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",
52 [FAULT_EVICT_INODE
] = "evict_inode fail",
55 static void f2fs_build_fault_attr(unsigned int rate
)
58 atomic_set(&f2fs_fault
.inject_ops
, 0);
59 f2fs_fault
.inject_rate
= rate
;
60 f2fs_fault
.inject_type
= (1 << FAULT_MAX
) - 1;
62 memset(&f2fs_fault
, 0, sizeof(struct f2fs_fault_info
));
67 /* f2fs-wide shrinker description */
68 static struct shrinker f2fs_shrinker_info
= {
69 .scan_objects
= f2fs_shrink_scan
,
70 .count_objects
= f2fs_shrink_count
,
71 .seeks
= DEFAULT_SEEKS
,
76 Opt_disable_roll_forward
,
86 Opt_disable_ext_identify
,
105 static match_table_t f2fs_tokens
= {
106 {Opt_gc_background
, "background_gc=%s"},
107 {Opt_disable_roll_forward
, "disable_roll_forward"},
108 {Opt_norecovery
, "norecovery"},
109 {Opt_discard
, "discard"},
110 {Opt_nodiscard
, "nodiscard"},
111 {Opt_noheap
, "no_heap"},
112 {Opt_user_xattr
, "user_xattr"},
113 {Opt_nouser_xattr
, "nouser_xattr"},
115 {Opt_noacl
, "noacl"},
116 {Opt_active_logs
, "active_logs=%u"},
117 {Opt_disable_ext_identify
, "disable_ext_identify"},
118 {Opt_inline_xattr
, "inline_xattr"},
119 {Opt_inline_data
, "inline_data"},
120 {Opt_inline_dentry
, "inline_dentry"},
121 {Opt_flush_merge
, "flush_merge"},
122 {Opt_noflush_merge
, "noflush_merge"},
123 {Opt_nobarrier
, "nobarrier"},
124 {Opt_fastboot
, "fastboot"},
125 {Opt_extent_cache
, "extent_cache"},
126 {Opt_noextent_cache
, "noextent_cache"},
127 {Opt_noinline_data
, "noinline_data"},
128 {Opt_data_flush
, "data_flush"},
129 {Opt_mode
, "mode=%s"},
130 {Opt_fault_injection
, "fault_injection=%u"},
131 {Opt_lazytime
, "lazytime"},
132 {Opt_nolazytime
, "nolazytime"},
136 /* Sysfs support for f2fs */
138 GC_THREAD
, /* struct f2fs_gc_thread */
139 SM_INFO
, /* struct f2fs_sm_info */
140 NM_INFO
, /* struct f2fs_nm_info */
141 F2FS_SBI
, /* struct f2fs_sb_info */
142 #ifdef CONFIG_F2FS_FAULT_INJECTION
143 FAULT_INFO_RATE
, /* struct f2fs_fault_info */
144 FAULT_INFO_TYPE
, /* struct f2fs_fault_info */
149 struct attribute attr
;
150 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
151 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
152 const char *, size_t);
157 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
159 if (struct_type
== GC_THREAD
)
160 return (unsigned char *)sbi
->gc_thread
;
161 else if (struct_type
== SM_INFO
)
162 return (unsigned char *)SM_I(sbi
);
163 else if (struct_type
== NM_INFO
)
164 return (unsigned char *)NM_I(sbi
);
165 else if (struct_type
== F2FS_SBI
)
166 return (unsigned char *)sbi
;
167 #ifdef CONFIG_F2FS_FAULT_INJECTION
168 else if (struct_type
== FAULT_INFO_RATE
||
169 struct_type
== FAULT_INFO_TYPE
)
170 return (unsigned char *)&f2fs_fault
;
175 static ssize_t
lifetime_write_kbytes_show(struct f2fs_attr
*a
,
176 struct f2fs_sb_info
*sbi
, char *buf
)
178 struct super_block
*sb
= sbi
->sb
;
180 if (!sb
->s_bdev
->bd_part
)
181 return snprintf(buf
, PAGE_SIZE
, "0\n");
183 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
184 (unsigned long long)(sbi
->kbytes_written
+
185 BD_PART_WRITTEN(sbi
)));
188 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
189 struct f2fs_sb_info
*sbi
, char *buf
)
191 unsigned char *ptr
= NULL
;
194 ptr
= __struct_ptr(sbi
, a
->struct_type
);
198 ui
= (unsigned int *)(ptr
+ a
->offset
);
200 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
203 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
204 struct f2fs_sb_info
*sbi
,
205 const char *buf
, size_t count
)
212 ptr
= __struct_ptr(sbi
, a
->struct_type
);
216 ui
= (unsigned int *)(ptr
+ a
->offset
);
218 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
221 #ifdef CONFIG_F2FS_FAULT_INJECTION
222 if (a
->struct_type
== FAULT_INFO_TYPE
&& t
>= (1 << FAULT_MAX
))
229 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
230 struct attribute
*attr
, char *buf
)
232 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
234 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
236 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
239 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
240 const char *buf
, size_t len
)
242 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
244 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
246 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
249 static void f2fs_sb_release(struct kobject
*kobj
)
251 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
253 complete(&sbi
->s_kobj_unregister
);
256 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
257 static struct f2fs_attr f2fs_attr_##_name = { \
258 .attr = {.name = __stringify(_name), .mode = _mode }, \
261 .struct_type = _struct_type, \
265 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
266 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
267 f2fs_sbi_show, f2fs_sbi_store, \
268 offsetof(struct struct_name, elname))
270 #define F2FS_GENERAL_RO_ATTR(name) \
271 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
273 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
274 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
275 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
276 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
277 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
278 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
279 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, batched_trim_sections
, trim_sections
);
280 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
281 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
282 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
283 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
284 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ra_nid_pages
, ra_nid_pages
);
285 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, dirty_nats_ratio
, dirty_nats_ratio
);
286 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
287 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
288 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, cp_interval
, interval_time
[CP_TIME
]);
289 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, idle_interval
, interval_time
[REQ_TIME
]);
290 #ifdef CONFIG_F2FS_FAULT_INJECTION
291 F2FS_RW_ATTR(FAULT_INFO_RATE
, f2fs_fault_info
, inject_rate
, inject_rate
);
292 F2FS_RW_ATTR(FAULT_INFO_TYPE
, f2fs_fault_info
, inject_type
, inject_type
);
294 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes
);
296 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
297 static struct attribute
*f2fs_attrs
[] = {
298 ATTR_LIST(gc_min_sleep_time
),
299 ATTR_LIST(gc_max_sleep_time
),
300 ATTR_LIST(gc_no_gc_sleep_time
),
302 ATTR_LIST(reclaim_segments
),
303 ATTR_LIST(max_small_discards
),
304 ATTR_LIST(batched_trim_sections
),
305 ATTR_LIST(ipu_policy
),
306 ATTR_LIST(min_ipu_util
),
307 ATTR_LIST(min_fsync_blocks
),
308 ATTR_LIST(max_victim_search
),
309 ATTR_LIST(dir_level
),
310 ATTR_LIST(ram_thresh
),
311 ATTR_LIST(ra_nid_pages
),
312 ATTR_LIST(dirty_nats_ratio
),
313 ATTR_LIST(cp_interval
),
314 ATTR_LIST(idle_interval
),
315 ATTR_LIST(lifetime_write_kbytes
),
319 static const struct sysfs_ops f2fs_attr_ops
= {
320 .show
= f2fs_attr_show
,
321 .store
= f2fs_attr_store
,
324 static struct kobj_type f2fs_ktype
= {
325 .default_attrs
= f2fs_attrs
,
326 .sysfs_ops
= &f2fs_attr_ops
,
327 .release
= f2fs_sb_release
,
330 #ifdef CONFIG_F2FS_FAULT_INJECTION
331 /* sysfs for f2fs fault injection */
332 static struct kobject f2fs_fault_inject
;
334 static struct attribute
*f2fs_fault_attrs
[] = {
335 ATTR_LIST(inject_rate
),
336 ATTR_LIST(inject_type
),
340 static struct kobj_type f2fs_fault_ktype
= {
341 .default_attrs
= f2fs_fault_attrs
,
342 .sysfs_ops
= &f2fs_attr_ops
,
346 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
348 struct va_format vaf
;
354 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
358 static void init_once(void *foo
)
360 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
362 inode_init_once(&fi
->vfs_inode
);
365 static int parse_options(struct super_block
*sb
, char *options
)
367 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
368 struct request_queue
*q
;
369 substring_t args
[MAX_OPT_ARGS
];
373 #ifdef CONFIG_F2FS_FAULT_INJECTION
374 f2fs_build_fault_attr(0);
380 while ((p
= strsep(&options
, ",")) != NULL
) {
385 * Initialize args struct so we know whether arg was
386 * found; some options take optional arguments.
