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
);
629 * f2fs_dirty_inode() is called from __mark_inode_dirty()
631 * We should call set_dirty_inode to write the dirty inode through write_inode.
633 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
635 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
637 if (inode
->i_ino
== F2FS_NODE_INO(sbi
) ||
638 inode
->i_ino
== F2FS_META_INO(sbi
))
641 if (flags
== I_DIRTY_TIME
)
644 if (is_inode_flag_set(inode
, FI_AUTO_RECOVER
))
645 clear_inode_flag(inode
, FI_AUTO_RECOVER
);
647 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
648 if (is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
649 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
653 set_inode_flag(inode
, FI_DIRTY_INODE
);
654 list_add_tail(&F2FS_I(inode
)->gdirty_list
,
655 &sbi
->inode_list
[DIRTY_META
]);
656 inc_page_count(sbi
, F2FS_DIRTY_IMETA
);
657 stat_inc_dirty_inode(sbi
, DIRTY_META
);
658 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
661 void f2fs_inode_synced(struct inode
*inode
)
663 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
665 spin_lock(&sbi
->inode_lock
[DIRTY_META
]);
666 if (!is_inode_flag_set(inode
, FI_DIRTY_INODE
)) {
667 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
670 list_del_init(&F2FS_I(inode
)->gdirty_list
);
671 clear_inode_flag(inode
, FI_DIRTY_INODE
);
672 clear_inode_flag(inode
, FI_AUTO_RECOVER
);
673 dec_page_count(sbi
, F2FS_DIRTY_IMETA
);
674 stat_dec_dirty_inode(F2FS_I_SB(inode
), DIRTY_META
);
675 spin_unlock(&sbi
->inode_lock
[DIRTY_META
]);
678 static void f2fs_i_callback(struct rcu_head
*head
)
680 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
681 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
684 static void f2fs_destroy_inode(struct inode
*inode
)
686 percpu_counter_destroy(&F2FS_I(inode
)->dirty_pages
);
687 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
690 static void destroy_percpu_info(struct f2fs_sb_info
*sbi
)
694 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
695 percpu_counter_destroy(&sbi
->nr_pages
[i
]);
696 percpu_counter_destroy(&sbi
->alloc_valid_block_count
);
697 percpu_counter_destroy(&sbi
->total_valid_inode_count
);
700 static void f2fs_put_super(struct super_block
*sb
)
702 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
705 remove_proc_entry("segment_info", sbi
->s_proc
);
706 remove_proc_entry("segment_bits", sbi
->s_proc
);
707 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
709 kobject_del(&sbi
->s_kobj
);
713 /* prevent remaining shrinker jobs */
714 mutex_lock(&sbi
->umount_mutex
);
717 * We don't need to do checkpoint when superblock is clean.
718 * But, the previous checkpoint was not done by umount, it needs to do
719 * clean checkpoint again.
721 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
722 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
723 struct cp_control cpc
= {
726 write_checkpoint(sbi
, &cpc
);
729 /* write_checkpoint can update stat informaion */
730 f2fs_destroy_stats(sbi
);
733 * normally superblock is clean, so we need to release this.
734 * In addition, EIO will skip do checkpoint, we need this as well.
736 release_ino_entry(sbi
, true);
737 release_discard_addrs(sbi
);
739 f2fs_leave_shrinker(sbi
);
740 mutex_unlock(&sbi
->umount_mutex
);
742 /* our cp_error case, we can wait for any writeback page */
743 f2fs_flush_merged_bios(sbi
);
745 iput(sbi
->node_inode
);
746 iput(sbi
->meta_inode
);
748 /* destroy f2fs internal modules */
749 destroy_node_manager(sbi
);
750 destroy_segment_manager(sbi
);
753 kobject_put(&sbi
->s_kobj
);
754 wait_for_completion(&sbi
->s_kobj_unregister
);
756 sb
->s_fs_info
= NULL
;
757 if (sbi
->s_chksum_driver
)
758 crypto_free_shash(sbi
->s_chksum_driver
);
759 kfree(sbi
->raw_super
);
761 destroy_percpu_info(sbi
);
765 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
767 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
770 trace_f2fs_sync_fs(sb
, sync
);
773 struct cp_control cpc
;
775 cpc
.reason
= __get_cp_reason(sbi
);
777 mutex_lock(&sbi
->gc_mutex
);
778 err
= write_checkpoint(sbi
, &cpc
);
779 mutex_unlock(&sbi
->gc_mutex
);
781 f2fs_trace_ios(NULL
, 1);
786 static int f2fs_freeze(struct super_block
*sb
)
790 if (f2fs_readonly(sb
))
793 err
= f2fs_sync_fs(sb
, 1);
797 static int f2fs_unfreeze(struct super_block
*sb
)
802 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
804 struct super_block
*sb
= dentry
->d_sb
;
805 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
806 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
807 block_t total_count
, user_block_count
, start_count
, ovp_count
;
809 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
810 user_block_count
= sbi
->user_block_count
;
811 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
812 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
813 buf
->f_type
= F2FS_SUPER_MAGIC
;
814 buf
->f_bsize
= sbi
->blocksize
;
816 buf
->f_blocks
= total_count
- start_count
