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 if (is_inode_flag_set(inode
, FI_AUTO_RECOVER
))
617 clear_inode_flag(inode
, FI_AUTO_RECOVER
);
619 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
620 if (is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
621 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
625 set_inode_flag(inode
, FI_DIRTY_INODE
);
626 list_add_tail(&F2FS_I(inode
)->gdirty_list
,
627 &sbi
->inode_list
[DIRTY_META
]);
628 inc_page_count(sbi
, F2FS_DIRTY_IMETA
);
629 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
630 stat_inc_dirty_inode(sbi
, DIRTY_META
);
633 void f2fs_inode_synced(struct inode
*inode
)
635 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
637 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
638 if (!is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
639 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
642 list_del_init(&F2FS_I(inode
)->gdirty_list
);
643 clear_inode_flag(inode
, FI_DIRTY_INODE
);
644 clear_inode_flag(inode
, FI_AUTO_RECOVER
);
645 dec_page_count(sbi
, F2FS_DIRTY_IMETA
);
646 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
647 stat_dec_dirty_inode(F2FS_I_SB(inode
), DIRTY_META
);
650 static void f2fs_i_callback(struct rcu_head
*head
)
652 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
653 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
656 static void f2fs_destroy_inode(struct inode
*inode
)
658 percpu_counter_destroy(&F2FS_I(inode
)->dirty_pages
);
659 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
662 static void destroy_percpu_info(struct f2fs_sb_info
*sbi
)
666 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
667 percpu_counter_destroy(&sbi
->nr_pages
[i
]);
668 percpu_counter_destroy(&sbi
->alloc_valid_block_count
);
669 percpu_counter_destroy(&sbi
->total_valid_inode_count
);
672 static void f2fs_put_super(struct super_block
*sb
)
674 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
677 remove_proc_entry("segment_info", sbi
->s_proc
);
678 remove_proc_entry("segment_bits", sbi
->s_proc
);
679 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
681 kobject_del(&sbi
->s_kobj
);
685 /* prevent remaining shrinker jobs */
686 mutex_lock(&sbi
->umount_mutex
);
689 * We don't need to do checkpoint when superblock is clean.
690 * But, the previous checkpoint was not done by umount, it needs to do
691 * clean checkpoint again.
693 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
694 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
695 struct cp_control cpc
= {
698 write_checkpoint(sbi
, &cpc
);
701 /* write_checkpoint can update stat informaion */
702 f2fs_destroy_stats(sbi
);
705 * normally superblock is clean, so we need to release this.
706 * In addition, EIO will skip do checkpoint, we need this as well.
708 release_ino_entry(sbi
, true);
709 release_discard_addrs(sbi
);
711 f2fs_leave_shrinker(sbi
);
712 mutex_unlock(&sbi
->umount_mutex
);
714 /* our cp_error case, we can wait for any writeback page */
715 f2fs_flush_merged_bios(sbi
);
717 iput(sbi
->node_inode
);
718 iput(sbi
->meta_inode
);
720 /* destroy f2fs internal modules */
721 destroy_node_manager(sbi
);
722 destroy_segment_manager(sbi
);
725 kobject_put(&sbi
->s_kobj
);
726 wait_for_completion(&sbi
->s_kobj_unregister
);
728 sb
->s_fs_info
= NULL
;
729 if (sbi
->s_chksum_driver
)
730 crypto_free_shash(sbi
->s_chksum_driver
);
731 kfree(sbi
->raw_super
);
733 destroy_percpu_info(sbi
);
737 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
739 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
742 trace_f2fs_sync_fs(sb
, sync
);
745 struct cp_control cpc
;
747 cpc
.reason
= __get_cp_reason(sbi
);
749 mutex_lock(&sbi
->gc_mutex
);
750 err
= write_checkpoint(sbi
, &cpc
);
751 mutex_unlock(&sbi
->gc_mutex
);
753 f2fs_trace_ios(NULL
, 1);
758 static int f2fs_freeze(struct super_block
*sb
)
762 if (f2fs_readonly(sb
))
765 err
= f2fs_sync_fs(sb
, 1);
769 static int f2fs_unfreeze(struct super_block
*sb
)
774 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
776 struct super_block
*sb
= dentry
->d_sb
;
777 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
778 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
779 block_t total_count
, user_block_count
, start_count
, ovp_count
;
781 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
782 user_block_count
= sbi
->user_block_count
;
783 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
784 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
785 buf
->f_type
= F2FS_SUPER_MAGIC
;
786 buf
->f_bsize
= sbi
->blocksize
;
788 buf
->f_blocks