388 args
[0].to
= args
[0].from
= NULL
;
389 token
= match_token(p
, f2fs_tokens
, args
);
392 case Opt_gc_background
:
393 name
= match_strdup(&args
[0]);
397 if (strlen(name
) == 2 && !strncmp(name
, "on", 2)) {
399 clear_opt(sbi
, FORCE_FG_GC
);
400 } else if (strlen(name
) == 3 && !strncmp(name
, "off", 3)) {
401 clear_opt(sbi
, BG_GC
);
402 clear_opt(sbi
, FORCE_FG_GC
);
403 } else if (strlen(name
) == 4 && !strncmp(name
, "sync", 4)) {
405 set_opt(sbi
, FORCE_FG_GC
);
412 case Opt_disable_roll_forward
:
413 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
416 /* this option mounts f2fs with ro */
417 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
418 if (!f2fs_readonly(sb
))
422 q
= bdev_get_queue(sb
->s_bdev
);
423 if (blk_queue_discard(q
)) {
424 set_opt(sbi
, DISCARD
);
426 f2fs_msg(sb
, KERN_WARNING
,
427 "mounting with \"discard\" option, but "
428 "the device does not support discard");
432 clear_opt(sbi
, DISCARD
);
434 set_opt(sbi
, NOHEAP
);
436 #ifdef CONFIG_F2FS_FS_XATTR
438 set_opt(sbi
, XATTR_USER
);
440 case Opt_nouser_xattr
:
441 clear_opt(sbi
, XATTR_USER
);
443 case Opt_inline_xattr
:
444 set_opt(sbi
, INLINE_XATTR
);
448 f2fs_msg(sb
, KERN_INFO
,
449 "user_xattr options not supported");
451 case Opt_nouser_xattr
:
452 f2fs_msg(sb
, KERN_INFO
,
453 "nouser_xattr options not supported");
455 case Opt_inline_xattr
:
456 f2fs_msg(sb
, KERN_INFO
,
457 "inline_xattr options not supported");
460 #ifdef CONFIG_F2FS_FS_POSIX_ACL
462 set_opt(sbi
, POSIX_ACL
);
465 clear_opt(sbi
, POSIX_ACL
);
469 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
472 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
475 case Opt_active_logs
:
476 if (args
->from
&& match_int(args
, &arg
))
478 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
480 sbi
->active_logs
= arg
;
482 case Opt_disable_ext_identify
:
483 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
485 case Opt_inline_data
:
486 set_opt(sbi
, INLINE_DATA
);
488 case Opt_inline_dentry
:
489 set_opt(sbi
, INLINE_DENTRY
);
491 case Opt_flush_merge
:
492 set_opt(sbi
, FLUSH_MERGE
);
494 case Opt_noflush_merge
:
495 clear_opt(sbi
, FLUSH_MERGE
);
498 set_opt(sbi
, NOBARRIER
);
501 set_opt(sbi
, FASTBOOT
);
503 case Opt_extent_cache
:
504 set_opt(sbi
, EXTENT_CACHE
);
506 case Opt_noextent_cache
:
507 clear_opt(sbi
, EXTENT_CACHE
);
509 case Opt_noinline_data
:
510 clear_opt(sbi
, INLINE_DATA
);
513 set_opt(sbi
, DATA_FLUSH
);
516 name
= match_strdup(&args
[0]);
520 if (strlen(name
) == 8 &&
521 !strncmp(name
, "adaptive", 8)) {
522 set_opt_mode(sbi
, F2FS_MOUNT_ADAPTIVE
);
523 } else if (strlen(name
) == 3 &&
524 !strncmp(name
, "lfs", 3)) {
525 set_opt_mode(sbi
, F2FS_MOUNT_LFS
);
532 case Opt_fault_injection
:
533 if (args
->from
&& match_int(args
, &arg
))
535 #ifdef CONFIG_F2FS_FAULT_INJECTION
536 f2fs_build_fault_attr(arg
);
538 f2fs_msg(sb
, KERN_INFO
,
539 "FAULT_INJECTION was not selected");
543 sb
->s_flags
|= MS_LAZYTIME
;
546 sb
->s_flags
&= ~MS_LAZYTIME
;
549 f2fs_msg(sb
, KERN_ERR
,
550 "Unrecognized mount option \"%s\" or missing value",
558 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
560 struct f2fs_inode_info
*fi
;
562 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
566 init_once((void *) fi
);
568 if (percpu_counter_init(&fi
->dirty_pages
, 0, GFP_NOFS
)) {
569 kmem_cache_free(f2fs_inode_cachep
, fi
);
573 /* Initialize f2fs-specific inode info */
574 fi
->vfs_inode
.i_version
= 1;
575 fi
->i_current_depth
= 1;
577 init_rwsem(&fi
->i_sem
);
578 INIT_LIST_HEAD(&fi
->dirty_list
);
579 INIT_LIST_HEAD(&fi
->gdirty_list
);
580 INIT_LIST_HEAD(&fi
->inmem_pages
);
581 mutex_init(&fi
->inmem_lock
);
583 /* Will be used by directory only */
584 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
585 return &fi
->vfs_inode
;
588 static int f2fs_drop_inode(struct inode
*inode
)
591 * This is to avoid a deadlock condition like below.
