1650 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1651 if (IS_ERR(sbi->node_inode)) {
1652 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1653 err = PTR_ERR(sbi->node_inode);
1654 goto free_nm;
1657 f2fs_join_shrinker(sbi);
1659 /* if there are nt orphan nodes free them */
1660 err = recover_orphan_inodes(sbi);
1661 if (err)
1662 goto free_node_inode;
1664 /* read root inode and dentry */
1665 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1666 if (IS_ERR(root)) {
1667 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1668 err = PTR_ERR(root);
1669 goto free_node_inode;
1671 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1672 iput(root);
1673 err = -EINVAL;
1674 goto free_node_inode;
1677 sb->s_root = d_make_root(root); /* allocate root dentry */
1678 if (!sb->s_root) {
1679 err = -ENOMEM;
1680 goto free_root_inode;
1683 err = f2fs_build_stats(sbi);
1684 if (err)
1685 goto free_root_inode;
1687 if (f2fs_proc_root)
1688 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1690 if (sbi->s_proc) {
1691 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1692 &f2fs_seq_segment_info_fops, sb);
1693 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc,
1694 &f2fs_seq_segment_bits_fops, sb);
1697 sbi->s_kobj.kset = f2fs_kset;
1698 init_completion(&sbi->s_kobj_unregister);
1699 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1700 "%s", sb->s_id);
1701 if (err)
1702 goto free_proc;
1704 /* recover fsynced data */
1705 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1707 * mount should be failed, when device has readonly mode, and
1708 * previous checkpoint was not done by clean system shutdown.
1710 if (bdev_read_only(sb->s_bdev) &&
1711 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1712 err = -EROFS;
1713 goto free_kobj;
1716 if (need_fsck)
1717 set_sbi_flag(sbi, SBI_NEED_FSCK);
1719 err = recover_fsync_data(sbi, false);
1720 if (err < 0) {
1721 need_fsck = true;
1722 f2fs_msg(sb, KERN_ERR,
1723 "Cannot recover all fsync data errno=%d", err);
1724 goto free_kobj;
1726 } else {
1727 err = recover_fsync_data(sbi, true);
1729 if (!f2fs_readonly(sb) && err > 0) {
1730 err = -EINVAL;
1731 f2fs_msg(sb, KERN_ERR,
1732 "Need to recover fsync data");
1733 goto free_kobj;
1737 /* recover_fsync_data() cleared this already */
1738 clear_sbi_flag(sbi, SBI_POR_DOING);
1741 * If filesystem is not mounted as read-only then
1742 * do start the gc_thread.
1744 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1745 /* After POR, we can run background GC thread.*/
1746 err = start_gc_thread(sbi);
1747 if (err)
1748 goto free_kobj;
1750 kfree(options);
1752 /* recover broken superblock */
1753 if (recovery) {
1754 err = f2fs_commit_super(sbi, true);
1755 f2fs_msg(sb, KERN_INFO,
1756 "Try to recover %dth superblock, ret: %d",
1757 sbi->valid_super_block ? 1 : 2, err);
1760 f2fs_update_time(sbi, CP_TIME);
1761 f2fs_update_time(sbi, REQ_TIME);
1762 return 0;
1764 free_kobj:
1765 kobject_del(&sbi->s_kobj);
1766 kobject_put(&sbi->s_kobj);
1767 wait_for_completion(&sbi->s_kobj_unregister);
1768 free_proc:
1769 if (sbi->s_proc) {
1770 remove_proc_entry("segment_info", sbi->s_proc);
1771 remove_proc_entry("segment_bits", sbi->s_proc);
1772 remove_proc_entry(sb->s_id, f2fs_proc_root);
1774 f2fs_destroy_stats(sbi);
1775 free_root_inode:
1776 dput(sb->s_root);
1777 sb->s_root = NULL;
1778 free_node_inode:
1779 mutex_lock(&sbi->umount_mutex);
1780 f2fs_leave_shrinker(sbi);
1781 iput(sbi->node_inode);
1782 mutex_unlock(&sbi->umount_mutex);
1783 free_nm:
1784 destroy_node_manager(sbi);
1785 free_sm:
1786 destroy_segment_manager(sbi);
1787 kfree(sbi->ckpt);
1788 free_meta_inode:
1789 make_bad_inode(sbi->meta_inode);
1790 iput(sbi->meta_inode);
1791 free_options:
1792 destroy_percpu_info(sbi);
1793 kfree(options);
1794 free_sb_buf:
1795 kfree(raw_super);
1796 free_sbi:
1797 if (sbi->s_chksum_driver)
1798 crypto_free_shash(sbi->s_chksum_driver);
1799 kfree(sbi);
1801 /* give only one another chance */
1802 if (retry) {
1803 retry = false;
1804 shrink_dcache_sb(sb);
1805 goto try_onemore;
1807 return err;
1810 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1811 const char *dev_name, void *data)
1813 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1816 static void kill_f2fs_super(struct super_block *sb)
1818 if (sb->s_root)
1819 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1820 kill_block_super(sb);
1823 static struct file_system_type f2fs_fs_type = {
1824 .owner = THIS_MODULE,
1825 .name = "f2fs",
1826 .mount = f2fs_mount,
1827 .kill_sb = kill_f2fs_super,
1828 .fs_flags = FS_REQUIRES_DEV,
1830 MODULE_ALIAS_FS("f2fs");
1832 static int __init init_inodecache(void)
1834 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
1835 sizeof(struct f2fs_inode_info), 0,
1836 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
1837 if (!f2fs_inode_cachep)
1838 return -ENOMEM;
1839 return 0;
1842 static void destroy_inodecache(void)
1845 * Make sure all delayed rcu free inodes are flushed before we
1846 * destroy cache.