= total_count
- start_count
;
789 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
790 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
792 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
793 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
795 buf
->f_namelen
= F2FS_NAME_LEN
;
796 buf
->f_fsid
.val
[0] = (u32
)id
;
797 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
802 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
804 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
806 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
)) {
807 if (test_opt(sbi
, FORCE_FG_GC
))
808 seq_printf(seq
, ",background_gc=%s", "sync");
810 seq_printf(seq
, ",background_gc=%s", "on");
812 seq_printf(seq
, ",background_gc=%s", "off");
814 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
815 seq_puts(seq
, ",disable_roll_forward");
816 if (test_opt(sbi
, DISCARD
))
817 seq_puts(seq
, ",discard");
818 if (test_opt(sbi
, NOHEAP
))
819 seq_puts(seq
, ",no_heap_alloc");
820 #ifdef CONFIG_F2FS_FS_XATTR
821 if (test_opt(sbi
, XATTR_USER
))
822 seq_puts(seq
, ",user_xattr");
824 seq_puts(seq
, ",nouser_xattr");
825 if (test_opt(sbi
, INLINE_XATTR
))
826 seq_puts(seq
, ",inline_xattr");
828 #ifdef CONFIG_F2FS_FS_POSIX_ACL
829 if (test_opt(sbi
, POSIX_ACL
))
830 seq_puts(seq
, ",acl");
832 seq_puts(seq
, ",noacl");
834 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
835 seq_puts(seq
, ",disable_ext_identify");
836 if (test_opt(sbi
, INLINE_DATA
))
837 seq_puts(seq
, ",inline_data");
839 seq_puts(seq
, ",noinline_data");
840 if (test_opt(sbi
, INLINE_DENTRY
))
841 seq_puts(seq
, ",inline_dentry");
842 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
843 seq_puts(seq
, ",flush_merge");
844 if (test_opt(sbi
, NOBARRIER
))
845 seq_puts(seq
, ",nobarrier");
846 if (test_opt(sbi
, FASTBOOT
))
847 seq_puts(seq
, ",fastboot");
848 if (test_opt(sbi
, EXTENT_CACHE
))
849 seq_puts(seq
, ",extent_cache");
851 seq_puts(seq
, ",noextent_cache");
852 if (test_opt(sbi
, DATA_FLUSH
))
853 seq_puts(seq
, ",data_flush");
854 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
859 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
861 struct super_block
*sb
= seq
->private;
862 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
863 unsigned int total_segs
=
864 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
867 seq_puts(seq
, "format: segment_type|valid_blocks\n"
868 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
870 for (i
= 0; i
< total_segs
; i
++) {
871 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
874 seq_printf(seq
, "%-10d", i
);
875 seq_printf(seq
, "%d|%-3u", se
->type
,
876 get_valid_blocks(sbi
, i
, 1));
877 if ((i
% 10) == 9 || i
== (total_segs
- 1))
886 static int segment_bits_seq_show(struct seq_file
*seq
, void *offset
)
888 struct super_block
*sb
= seq
->private;
889 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
890 unsigned int total_segs
=
891 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
894 seq_puts(seq
, "format: segment_type|valid_blocks|bitmaps\n"
895 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
897 for (i
= 0; i
< total_segs
; i
++) {
898 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
900 seq_printf(seq
, "%-10d", i
);
901 seq_printf(seq
, "%d|%-3u|", se
->type
,
902 get_valid_blocks(sbi
, i
, 1));
903 for (j
= 0; j
< SIT_VBLOCK_MAP_SIZE
; j
++)
904 seq_printf(seq
, "%x ", se
->cur_valid_map
[j
]);
910 #define F2FS_PROC_FILE_DEF(_name) \
911 static int _name##_open_fs(struct inode *inode, struct file *file) \
913 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
916 static const struct file_operations f2fs_seq_##_name##_fops = { \
917 .owner = THIS_MODULE, \
918 .open = _name##_open_fs, \
920 .llseek = seq_lseek, \
921 .release = single_release, \
924 F2FS_PROC_FILE_DEF(segment_info
);
925 F2FS_PROC_FILE_DEF(segment_bits
);
927 static void default_options(struct f2fs_sb_info
*sbi
)
929 /* init some FS parameters */
930 sbi
->active_logs
= NR_CURSEG_TYPE
;
933 set_opt(sbi
, INLINE_DATA
);
934 set_opt(sbi
, EXTENT_CACHE
);
936 #ifdef CONFIG_F2FS_FS_XATTR
937 set_opt(sbi
, XATTR_USER
);
939 #ifdef CONFIG_F2FS_FS_POSIX_ACL
940 set_opt(sbi
, POSIX_ACL
);
944 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
946 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
947 struct f2fs_mount_info org_mount_opt
;
948 int err
, active_logs
;
949 bool need_restart_gc
= false;
950 bool need_stop_gc
= false;
951 bool no_extent_cache
= !test_opt(sbi
, EXTENT_CACHE
);
954 * Save the old mount options in case we
955 * need to restore them.
