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>
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>
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
)
57 atomic_set(&f2fs_fault
.inject_ops
, 0);
58 f2fs_fault
.inject_rate
= rate
;
59 f2fs_fault
.inject_type
= (1 << FAULT_MAX
) - 1;
61 memset(&f2fs_fault
, 0, sizeof(struct f2fs_fault_info
));
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
,
75 Opt_disable_roll_forward
,
84 Opt_disable_ext_identify
,
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"},
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"},
125 /* Sysfs support for f2fs */
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 */
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);
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
;
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
;
183 ptr
= __struct_ptr(sbi
, a
->struct_type
);
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
)
201 ptr
= __struct_ptr(sbi
, a
->struct_type
);
205 ui
= (unsigned int *)(ptr
+ a
->offset
);
207 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
210 #ifdef CONFIG_F2FS_FAULT_INJECTION
211 if (a
->struct_type
== FAULT_INFO_TYPE
&& t
>= (1 << FAULT_MAX
))
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
,
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
,
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
,
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 }, \
250 .struct_type = _struct_type, \
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
);
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
),
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
),
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
),
329 static struct kobj_type f2fs_fault_ktype
= {
330 .default_attrs
= f2fs_fault_attrs
,
331 .sysfs_ops
= &f2fs_attr_ops
,
335 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
337 struct va_format vaf
;
343 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
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
];
362 #ifdef CONFIG_F2FS_FAULT_INJECTION
363 f2fs_build_fault_attr(0);
369 while ((p
= strsep(&options
, ",")) != NULL
) {
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
);
381 case Opt_gc_background
:
382 name
= match_strdup(&args
[0]);
386 if (strlen(name
) == 2 && !strncmp(name
, "on", 2)) {
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)) {
394 set_opt(sbi
, FORCE_FG_GC
);
401 case Opt_disable_roll_forward
:
402 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
405 /* this option mounts f2fs with ro */
406 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
407 if (!f2fs_readonly(sb
))
411 q
= bdev_get_queue(sb
->s_bdev
);
412 if (blk_queue_discard(q
)) {
413 set_opt(sbi
, DISCARD
);
415 f2fs_msg(sb
, KERN_WARNING
,
416 "mounting with \"discard\" option, but "
417 "the device does not support discard");
421 set_opt(sbi
, NOHEAP
);
423 #ifdef CONFIG_F2FS_FS_XATTR
425 set_opt(sbi
, XATTR_USER
);
427 case Opt_nouser_xattr
:
428 clear_opt(sbi
, XATTR_USER
);
430 case Opt_inline_xattr
:
431 set_opt(sbi
, INLINE_XATTR
);
435 f2fs_msg(sb
, KERN_INFO
,
436 "user_xattr options not supported");
438 case Opt_nouser_xattr
:
439 f2fs_msg(sb
, KERN_INFO
,
440 "nouser_xattr options not supported");
442 case Opt_inline_xattr
:
443 f2fs_msg(sb
, KERN_INFO
,
444 "inline_xattr options not supported");
447 #ifdef CONFIG_F2FS_FS_POSIX_ACL
449 set_opt(sbi
, POSIX_ACL
);
452 clear_opt(sbi
, POSIX_ACL
);
456 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
459 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
462 case Opt_active_logs
:
463 if (args
->from
&& match_int(args
, &arg
))
465 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
467 sbi
->active_logs
= arg
;
469 case Opt_disable_ext_identify
:
470 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
472 case Opt_inline_data
:
473 set_opt(sbi
, INLINE_DATA
);
475 case Opt_inline_dentry
:
476 set_opt(sbi
, INLINE_DENTRY
);
478 case Opt_flush_merge
:
479 set_opt(sbi
, FLUSH_MERGE
);
482 set_opt(sbi
, NOBARRIER
);
485 set_opt(sbi
, FASTBOOT
);
487 case Opt_extent_cache
:
488 set_opt(sbi
, EXTENT_CACHE
);
490 case Opt_noextent_cache
:
491 clear_opt(sbi
, EXTENT_CACHE
);
493 case Opt_noinline_data
:
494 clear_opt(sbi
, INLINE_DATA
);
497 set_opt(sbi
, DATA_FLUSH
);
499 case Opt_fault_injection
:
500 if (args
->from
&& match_int(args
, &arg
))
502 #ifdef CONFIG_F2FS_FAULT_INJECTION
503 f2fs_build_fault_attr(arg
);
505 f2fs_msg(sb
, KERN_INFO
,
506 "FAULT_INJECTION was not selected");
510 f2fs_msg(sb
, KERN_ERR
,
511 "Unrecognized mount option \"%s\" or missing value",
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
);
527 init_once((void *) fi
);
529 if (percpu_counter_init(&fi
->dirty_pages
, 0, GFP_NOFS
)) {
530 kmem_cache_free(f2fs_inode_cachep
, fi
);
534 /* Initialize f2fs-specific inode info */
535 fi
->vfs_inode
.i_version
= 1;
536 fi
->i_current_depth
= 1;
538 init_rwsem(&fi
->i_sem
);
539 INIT_LIST_HEAD(&fi
->dirty_list
);
540 INIT_LIST_HEAD(&fi
->gdirty_list
);
541 INIT_LIST_HEAD(&fi
->inmem_pages
);
542 mutex_init(&fi
->inmem_lock
);
544 /* Will be used by directory only */
545 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
546 return &fi
->vfs_inode
;
549 static int f2fs_drop_inode(struct inode
*inode
)
554 * This is to avoid a deadlock condition like below.