592 * writeback_single_inode(inode)
593 * - f2fs_write_data_page
594 * - f2fs_gc -> iput -> evict
595 * - inode_wait_for_writeback(inode)
597 if ((!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
)) {
598 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
599 /* to avoid evict_inode call simultaneously */
600 atomic_inc(&inode
->i_count
);
601 spin_unlock(&inode
->i_lock
);
603 /* some remained atomic pages should discarded */
604 if (f2fs_is_atomic_file(inode
))
605 drop_inmem_pages(inode
);
607 /* should remain fi->extent_tree for writepage */
608 f2fs_destroy_extent_node(inode
);
610 sb_start_intwrite(inode
->i_sb
);
611 f2fs_i_size_write(inode
, 0);
613 if (F2FS_HAS_BLOCKS(inode
))
614 f2fs_truncate(inode
);
616 sb_end_intwrite(inode
->i_sb
);
618 fscrypt_put_encryption_info(inode
, NULL
);
619 spin_lock(&inode
->i_lock
);
620 atomic_dec(&inode
->i_count
);
625 return generic_drop_inode(inode
);
628 int f2fs_inode_dirtied(struct inode
*inode
)
630 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
632 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
633 if (is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
634 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
638 set_inode_flag(inode
, FI_DIRTY_INODE
);
639 list_add_tail(&F2FS_I(inode
)->gdirty_list
,
640 &sbi
->inode_list
[DIRTY_META
]);
641 inc_page_count(sbi
, F2FS_DIRTY_IMETA
);
642 stat_inc_dirty_inode(sbi
, DIRTY_META
);
643 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
648 void f2fs_inode_synced(struct inode
*inode
)
650 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
652 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
653 if (!is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
654 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
657 list_del_init(&F2FS_I(inode
)->gdirty_list
);
658 clear_inode_flag(inode
, FI_DIRTY_INODE
);
659 clear_inode_flag(inode
, FI_AUTO_RECOVER
);
660 dec_page_count(sbi
, F2FS_DIRTY_IMETA
);
661 stat_dec_dirty_inode(F2FS_I_SB(inode
), DIRTY_META
);
662 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
666 * f2fs_dirty_inode() is called from __mark_inode_dirty()
668 * We should call set_dirty_inode to write the dirty inode through write_inode.
670 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
672 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
674 if (inode
->i_ino
== F2FS_NODE_INO(sbi
) ||
675 inode
->i_ino
== F2FS_META_INO(sbi
))
678 if (flags
== I_DIRTY_TIME
)
681 if (is_inode_flag_set(inode
, FI_AUTO_RECOVER
))
682 clear_inode_flag(inode
, FI_AUTO_RECOVER
);
684 f2fs_inode_dirtied(inode
);
687 static void f2fs_i_callback(struct rcu_head
*head
)
689 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
690 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
693 static void f2fs_destroy_inode(struct inode
*inode
)
695 percpu_counter_destroy(&F2FS_I(inode
)->dirty_pages
);
696 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
699 static void destroy_percpu_info(struct f2fs_sb_info
*sbi
)
703 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
704 percpu_counter_destroy(&sbi
->nr_pages
[i
]);
705 percpu_counter_destroy(&sbi
->alloc_valid_block_count
);
706 percpu_counter_destroy(&sbi
->total_valid_inode_count
);
708 percpu_free_rwsem(&sbi
->cp_rwsem
);
711 static void f2fs_put_super(struct super_block
*sb
)
713 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
716 remove_proc_entry("segment_info", sbi
->s_proc
);
717 remove_proc_entry("segment_bits", sbi
->s_proc
);
718 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
720 kobject_del(&sbi
->s_kobj
);
724 /* prevent remaining shrinker jobs */
725 mutex_lock(&sbi
->umount_mutex
);
728 * We don't need to do checkpoint when superblock is clean.
729 * But, the previous checkpoint was not done by umount, it needs to do
730 * clean checkpoint again.
732 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
733 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
734 struct cp_control cpc
= {
737 write_checkpoint(sbi
, &cpc
);
740 /* write_checkpoint can update stat informaion */
741 f2fs_destroy_stats(sbi
);
744 * normally superblock is clean, so we need to release this.
745 * In addition, EIO will skip do checkpoint, we need this as well.
747 release_ino_entry(sbi
, true);
748 release_discard_addrs(sbi
);
750 f2fs_leave_shrinker(sbi
);
751 mutex_unlock(&sbi
->umount_mutex
);
753 /* our cp_error case, we can wait for any writeback page */
754 f2fs_flush_merged_bios(sbi
);
756 iput(sbi
->node_inode
);
757 iput(sbi
->meta_inode
);
759 /* destroy f2fs internal modules */
760 destroy_node_manager(sbi
);
761 destroy_segment_manager(sbi
);
764 kobject_put(&sbi
->s_kobj
);
765 wait_for_completion(&sbi
->s_kobj_unregister
);
767 sb
->s_fs_info
= NULL
;
768 if (sbi
->s_chksum_driver
)
769 crypto_free_shash(sbi
->s_chksum_driver
);
770 kfree(sbi
->raw_super
);
772 destroy_percpu_info(sbi
);
776 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
778 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
781 trace_f2fs_sync_fs(sb
, sync
);
784 struct cp_control cpc
;
786 cpc
.reason
= __get_cp_reason(sbi
);
788 mutex_lock(&sbi
->gc_mutex
);
789 err
= write_checkpoint(sbi
, &cpc
);
790 mutex_unlock(&sbi
->gc_mutex
);
792 f2fs_trace_ios(NULL
, 1);
797 static int f2fs_freeze(struct super_block
*sb
)
801 if (f2fs_readonly(sb
))
804 err
= f2fs_sync_fs(sb
, 1);
808 static int f2fs_unfreeze(struct super_block
*sb
)
813 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
815 struct super_block
*sb
= dentry
->d_sb
;
816 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
817 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
818 block_t total_count
, user_block_count
, start_count
, ovp_count
;
820 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
821 user_block_count
= sbi
->user_block_count
;
822 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
823 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
824 buf
->f_type
= F2FS_SUPER_MAGIC
;
825 buf
->f_bsize
= sbi
->blocksize
;
827 buf
->f_blocks
= total_count
- start_count
;
828 buf
->f_bfree
= user_block_count
- valid_user_blocks(sbi
) + ovp_count
;
829 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
831 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
832 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
834 buf
->f_namelen
= F2FS_NAME_LEN
;
835 buf