957 org_mount_opt
= sbi
->mount_opt
;
958 active_logs
= sbi
->active_logs
;
960 /* recover superblocks we couldn't write due to previous RO mount */
961 if (!(*flags
& MS_RDONLY
) && is_sbi_flag_set(sbi
, SBI_NEED_SB_WRITE
)) {
962 err
= f2fs_commit_super(sbi
, false);
963 f2fs_msg(sb
, KERN_INFO
,
964 "Try to recover all the superblocks, ret: %d", err
);
966 clear_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
969 sbi
->mount_opt
.opt
= 0;
970 default_options(sbi
);
972 /* parse mount options */
973 err
= parse_options(sb
, data
);
978 * Previous and new state of filesystem is RO,
979 * so skip checking GC and FLUSH_MERGE conditions.
981 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
984 /* disallow enable/disable extent_cache dynamically */
985 if (no_extent_cache
== !!test_opt(sbi
, EXTENT_CACHE
)) {
987 f2fs_msg(sbi
->sb
, KERN_WARNING
,
988 "switch extent_cache option is not allowed");
993 * We stop the GC thread if FS is mounted as RO
994 * or if background_gc = off is passed in mount
995 * option. Also sync the filesystem.
997 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
998 if (sbi
->gc_thread
) {
1000 need_restart_gc
= true;
1002 } else if (!sbi
->gc_thread
) {
1003 err
= start_gc_thread(sbi
);
1006 need_stop_gc
= true;
1009 if (*flags
& MS_RDONLY
) {
1010 writeback_inodes_sb(sb
, WB_REASON_SYNC
);
1013 set_sbi_flag(sbi
, SBI_IS_DIRTY
);
1014 set_sbi_flag(sbi
, SBI_IS_CLOSE
);
1015 f2fs_sync_fs(sb
, 1);
1016 clear_sbi_flag(sbi
, SBI_IS_CLOSE
);
1020 * We stop issue flush thread if FS is mounted as RO
1021 * or if flush_merge is not passed in mount option.
1023 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
1024 destroy_flush_cmd_control(sbi
);
1025 } else if (!SM_I(sbi
)->cmd_control_info
) {
1026 err
= create_flush_cmd_control(sbi
);
1031 /* Update the POSIXACL Flag */
1032 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1033 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1037 if (need_restart_gc
) {
1038 if (start_gc_thread(sbi
))
1039 f2fs_msg(sbi
->sb
, KERN_WARNING
,
1040 "background gc thread has stopped");
1041 } else if (need_stop_gc
) {
1042 stop_gc_thread(sbi
);
1045 sbi
->mount_opt
= org_mount_opt
;
1046 sbi
->active_logs
= active_logs
;
1050 static struct super_operations f2fs_sops
= {
1051 .alloc_inode
= f2fs_alloc_inode
,
1052 .drop_inode
= f2fs_drop_inode
,
1053 .destroy_inode
= f2fs_destroy_inode
,
1054 .write_inode
= f2fs_write_inode
,
1055 .dirty_inode
= f2fs_dirty_inode
,
1056 .show_options
= f2fs_show_options
,
1057 .evict_inode
= f2fs_evict_inode
,
1058 .put_super
= f2fs_put_super
,
1059 .sync_fs
= f2fs_sync_fs
,
1060 .freeze_fs
= f2fs_freeze
,
1061 .unfreeze_fs
= f2fs_unfreeze
,
1062 .statfs
= f2fs_statfs
,
1063 .remount_fs
= f2fs_remount
,
1066 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1067 static int f2fs_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1069 return f2fs_getxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1070 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1074 static int f2fs_key_prefix(struct inode
*inode
, u8
**key
)
1076 *key
= F2FS_I_SB(inode
)->key_prefix
;
1077 return F2FS_I_SB(inode
)->key_prefix_size
;
1080 static int f2fs_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1083 return f2fs_setxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1084 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1085 ctx
, len
, fs_data
, XATTR_CREATE
);
1088 static unsigned f2fs_max_namelen(struct inode
*inode
)
1090 return S_ISLNK(inode
->i_mode
) ?
1091 inode
->i_sb
->s_blocksize
: F2FS_NAME_LEN
;
1094 static struct fscrypt_operations f2fs_cryptops
= {
1095 .get_context
= f2fs_get_context
,
1096 .key_prefix
= f2fs_key_prefix
,
1097 .set_context
= f2fs_set_context
,
1098 .is_encrypted
= f2fs_encrypted_inode
,
1099 .empty_dir
= f2fs_empty_dir
,
1100 .max_namelen
= f2fs_max_namelen
,
1103 static struct fscrypt_operations f2fs_cryptops
= {
1104 .is_encrypted
= f2fs_encrypted_inode
,
1108 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
1109 u64 ino
, u32 generation
)
1111 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
1112 struct inode
*inode
;
1114 if (check_nid_range(sbi
, ino
))
1115 return ERR_PTR(-ESTALE
);
1118 * f2fs_iget isn't quite right if the inode is currently unallocated!
1119 * However f2fs_iget currently does appropriate checks to handle stale
1120 * inodes so everything is OK.