555 * writeback_single_inode(inode)
556 * - f2fs_write_data_page
557 * - f2fs_gc -> iput -> evict
558 * - inode_wait_for_writeback(inode)
560 if ((!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
)) {
561 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
562 /* to avoid evict_inode call simultaneously */
563 atomic_inc(&inode
->i_count
);
564 spin_unlock(&inode
->i_lock
);
566 /* some remained atomic pages should discarded */
567 if (f2fs_is_atomic_file(inode
))
568 drop_inmem_pages(inode
);
570 /* should remain fi->extent_tree for writepage */
571 f2fs_destroy_extent_node(inode
);
573 sb_start_intwrite(inode
->i_sb
);
574 f2fs_i_size_write(inode
, 0);
576 if (F2FS_HAS_BLOCKS(inode
))
577 f2fs_truncate(inode
, true);
579 sb_end_intwrite(inode
->i_sb
);
581 fscrypt_put_encryption_info(inode
, NULL
);
582 spin_lock(&inode
->i_lock
);
583 atomic_dec(&inode
->i_count
);
588 ret
= generic_drop_inode(inode
);
589 if (is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
591 inode
->i_state
|= I_WILL_FREE
;
592 spin_unlock(&inode
->i_lock
);
594 update_inode_page(inode
);
596 spin_lock(&inode
->i_lock
);
598 inode
->i_state
&= ~I_WILL_FREE
;
604 * f2fs_dirty_inode() is called from __mark_inode_dirty()
606 * We should call set_dirty_inode to write the dirty inode through write_inode.
608 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
610 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
612 if (inode
->i_ino
== F2FS_NODE_INO(sbi
) ||
613 inode
->i_ino
== F2FS_META_INO(sbi
))
616 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
617 if (is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
618 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
622 set_inode_flag(inode
, FI_DIRTY_INODE
);
623 list_add_tail(&F2FS_I(inode
)->gdirty_list
,
624 &sbi
->inode_list
[DIRTY_META
]);
625 inc_page_count(sbi
, F2FS_DIRTY_IMETA
);
626 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
627 stat_inc_dirty_inode(sbi
, DIRTY_META
);
630 void f2fs_inode_synced(struct inode
*inode
)
632 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
634 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
635 if (!is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
636 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
639 list_del_init(&F2FS_I(inode
)->gdirty_list
);
640 clear_inode_flag(inode
, FI_DIRTY_INODE
);
641 dec_page_count(sbi
, F2FS_DIRTY_IMETA
);
642 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
643 stat_dec_dirty_inode(F2FS_I_SB(inode
), DIRTY_META
);
646 static void f2fs_i_callback(struct rcu_head
*head
)
648 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
649 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
652 static void f2fs_destroy_inode(struct inode
*inode
)
654 percpu_counter_destroy(&F2FS_I(inode
)->dirty_pages
);
655 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
658 static void destroy_percpu_info(struct f2fs_sb_info
*sbi
)
662 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
663 percpu_counter_destroy(&sbi
->nr_pages
[i
]);
664 percpu_counter_destroy(&sbi
->alloc_valid_block_count
);
665 percpu_counter_destroy(&sbi
->total_valid_inode_count
);
668 static void f2fs_put_super(struct super_block
*sb
)
670 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
673 remove_proc_entry("segment_info", sbi
->s_proc
);
674 remove_proc_entry("segment_bits", sbi
->s_proc
);
675 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
677 kobject_del(&sbi
->s_kobj
);
681 /* prevent remaining shrinker jobs */
682 mutex_lock(&sbi
->umount_mutex
);
685 * We don't need to do checkpoint when superblock is clean.
686 * But, the previous checkpoint was not done by umount, it needs to do
687 * clean checkpoint again.
689 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
690 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
691 struct cp_control cpc
= {
694 write_checkpoint(sbi
, &cpc
);
697 /* write_checkpoint can update stat informaion */
698 f2fs_destroy_stats(sbi
);
701 * normally superblock is clean, so we need to release this.
702 * In addition, EIO will skip do checkpoint, we need this as well.
704 release_ino_entry(sbi
, true);
705 release_discard_addrs(sbi
);
707 f2fs_leave_shrinker(sbi
);
708 mutex_unlock(&sbi
->umount_mutex
);
710 /* our cp_error case, we can wait for any writeback page */
711 f2fs_flush_merged_bios(sbi
);
713 iput(sbi
->node_inode
);
714 iput(sbi
->meta_inode
);
716 /* destroy f2fs internal modules */
717 destroy_node_manager(sbi
);
718 destroy_segment_manager(sbi
);
721 kobject_put(&sbi
->s_kobj
);
722 wait_for_completion(&sbi
->s_kobj_unregister
);
724 sb
->s_fs_info
= NULL
;
725 if (sbi
->s_chksum_driver
)
726 crypto_free_shash(sbi
->s_chksum_driver
);
727 kfree(sbi
->raw_super
);
729 destroy_percpu_info(sbi
);
733 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
735 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
738 trace_f2fs_sync_fs(sb
, sync
);
741 struct cp_control cpc
;
743 cpc
.reason
= __get_cp_reason(sbi
);
745 mutex_lock(&sbi
->gc_mutex
);
746 err
= write_checkpoint(sbi
, &cpc
);
747 mutex_unlock(&sbi
->gc_mutex
);
749 f2fs_trace_ios(NULL
, 1);
754 static int f2fs_freeze(struct super_block
*sb
)
758 if (f2fs_readonly(sb
))
761 err
= f2fs_sync_fs(sb
, 1);
765 static int f2fs_unfreeze(struct super_block
*sb
)
770 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
772 struct super_block
*sb
= dentry
->d_sb
;
773 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
774 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
775 block_t total_count
, user_block_count
, start_count