;
817 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
818 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
820 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
821 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
823 buf
->f_namelen
= F2FS_NAME_LEN
;
824 buf
->f_fsid
.val
[0] = (u32
)id
;
825 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
830 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
832 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
834 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
)) {
835 if (test_opt(sbi
, FORCE_FG_GC
))
836 seq_printf(seq
, ",background_gc=%s", "sync");
838 seq_printf(seq
, ",background_gc=%s", "on");
840 seq_printf(seq
, ",background_gc=%s", "off");
842 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
843 seq_puts(seq
, ",disable_roll_forward");
844 if (test_opt(sbi
, DISCARD
))
845 seq_puts(seq
, ",discard");
846 if (test_opt(sbi
, NOHEAP
))
847 seq_puts(seq
, ",no_heap_alloc");
848 #ifdef CONFIG_F2FS_FS_XATTR
849 if (test_opt(sbi
, XATTR_USER
))
850 seq_puts(seq
, ",user_xattr");
852 seq_puts(seq
, ",nouser_xattr");
853 if (test_opt(sbi
, INLINE_XATTR
))
854 seq_puts(seq
, ",inline_xattr");
856 #ifdef CONFIG_F2FS_FS_POSIX_ACL
857 if (test_opt(sbi
, POSIX_ACL
))
858 seq_puts(seq
, ",acl");
860 seq_puts(seq
, ",noacl");
862 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
863 seq_puts(seq
, ",disable_ext_identify");
864 if (test_opt(sbi
, INLINE_DATA
))
865 seq_puts(seq
, ",inline_data");
867 seq_puts(seq
, ",noinline_data");
868 if (test_opt(sbi
, INLINE_DENTRY
))
869 seq_puts(seq
, ",inline_dentry");
870 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
871 seq_puts(seq
, ",flush_merge");
872 if (test_opt(sbi
, NOBARRIER
))
873 seq_puts(seq
, ",nobarrier");
874 if (test_opt(sbi
, FASTBOOT
))
875 seq_puts(seq
, ",fastboot");
876 if (test_opt(sbi
, EXTENT_CACHE
))
877 seq_puts(seq
, ",extent_cache");
879 seq_puts(seq
, ",noextent_cache");
880 if (test_opt(sbi
, DATA_FLUSH
))
881 seq_puts(seq
, ",data_flush");
883 seq_puts(seq
, ",mode=");
884 if (test_opt(sbi
, ADAPTIVE
))
885 seq_puts(seq
, "adaptive");
886 else if (test_opt(sbi
, LFS
))
887 seq_puts(seq
, "lfs");
888 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
893 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
895 struct super_block
*sb
= seq
->private;
896 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
897 unsigned int total_segs
=
898 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
901 seq_puts(seq
, "format: segment_type|valid_blocks\n"
902 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
904 for (i
= 0; i
< total_segs
; i
++) {
905 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
908 seq_printf(seq
, "%-10d", i
);
909 seq_printf(seq
, "%d|%-3u", se
->type
,
910 get_valid_blocks(sbi
, i
, 1));
911 if ((i
% 10) == 9 || i
== (total_segs
- 1))
920 static int segment_bits_seq_show(struct seq_file
*seq
, void *offset
)
922 struct super_block
*sb
= seq
->private;
923 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
924 unsigned int total_segs
=
925 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
928 seq_puts(seq
, "format: segment_type|valid_blocks|bitmaps\n"
929 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
931 for (i
= 0; i
< total_segs
; i
++) {
932 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
934 seq_printf(seq
, "%-10d", i
);
935 seq_printf(seq
, "%d|%-3u|", se
->type
,
936 get_valid_blocks(sbi
, i
, 1));
937 for (j
= 0; j
< SIT_VBLOCK_MAP_SIZE
; j
++)
938 seq_printf(seq
, "%x ", se
->cur_valid_map
[j
]);
944 #define F2FS_PROC_FILE_DEF(_name) \
945 static int _name##_open_fs(struct inode *inode, struct file *file) \
947 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
950 static const struct file_operations f2fs_seq_##_name##_fops = { \
951 .owner = THIS_MODULE, \
952 .open = _name##_open_fs, \
954 .llseek = seq_lseek, \
955 .release = single_release, \
958 F2FS_PROC_FILE_DEF(segment_info
);
959 F2FS_PROC_FILE_DEF(segment_bits
);
961 static void default_options(struct f2fs_sb_info
*sbi
)
963 /* init some FS parameters */
964 sbi
->active_logs
= NR_CURSEG_TYPE
;
967 set_opt(sbi
, INLINE_DATA
);
968 set_opt(sbi
, EXTENT_CACHE
);
969 sbi
->sb
->s_flags
|= MS_LAZYTIME
;
970 set_opt(sbi
, FLUSH_MERGE
);
971 if (f2fs_sb_mounted_hmsmr(sbi
->sb
)) {
972 set_opt_mode(sbi
, F2FS_MOUNT_LFS
);
973 set_opt(sbi
, DISCARD
);
975 set_opt_mode(sbi
, F2FS_MOUNT_ADAPTIVE
);
978 #ifdef CONFIG_F2FS_FS_XATTR
979 set_opt(sbi
, XATTR_USER
);
981 #ifdef CONFIG_F2FS_FS_POSIX_ACL
982 set_opt(sbi
, POSIX_ACL
);
986 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
988 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
989 struct f2fs_mount_info org_mount_opt
;
990 int err
, active_logs
;
991 bool need_restart_gc
= false;
992 bool need_stop_gc
= false;
993 bool no_extent_cache
= !test_opt(sbi
, EXTENT_CACHE
);
996 * Save the old mount options in case we
997 * need to restore them.