1122 inode
= f2fs_iget(sb
, ino
);
1124 return ERR_CAST(inode
);
1125 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
1126 /* we didn't find the right inode.. */
1128 return ERR_PTR(-ESTALE
);
1133 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1134 int fh_len
, int fh_type
)
1136 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1137 f2fs_nfs_get_inode
);
1140 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1141 int fh_len
, int fh_type
)
1143 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1144 f2fs_nfs_get_inode
);
1147 static const struct export_operations f2fs_export_ops
= {
1148 .fh_to_dentry
= f2fs_fh_to_dentry
,
1149 .fh_to_parent
= f2fs_fh_to_parent
,
1150 .get_parent
= f2fs_get_parent
,
1153 static loff_t
max_file_blocks(void)
1155 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
1156 loff_t leaf_count
= ADDRS_PER_BLOCK
;
1158 /* two direct node blocks */
1159 result
+= (leaf_count
* 2);
1161 /* two indirect node blocks */
1162 leaf_count
*= NIDS_PER_BLOCK
;
1163 result
+= (leaf_count
* 2);
1165 /* one double indirect node block */
1166 leaf_count
*= NIDS_PER_BLOCK
;
1167 result
+= leaf_count
;
1172 static int __f2fs_commit_super(struct buffer_head
*bh
,
1173 struct f2fs_super_block
*super
)
1177 memcpy(bh
->b_data
+ F2FS_SUPER_OFFSET
, super
, sizeof(*super
));
1178 set_buffer_uptodate(bh
);
1179 set_buffer_dirty(bh
);
1182 /* it's rare case, we can do fua all the time */
1183 return __sync_dirty_buffer(bh
, WRITE_FLUSH_FUA
);
1186 static inline bool sanity_check_area_boundary(struct f2fs_sb_info
*sbi
,
1187 struct buffer_head
*bh
)
1189 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1190 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1191 struct super_block
*sb
= sbi
->sb
;
1192 u32 segment0_blkaddr
= le32_to_cpu(raw_super
->segment0_blkaddr
);
1193 u32 cp_blkaddr
= le32_to_cpu(raw_super
->cp_blkaddr
);
1194 u32 sit_blkaddr
= le32_to_cpu(raw_super
->sit_blkaddr
);
1195 u32 nat_blkaddr
= le32_to_cpu(raw_super
->nat_blkaddr
);
1196 u32 ssa_blkaddr
= le32_to_cpu(raw_super
->ssa_blkaddr
);
1197 u32 main_blkaddr
= le32_to_cpu(raw_super
->main_blkaddr
);
1198 u32 segment_count_ckpt
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1199 u32 segment_count_sit
= le32_to_cpu(raw_super
->segment_count_sit
);
1200 u32 segment_count_nat
= le32_to_cpu(raw_super
->segment_count_nat
);
1201 u32 segment_count_ssa
= le32_to_cpu(raw_super
->segment_count_ssa
);
1202 u32 segment_count_main
= le32_to_cpu(raw_super
->segment_count_main
);
1203 u32 segment_count
= le32_to_cpu(raw_super
->segment_count
);
1204 u32 log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1205 u64 main_end_blkaddr
= main_blkaddr
+
1206 (segment_count_main
<< log_blocks_per_seg
);
1207 u64 seg_end_blkaddr
= segment0_blkaddr
+
1208 (segment_count
<< log_blocks_per_seg
);
1210 if (segment0_blkaddr
!= cp_blkaddr
) {
1211 f2fs_msg(sb
, KERN_INFO
,
1212 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1213 segment0_blkaddr
, cp_blkaddr
);
1217 if (cp_blkaddr
+ (segment_count_ckpt
<< log_blocks_per_seg
) !=
1219 f2fs_msg(sb
, KERN_INFO
,
1220 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1221 cp_blkaddr
, sit_blkaddr
,
1222 segment_count_ckpt
<< log_blocks_per_seg
);
1226 if (sit_blkaddr
+ (segment_count_sit
<< log_blocks_per_seg
) !=
1228 f2fs_msg(sb
, KERN_INFO
,
1229 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1230 sit_blkaddr
, nat_blkaddr
,
1231 segment_count_sit
<< log_blocks_per_seg
);
1235 if (nat_blkaddr
+ (segment_count_nat
<< log_blocks_per_seg
) !=
1237 f2fs_msg(sb
, KERN_INFO
,
1238 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1239 nat_blkaddr
, ssa_blkaddr
,
1240 segment_count_nat
<< log_blocks_per_seg
);
1244 if (ssa_blkaddr
+ (segment_count_ssa
<< log_blocks_per_seg
) !=
1246 f2fs_msg(sb
, KERN_INFO
,
1247 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1248 ssa_blkaddr
, main_blkaddr
,
1249 segment_count_ssa
<< log_blocks_per_seg
);
1253 if (main_end_blkaddr
> seg_end_blkaddr
) {
1254 f2fs_msg(sb
, KERN_INFO
,
1255 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1258 (segment_count
<< log_blocks_per_seg
),
1259 segment_count_main
<< log_blocks_per_seg
);
1261 } else if (main_end_blkaddr
< seg_end_blkaddr
) {
1265 /* fix in-memory information all the time */
1266 raw_super
->segment_count
= cpu_to_le32((main_end_blkaddr
-
1267 segment0_blkaddr
) >> log_blocks_per_seg
);
1269 if (f2fs_readonly(sb
) || bdev_read_only(sb
->s_bdev
)) {
1270 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1273 err
= __f2fs_commit_super(bh
, NULL
);
1274 res
= err
? "failed" : "done";
1276 f2fs_msg(sb
, KERN_INFO
,
1277 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1280 (segment_count
<< log_blocks_per_seg
),
1281 segment_count_main
<< log_blocks_per_seg
);
1288 static int sanity_check_raw_super(struct f2fs_sb_info