, ovp_count
;
777 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
778 user_block_count
= sbi
->user_block_count
;
779 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
780 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
781 buf
->f_type
= F2FS_SUPER_MAGIC
;
782 buf
->f_bsize
= sbi
->blocksize
;
784 buf
->f_blocks
= total_count
- start_count
;
785 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
786 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
788 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
789 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
791 buf
->f_namelen
= F2FS_NAME_LEN
;
792 buf
->f_fsid
.val
[0] = (u32
)id
;
793 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
798 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
800 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
802 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
)) {
803 if (test_opt(sbi
, FORCE_FG_GC
))
804 seq_printf(seq
, ",background_gc=%s", "sync");
806 seq_printf(seq
, ",background_gc=%s", "on");
808 seq_printf(seq
, ",background_gc=%s", "off");
810 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
811 seq_puts(seq
, ",disable_roll_forward");
812 if (test_opt(sbi
, DISCARD
))
813 seq_puts(seq
, ",discard");
814 if (test_opt(sbi
, NOHEAP
))
815 seq_puts(seq
, ",no_heap_alloc");
816 #ifdef CONFIG_F2FS_FS_XATTR
817 if (test_opt(sbi
, XATTR_USER
))
818 seq_puts(seq
, ",user_xattr");
820 seq_puts(seq
, ",nouser_xattr");
821 if (test_opt(sbi
, INLINE_XATTR
))
822 seq_puts(seq
, ",inline_xattr");
824 #ifdef CONFIG_F2FS_FS_POSIX_ACL
825 if (test_opt(sbi
, POSIX_ACL
))
826 seq_puts(seq
, ",acl");
828 seq_puts(seq
, ",noacl");
830 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
831 seq_puts(seq
, ",disable_ext_identify");
832 if (test_opt(sbi
, INLINE_DATA
))
833 seq_puts(seq
, ",inline_data");
835 seq_puts(seq
, ",noinline_data");
836 if (test_opt(sbi
, INLINE_DENTRY
))
837 seq_puts(seq
, ",inline_dentry");
838 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
839 seq_puts(seq
, ",flush_merge");
840 if (test_opt(sbi
, NOBARRIER
))
841 seq_puts(seq
, ",nobarrier");
842 if (test_opt(sbi
, FASTBOOT
))
843 seq_puts(seq
, ",fastboot");
844 if (test_opt(sbi
, EXTENT_CACHE
))
845 seq_puts(seq
, ",extent_cache");
847 seq_puts(seq
, ",noextent_cache");
848 if (test_opt(sbi
, DATA_FLUSH
))
849 seq_puts(seq
, ",data_flush");
850 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
855 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
857 struct super_block
*sb
= seq
->private;
858 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
859 unsigned int total_segs
=
860 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
863 seq_puts(seq
, "format: segment_type|valid_blocks\n"
864 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
866 for (i
= 0; i
< total_segs
; i
++) {
867 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
870 seq_printf(seq
, "%-10d", i
);
871 seq_printf(seq
, "%d|%-3u", se
->type
,
872 get_valid_blocks(sbi
, i
, 1));
873 if ((i
% 10) == 9 || i
== (total_segs
- 1))
882 static int segment_bits_seq_show(struct seq_file
*seq
, void *offset
)
884 struct super_block
*sb
= seq
->private;
885 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
886 unsigned int total_segs
=
887 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
890 seq_puts(seq
, "format: segment_type|valid_blocks|bitmaps\n"
891 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
893 for (i
= 0; i
< total_segs
; i
++) {
894 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
896 seq_printf(seq
, "%-10d", i
);
897 seq_printf(seq
, "%d|%-3u|", se
->type
,
898 get_valid_blocks(sbi
, i
, 1));
899 for (j
= 0; j
< SIT_VBLOCK_MAP_SIZE
; j
++)
900 seq_printf(seq
, "%x ", se
->cur_valid_map
[j
]);
906 #define F2FS_PROC_FILE_DEF(_name) \
907 static int _name##_open_fs(struct inode *inode, struct file *file) \
909 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
912 static const struct file_operations f2fs_seq_##_name##_fops = { \
913 .owner = THIS_MODULE, \
914 .open = _name##_open_fs, \
916 .llseek = seq_lseek, \
917 .release = single_release, \
920 F2FS_PROC_FILE_DEF(segment_info
);
921 F2FS_PROC_FILE_DEF(segment_bits
);
923 static void default_options(struct f2fs_sb_info
*sbi
)
925 /* init some FS parameters */
926 sbi
->active_logs
= NR_CURSEG_TYPE
;
929 set_opt(sbi
, INLINE_DATA
);
930 set_opt(sbi
, EXTENT_CACHE
);
932 #ifdef CONFIG_F2FS_FS_XATTR
933 set_opt(sbi
, XATTR_USER
);
935 #ifdef CONFIG_F2FS_FS_POSIX_ACL
936 set_opt(sbi
, POSIX_ACL
);
940 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
942 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
943 struct f2fs_mount_info org_mount_opt
;
944 int err
, active_logs
;
945 bool need_restart_gc
= false;
946 bool need_stop_gc
= false;
947 bool no_extent_cache
= !test_opt(sbi
, EXTENT_CACHE
);
950 * Save the old mount options in case we
951 * need to restore them.
953 org_mount_opt
= sbi
->mount_opt
;
954 active_logs
= sbi
->active_logs
;
956 /* recover superblocks we couldn't write due to previous RO mount */
957 if (!(*flags
& MS_RDONLY
) && is_sbi_flag_set(sbi
, SBI_NEED_SB_WRITE
)) {
958 err
= f2fs_commit_super(sbi
, false);
959 f2fs_msg(sb
, KERN_INFO
,
960 "Try to recover all the superblocks, ret: %d", err
);
962 clear_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
965 sbi
->mount_opt
.opt
= 0;
966 default_options(sbi
);
968 /* parse mount options */
969 err
= parse_options(sb
, data
);
974 * Previous and new state of filesystem is RO,
975 * so skip checking GC and FLUSH_MERGE conditions.