999 org_mount_opt
= sbi
->mount_opt
;
1000 active_logs
= sbi
->active_logs
;
1002 /* recover superblocks we couldn't write due to previous RO mount */
1003 if (!(*flags
& MS_RDONLY
) && is_sbi_flag_set(sbi
, SBI_NEED_SB_WRITE
)) {
1004 err
= f2fs_commit_super(sbi
, false);
1005 f2fs_msg(sb
, KERN_INFO
,
1006 "Try to recover all the superblocks, ret: %d", err
);
1008 clear_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1011 sbi
->mount_opt
.opt
= 0;
1012 default_options(sbi
);
1014 /* parse mount options */
1015 err
= parse_options(sb
, data
);
1020 * Previous and new state of filesystem is RO,
1021 * so skip checking GC and FLUSH_MERGE conditions.
1023 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
1026 /* disallow enable/disable extent_cache dynamically */
1027 if (no_extent_cache
== !!test_opt(sbi
, EXTENT_CACHE
)) {
1029 f2fs_msg(sbi
->sb
, KERN_WARNING
,
1030 "switch extent_cache option is not allowed");
1035 * We stop the GC thread if FS is mounted as RO
1036 * or if background_gc = off is passed in mount
1037 * option. Also sync the filesystem.
1039 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
1040 if (sbi
->gc_thread
) {
1041 stop_gc_thread(sbi
);
1042 need_restart_gc
= true;
1044 } else if (!sbi
->gc_thread
) {
1045 err
= start_gc_thread(sbi
);
1048 need_stop_gc
= true;
1051 if (*flags
& MS_RDONLY
) {
1052 writeback_inodes_sb(sb
, WB_REASON_SYNC
);
1055 set_sbi_flag(sbi
, SBI_IS_DIRTY
);
1056 set_sbi_flag(sbi
, SBI_IS_CLOSE
);
1057 f2fs_sync_fs(sb
, 1);
1058 clear_sbi_flag(sbi
, SBI_IS_CLOSE
);
1062 * We stop issue flush thread if FS is mounted as RO
1063 * or if flush_merge is not passed in mount option.
1065 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
1066 destroy_flush_cmd_control(sbi
);
1067 } else if (!SM_I(sbi
)->cmd_control_info
) {
1068 err
= create_flush_cmd_control(sbi
);
1073 /* Update the POSIXACL Flag */
1074 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1075 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1079 if (need_restart_gc
) {
1080 if (start_gc_thread(sbi
))
1081 f2fs_msg(sbi
->sb
, KERN_WARNING
,
1082 "background gc thread has stopped");
1083 } else if (need_stop_gc
) {
1084 stop_gc_thread(sbi
);
1087 sbi
->mount_opt
= org_mount_opt
;
1088 sbi
->active_logs
= active_logs
;
1092 static struct super_operations f2fs_sops
= {
1093 .alloc_inode
= f2fs_alloc_inode
,
1094 .drop_inode
= f2fs_drop_inode
,
1095 .destroy_inode
= f2fs_destroy_inode
,
1096 .write_inode
= f2fs_write_inode
,
1097 .dirty_inode
= f2fs_dirty_inode
,
1098 .show_options
= f2fs_show_options
,
1099 .evict_inode
= f2fs_evict_inode
,
1100 .put_super
= f2fs_put_super
,
1101 .sync_fs
= f2fs_sync_fs
,
1102 .freeze_fs
= f2fs_freeze
,
1103 .unfreeze_fs
= f2fs_unfreeze
,
1104 .statfs
= f2fs_statfs
,
1105 .remount_fs
= f2fs_remount
,
1108 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1109 static int f2fs_get_context(struct inode
*inode
, void *ctx
, size_t len
)
1111 return f2fs_getxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1112 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1116 static int f2fs_key_prefix(struct inode
*inode
, u8
**key
)
1118 *key
= F2FS_I_SB(inode
)->key_prefix
;
1119 return F2FS_I_SB(inode
)->key_prefix_size
;
1122 static int f2fs_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
1125 return f2fs_setxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
1126 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
1127 ctx
, len
, fs_data
, XATTR_CREATE
);
1130 static unsigned f2fs_max_namelen(struct inode
*inode
)
1132 return S_ISLNK(inode
->i_mode
) ?