*sbi
,
1289 struct buffer_head
*bh
)
1291 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1292 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1293 struct super_block
*sb
= sbi
->sb
;
1294 unsigned int blocksize
;
1296 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
1297 f2fs_msg(sb
, KERN_INFO
,
1298 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1299 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
1303 /* Currently, support only 4KB page cache size */
1304 if (F2FS_BLKSIZE
!= PAGE_SIZE
) {
1305 f2fs_msg(sb
, KERN_INFO
,
1306 "Invalid page_cache_size (%lu), supports only 4KB\n",
1311 /* Currently, support only 4KB block size */
1312 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
1313 if (blocksize
!= F2FS_BLKSIZE
) {
1314 f2fs_msg(sb
, KERN_INFO
,
1315 "Invalid blocksize (%u), supports only 4KB\n",
1320 /* check log blocks per segment */
1321 if (le32_to_cpu(raw_super
->log_blocks_per_seg
) != 9) {
1322 f2fs_msg(sb
, KERN_INFO
,
1323 "Invalid log blocks per segment (%u)\n",
1324 le32_to_cpu(raw_super
->log_blocks_per_seg
));
1328 /* Currently, support 512/1024/2048/4096 bytes sector size */
1329 if (le32_to_cpu(raw_super
->log_sectorsize
) >
1330 F2FS_MAX_LOG_SECTOR_SIZE
||
1331 le32_to_cpu(raw_super
->log_sectorsize
) <
1332 F2FS_MIN_LOG_SECTOR_SIZE
) {
1333 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
1334 le32_to_cpu(raw_super
->log_sectorsize
));
1337 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
1338 le32_to_cpu(raw_super
->log_sectorsize
) !=
1339 F2FS_MAX_LOG_SECTOR_SIZE
) {
1340 f2fs_msg(sb
, KERN_INFO
,
1341 "Invalid log sectors per block(%u) log sectorsize(%u)",
1342 le32_to_cpu(raw_super
->log_sectors_per_block
),
1343 le32_to_cpu(raw_super
->log_sectorsize
));
1347 /* check reserved ino info */
1348 if (le32_to_cpu(raw_super
->node_ino
) != 1 ||
1349 le32_to_cpu(raw_super
->meta_ino
) != 2 ||
1350 le32_to_cpu(raw_super
->root_ino
) != 3) {
1351 f2fs_msg(sb
, KERN_INFO
,
1352 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1353 le32_to_cpu(raw_super
->node_ino
),
1354 le32_to_cpu(raw_super
->meta_ino
),
1355 le32_to_cpu(raw_super
->root_ino
));
1359 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1360 if (sanity_check_area_boundary(sbi
, bh
))
1366 int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
1368 unsigned int total
, fsmeta
;
1369 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
1370 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
1372 total
= le32_to_cpu(raw_super
->segment_count
);
1373 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1374 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
1375 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
1376 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
1377 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
1379 if (unlikely(fsmeta
>= total
))
1382 if (unlikely(f2fs_cp_error(sbi
))) {
1383 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
1389 static void init_sb_info(struct f2fs_sb_info
*sbi
)
1391 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
1393 sbi
->log_sectors_per_block
=
1394 le32_to_cpu(raw_super
->log_sectors_per_block
);
1395 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
1396 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
1397 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1398 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
1399 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
1400 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
1401 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
1402 sbi
->total_node_count
=
1403 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
1404 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
1405 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
1406 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
1407 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
1408 sbi
->cur_victim_sec
= NULL_SECNO
;
1409 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
1411 sbi
->dir_level
= DEF_DIR_LEVEL
;
1412 sbi
->interval_time
[CP_TIME
] = DEF_CP_INTERVAL
;
1413 sbi
->interval_time
[REQ_TIME
] = DEF_IDLE_INTERVAL
;
1414 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
1416 INIT_LIST_HEAD(&sbi
->s_list
);
1417 mutex_init(&sbi
->umount_mutex
);
1419 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1420 memcpy(sbi
->key_prefix
, F2FS_KEY_DESC_PREFIX
,
1421 F2FS_KEY_DESC_PREFIX_SIZE
);
1422 sbi
->key_prefix_size
= F2FS_KEY_DESC_PREFIX_SIZE
;
1426 static int init_percpu_info(struct f2fs_sb_info
*sbi
)
1430 for (i
= 0; i
< NR_COUNT_TYPE
; i
++) {
1431 err
= percpu_counter_init(&sbi
->nr_pages
[i
], 0, GFP_KERNEL
);
1436 err
= percpu_counter_init(&sbi
->alloc_valid_block_count
, 0, GFP_KERNEL
);
1440 return percpu_counter_init(&sbi
->total_valid_inode_count
, 0,
1445 * Read f2fs raw super block.