977 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
980 /* disallow enable/disable extent_cache dynamically */
981 if (no_extent_cache
== !!test_opt(sbi
, EXTENT_CACHE
)) {
983 f2fs_msg(sbi
->sb
, KERN_WARNING
,
984 "switch extent_cache option is not allowed");
989 * We stop the GC thread if FS is mounted as RO
990 * or if background_gc = off is passed in mount
991 * option. Also sync the filesystem.
993 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
994 if (sbi
->gc_thread
) {
996 need_restart_gc
= true;
998 } else if (!sbi
->gc_thread
) {
999 err
= start_gc_thread(sbi
);
1002 need_stop_gc
= true;
1005 if (*flags
& MS_RDONLY
) {
1006 writeback_inodes_sb(sb
, WB_REASON_SYNC
);
1009 set_sbi_flag(sbi
, SBI_IS_DIRTY
);
1010 set_sbi_flag(sbi
, SBI_IS_CLOSE
);
1011 f2fs_sync_fs(sb
, 1);
1012 clear_sbi_flag(sbi
, SBI_IS_CLOSE
);
1016 * We stop issue flush thread if FS is mounted as RO
1017 * or if flush_merge is not passed in mount option.
1019 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
1020 destroy_flush_cmd_control(sbi
);
1021 } else if (!SM_I(sbi
)->cmd_control_info
) {
1022 err
= create_flush_cmd_control(sbi
);
1027 /* Update the POSIXACL Flag */
1028 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1029 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1033 if (need_restart_gc
) {
1034 if (start_gc_thread(sbi
))
1035 f2fs_msg(sbi
->sb
, KERN_WARNING
,
1036 "background gc thread has stopped");
1037 } else if (need_stop_gc
) {
1038 stop_gc_thread(sbi
);
1041 sbi
->mount_opt
= org_mount_opt
;
1042 sbi
->active_logs
= active_logs
;
1046 static struct super_operations f2fs_sops
= {
1047 .alloc_inode
= f2fs_alloc_inode
,
1048 .drop_inode
= f2fs_drop_inode
,
1049 .destroy_inode
= f2fs_destroy_inode
,
1050 .write_inode
= f2fs_write_inode
,
1051 .dirty_inode
= f2fs_dirty_inode
,
1052 .show_options
= f2fs_show_options
,
1053 .evict_inode
= f2fs_evict_inode
,
1054 .put_super
= f2fs_put_super
,
1055 .sync_fs
= f2fs_sync_fs
,
1056 .freeze_fs
= f2fs_freeze
,
1057 .unfreeze_fs
= f2fs_unfreeze
,
1058 .statfs
= f2fs_statfs
,
1059 .remount_fs
= f2fs_remount
,
1062 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1063 static int f2fs_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1065 return f2fs_getxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1066 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1070 static int f2fs_key_prefix(struct inode
*inode
, u8
**key
)
1072 *key
= F2FS_I_SB(inode
)->key_prefix
;
1073 return F2FS_I_SB(inode
)->key_prefix_size
;
1076 static int f2fs_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1079 return f2fs_setxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1080 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1081 ctx
, len
, fs_data
, XATTR_CREATE
);
1084 static unsigned f2fs_max_namelen(struct inode
*inode
)
1086 return S_ISLNK(inode
->i_mode
) ?
1087 inode
->i_sb
->s_blocksize
: F2FS_NAME_LEN
;
1090 static struct fscrypt_operations f2fs_cryptops
= {
1091 .get_context
= f2fs_get_context
,
1092 .key_prefix
= f2fs_key_prefix
,
1093 .set_context
= f2fs_set_context
,
1094 .is_encrypted
= f2fs_encrypted_inode
,
1095 .empty_dir
= f2fs_empty_dir
,
1096 .max_namelen
= f2fs_max_namelen
,
1099 static struct fscrypt_operations f2fs_cryptops
= {
1100 .is_encrypted
= f2fs_encrypted_inode
,
1104 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
1105 u64 ino
, u32 generation
)
1107 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
1108 struct inode
*inode
;
1110 if (check_nid_range(sbi
, ino
))
1111 return ERR_PTR(-ESTALE
);
1114 * f2fs_iget isn't quite right if the inode is currently unallocated!
1115 * However f2fs_iget currently does appropriate checks to handle stale
1116 * inodes so everything is OK.