1133 inode
->i_sb
->s_blocksize
: F2FS_NAME_LEN
;
1136 static struct fscrypt_operations f2fs_cryptops
= {
1137 .get_context
= f2fs_get_context
,
1138 .key_prefix
= f2fs_key_prefix
,
1139 .set_context
= f2fs_set_context
,
1140 .is_encrypted
= f2fs_encrypted_inode
,
1141 .empty_dir
= f2fs_empty_dir
,
1142 .max_namelen
= f2fs_max_namelen
,
1145 static struct fscrypt_operations f2fs_cryptops
= {
1146 .is_encrypted
= f2fs_encrypted_inode
,
1150 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
1151 u64 ino
, u32 generation
)
1153 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
1154 struct inode
*inode
;
1156 if (check_nid_range(sbi
, ino
))
1157 return ERR_PTR(-ESTALE
);
1160 * f2fs_iget isn't quite right if the inode is currently unallocated!
1161 * However f2fs_iget currently does appropriate checks to handle stale
1162 * inodes so everything is OK.
1164 inode
= f2fs_iget(sb
, ino
);
1166 return ERR_CAST(inode
);
1167 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
1168 /* we didn't find the right inode.. */
1170 return ERR_PTR(-ESTALE
);
1175 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1176 int fh_len
, int fh_type
)
1178 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1179 f2fs_nfs_get_inode
);
1182 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1183 int fh_len
, int fh_type
)
1185 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1186 f2fs_nfs_get_inode
);
1189 static const struct export_operations f2fs_export_ops
= {
1190 .fh_to_dentry
= f2fs_fh_to_dentry
,
1191 .fh_to_parent
= f2fs_fh_to_parent
,
1192 .get_parent
= f2fs_get_parent
,
1195 static loff_t
max_file_blocks(void)
1197 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
1198 loff_t leaf_count
= ADDRS_PER_BLOCK
;
1200 /* two direct node blocks */
1201 result
+= (leaf_count
* 2);
1203 /* two indirect node blocks */
1204 leaf_count
*= NIDS_PER_BLOCK
;
1205 result
+= (leaf_count
* 2);
1207 /* one double indirect node block */
1208 leaf_count
*= NIDS_PER_BLOCK
;
1209 result
+= leaf_count
;
1214 static int __f2fs_commit_super(struct buffer_head
*bh
,
1215 struct f2fs_super_block
*super
)
1219 memcpy(bh
->b_data
+ F2FS_SUPER_OFFSET
, super
, sizeof(*super
));
1220 set_buffer_uptodate(bh
);
1221 set_buffer_dirty(bh
);
1224 /* it's rare case, we can do fua all the time */
1225 return __sync_dirty_buffer(bh
, WRITE_FLUSH_FUA
);
1228 static inline bool sanity_check_area_boundary(struct f2fs_sb_info
*sbi
,
1229 struct buffer_head
*bh
)
1231 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1232 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1233 struct super_block
*sb
= sbi
->sb
;
1234 u32 segment0_blkaddr
= le32_to_cpu(raw_super
->segment0_blkaddr
);
1235 u32 cp_blkaddr
= le32_to_cpu(raw_super
->cp_blkaddr
);
1236 u32 sit_blkaddr
= le32_to_cpu(raw_super
->sit_blkaddr
);
1237 u32 nat_blkaddr
= le32_to_cpu(raw_super
->nat_blkaddr
);
1238 u32 ssa_blkaddr
= le32_to_cpu(raw_super
->ssa_blkaddr
);
1239 u32 main_blkaddr
= le32_to_cpu(raw_super
->main_blkaddr
);
1240 u32 segment_count_ckpt
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1241 u32 segment_count_sit
= le32_to_cpu(raw_super
->segment_count_sit
);
1242 u32 segment_count_nat
= le32_to_cpu(raw_super
->segment_count_nat
);
1243 u32 segment_count_ssa
= le32_to_cpu(raw_super
->segment_count_ssa
);
1244 u32 segment_count_main
= le32_to_cpu(raw_super
->segment_count_main
);
1245 u32 segment_count
= le32_to_cpu(raw_super
->segment_count
);
1246 u32 log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1247 u64 main_end_blkaddr
= main_blkaddr
+
1248 (segment_count_main
<< log_blocks_per_seg
);
1249 u64 seg_end_blkaddr
= segment0_blkaddr
+
1250 (segment_count
<< log_blocks_per_seg
);
1252 if (segment0_blkaddr
!= cp_blkaddr
) {
1253 f2fs_msg(sb
, KERN_INFO
,
1254 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1255 segment0_blkaddr
, cp_blkaddr
);
1259 if (cp_blkaddr
+ (segment_count_ckpt
<< log_blocks_per_seg
) !=
1261 f2fs_msg(sb
, KERN_INFO
,
1262 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1263 cp_blkaddr
, sit_blkaddr
,
1264 segment_count_ckpt
<< log_blocks_per_seg
);
1268 if (sit_blkaddr
+ (segment_count_sit
<< log_blocks_per_seg
) !=
1270 f2fs_msg(sb
, KERN_INFO
,
1271 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1272 sit_blkaddr
, nat_blkaddr
,
1273 segment_count_sit
<< log_blocks_per_seg
);
1277 if (nat_blkaddr
+ (segment_count_nat
<< log_blocks_per_seg
) !=
1279 f2fs_msg(sb
, KERN_INFO
,
1280 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1281 nat_blkaddr