1446 * Because we have two copies of super block, so read both of them
1447 * to get the first valid one. If any one of them is broken, we pass
1448 * them recovery flag back to the caller.
1450 static int read_raw_super_block(struct f2fs_sb_info
*sbi
,
1451 struct f2fs_super_block
**raw_super
,
1452 int *valid_super_block
, int *recovery
)
1454 struct super_block
*sb
= sbi
->sb
;
1456 struct buffer_head
*bh
;
1457 struct f2fs_super_block
*super
;
1460 super
= kzalloc(sizeof(struct f2fs_super_block
), GFP_KERNEL
);
1464 for (block
= 0; block
< 2; block
++) {
1465 bh
= sb_bread(sb
, block
);
1467 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1473 /* sanity checking of raw super */
1474 if (sanity_check_raw_super(sbi
, bh
)) {
1475 f2fs_msg(sb
, KERN_ERR
,
1476 "Can't find valid F2FS filesystem in %dth superblock",
1484 memcpy(super
, bh
->b_data
+ F2FS_SUPER_OFFSET
,
1486 *valid_super_block
= block
;
1492 /* Fail to read any one of the superblocks*/
1496 /* No valid superblock */
1505 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1507 struct buffer_head
*bh
;
1510 if ((recover
&& f2fs_readonly(sbi
->sb
)) ||
1511 bdev_read_only(sbi
->sb
->s_bdev
)) {
1512 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1516 /* write back-up superblock first */
1517 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
? 0: 1);
1520 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1523 /* if we are in recovery path, skip writing valid superblock */
1527 /* write current valid superblock */
1528 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
);
1531 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1536 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1538 struct f2fs_sb_info
*sbi
;
1539 struct f2fs_super_block
*raw_super
;
1542 bool retry
= true, need_fsck
= false;
1543 char *options
= NULL
;
1544 int recovery
, i
, valid_super_block
;
1545 struct curseg_info
*seg_i
;
1550 valid_super_block
= -1;
1553 /* allocate memory for f2fs-specific super block info */
1554 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1560 /* Load the checksum driver */
1561 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32", 0, 0);
1562 if (IS_ERR(sbi
->s_chksum_driver
)) {
1563 f2fs_msg(sb
, KERN_ERR
, "Cannot load crc32 driver.");
1564 err
= PTR_ERR(sbi
->s_chksum_driver
);
1565 sbi
->s_chksum_driver
= NULL
;
1569 /* set a block size */
1570 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1571 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1575 err
= read_raw_super_block(sbi
, &raw_super
, &valid_super_block
,
1580 sb
->s_fs_info
= sbi
;
1581 default_options(sbi
);
1582 /* parse mount options */
1583 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1584 if (data
&& !options
) {
1589 err
= parse_options(sb
, options
);
1593 sbi
->max_file_blocks
= max_file_blocks();
1594 sb
->s_maxbytes
= sbi
->max_file_blocks
<<
1595 le32_to_cpu(raw_super
->log_blocksize
);
1596 sb
->s_max_links
= F2FS_LINK_MAX
;
1597 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1599 sb
->s_op
= &f2fs_sops
;
1600 sb
->s_cop
= &f2fs_cryptops
;
1601 sb
->s_xattr
= f2fs_xattr_handlers
;
1602 sb
->s_export_op
= &f2fs_export_ops
;
1603 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1604 sb
->s_time_gran
= 1;
1605 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1606 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1607 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1609 /* init f2fs-specific super block info */
1610 sbi
->raw_super
= raw_super
;
1611 sbi
->valid_super_block
= valid_super_block
;
1612 mutex_init(&sbi
->gc_mutex
);
1613 mutex_init(&sbi
->writepages
);
1614 mutex_init(&sbi
->cp_mutex
);
1615 init_rwsem(&sbi
->node_write
);
1617 /* disallow all the data/node/meta page writes */
1618 set_sbi_flag(sbi
, SBI_POR_DOING
);
1619 spin_lock_init(&sbi
->stat_lock
);
1621 init_rwsem(&sbi
->read_io
.io_rwsem
);
1622 sbi
->read_io
.sbi
= sbi
;
1623 sbi
->read_io
.bio
= NULL
;