1118 inode
= f2fs_iget(sb
, ino
);
1120 return ERR_CAST(inode
);
1121 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
1122 /* we didn't find the right inode.. */
1124 return ERR_PTR(-ESTALE
);
1129 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1130 int fh_len
, int fh_type
)
1132 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1133 f2fs_nfs_get_inode
);
1136 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1137 int fh_len
, int fh_type
)
1139 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1140 f2fs_nfs_get_inode
);
1143 static const struct export_operations f2fs_export_ops
= {
1144 .fh_to_dentry
= f2fs_fh_to_dentry
,
1145 .fh_to_parent
= f2fs_fh_to_parent
,
1146 .get_parent
= f2fs_get_parent
,
1149 static loff_t
max_file_blocks(void)
1151 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
1152 loff_t leaf_count
= ADDRS_PER_BLOCK
;
1154 /* two direct node blocks */
1155 result
+= (leaf_count
* 2);
1157 /* two indirect node blocks */
1158 leaf_count
*= NIDS_PER_BLOCK
;
1159 result
+= (leaf_count
* 2);
1161 /* one double indirect node block */
1162 leaf_count
*= NIDS_PER_BLOCK
;
1163 result
+= leaf_count
;
1168 static int __f2fs_commit_super(struct buffer_head
*bh
,
1169 struct f2fs_super_block
*super
)
1173 memcpy(bh
->b_data
+ F2FS_SUPER_OFFSET
, super
, sizeof(*super
));
1174 set_buffer_uptodate(bh
);
1175 set_buffer_dirty(bh
);
1178 /* it's rare case, we can do fua all the time */
1179 return __sync_dirty_buffer(bh
, WRITE_FLUSH_FUA
);
1182 static inline bool sanity_check_area_boundary(struct f2fs_sb_info
*sbi
,
1183 struct buffer_head
*bh
)
1185 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1186 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1187 struct super_block
*sb
= sbi
->sb
;
1188 u32 segment0_blkaddr
= le32_to_cpu(raw_super
->segment0_blkaddr
);
1189 u32 cp_blkaddr
= le32_to_cpu(raw_super
->cp_blkaddr
);
1190 u32 sit_blkaddr
= le32_to_cpu(raw_super
->sit_blkaddr
);
1191 u32 nat_blkaddr
= le32_to_cpu(raw_super
->nat_blkaddr
);
1192 u32 ssa_blkaddr
= le32_to_cpu(raw_super
->ssa_blkaddr
);
1193 u32 main_blkaddr
= le32_to_cpu(raw_super
->main_blkaddr
);
1194 u32 segment_count_ckpt
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1195 u32 segment_count_sit
= le32_to_cpu(raw_super
->segment_count_sit
);
1196 u32 segment_count_nat
= le32_to_cpu(raw_super
->segment_count_nat
);
1197 u32 segment_count_ssa
= le32_to_cpu(raw_super
->segment_count_ssa
);
1198 u32 segment_count_main
= le32_to_cpu(raw_super
->segment_count_main
);
1199 u32 segment_count
= le32_to_cpu(raw_super
->segment_count
);
1200 u32 log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1201 u64 main_end_blkaddr
= main_blkaddr
+
1202 (segment_count_main
<< log_blocks_per_seg
);
1203 u64 seg_end_blkaddr
= segment0_blkaddr
+
1204 (segment_count
<< log_blocks_per_seg
);
1206 if (segment0_blkaddr
!= cp_blkaddr
) {
1207 f2fs_msg(sb
, KERN_INFO
,
1208 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1209 segment0_blkaddr
, cp_blkaddr
);
1213 if (cp_blkaddr
+ (segment_count_ckpt
<< log_blocks_per_seg
) !=
1215 f2fs_msg(sb
, KERN_INFO
,
1216 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1217 cp_blkaddr
, sit_blkaddr
,
1218 segment_count_ckpt
<< log_blocks_per_seg
);
1222 if (sit_blkaddr
+ (segment_count_sit
<< log_blocks_per_seg
) !=
1224 f2fs_msg(sb
, KERN_INFO
,
1225 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1226 sit_blkaddr
, nat_blkaddr
,
1227 segment_count_sit
<< log_blocks_per_seg
);
1231 if (nat_blkaddr
+ (segment_count_nat
<< log_blocks_per_seg
) !=
1233 f2fs_msg(sb
, KERN_INFO
,
1234 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1235 nat_blkaddr
, ssa_blkaddr
,
1236 segment_count_nat
<< log_blocks_per_seg
);
1240 if (ssa_blkaddr
+ (segment_count_ssa
<< log_blocks_per_seg
) !=
1242 f2fs_msg(sb
, KERN_INFO
,
1243 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1244 ssa_blkaddr
, main_blkaddr
,
1245 segment_count_ssa
<< log_blocks_per_seg
);
1249 if (main_end_blkaddr
> seg_end_blkaddr
) {
1250 f2fs_msg(sb
, KERN_INFO
,
1251 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1254 (segment_count
<< log_blocks_per_seg
),
1255 segment_count_main
<< log_blocks_per_seg
);
1257 } else if (main_end_blkaddr
< seg_end_blkaddr
) {
1261 /* fix in-memory information all the time */
1262 raw_super
->segment_count
= cpu_to_le32((main_end_blkaddr
-
1263 segment0_blkaddr
) >> log_blocks_per_seg
);
1265 if (f2fs_readonly(sb
) || bdev_read_only(sb
->s_bdev
)) {
1266 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1269 err
= __f2fs_commit_super(bh
, NULL
);
1270 res
= err
? "failed" : "done";
1272 f2fs_msg(sb
, KERN_INFO
,
1273 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1276 (segment_count
<< log_blocks_per_seg
),
1277 segment_count_main
<< log_blocks_per_seg
);
1284 static int sanity_check_raw_super(struct f2fs_sb_info
*sbi
,
1285 struct buffer_head
*bh
)
1287 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1288 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1289 struct super_block
*sb
= sbi
->sb
;
1290 unsigned int blocksize
;
1292 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
1293 f2fs_msg(sb
, KERN_INFO
,
1294 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1295 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
1299 /* Currently, support only 4KB page cache size */
1300 if (F2FS_BLKSIZE
!= PAGE_SIZE
) {
1301 f2fs_msg(sb
, KERN_INFO
,
1302 "Invalid page_cache_size (%lu), supports only 4KB\n",
1307 /* Currently, support only 4KB block size */
1308 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
1309 if (blocksize
!= F2FS_BLKSIZE
) {
1310 f2fs_msg(sb
, KERN_INFO
,
1311 "Invalid blocksize (%u), supports only 4KB\n",
1316 /* check log blocks per segment */
1317 if (le32_to_cpu(raw_super
->log_blocks_per_seg
) != 9) {
1318 f2fs_msg(sb
, KERN_INFO
,
1319 "Invalid log blocks per segment (%u)\n",
1320 le32_to_cpu(raw_super