->f_fsid
.val
[0] = (u32
)id
;
836 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
841 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
843 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
845 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
)) {
846 if (test_opt(sbi
, FORCE_FG_GC
))
847 seq_printf(seq
, ",background_gc=%s", "sync");
849 seq_printf(seq
, ",background_gc=%s", "on");
851 seq_printf(seq
, ",background_gc=%s", "off");
853 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
854 seq_puts(seq
, ",disable_roll_forward");
855 if (test_opt(sbi
, DISCARD
))
856 seq_puts(seq
, ",discard");
857 if (test_opt(sbi
, NOHEAP
))
858 seq_puts(seq
, ",no_heap_alloc");
859 #ifdef CONFIG_F2FS_FS_XATTR
860 if (test_opt(sbi
, XATTR_USER
))
861 seq_puts(seq
, ",user_xattr");
863 seq_puts(seq
, ",nouser_xattr");
864 if (test_opt(sbi
, INLINE_XATTR
))
865 seq_puts(seq
, ",inline_xattr");
867 #ifdef CONFIG_F2FS_FS_POSIX_ACL
868 if (test_opt(sbi
, POSIX_ACL
))
869 seq_puts(seq
, ",acl");
871 seq_puts(seq
, ",noacl");
873 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
874 seq_puts(seq
, ",disable_ext_identify");
875 if (test_opt(sbi
, INLINE_DATA
))
876 seq_puts(seq
, ",inline_data");
878 seq_puts(seq
, ",noinline_data");
879 if (test_opt(sbi
, INLINE_DENTRY
))
880 seq_puts(seq
, ",inline_dentry");
881 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
882 seq_puts(seq
, ",flush_merge");
883 if (test_opt(sbi
, NOBARRIER
))
884 seq_puts(seq
, ",nobarrier");
885 if (test_opt(sbi
, FASTBOOT
))
886 seq_puts(seq
, ",fastboot");
887 if (test_opt(sbi
, EXTENT_CACHE
))
888 seq_puts(seq
, ",extent_cache");
890 seq_puts(seq
, ",noextent_cache");
891 if (test_opt(sbi
, DATA_FLUSH
))
892 seq_puts(seq
, ",data_flush");
894 seq_puts(seq
, ",mode=");
895 if (test_opt(sbi
, ADAPTIVE
))
896 seq_puts(seq
, "adaptive");
897 else if (test_opt(sbi
, LFS
))
898 seq_puts(seq
, "lfs");
899 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
904 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
906 struct super_block
*sb
= seq
->private;
907 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
908 unsigned int total_segs
=
909 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
912 seq_puts(seq
, "format: segment_type|valid_blocks\n"
913 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
915 for (i
= 0; i
< total_segs
; i
++) {
916 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
919 seq_printf(seq
, "%-10d", i
);
920 seq_printf(seq
, "%d|%-3u", se
->type
,
921 get_valid_blocks(sbi
, i
, 1));
922 if ((i
% 10) == 9 || i
== (total_segs
- 1))
931 static int segment_bits_seq_show(struct seq_file
*seq
, void *offset
)
933 struct super_block
*sb
= seq
->private;
934 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
935 unsigned int total_segs
=
936 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
939 seq_puts(seq
, "format: segment_type|valid_blocks|bitmaps\n"
940 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
942 for (i
= 0; i
< total_segs
; i
++) {
943 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
945 seq_printf(seq
, "%-10d", i
);
946 seq_printf(seq
, "%d|%-3u|", se
->type
,
947 get_valid_blocks(sbi
, i
, 1));
948 for (j
= 0; j
< SIT_VBLOCK_MAP_SIZE
; j
++)
949 seq_printf(seq
, "%x ", se
->cur_valid_map
[j
]);
955 #define F2FS_PROC_FILE_DEF(_name) \
956 static int _name##_open_fs(struct inode *inode, struct file *file) \
958 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
961 static const struct file_operations f2fs_seq_##_name##_fops = { \
962 .owner = THIS_MODULE, \
963 .open = _name##_open_fs, \
965 .llseek = seq_lseek, \
966 .release = single_release, \
969 F2FS_PROC_FILE_DEF(segment_info
);
970 F2FS_PROC_FILE_DEF(segment_bits
);
972 static void default_options(struct f2fs_sb_info
*sbi
)
974 /* init some FS parameters */
975 sbi
->active_logs
= NR_CURSEG_TYPE
;
978 set_opt(sbi
, INLINE_DATA
);
979 set_opt(sbi
, EXTENT_CACHE
);
980 sbi
->sb
->s_flags
|= MS_LAZYTIME
;
981 set_opt(sbi
, FLUSH_MERGE
);
982 if (f2fs_sb_mounted_hmsmr(sbi
->sb
)) {
983 set_opt_mode(sbi
, F2FS_MOUNT_LFS
);
984 set_opt(sbi
, DISCARD
);
986 set_opt_mode(sbi
, F2FS_MOUNT_ADAPTIVE
);
989 #ifdef CONFIG_F2FS_FS_XATTR
990 set_opt(sbi
, XATTR_USER
);
992 #ifdef CONFIG_F2FS_FS_POSIX_ACL
993 set_opt(sbi
, POSIX_ACL
);
997 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
999 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
1000 struct f2fs_mount_info org_mount_opt
;
1001 int err
, active_logs
;
1002 bool need_restart_gc
= false;
1003 bool need_stop_gc
= false;
1004 bool no_extent_cache
= !test_opt(sbi
, EXTENT_CACHE
);
1007 * Save the old mount options in case we
1008 * need to restore them.
1010 org_mount_opt
= sbi
->mount_opt
;
1011 active_logs
= sbi
->active_logs
;
1013 /* recover superblocks we couldn't write due to previous RO mount */
1014 if (!(*flags
& MS_RDONLY
) && is_sbi_flag_set(sbi
, SBI_NEED_SB_WRITE
)) {
1015 err
= f2fs_commit_super(sbi
, false);
1016 f2fs_msg(sb
, KERN_INFO
,
1017 "Try to recover all the superblocks, ret: %d", err
);
1019 clear_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1022 sbi
->mount_opt
.opt
= 0;
1023 default_options(sbi
);
1025 /* parse mount options */
1026 err
= parse_options(sb
, data
);
1031 * Previous and new state of filesystem is RO,
1032 * so skip checking GC and FLUSH_MERGE conditions.
1034 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
1037 /* disallow enable/disable extent_cache dynamically */
1038 if (no_extent_cache
== !!test_opt(sbi
, EXTENT_CACHE
)) {
1040 f2fs_msg(sbi
->sb
, KERN_WARNING
,
1041 "switch extent_cache option is not allowed");
1046 * We stop the GC thread if FS is mounted as RO
1047 * or if background_gc = off is passed in mount
1048 * option. Also sync the filesystem.
1050 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
1051 if (sbi
->gc_thread
) {
1052 stop_gc_thread(sbi
);
1053 need_restart_gc
= true;
1055 } else if (!sbi
->gc_thread
) {
1056 err
= start_gc_thread(sbi
);
1059 need_stop_gc
= true;
1062 if (*flags
& MS_RDONLY
) {
1063 writeback_inodes_sb(sb
, WB_REASON_SYNC
);
1066 set_sbi_flag(sbi
, SBI_IS_DIRTY
);
1067 set_sbi_flag(sbi
, SBI_IS_CLOSE
);
1068 f2fs_sync_fs(sb
, 1);
1069 clear_sbi_flag(sbi
, SBI_IS_CLOSE
);
1073 * We stop issue flush thread if FS is mounted as RO
1074 * or if flush_merge is not passed in mount option.