1624 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1625 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1626 sbi
->write_io
[i
].sbi
= sbi
;
1627 sbi
->write_io
[i
].bio
= NULL
;
1630 init_rwsem(&sbi
->cp_rwsem
);
1631 init_waitqueue_head(&sbi
->cp_wait
);
1634 err
= init_percpu_info(sbi
);
1638 /* get an inode for meta space */
1639 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1640 if (IS_ERR(sbi
->meta_inode
)) {
1641 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1642 err
= PTR_ERR(sbi
->meta_inode
);
1646 err
= get_valid_checkpoint(sbi
);
1648 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1649 goto free_meta_inode
;
1652 sbi
->total_valid_node_count
=
1653 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1654 percpu_counter_set(&sbi
->total_valid_inode_count
,
1655 le32_to_cpu(sbi
->ckpt
->valid_inode_count
));
1656 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1657 sbi
->total_valid_block_count
=
1658 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1659 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1661 for (i
= 0; i
< NR_INODE_TYPE
; i
++) {
1662 INIT_LIST_HEAD(&sbi
->inode_list
[i
]);
1663 spin_lock_init(&sbi
->inode_lock
[i
]);
1666 init_extent_cache_info(sbi
);
1668 init_ino_entry_info(sbi
);
1670 /* setup f2fs internal modules */
1671 err
= build_segment_manager(sbi
);
1673 f2fs_msg(sb
, KERN_ERR
,
1674 "Failed to initialize F2FS segment manager");
1677 err
= build_node_manager(sbi
);
1679 f2fs_msg(sb
, KERN_ERR
,
1680 "Failed to initialize F2FS node manager");
1684 /* For write statistics */
1685 if (sb
->s_bdev
->bd_part
)
1686 sbi
->sectors_written_start
=
1687 (u64
)part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
1689 /* Read accumulated write IO statistics if exists */
1690 seg_i
= CURSEG_I(sbi
, CURSEG_HOT_NODE
);
1691 if (__exist_node_summaries(sbi
))
1692 sbi
->kbytes_written
=
1693 le64_to_cpu(seg_i
->journal
->info
.kbytes_written
);
1695 build_gc_manager(sbi
);
1697 /* get an inode for node space */
1698 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1699 if (IS_ERR(sbi
->node_inode
)) {
1700 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1701 err
= PTR_ERR(sbi
->node_inode
);
1705 f2fs_join_shrinker(sbi
);
1707 /* if there are nt orphan nodes free them */
1708 err
= recover_orphan_inodes(sbi
);
1710 goto free_node_inode
;
1712 /* read root inode and dentry */
1713 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1715 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1716 err
= PTR_ERR(root
);
1717 goto free_node_inode
;
1719 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1722 goto free_node_inode
;
1725 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1728 goto free_root_inode
;
1731 err
= f2fs_build_stats(sbi
);
1733 goto free_root_inode
;
1736 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1739 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1740 &f2fs_seq_segment_info_fops
, sb
);
1741 proc_create_data("segment_bits", S_IRUGO
, sbi
->s_proc
,
1742 &f2fs_seq_segment_bits_fops
, sb
);
1745 sbi
->s_kobj
.kset
= f2fs_kset
;
1746 init_completion(&sbi
->s_kobj_unregister
);
1747 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1752 /* recover fsynced data */
1753 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1755 * mount should be failed, when device has readonly mode, and
1756 * previous checkpoint was not done by clean system shutdown.
1758 if (bdev_read_only(sb
->s_bdev
) &&
1759 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1765 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1767 err
= recover_fsync_data(sbi
, false);
1770 f2fs_msg(sb
, KERN_ERR
,
1771 "Cannot recover all fsync data errno=%d", err
);
1775 err
= recover_fsync_data(sbi
, true);
1777 if (!f2fs_readonly(sb
) && err
> 0) {
1779 f2fs_msg(sb
, KERN_ERR
,
1780 "Need to recover fsync data");
1785 /* recover_fsync_data() cleared this already */
1786 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1789 * If filesystem is not mounted as read-only then
1790 * do start the gc_thread.