->log_blocks_per_seg
));
1324 /* Currently, support 512/1024/2048/4096 bytes sector size */
1325 if (le32_to_cpu(raw_super
->log_sectorsize
) >
1326 F2FS_MAX_LOG_SECTOR_SIZE
||
1327 le32_to_cpu(raw_super
->log_sectorsize
) <
1328 F2FS_MIN_LOG_SECTOR_SIZE
) {
1329 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
1330 le32_to_cpu(raw_super
->log_sectorsize
));
1333 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
1334 le32_to_cpu(raw_super
->log_sectorsize
) !=
1335 F2FS_MAX_LOG_SECTOR_SIZE
) {
1336 f2fs_msg(sb
, KERN_INFO
,
1337 "Invalid log sectors per block(%u) log sectorsize(%u)",
1338 le32_to_cpu(raw_super
->log_sectors_per_block
),
1339 le32_to_cpu(raw_super
->log_sectorsize
));
1343 /* check reserved ino info */
1344 if (le32_to_cpu(raw_super
->node_ino
) != 1 ||
1345 le32_to_cpu(raw_super
->meta_ino
) != 2 ||
1346 le32_to_cpu(raw_super
->root_ino
) != 3) {
1347 f2fs_msg(sb
, KERN_INFO
,
1348 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1349 le32_to_cpu(raw_super
->node_ino
),
1350 le32_to_cpu(raw_super
->meta_ino
),
1351 le32_to_cpu(raw_super
->root_ino
));
1355 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1356 if (sanity_check_area_boundary(sbi
, bh
))
1362 int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
1364 unsigned int total
, fsmeta
;
1365 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
1366 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
1368 total
= le32_to_cpu(raw_super
->segment_count
);
1369 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1370 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
1371 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
1372 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
1373 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
1375 if (unlikely(fsmeta
>= total
))
1378 if (unlikely(f2fs_cp_error(sbi
))) {
1379 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
1385 static void init_sb_info(struct f2fs_sb_info
*sbi
)
1387 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
1389 sbi
->log_sectors_per_block
=
1390 le32_to_cpu(raw_super
->log_sectors_per_block
);
1391 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
1392 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
1393 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1394 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
1395 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
1396 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
1397 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
1398 sbi
->total_node_count
=
1399 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
1400 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
1401 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
1402 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
1403 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
1404 sbi
->cur_victim_sec
= NULL_SECNO
;
1405 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
1407 sbi
->dir_level
= DEF_DIR_LEVEL
;
1408 sbi
->interval_time
[CP_TIME
] = DEF_CP_INTERVAL
;
1409 sbi
->interval_time
[REQ_TIME
] = DEF_IDLE_INTERVAL
;
1410 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
1412 INIT_LIST_HEAD(&sbi
->s_list
);
1413 mutex_init(&sbi
->umount_mutex
);
1415 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1416 memcpy(sbi
->key_prefix
, F2FS_KEY_DESC_PREFIX
,
1417 F2FS_KEY_DESC_PREFIX_SIZE
);
1418 sbi
->key_prefix_size
= F2FS_KEY_DESC_PREFIX_SIZE
;
1422 static int init_percpu_info(struct f2fs_sb_info
*sbi
)
1426 for (i
= 0; i
< NR_COUNT_TYPE
; i
++) {
1427 err
= percpu_counter_init(&sbi
->nr_pages
[i
], 0, GFP_KERNEL
);
1432 err
= percpu_counter_init(&sbi
->alloc_valid_block_count
, 0, GFP_KERNEL
);
1436 return percpu_counter_init(&sbi
->total_valid_inode_count
, 0,
1441 * Read f2fs raw super block.
1442 * Because we have two copies of super block, so read both of them
1443 * to get the first valid one. If any one of them is broken, we pass
1444 * them recovery flag back to the caller.
1446 static int read_raw_super_block(struct f2fs_sb_info
*sbi
,
1447 struct f2fs_super_block
**raw_super
,
1448 int *valid_super_block
, int *recovery
)
1450 struct super_block
*sb
= sbi
->sb
;
1452 struct buffer_head
*bh
;
1453 struct f2fs_super_block
*super
;
1456 super
= kzalloc(sizeof(struct f2fs_super_block
), GFP_KERNEL
);
1460 for (block
= 0; block
< 2; block
++) {
1461 bh
= sb_bread(sb
, block
);
1463 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1469 /* sanity checking of raw super */
1470 if (sanity_check_raw_super(sbi
, bh
)) {
1471 f2fs_msg(sb
, KERN_ERR
,
1472 "Can't find valid F2FS filesystem in %dth superblock",
1480 memcpy(super
, bh
->b_data
+ F2FS_SUPER_OFFSET
,
1482 *valid_super_block
= block
;
1488 /* Fail to read any one of the superblocks*/
1492 /* No valid superblock */
1501 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1503 struct buffer_head
*bh
;
1506 if ((recover
&& f2fs_readonly(sbi
->sb
)) ||
1507 bdev_read_only(sbi
->sb
->s_bdev
)) {
1508 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1512 /* write back-up superblock first */
1513 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
? 0: 1);
1516 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1519 /* if we are in recovery path, skip writing valid superblock */
1523 /* write current valid superblock */
1524 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
);
1527 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1532 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1534 struct f2fs_sb_info
*sbi
;
1535 struct f2fs_super_block
*raw_super
;
1538 bool retry
= true, need_fsck
= false;
1539 char *options
= NULL
;
1540 int recovery
, i
, valid_super_block
;
1541 struct curseg_info
*seg_i
;
1546 valid_super_block
= -1;
1549 /* allocate memory for f2fs-specific super block info */
1550 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1556 /* Load the checksum driver */