1076 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
1077 destroy_flush_cmd_control(sbi
);
1078 } else if (!SM_I(sbi
)->cmd_control_info
) {
1079 err
= create_flush_cmd_control(sbi
);
1084 /* Update the POSIXACL Flag */
1085 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1086 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1090 if (need_restart_gc
) {
1091 if (start_gc_thread(sbi
))
1092 f2fs_msg(sbi
->sb
, KERN_WARNING
,
1093 "background gc thread has stopped");
1094 } else if (need_stop_gc
) {
1095 stop_gc_thread(sbi
);
1098 sbi
->mount_opt
= org_mount_opt
;
1099 sbi
->active_logs
= active_logs
;
1103 static struct super_operations f2fs_sops
= {
1104 .alloc_inode
= f2fs_alloc_inode
,
1105 .drop_inode
= f2fs_drop_inode
,
1106 .destroy_inode
= f2fs_destroy_inode
,
1107 .write_inode
= f2fs_write_inode
,
1108 .dirty_inode
= f2fs_dirty_inode
,
1109 .show_options
= f2fs_show_options
,
1110 .evict_inode
= f2fs_evict_inode
,
1111 .put_super
= f2fs_put_super
,
1112 .sync_fs
= f2fs_sync_fs
,
1113 .freeze_fs
= f2fs_freeze
,
1114 .unfreeze_fs
= f2fs_unfreeze
,
1115 .statfs
= f2fs_statfs
,
1116 .remount_fs
= f2fs_remount
,
1119 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1120 static int f2fs_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1122 return f2fs_getxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1123 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1127 static int f2fs_key_prefix(struct inode
*inode
, u8
**key
)
1129 *key
= F2FS_I_SB(inode
)->key_prefix
;
1130 return F2FS_I_SB(inode
)->key_prefix_size
;
1133 static int f2fs_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1136 return f2fs_setxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1137 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1138 ctx
, len
, fs_data
, XATTR_CREATE
);
1141 static unsigned f2fs_max_namelen(struct inode
*inode
)
1143 return S_ISLNK(inode
->i_mode
) ?
1144 inode
->i_sb
->s_blocksize
: F2FS_NAME_LEN
;
1147 static struct fscrypt_operations f2fs_cryptops
= {
1148 .get_context
= f2fs_get_context
,
1149 .key_prefix
= f2fs_key_prefix
,
1150 .set_context
= f2fs_set_context
,
1151 .is_encrypted
= f2fs_encrypted_inode
,
1152 .empty_dir
= f2fs_empty_dir
,
1153 .max_namelen
= f2fs_max_namelen
,
1156 static struct fscrypt_operations f2fs_cryptops
= {
1157 .is_encrypted
= f2fs_encrypted_inode
,
1161 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
1162 u64 ino
, u32 generation
)
1164 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
1165 struct inode
*inode
;
1167 if (check_nid_range(sbi
, ino
))
1168 return ERR_PTR(-ESTALE
);
1171 * f2fs_iget isn't quite right if the inode is currently unallocated!
1172 * However f2fs_iget currently does appropriate checks to handle stale
1173 * inodes so everything is OK.
1175 inode
= f2fs_iget(sb
, ino
);
1177 return ERR_CAST(inode
);
1178 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
1179 /* we didn't find the right inode.. */
1181 return ERR_PTR(-ESTALE
);
1186 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1187 int fh_len
, int fh_type
)
1189 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1190 f2fs_nfs_get_inode
);
1193 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1194 int fh_len
, int fh_type
)
1196 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1197 f2fs_nfs_get_inode
);
1200 static const struct export_operations f2fs_export_ops
= {
1201 .fh_to_dentry
= f2fs_fh_to_dentry
,
1202 .fh_to_parent
= f2fs_fh_to_parent
,
1203 .get_parent
= f2fs_get_parent
,
1206 static loff_t
max_file_blocks(void)
1208 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
1209 loff_t leaf_count
= ADDRS_PER_BLOCK
;
1211 /* two direct node blocks */
1212 result
+= (leaf_count
* 2);
1214 /* two indirect node blocks */
1215 leaf_count
*= NIDS_PER_BLOCK
;
1216 result
+= (leaf_count
* 2);
1218 /* one double indirect node block */
1219 leaf_count
*= NIDS_PER_BLOCK
;
1220 result
+= leaf_count
;
1225 static int __f2fs_commit_super(struct buffer_head
*bh
,
1226 struct f2fs_super_block
*super
)
1230 memcpy(bh
->b_data
+ F2FS_SUPER_OFFSET
, super
, sizeof(*super
));
1231 set_buffer_uptodate(bh
);
1232 set_buffer_dirty(bh
);
1235 /* it's rare case, we can do fua all the time */
1236 return __sync_dirty_buffer(bh
, WRITE_FLUSH_FUA
);
1239 static inline bool sanity_check_area_boundary(struct f2fs_sb_info
*sbi
,
1240 struct buffer_head
*bh
)
1242 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1243 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1244 struct super_block
*sb
= sbi
->sb
;
1245 u32 segment0_blkaddr
= le32_to_cpu(raw_super
->segment0_blkaddr
);
1246 u32 cp_blkaddr
= le32_to_cpu(raw_super
->cp_blkaddr
);
1247 u32 sit_blkaddr
= le32_to_cpu(raw_super
->sit_blkaddr
);
1248 u32 nat_blkaddr
= le32_to_cpu(raw_super
->nat_blkaddr
);
1249 u32 ssa_blkaddr
= le32_to_cpu(raw_super
->ssa_blkaddr
);
1250 u32 main_blkaddr
= le32_to_cpu(raw_super
->main_blkaddr
);
1251 u32 segment_count_ckpt
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1252 u32 segment_count_sit
= le32_to_cpu(raw_super
->segment_count_sit
);
1253 u32 segment_count_nat
= le32_to_cpu(raw_super
->segment_count_nat
);
1254 u32 segment_count_ssa
= le32_to_cpu(raw_super
->segment_count_ssa
);
1255 u32 segment_count_main
= le32_to_cpu(raw_super
->segment_count_main
);
1256 u32 segment_count
= le32_to_cpu(raw_super
->segment_count
);
1257 u32 log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1258 u64 main_end_blkaddr
= main_blkaddr
+
1259 (segment_count_main
<< log_blocks_per_seg
);
1260 u64 seg_end_blkaddr
= segment0_blkaddr
+
1261 (segment_count
<< log_blocks_per_seg
);
1263 if (segment0_blkaddr
!= cp_blkaddr
) {
1264 f2fs_msg(sb
, KERN_INFO
,
1265 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1266 segment0_blkaddr
, cp_blkaddr
);
1270 if (cp_blkaddr
+ (segment_count_ckpt
<< log_blocks_per_seg
) !=
1272 f2fs_msg(sb
, KERN_INFO
,
1273 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1274 cp_blkaddr
, sit_blkaddr
,
1275 segment_count_ckpt
<< log_blocks_per_seg
);
1279 if (sit_blkaddr
+ (segment_count_sit
<< log_blocks_per_seg
) !=
1281 f2fs_msg(sb
, KERN_INFO
,
1282 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1283 sit_blkaddr
, nat_blkaddr
,
1284 segment_count_sit
<< log_blocks_per_seg
);
1288 if (nat_blkaddr
+ (segment_count_nat