, ssa_blkaddr
,
1282 segment_count_nat
<< log_blocks_per_seg
);
1286 if (ssa_blkaddr
+ (segment_count_ssa
<< log_blocks_per_seg
) !=
1288 f2fs_msg(sb
, KERN_INFO
,
1289 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1290 ssa_blkaddr
, main_blkaddr
,
1291 segment_count_ssa
<< log_blocks_per_seg
);
1295 if (main_end_blkaddr
> seg_end_blkaddr
) {
1296 f2fs_msg(sb
, KERN_INFO
,
1297 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1300 (segment_count
<< log_blocks_per_seg
),
1301 segment_count_main
<< log_blocks_per_seg
);
1303 } else if (main_end_blkaddr
< seg_end_blkaddr
) {
1307 /* fix in-memory information all the time */
1308 raw_super
->segment_count
= cpu_to_le32((main_end_blkaddr
-
1309 segment0_blkaddr
) >> log_blocks_per_seg
);
1311 if (f2fs_readonly(sb
) || bdev_read_only(sb
->s_bdev
)) {
1312 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1315 err
= __f2fs_commit_super(bh
, NULL
);
1316 res
= err
? "failed" : "done";
1318 f2fs_msg(sb
, KERN_INFO
,
1319 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1322 (segment_count
<< log_blocks_per_seg
),
1323 segment_count_main
<< log_blocks_per_seg
);
1330 static int sanity_check_raw_super(struct f2fs_sb_info
*sbi
,
1331 struct buffer_head
*bh
)
1333 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1334 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1335 struct super_block
*sb
= sbi
->sb
;
1336 unsigned int blocksize
;
1338 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
1339 f2fs_msg(sb
, KERN_INFO
,
1340 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1341 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
1345 /* Currently, support only 4KB page cache size */
1346 if (F2FS_BLKSIZE
!= PAGE_SIZE
) {
1347 f2fs_msg(sb
, KERN_INFO
,
1348 "Invalid page_cache_size (%lu), supports only 4KB\n",
1353 /* Currently, support only 4KB block size */
1354 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
1355 if (blocksize
!= F2FS_BLKSIZE
) {
1356 f2fs_msg(sb
, KERN_INFO
,
1357 "Invalid blocksize (%u), supports only 4KB\n",
1362 /* check log blocks per segment */
1363 if (le32_to_cpu(raw_super
->log_blocks_per_seg
) != 9) {
1364 f2fs_msg(sb
, KERN_INFO
,
1365 "Invalid log blocks per segment (%u)\n",
1366 le32_to_cpu(raw_super
->log_blocks_per_seg
));
1370 /* Currently, support 512/1024/2048/4096 bytes sector size */
1371 if (le32_to_cpu(raw_super
->log_sectorsize
) >
1372 F2FS_MAX_LOG_SECTOR_SIZE
||
1373 le32_to_cpu(raw_super
->log_sectorsize
) <
1374 F2FS_MIN_LOG_SECTOR_SIZE
) {
1375 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
1376 le32_to_cpu(raw_super
->log_sectorsize
));
1379 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
1380 le32_to_cpu(raw_super
->log_sectorsize
) !=
1381 F2FS_MAX_LOG_SECTOR_SIZE
) {
1382 f2fs_msg(sb
, KERN_INFO
,
1383 "Invalid log sectors per block(%u) log sectorsize(%u)",
1384 le32_to_cpu(raw_super
->log_sectors_per_block
),
1385 le32_to_cpu(raw_super
->log_sectorsize
));
1389 /* check reserved ino info */
1390 if (le32_to_cpu(raw_super
->node_ino
) != 1 ||
1391 le32_to_cpu(raw_super
->meta_ino
) != 2 ||
1392 le32_to_cpu(raw_super
->root_ino
) != 3) {
1393 f2fs_msg(sb
, KERN_INFO
,
1394 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1395 le32_to_cpu(raw_super
->node_ino
),
1396 le32_to_cpu(raw_super
->meta_ino
),
1397 le32_to_cpu(raw_super
->root_ino
));
1401 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1402 if (sanity_check_area_boundary(sbi
, bh
))
1408 int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
1410 unsigned int total
, fsmeta
;
1411 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
1412 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
1414 total
= le32_to_cpu(raw_super
->segment_count
);
1415 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1416 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
1417 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
1418 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
1419 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
1421 if (unlikely(fsmeta
>= total
))
1424 if (unlikely(f2fs_cp_error(sbi
))) {
1425 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
1431 static void init_sb_info(struct f2fs_sb_info
*sbi
)
1433 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
1435 sbi
->log_sectors_per_block
=
1436 le32_to_cpu(raw_super
->log_sectors_per_block
);
1437 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
1438 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