1792 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1793 /* After POR, we can run background GC thread.*/
1794 err
= start_gc_thread(sbi
);
1800 /* recover broken superblock */
1802 err
= f2fs_commit_super(sbi
, true);
1803 f2fs_msg(sb
, KERN_INFO
,
1804 "Try to recover %dth superblock, ret: %d",
1805 sbi
->valid_super_block
? 1 : 2, err
);
1808 f2fs_update_time(sbi
, CP_TIME
);
1809 f2fs_update_time(sbi
, REQ_TIME
);
1813 f2fs_sync_inode_meta(sbi
);
1814 kobject_del(&sbi
->s_kobj
);
1815 kobject_put(&sbi
->s_kobj
);
1816 wait_for_completion(&sbi
->s_kobj_unregister
);
1819 remove_proc_entry("segment_info", sbi
->s_proc
);
1820 remove_proc_entry("segment_bits", sbi
->s_proc
);
1821 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1823 f2fs_destroy_stats(sbi
);
1828 mutex_lock(&sbi
->umount_mutex
);
1829 f2fs_leave_shrinker(sbi
);
1830 iput(sbi
->node_inode
);
1831 mutex_unlock(&sbi
->umount_mutex
);
1833 destroy_node_manager(sbi
);
1835 destroy_segment_manager(sbi
);
1838 make_bad_inode(sbi
->meta_inode
);
1839 iput(sbi
->meta_inode
);
1841 destroy_percpu_info(sbi
);
1846 if (sbi
->s_chksum_driver
)
1847 crypto_free_shash(sbi
->s_chksum_driver
);
1850 /* give only one another chance */
1853 shrink_dcache_sb(sb
);
1859 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1860 const char *dev_name
, void *data
)
1862 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1865 static void kill_f2fs_super(struct super_block
*sb
)
1868 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1869 kill_block_super(sb
);
1872 static struct file_system_type f2fs_fs_type
= {
1873 .owner
= THIS_MODULE
,
1875 .mount
= f2fs_mount
,
1876 .kill_sb
= kill_f2fs_super
,
1877 .fs_flags
= FS_REQUIRES_DEV
,
1879 MODULE_ALIAS_FS("f2fs");
1881 static int __init
init_inodecache(void)
1883 f2fs_inode_cachep
= kmem_cache_create("f2fs_inode_cache",
1884 sizeof(struct f2fs_inode_info
), 0,
1885 SLAB_RECLAIM_ACCOUNT
|SLAB_ACCOUNT
, NULL
);
1886 if (!f2fs_inode_cachep
)
1891 static void destroy_inodecache(void)
1894 * Make sure all delayed rcu free inodes are flushed before we
1898 kmem_cache_destroy(f2fs_inode_cachep
);
1901 static int __init
init_f2fs_fs(void)
1905 f2fs_build_trace_ios();
1907 err
= init_inodecache();
1910 err
= create_node_manager_caches();
1912 goto free_inodecache
;
1913 err
= create_segment_manager_caches();
1915 goto free_node_manager_caches
;
1916 err
= create_checkpoint_caches();
1918 goto free_segment_manager_caches
;
1919 err
= create_extent_cache();
1921 goto free_checkpoint_caches
;
1922 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1925 goto free_extent_cache
;
1927 #ifdef CONFIG_F2FS_FAULT_INJECTION
1928 f2fs_fault_inject
.kset
= f2fs_kset
;
1929 f2fs_build_fault_attr(0);
1930 err
= kobject_init_and_add(&f2fs_fault_inject
, &f2fs_fault_ktype
,
1931 NULL
, "fault_injection");
1933 f2fs_fault_inject
.kset
= NULL
;
1937 err
= register_shrinker(&f2fs_shrinker_info
);
1941 err
= register_filesystem(&f2fs_fs_type
);
1944 err
= f2fs_create_root_stats();
1946 goto free_filesystem
;
1947 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1951 unregister_filesystem(&f2fs_fs_type
);
1953 unregister_shrinker(&f2fs_shrinker_info
);
1955 #ifdef CONFIG_F2FS_FAULT_INJECTION
1956 if (f2fs_fault_inject
.kset
)
1957 kobject_put(&f2fs_fault_inject
);
1959 kset_unregister(f2fs_kset
);
1961 destroy_extent_cache();
1962 free_checkpoint_caches
:
1963 destroy_checkpoint_caches();
1964 free_segment_manager_caches
:
1965 destroy_segment_manager_caches();
1966 free_node_manager_caches
:
1967 destroy_node_manager_caches();
1969 destroy_inodecache();
1974 static void __exit
exit_f2fs_fs(void)
1976 remove_proc_entry("fs/f2fs", NULL
);
1977 f2fs_destroy_root_stats();
1978 unregister_filesystem(&f2fs_fs_type
);
1979 unregister_shrinker(&f2fs_shrinker_info
);
1980 #ifdef CONFIG_F2FS_FAULT_INJECTION
1981 kobject_put(&f2fs_fault_inject
);
1983 kset_unregister(f2fs_kset
);
1984 destroy_extent_cache();
1985 destroy_checkpoint_caches();
1986 destroy_segment_manager_caches();
1987 destroy_node_manager_caches();
1988 destroy_inodecache();
1989 f2fs_destroy_trace_ios();
1992 module_init(init_f2fs_fs
)
1993 module_exit(exit_f2fs_fs
)
1995 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1996 MODULE_DESCRIPTION("Flash Friendly File System");
1997 MODULE_LICENSE("GPL");