<< log_blocks_per_seg
) !=
1290 f2fs_msg(sb
, KERN_INFO
,
1291 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1292 nat_blkaddr
, ssa_blkaddr
,
1293 segment_count_nat
<< log_blocks_per_seg
);
1297 if (ssa_blkaddr
+ (segment_count_ssa
<< log_blocks_per_seg
) !=
1299 f2fs_msg(sb
, KERN_INFO
,
1300 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1301 ssa_blkaddr
, main_blkaddr
,
1302 segment_count_ssa
<< log_blocks_per_seg
);
1306 if (main_end_blkaddr
> seg_end_blkaddr
) {
1307 f2fs_msg(sb
, KERN_INFO
,
1308 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1311 (segment_count
<< log_blocks_per_seg
),
1312 segment_count_main
<< log_blocks_per_seg
);
1314 } else if (main_end_blkaddr
< seg_end_blkaddr
) {
1318 /* fix in-memory information all the time */
1319 raw_super
->segment_count
= cpu_to_le32((main_end_blkaddr
-
1320 segment0_blkaddr
) >> log_blocks_per_seg
);
1322 if (f2fs_readonly(sb
) || bdev_read_only(sb
->s_bdev
)) {
1323 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1326 err
= __f2fs_commit_super(bh
, NULL
);
1327 res
= err
? "failed" : "done";
1329 f2fs_msg(sb
, KERN_INFO
,
1330 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1333 (segment_count
<< log_blocks_per_seg
),
1334 segment_count_main
<< log_blocks_per_seg
);
1341 static int sanity_check_raw_super(struct f2fs_sb_info
*sbi
,
1342 struct buffer_head
*bh
)
1344 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1345 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1346 struct super_block
*sb
= sbi
->sb
;
1347 unsigned int blocksize
;
1349 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
1350 f2fs_msg(sb
, KERN_INFO
,
1351 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1352 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
1356 /* Currently, support only 4KB page cache size */
1357 if (F2FS_BLKSIZE
!= PAGE_SIZE
) {
1358 f2fs_msg(sb
, KERN_INFO
,
1359 "Invalid page_cache_size (%lu), supports only 4KB\n",
1364 /* Currently, support only 4KB block size */
1365 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
1366 if (blocksize
!= F2FS_BLKSIZE
) {
1367 f2fs_msg(sb
, KERN_INFO
,
1368 "Invalid blocksize (%u), supports only 4KB\n",
1373 /* check log blocks per segment */
1374 if (le32_to_cpu(raw_super
->log_blocks_per_seg
) != 9) {
1375 f2fs_msg(sb
, KERN_INFO
,
1376 "Invalid log blocks per segment (%u)\n",
1377 le32_to_cpu(raw_super
->log_blocks_per_seg
));
1381 /* Currently, support 512/1024/2048/4096 bytes sector size */
1382 if (le32_to_cpu(raw_super
->log_sectorsize
) >
1383 F2FS_MAX_LOG_SECTOR_SIZE
||
1384 le32_to_cpu(raw_super
->log_sectorsize
) <
1385 F2FS_MIN_LOG_SECTOR_SIZE
) {
1386 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
1387 le32_to_cpu(raw_super
->log_sectorsize
));
1390 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
1391 le32_to_cpu(raw_super
->log_sectorsize
) !=
1392 F2FS_MAX_LOG_SECTOR_SIZE
) {
1393 f2fs_msg(sb
, KERN_INFO
,
1394 "Invalid log sectors per block(%u) log sectorsize(%u)",
1395 le32_to_cpu(raw_super
->log_sectors_per_block
),
1396 le32_to_cpu(raw_super
->log_sectorsize
));
1400 /* check reserved ino info */
1401 if (le32_to_cpu(raw_super
->node_ino
) != 1 ||
1402 le32_to_cpu(raw_super
->meta_ino
) != 2 ||
1403 le32_to_cpu(raw_super
->root_ino
) != 3) {
1404 f2fs_msg(sb
, KERN_INFO
,
1405 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1406 le32_to_cpu(raw_super
->node_ino
),
1407 le32_to_cpu(raw_super
->meta_ino
),
1408 le32_to_cpu(raw_super
->root_ino
));
1412 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1413 if (sanity_check_area_boundary(sbi
, bh
))
1419 int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
1421 unsigned int total
, fsmeta
;
1422 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
1423 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
1425 total
= le32_to_cpu(raw_super
->segment_count
);
1426 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1427 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
1428 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
1429 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
1430 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
1432 if (unlikely(fsmeta
>= total
))
1435 if (unlikely(f2fs_cp_error(sbi
))) {
1436 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
1442 static void init_sb_info(struct f2fs_sb_info
*sbi
)
1444 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
1446 sbi
->log_sectors_per_block
=
1447 le32_to_cpu(raw_super
->log_sectors_per_block
);
1448 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
1449 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
1450 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1451 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
1452 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
1453 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
1454 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
1455 sbi
->total_node_count
=
1456 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
1457 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
1458 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
1459 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
1460 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
1461 sbi
->cur_victim_sec
= NULL_SECNO
;
1462 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
1464 sbi
->dir_level
= DEF_DIR_LEVEL
;
1465 sbi
->interval_time
[CP_TIME
] = DEF_CP_INTERVAL
;
1466 sbi
->interval_time
[REQ_TIME
] = DEF_IDLE_INTERVAL
;
1467 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
1469 INIT_LIST_HEAD(&sbi
->s_list
);
1470 mutex_init(&sbi
->umount_mutex
);
1471 mutex_init(&sbi
->wio_mutex
[NODE
]);
1472 mutex_init(&sbi
->wio_mutex
[DATA
]);
1474 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1475 memcpy(sbi
->key_prefix
, F2FS_KEY_DESC_PREFIX
,
1476 F2FS_KEY_DESC_PREFIX_SIZE
);
1477 sbi
->key_prefix_size
= F2FS_KEY_DESC_PREFIX_SIZE
;
1481 static int init_percpu_info(struct f2fs_sb_info
*sbi
)
1485 if (percpu_init_rwsem(&sbi
->cp_rwsem
))
1488 for (i
= 0; i
< NR_COUNT_TYPE
; i
++) {
1489 err
= percpu_counter_init(&sbi
->nr_pages
[i
], 0, GFP_KERNEL
);
1494 err
= percpu_counter_init(&sbi
->alloc_valid_block_count
, 0, GFP_KERNEL
);
1498 return percpu_counter_init(&sbi
->total_valid_inode_count
, 0,
1503 * Read f2fs raw super block.