1557 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32", 0, 0);
1558 if (IS_ERR(sbi
->s_chksum_driver
)) {
1559 f2fs_msg(sb
, KERN_ERR
, "Cannot load crc32 driver.");
1560 err
= PTR_ERR(sbi
->s_chksum_driver
);
1561 sbi
->s_chksum_driver
= NULL
;
1565 /* set a block size */
1566 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1567 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1571 err
= read_raw_super_block(sbi
, &raw_super
, &valid_super_block
,
1576 sb
->s_fs_info
= sbi
;
1577 default_options(sbi
);
1578 /* parse mount options */
1579 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1580 if (data
&& !options
) {
1585 err
= parse_options(sb
, options
);
1589 sbi
->max_file_blocks
= max_file_blocks();
1590 sb
->s_maxbytes
= sbi
->max_file_blocks
<<
1591 le32_to_cpu(raw_super
->log_blocksize
);
1592 sb
->s_max_links
= F2FS_LINK_MAX
;
1593 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1595 sb
->s_op
= &f2fs_sops
;
1596 sb
->s_cop
= &f2fs_cryptops
;
1597 sb
->s_xattr
= f2fs_xattr_handlers
;
1598 sb
->s_export_op
= &f2fs_export_ops
;
1599 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1600 sb
->s_time_gran
= 1;
1601 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1602 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1603 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1605 /* init f2fs-specific super block info */
1606 sbi
->raw_super
= raw_super
;
1607 sbi
->valid_super_block
= valid_super_block
;
1608 mutex_init(&sbi
->gc_mutex
);
1609 mutex_init(&sbi
->writepages
);
1610 mutex_init(&sbi
->cp_mutex
);
1611 init_rwsem(&sbi
->node_write
);
1613 /* disallow all the data/node/meta page writes */
1614 set_sbi_flag(sbi
, SBI_POR_DOING
);
1615 spin_lock_init(&sbi
->stat_lock
);
1617 init_rwsem(&sbi
->read_io
.io_rwsem
);
1618 sbi
->read_io
.sbi
= sbi
;
1619 sbi
->read_io
.bio
= NULL
;
1620 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1621 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1622 sbi
->write_io
[i
].sbi
= sbi
;
1623 sbi
->write_io
[i
].bio
= NULL
;
1626 init_rwsem(&sbi
->cp_rwsem
);
1627 init_waitqueue_head(&sbi
->cp_wait
);
1630 err
= init_percpu_info(sbi
);
1634 /* get an inode for meta space */
1635 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1636 if (IS_ERR(sbi
->meta_inode
)) {
1637 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1638 err
= PTR_ERR(sbi
->meta_inode
);
1642 err
= get_valid_checkpoint(sbi
);
1644 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1645 goto free_meta_inode
;
1648 sbi
->total_valid_node_count
=
1649 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1650 percpu_counter_set(&sbi
->total_valid_inode_count
,
1651 le32_to_cpu(sbi
->ckpt
->valid_inode_count
));
1652 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1653 sbi
->total_valid_block_count
=
1654 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1655 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1657 for (i
= 0; i
< NR_INODE_TYPE
; i
++) {
1658 INIT_LIST_HEAD(&sbi
->inode_list
[i
]);
1659 spin_lock_init(&sbi
->inode_lock
[i
]);
1662 init_extent_cache_info(sbi
);
1664 init_ino_entry_info(sbi
);
1666 /* setup f2fs internal modules */
1667 err
= build_segment_manager(sbi
);
1669 f2fs_msg(sb
, KERN_ERR
,
1670 "Failed to initialize F2FS segment manager");
1673 err
= build_node_manager(sbi
);
1675 f2fs_msg(sb
, KERN_ERR
,
1676 "Failed to initialize F2FS node manager");
1680 /* For write statistics */
1681 if (sb
->s_bdev
->bd_part
)
1682 sbi
->sectors_written_start
=
1683 (u64
)part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
1685 /* Read accumulated write IO statistics if exists */
1686 seg_i
= CURSEG_I(sbi
, CURSEG_HOT_NODE
);
1687 if (__exist_node_summaries(sbi
))
1688 sbi
->kbytes_written
=
1689 le64_to_cpu(seg_i
->journal
->info
.kbytes_written
);
1691 build_gc_manager(sbi
);
1693 /* get an inode for node space */
1694 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1695 if (IS_ERR(sbi
->node_inode
)) {
1696 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1697 err
= PTR_ERR(sbi
->node_inode
);
1701 f2fs_join_shrinker(sbi
);
1703 /* if there are nt orphan nodes free them */
1704 err
= recover_orphan_inodes(sbi
);
1706 goto free_node_inode
;
1708 /* read root inode and dentry */
1709 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1711 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1712 err
= PTR_ERR(root
);
1713 goto free_node_inode
;
1715 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1718 goto free_node_inode
;
1721 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1724 goto free_root_inode
;
1727 err
= f2fs_build_stats(sbi
);
1729 goto free_root_inode
;
1732 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1735 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1736 &f2fs_seq_segment_info_fops
, sb
);
1737 proc_create_data("segment_bits", S_IRUGO
, sbi
->s_proc
,
1738 &f2fs_seq_segment_bits_fops
, sb
);
1741 sbi
->s_kobj
.kset
= f2fs_kset
;
1742 init_completion(&sbi
->s_kobj_unregister
);
1743 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1748 /* recover fsynced data */
1749 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1751 * mount should be failed, when device has readonly mode, and
1752 * previous checkpoint was not done by clean system shutdown.
1754 if (bdev_read_only(sb
->s_bdev
) &&
1755 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1761 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1763 err
= recover_fsync_data(sbi
, false);
1766 f2fs_msg(sb
, KERN_ERR
,
1767 "Cannot recover all fsync data errno=%d", err
);
1771 err
= recover_fsync_data(sbi
, true);
1773 if (!f2fs_readonly(sb
) && err
> 0) {
1775 f2fs_msg(sb
, KERN_ERR
,
1776 "Need to recover fsync data");
1781 /* recover_fsync_data() cleared this already */
1782 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1785 * If filesystem is not mounted as read-only then
1786 * do start the gc_thread.