1439 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1440 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
1441 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
1442 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
1443 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
1444 sbi
->total_node_count
=
1445 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
1446 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
1447 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
1448 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
1449 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
1450 sbi
->cur_victim_sec
= NULL_SECNO
;
1451 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
1453 sbi
->dir_level
= DEF_DIR_LEVEL
;
1454 sbi
->interval_time
[CP_TIME
] = DEF_CP_INTERVAL
;
1455 sbi
->interval_time
[REQ_TIME
] = DEF_IDLE_INTERVAL
;
1456 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
1458 INIT_LIST_HEAD(&sbi
->s_list
);
1459 mutex_init(&sbi
->umount_mutex
);
1460 mutex_init(&sbi
->wio_mutex
[NODE
]);
1461 mutex_init(&sbi
->wio_mutex
[DATA
]);
1463 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1464 memcpy(sbi
->key_prefix
, F2FS_KEY_DESC_PREFIX
,
1465 F2FS_KEY_DESC_PREFIX_SIZE
);
1466 sbi
->key_prefix_size
= F2FS_KEY_DESC_PREFIX_SIZE
;
1470 static int init_percpu_info(struct f2fs_sb_info
*sbi
)
1474 for (i
= 0; i
< NR_COUNT_TYPE
; i
++) {
1475 err
= percpu_counter_init(&sbi
->nr_pages
[i
], 0, GFP_KERNEL
);
1480 err
= percpu_counter_init(&sbi
->alloc_valid_block_count
, 0, GFP_KERNEL
);
1484 return percpu_counter_init(&sbi
->total_valid_inode_count
, 0,
1489 * Read f2fs raw super block.
1490 * Because we have two copies of super block, so read both of them
1491 * to get the first valid one. If any one of them is broken, we pass
1492 * them recovery flag back to the caller.
1494 static int read_raw_super_block(struct f2fs_sb_info
*sbi
,
1495 struct f2fs_super_block
**raw_super
,
1496 int *valid_super_block
, int *recovery
)
1498 struct super_block
*sb
= sbi
->sb
;
1500 struct buffer_head
*bh
;
1501 struct f2fs_super_block
*super
;
1504 super
= kzalloc(sizeof(struct f2fs_super_block
), GFP_KERNEL
);
1508 for (block
= 0; block
< 2; block
++) {
1509 bh
= sb_bread(sb
, block
);
1511 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1517 /* sanity checking of raw super */
1518 if (sanity_check_raw_super(sbi
, bh
)) {
1519 f2fs_msg(sb
, KERN_ERR
,
1520 "Can't find valid F2FS filesystem in %dth superblock",
1528 memcpy(super
, bh
->b_data
+ F2FS_SUPER_OFFSET
,
1530 *valid_super_block
= block
;
1536 /* Fail to read any one of the superblocks*/
1540 /* No valid superblock */
1549 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1551 struct buffer_head
*bh
;
1554 if ((recover
&& f2fs_readonly(sbi
->sb
)) ||
1555 bdev_read_only(sbi
->sb
->s_bdev
)) {
1556 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1560 /* write back-up superblock first */
1561 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
? 0: 1);
1564 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1567 /* if we are in recovery path, skip writing valid superblock */
1571 /* write current valid superblock */
1572 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
);
1575 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1580 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1582 struct f2fs_sb_info
*sbi
;
1583 struct f2fs_super_block
*raw_super
;
1586 bool retry
= true, need_fsck
= false;
1587 char *options
= NULL
;
1588 int recovery
, i
, valid_super_block
;
1589 struct curseg_info
*seg_i
;
1594 valid_super_block
= -1;
1597 /* allocate memory for f2fs-specific super block info */
1598 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1604 /* Load the checksum driver */
1605 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32", 0, 0);
1606 if (IS_ERR(sbi
->s_chksum_driver
)) {
1607 f2fs_msg(sb
, KERN_ERR
, "Cannot load crc32 driver.");
1608 err
= PTR_ERR(sbi
->s_chksum_driver
);
1609 sbi
->s_chksum_driver
= NULL
;
1613 /* set a block size */
1614 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1615 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1619 err
= read_raw_super_block(sbi
, &raw_super
, &valid_super_block
,
1624 sb
->s_fs_info
= sbi
;
1625 sbi
->raw_super
= raw_super
;
1627 default_options(sbi
);
1628 /* parse mount options */
1629 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1630 if (data
&& !options
) {
1635 err
= parse_options(sb
, options
);
1639 sbi
->max_file_blocks