1848 rcu_barrier();
1849 kmem_cache_destroy(f2fs_inode_cachep);
1852 static int __init init_f2fs_fs(void)
1854 int err;
1856 f2fs_build_trace_ios();
1858 err = init_inodecache();
1859 if (err)
1860 goto fail;
1861 err = create_node_manager_caches();
1862 if (err)
1863 goto free_inodecache;
1864 err = create_segment_manager_caches();
1865 if (err)
1866 goto free_node_manager_caches;
1867 err = create_checkpoint_caches();
1868 if (err)
1869 goto free_segment_manager_caches;
1870 err = create_extent_cache();
1871 if (err)
1872 goto free_checkpoint_caches;
1873 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1874 if (!f2fs_kset) {
1875 err = -ENOMEM;
1876 goto free_extent_cache;
1878 #ifdef CONFIG_F2FS_FAULT_INJECTION
1879 f2fs_fault_inject.kset = f2fs_kset;
1880 f2fs_build_fault_attr(0);
1881 err = kobject_init_and_add(&f2fs_fault_inject, &f2fs_fault_ktype,
1882 NULL, "fault_injection");
1883 if (err) {
1884 f2fs_fault_inject.kset = NULL;
1885 goto free_kset;
1887 #endif
1888 err = register_shrinker(&f2fs_shrinker_info);
1889 if (err)
1890 goto free_kset;
1892 err = register_filesystem(&f2fs_fs_type);
1893 if (err)
1894 goto free_shrinker;
1895 err = f2fs_create_root_stats();
1896 if (err)
1897 goto free_filesystem;
1898 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1899 return 0;
1901 free_filesystem:
1902 unregister_filesystem(&f2fs_fs_type);
1903 free_shrinker:
1904 unregister_shrinker(&f2fs_shrinker_info);
1905 free_kset:
1906 #ifdef CONFIG_F2FS_FAULT_INJECTION
1907 if (f2fs_fault_inject.kset)
1908 kobject_put(&f2fs_fault_inject);
1909 #endif
1910 kset_unregister(f2fs_kset);
1911 free_extent_cache:
1912 destroy_extent_cache();
1913 free_checkpoint_caches:
1914 destroy_checkpoint_caches();
1915 free_segment_manager_caches:
1916 destroy_segment_manager_caches();
1917 free_node_manager_caches:
1918 destroy_node_manager_caches();
1919 free_inodecache:
1920 destroy_inodecache();
1921 fail:
1922 return err;
1925 static void __exit exit_f2fs_fs(void)
1927 remove_proc_entry("fs/f2fs", NULL);
1928 f2fs_destroy_root_stats();
1929 unregister_filesystem(&f2fs_fs_type);
1930 unregister_shrinker(&f2fs_shrinker_info);
1931 #ifdef CONFIG_F2FS_FAULT_INJECTION
1932 kobject_put(&f2fs_fault_inject);
1933 #endif
1934 kset_unregister(f2fs_kset);
1935 destroy_extent_cache();
1936 destroy_checkpoint_caches();
1937 destroy_segment_manager_caches();
1938 destroy_node_manager_caches();
1939 destroy_inodecache();
1940 f2fs_destroy_trace_ios();
1943 module_init(init_f2fs_fs)
1944 module_exit(exit_f2fs_fs)
1946 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1947 MODULE_DESCRIPTION("Flash Friendly File System");
1948 MODULE_LICENSE("GPL");