1504 * Because we have two copies of super block, so read both of them
1505 * to get the first valid one. If any one of them is broken, we pass
1506 * them recovery flag back to the caller.
1508 static int read_raw_super_block(struct f2fs_sb_info
*sbi
,
1509 struct f2fs_super_block
**raw_super
,
1510 int *valid_super_block
, int *recovery
)
1512 struct super_block
*sb
= sbi
->sb
;
1514 struct buffer_head
*bh
;
1515 struct f2fs_super_block
*super
;
1518 super
= kzalloc(sizeof(struct f2fs_super_block
), GFP_KERNEL
);
1522 for (block
= 0; block
< 2; block
++) {
1523 bh
= sb_bread(sb
, block
);
1525 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1531 /* sanity checking of raw super */
1532 if (sanity_check_raw_super(sbi
, bh
)) {
1533 f2fs_msg(sb
, KERN_ERR
,
1534 "Can't find valid F2FS filesystem in %dth superblock",
1542 memcpy(super
, bh
->b_data
+ F2FS_SUPER_OFFSET
,
1544 *valid_super_block
= block
;
1550 /* Fail to read any one of the superblocks*/
1554 /* No valid superblock */
1563 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1565 struct buffer_head
*bh
;
1568 if ((recover
&& f2fs_readonly(sbi
->sb
)) ||
1569 bdev_read_only(sbi
->sb
->s_bdev
)) {
1570 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1574 /* write back-up superblock first */
1575 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
? 0: 1);
1578 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1581 /* if we are in recovery path, skip writing valid superblock */
1585 /* write current valid superblock */
1586 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
);
1589 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1594 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1596 struct f2fs_sb_info
*sbi
;
1597 struct f2fs_super_block
*raw_super
;
1600 bool retry
= true, need_fsck
= false;
1601 char *options
= NULL
;
1602 int recovery
, i
, valid_super_block
;
1603 struct curseg_info
*seg_i
;
1608 valid_super_block
= -1;
1611 /* allocate memory for f2fs-specific super block info */
1612 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1618 /* Load the checksum driver */
1619 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32", 0, 0);
1620 if (IS_ERR(sbi
->s_chksum_driver
)) {
1621 f2fs_msg(sb
, KERN_ERR
, "Cannot load crc32 driver.");
1622 err
= PTR_ERR(sbi
->s_chksum_driver
);
1623 sbi
->s_chksum_driver
= NULL
;
1627 /* set a block size */
1628 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1629 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1633 err
= read_raw_super_block(sbi
, &raw_super
, &valid_super_block
,
1638 sb
->s_fs_info
= sbi
;
1639 sbi
->raw_super
= raw_super
;
1641 default_options(sbi
);
1642 /* parse mount options */
1643 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1644 if (data
&& !options
) {
1649 err
= parse_options(sb
, options
);
1653 sbi
->max_file_blocks
= max_file_blocks();
1654 sb
->s_maxbytes
= sbi
->max_file_blocks
<<
1655 le32_to_cpu(raw_super
->log_blocksize
);
1656 sb
->s_max_links
= F2FS_LINK_MAX
;
1657 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1659 sb
->s_op
= &f2fs_sops
;
1660 sb
->s_cop
= &f2fs_cryptops
;
1661 sb
->s_xattr
= f2fs_xattr_handlers
;
1662 sb
->s_export_op
= &f2fs_export_ops
;
1663 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1664 sb
->s_time_gran
= 1;
1665 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1666 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1667 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1669 /* init f2fs-specific super block info */
1670 sbi
->valid_super_block
= valid_super_block
;
1671 mutex_init(&sbi
->gc_mutex
);
1672 mutex_init(&sbi
->cp_mutex
);
1673 init_rwsem(&sbi
->node_write
);
1675 /* disallow all the data/node/meta page writes */
1676 set_sbi_flag(sbi
, SBI_POR_DOING
);
1677 spin_lock_init(&sbi
->stat_lock
);
1679 init_rwsem(&sbi
->read_io
.io_rwsem
);
1680 sbi
->read_io
.sbi
= sbi
;
1681 sbi
->read_io
.bio
= NULL
;
1682 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1683 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1684 sbi
->write_io
[i
].sbi
= sbi
;
1685 sbi
->write_io
[i
].bio
= NULL
;
1688 init_waitqueue_head(&sbi
->cp_wait
);
1691 err
= init_percpu_info(sbi
);
1695 /* get an inode for meta space */
1696 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1697 if (IS_ERR(sbi
->meta_inode
)) {
1698 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1699 err
= PTR_ERR(sbi
->meta_inode
);
1703 err
= get_valid_checkpoint(sbi
);
1705 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1706 goto free_meta_inode
;
1709 sbi
->total_valid_node_count
=
1710 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1711 percpu_counter_set(&sbi
->total_valid_inode_count
,
1712 le32_to_cpu(sbi
->ckpt
->valid_inode_count
));
1713 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1714 sbi
->total_valid_block_count
=
1715 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1716 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1718 for (i
= 0; i
< NR_INODE_TYPE
; i
++) {
1719 INIT_LIST_HEAD(&sbi
->inode_list
[i
]);
1720 spin_lock_init(&sbi
->inode_lock
[i
]);
1723 init_extent_cache_info(sbi
);
1725 init_ino_entry_info(sbi
);
1727 /* setup f2fs internal modules */
1728 err
= build_segment_manager(sbi
);
1730 f2fs_msg(sb
, KERN_ERR
,
1731 "Failed to initialize F2FS segment manager");
1734 err
= build_node_manager(sbi
);
1736 f2fs_msg(sb
, KERN_ERR
,
1737 "Failed to initialize F2FS node manager");
1741 /* For write statistics */
1742 if (sb
->s_bdev
->bd_part
)
1743 sbi
->sectors_written_start
=
1744 (u64
)part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
1746 /* Read accumulated write IO statistics if exists */
1747 seg_i
= CURSEG_I(sbi
, CURSEG_HOT_NODE
);
1748 if (__exist_node_summaries(sbi
))
1749 sbi
->kbytes_written
=
1750 le64_to_cpu(seg_i
->journal
->info
.kbytes_written
);
1752 build_gc_manager(sbi
);
1754 /* get an inode for node space */
1755 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1756 if (IS_ERR(sbi
->node_inode
)) {
1757 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1758 err
= PTR_ERR(sbi
->node_inode
);
1762 f2fs_join_shrinker(sbi
);
1764 /* if there are nt orphan nodes free them */
1765 err
= recover_orphan_inodes(sbi
);
1767 goto free_node_inode
;
1769 /* read root inode and dentry */
1770 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1772 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1773 err
= PTR_ERR(root
);
1774 goto free_node_inode
;
1776 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1779 goto free_node_inode
;
1782 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1785 goto free_root_inode
;
1788 err
= f2fs_build_stats(sbi
);
1790 goto free_root_inode
;
1793 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1796 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1797 &f2fs_seq_segment_info_fops
, sb
);
1798 proc_create_data("segment_bits", S_IRUGO
, sbi
->s_proc
,
1799 &f2fs_seq_segment_bits_fops
, sb
);
1802 sbi
->s_kobj
.kset
= f2fs_kset
;
1803 init_completion(&sbi
->s_kobj_unregister
);
1804 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1809 /* recover fsynced data */
1810 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1812 * mount should be failed, when device has readonly mode, and
1813 * previous checkpoint was not done by clean system shutdown.