= max_file_blocks();
1640 sb
->s_maxbytes
= sbi
->max_file_blocks
<<
1641 le32_to_cpu(raw_super
->log_blocksize
);
1642 sb
->s_max_links
= F2FS_LINK_MAX
;
1643 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1645 sb
->s_op
= &f2fs_sops
;
1646 sb
->s_cop
= &f2fs_cryptops
;
1647 sb
->s_xattr
= f2fs_xattr_handlers
;
1648 sb
->s_export_op
= &f2fs_export_ops
;
1649 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1650 sb
->s_time_gran
= 1;
1651 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1652 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1653 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1655 /* init f2fs-specific super block info */
1656 sbi
->valid_super_block
= valid_super_block
;
1657 mutex_init(&sbi
->gc_mutex
);
1658 mutex_init(&sbi
->cp_mutex
);
1659 init_rwsem(&sbi
->node_write
);
1661 /* disallow all the data/node/meta page writes */
1662 set_sbi_flag(sbi
, SBI_POR_DOING
);
1663 spin_lock_init(&sbi
->stat_lock
);
1665 init_rwsem(&sbi
->read_io
.io_rwsem
);
1666 sbi
->read_io
.sbi
= sbi
;
1667 sbi
->read_io
.bio
= NULL
;
1668 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1669 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1670 sbi
->write_io
[i
].sbi
= sbi
;
1671 sbi
->write_io
[i
].bio
= NULL
;
1674 init_rwsem(&sbi
->cp_rwsem
);
1675 init_waitqueue_head(&sbi
->cp_wait
);
1678 err
= init_percpu_info(sbi
);
1682 /* get an inode for meta space */
1683 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1684 if (IS_ERR(sbi
->meta_inode
)) {
1685 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1686 err
= PTR_ERR(sbi
->meta_inode
);
1690 err
= get_valid_checkpoint(sbi
);
1692 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1693 goto free_meta_inode
;
1696 sbi
->total_valid_node_count
=
1697 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1698 percpu_counter_set(&sbi
->total_valid_inode_count
,
1699 le32_to_cpu(sbi
->ckpt
->valid_inode_count
));
1700 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1701 sbi
->total_valid_block_count
=
1702 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1703 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1705 for (i
= 0; i
< NR_INODE_TYPE
; i
++) {
1706 INIT_LIST_HEAD(&sbi
->inode_list
[i
]);
1707 spin_lock_init(&sbi
->inode_lock
[i
]);
1710 init_extent_cache_info(sbi
);
1712 init_ino_entry_info(sbi
);
1714 /* setup f2fs internal modules */
1715 err
= build_segment_manager(sbi
);
1717 f2fs_msg(sb
, KERN_ERR
,
1718 "Failed to initialize F2FS segment manager");
1721 err
= build_node_manager(sbi
);
1723 f2fs_msg(sb
, KERN_ERR
,
1724 "Failed to initialize F2FS node manager");
1728 /* For write statistics */
1729 if (sb
->s_bdev
->bd_part
)
1730 sbi
->sectors_written_start
=
1731 (u64
)part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
1733 /* Read accumulated write IO statistics if exists */
1734 seg_i
= CURSEG_I(sbi
, CURSEG_HOT_NODE
);
1735 if (__exist_node_summaries(sbi
))
1736 sbi
->kbytes_written
=
1737 le64_to_cpu(seg_i
->journal
->info
.kbytes_written
);
1739 build_gc_manager(sbi
);
1741 /* get an inode for node space */
1742 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1743 if (IS_ERR(sbi
->node_inode
)) {
1744 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1745 err
= PTR_ERR(sbi
->node_inode
);
1749 f2fs_join_shrinker(sbi
);
1751 /* if there are nt orphan nodes free them */
1752 err
= recover_orphan_inodes(sbi
);
1754 goto free_node_inode
;
1756 /* read root inode and dentry */
1757 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1759 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1760 err
= PTR_ERR(root
);
1761 goto free_node_inode
;
1763 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1766 goto free_node_inode
;
1769 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1772 goto free_root_inode
;
1775 err
= f2fs_build_stats(sbi
);
1777 goto free_root_inode
;
1780 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1783 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1784 &f2fs_seq_segment_info_fops
, sb
);
1785 proc_create_data("segment_bits", S_IRUGO
, sbi
->s_proc
,
1786 &f2fs_seq_segment_bits_fops
, sb
);
1789 sbi
->s_kobj
.kset
= f2fs_kset
;
1790 init_completion(&sbi
->s_kobj_unregister
);
1791 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1796 /* recover fsynced data */
1797 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1799 * mount should be failed, when device has readonly mode, and
1800 * previous checkpoint was not done by clean system shutdown.