1788 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1789 /* After POR, we can run background GC thread.*/
1790 err
= start_gc_thread(sbi
);
1796 /* recover broken superblock */
1798 err
= f2fs_commit_super(sbi
, true);
1799 f2fs_msg(sb
, KERN_INFO
,
1800 "Try to recover %dth superblock, ret: %d",
1801 sbi
->valid_super_block
? 1 : 2, err
);
1804 f2fs_update_time(sbi
, CP_TIME
);
1805 f2fs_update_time(sbi
, REQ_TIME
);
1809 f2fs_sync_inode_meta(sbi
);
1810 kobject_del(&sbi
->s_kobj
);
1811 kobject_put(&sbi
->s_kobj
);
1812 wait_for_completion(&sbi
->s_kobj_unregister
);
1815 remove_proc_entry("segment_info", sbi
->s_proc
);
1816 remove_proc_entry("segment_bits", sbi
->s_proc
);
1817 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1819 f2fs_destroy_stats(sbi
);
1824 mutex_lock(&sbi
->umount_mutex
);
1825 f2fs_leave_shrinker(sbi
);
1826 iput(sbi
->node_inode
);
1827 mutex_unlock(&sbi
->umount_mutex
);
1829 destroy_node_manager(sbi
);
1831 destroy_segment_manager(sbi
);
1834 make_bad_inode(sbi
->meta_inode
);
1835 iput(sbi
->meta_inode
);
1837 destroy_percpu_info(sbi
);
1842 if (sbi
->s_chksum_driver
)
1843 crypto_free_shash(sbi
->s_chksum_driver
);
1846 /* give only one another chance */
1849 shrink_dcache_sb(sb
);
1855 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1856 const char *dev_name
, void *data
)
1858 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1861 static void kill_f2fs_super(struct super_block
*sb
)
1864 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1865 kill_block_super(sb
);
1868 static struct file_system_type f2fs_fs_type
= {
1869 .owner
= THIS_MODULE
,
1871 .mount
= f2fs_mount
,
1872 .kill_sb
= kill_f2fs_super
,
1873 .fs_flags
= FS_REQUIRES_DEV
,
1875 MODULE_ALIAS_FS("f2fs");
1877 static int __init
init_inodecache(void)
1879 f2fs_inode_cachep
= kmem_cache_create("f2fs_inode_cache",
1880 sizeof(struct f2fs_inode_info
), 0,
1881 SLAB_RECLAIM_ACCOUNT
|SLAB_ACCOUNT
, NULL
);
1882 if (!f2fs_inode_cachep
)
1887 static void destroy_inodecache(void)
1890 * Make sure all delayed rcu free inodes are flushed before we
1894 kmem_cache_destroy(f2fs_inode_cachep
);
1897 static int __init
init_f2fs_fs(void)
1901 f2fs_build_trace_ios();
1903 err
= init_inodecache();
1906 err
= create_node_manager_caches();
1908 goto free_inodecache
;
1909 err
= create_segment_manager_caches();
1911 goto free_node_manager_caches
;
1912 err
= create_checkpoint_caches();
1914 goto free_segment_manager_caches
;
1915 err
= create_extent_cache();
1917 goto free_checkpoint_caches
;
1918 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1921 goto free_extent_cache
;
1923 #ifdef CONFIG_F2FS_FAULT_INJECTION
1924 f2fs_fault_inject
.kset
= f2fs_kset
;
1925 f2fs_build_fault_attr(0);
1926 err
= kobject_init_and_add(&f2fs_fault_inject
, &f2fs_fault_ktype
,
1927 NULL
, "fault_injection");
1929 f2fs_fault_inject
.kset
= NULL
;
1933 err
= register_shrinker(&f2fs_shrinker_info
);
1937 err
= register_filesystem(&f2fs_fs_type
);
1940 err
= f2fs_create_root_stats();
1942 goto free_filesystem
;
1943 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1947 unregister_filesystem(&f2fs_fs_type
);
1949 unregister_shrinker(&f2fs_shrinker_info
);
1951 #ifdef CONFIG_F2FS_FAULT_INJECTION
1952 if (f2fs_fault_inject
.kset
)
1953 kobject_put(&f2fs_fault_inject
);
1955 kset_unregister(f2fs_kset
);
1957 destroy_extent_cache();
1958 free_checkpoint_caches
:
1959 destroy_checkpoint_caches();
1960 free_segment_manager_caches
:
1961 destroy_segment_manager_caches();
1962 free_node_manager_caches
:
1963 destroy_node_manager_caches();
1965 destroy_inodecache();
1970 static void __exit
exit_f2fs_fs(void)
1972 remove_proc_entry("fs/f2fs", NULL
);
1973 f2fs_destroy_root_stats();
1974 unregister_filesystem(&f2fs_fs_type
);
1975 unregister_shrinker(&f2fs_shrinker_info
);
1976 #ifdef CONFIG_F2FS_FAULT_INJECTION
1977 kobject_put(&f2fs_fault_inject
);
1979 kset_unregister(f2fs_kset
);
1980 destroy_extent_cache();
1981 destroy_checkpoint_caches();
1982 destroy_segment_manager_caches();
1983 destroy_node_manager_caches();
1984 destroy_inodecache();
1985 f2fs_destroy_trace_ios();
1988 module_init(init_f2fs_fs
)
1989 module_exit(exit_f2fs_fs
)
1991 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1992 MODULE_DESCRIPTION("Flash Friendly File System");
1993 MODULE_LICENSE("GPL");