1815 if (bdev_read_only(sb
->s_bdev
) &&
1816 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1822 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1824 err
= recover_fsync_data(sbi
, false);
1827 f2fs_msg(sb
, KERN_ERR
,
1828 "Cannot recover all fsync data errno=%d", err
);
1832 err
= recover_fsync_data(sbi
, true);
1834 if (!f2fs_readonly(sb
) && err
> 0) {
1836 f2fs_msg(sb
, KERN_ERR
,
1837 "Need to recover fsync data");
1842 /* recover_fsync_data() cleared this already */
1843 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1846 * If filesystem is not mounted as read-only then
1847 * do start the gc_thread.
1849 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1850 /* After POR, we can run background GC thread.*/
1851 err
= start_gc_thread(sbi
);
1857 /* recover broken superblock */
1859 err
= f2fs_commit_super(sbi
, true);
1860 f2fs_msg(sb
, KERN_INFO
,
1861 "Try to recover %dth superblock, ret: %d",
1862 sbi
->valid_super_block
? 1 : 2, err
);
1865 f2fs_update_time(sbi
, CP_TIME
);
1866 f2fs_update_time(sbi
, REQ_TIME
);
1870 f2fs_sync_inode_meta(sbi
);
1871 kobject_del(&sbi
->s_kobj
);
1872 kobject_put(&sbi
->s_kobj
);
1873 wait_for_completion(&sbi
->s_kobj_unregister
);
1876 remove_proc_entry("segment_info", sbi
->s_proc
);
1877 remove_proc_entry("segment_bits", sbi
->s_proc
);
1878 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1880 f2fs_destroy_stats(sbi
);
1885 mutex_lock(&sbi
->umount_mutex
);
1886 f2fs_leave_shrinker(sbi
);
1887 iput(sbi
->node_inode
);
1888 mutex_unlock(&sbi
->umount_mutex
);
1890 destroy_node_manager(sbi
);
1892 destroy_segment_manager(sbi
);
1895 make_bad_inode(sbi
->meta_inode
);
1896 iput(sbi
->meta_inode
);
1898 destroy_percpu_info(sbi
);
1903 if (sbi
->s_chksum_driver
)
1904 crypto_free_shash(sbi
->s_chksum_driver
);
1907 /* give only one another chance */
1910 shrink_dcache_sb(sb
);
1916 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1917 const char *dev_name
, void *data
)
1919 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1922 static void kill_f2fs_super(struct super_block
*sb
)
1925 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1926 kill_block_super(sb
);
1929 static struct file_system_type f2fs_fs_type
= {
1930 .owner
= THIS_MODULE
,
1932 .mount
= f2fs_mount
,
1933 .kill_sb
= kill_f2fs_super
,
1934 .fs_flags
= FS_REQUIRES_DEV
,
1936 MODULE_ALIAS_FS("f2fs");
1938 static int __init
init_inodecache(void)
1940 f2fs_inode_cachep
= kmem_cache_create("f2fs_inode_cache",
1941 sizeof(struct f2fs_inode_info
), 0,
1942 SLAB_RECLAIM_ACCOUNT
|SLAB_ACCOUNT
, NULL
);
1943 if (!f2fs_inode_cachep
)
1948 static void destroy_inodecache(void)
1951 * Make sure all delayed rcu free inodes are flushed before we
1955 kmem_cache_destroy(f2fs_inode_cachep
);
1958 static int __init
init_f2fs_fs(void)
1962 f2fs_build_trace_ios();
1964 err
= init_inodecache();
1967 err
= create_node_manager_caches();
1969 goto free_inodecache
;
1970 err
= create_segment_manager_caches();
1972 goto free_node_manager_caches
;
1973 err
= create_checkpoint_caches();
1975 goto free_segment_manager_caches
;
1976 err
= create_extent_cache();
1978 goto free_checkpoint_caches
;
1979 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1982 goto free_extent_cache
;
1984 #ifdef CONFIG_F2FS_FAULT_INJECTION
1985 f2fs_fault_inject
.kset
= f2fs_kset
;
1986 f2fs_build_fault_attr(0);
1987 err
= kobject_init_and_add(&f2fs_fault_inject
, &f2fs_fault_ktype
,
1988 NULL
, "fault_injection");
1990 f2fs_fault_inject
.kset
= NULL
;
1994 err
= register_shrinker(&f2fs_shrinker_info
);
1998 err
= register_filesystem(&f2fs_fs_type
);
2001 err
= f2fs_create_root_stats();
2003 goto free_filesystem
;
2004 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
2008 unregister_filesystem(&f2fs_fs_type
);
2010 unregister_shrinker(&f2fs_shrinker_info
);
2012 #ifdef CONFIG_F2FS_FAULT_INJECTION
2013 if (f2fs_fault_inject
.kset
)
2014 kobject_put(&f2fs_fault_inject
);
2016 kset_unregister(f2fs_kset
);
2018 destroy_extent_cache();
2019 free_checkpoint_caches
:
2020 destroy_checkpoint_caches();
2021 free_segment_manager_caches
:
2022 destroy_segment_manager_caches();
2023 free_node_manager_caches
:
2024 destroy_node_manager_caches();
2026 destroy_inodecache();
2031 static void __exit
exit_f2fs_fs(void)
2033 remove_proc_entry("fs/f2fs", NULL
);
2034 f2fs_destroy_root_stats();
2035 unregister_filesystem(&f2fs_fs_type
);
2036 unregister_shrinker(&f2fs_shrinker_info
);
2037 #ifdef CONFIG_F2FS_FAULT_INJECTION
2038 kobject_put(&f2fs_fault_inject
);
2040 kset_unregister(f2fs_kset
);
2041 destroy_extent_cache();
2042 destroy_checkpoint_caches();
2043 destroy_segment_manager_caches();
2044 destroy_node_manager_caches();
2045 destroy_inodecache();
2046 f2fs_destroy_trace_ios();
2049 module_init(init_f2fs_fs
)
2050 module_exit(exit_f2fs_fs
)
2052 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2053 MODULE_DESCRIPTION("Flash Friendly File System");
2054 MODULE_LICENSE("GPL");