1802 if (bdev_read_only(sb
->s_bdev
) &&
1803 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1809 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1811 err
= recover_fsync_data(sbi
, false);
1814 f2fs_msg(sb
, KERN_ERR
,
1815 "Cannot recover all fsync data errno=%d", err
);
1819 err
= recover_fsync_data(sbi
, true);
1821 if (!f2fs_readonly(sb
) && err
> 0) {
1823 f2fs_msg(sb
, KERN_ERR
,
1824 "Need to recover fsync data");
1829 /* recover_fsync_data() cleared this already */
1830 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1833 * If filesystem is not mounted as read-only then
1834 * do start the gc_thread.
1836 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1837 /* After POR, we can run background GC thread.*/
1838 err
= start_gc_thread(sbi
);
1844 /* recover broken superblock */
1846 err
= f2fs_commit_super(sbi
, true);
1847 f2fs_msg(sb
, KERN_INFO
,
1848 "Try to recover %dth superblock, ret: %d",
1849 sbi
->valid_super_block
? 1 : 2, err
);
1852 f2fs_update_time(sbi
, CP_TIME
);
1853 f2fs_update_time(sbi
, REQ_TIME
);
1857 f2fs_sync_inode_meta(sbi
);
1858 kobject_del(&sbi
->s_kobj
);
1859 kobject_put(&sbi
->s_kobj
);
1860 wait_for_completion(&sbi
->s_kobj_unregister
);
1863 remove_proc_entry("segment_info", sbi
->s_proc
);
1864 remove_proc_entry("segment_bits", sbi
->s_proc
);
1865 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1867 f2fs_destroy_stats(sbi
);
1872 mutex_lock(&sbi
->umount_mutex
);
1873 f2fs_leave_shrinker(sbi
);
1874 iput(sbi
->node_inode
);
1875 mutex_unlock(&sbi
->umount_mutex
);
1877 destroy_node_manager(sbi
);
1879 destroy_segment_manager(sbi
);
1882 make_bad_inode(sbi
->meta_inode
);
1883 iput(sbi
->meta_inode
);
1885 destroy_percpu_info(sbi
);
1890 if (sbi
->s_chksum_driver
)
1891 crypto_free_shash(sbi
->s_chksum_driver
);
1894 /* give only one another chance */
1897 shrink_dcache_sb(sb
);
1903 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1904 const char *dev_name
, void *data
)
1906 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1909 static void kill_f2fs_super(struct super_block
*sb
)
1912 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1913 kill_block_super(sb
);
1916 static struct file_system_type f2fs_fs_type
= {
1917 .owner
= THIS_MODULE
,
1919 .mount
= f2fs_mount
,
1920 .kill_sb
= kill_f2fs_super
,
1921 .fs_flags
= FS_REQUIRES_DEV
,
1923 MODULE_ALIAS_FS("f2fs");
1925 static int __init
init_inodecache(void)
1927 f2fs_inode_cachep
= kmem_cache_create("f2fs_inode_cache",
1928 sizeof(struct f2fs_inode_info
), 0,
1929 SLAB_RECLAIM_ACCOUNT
|SLAB_ACCOUNT
, NULL
);
1930 if (!f2fs_inode_cachep
)
1935 static void destroy_inodecache(void)
1938 * Make sure all delayed rcu free inodes are flushed before we
1942 kmem_cache_destroy(f2fs_inode_cachep
);
1945 static int __init
init_f2fs_fs(void)
1949 f2fs_build_trace_ios();
1951 err
= init_inodecache();
1954 err
= create_node_manager_caches();
1956 goto free_inodecache
;
1957 err
= create_segment_manager_caches();
1959 goto free_node_manager_caches
;
1960 err
= create_checkpoint_caches();
1962 goto free_segment_manager_caches
;
1963 err
= create_extent_cache();
1965 goto free_checkpoint_caches
;
1966 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1969 goto free_extent_cache
;
1971 #ifdef CONFIG_F2FS_FAULT_INJECTION
1972 f2fs_fault_inject
.kset
= f2fs_kset
;
1973 f2fs_build_fault_attr(0);
1974 err
= kobject_init_and_add(&f2fs_fault_inject
, &f2fs_fault_ktype
,
1975 NULL
, "fault_injection");
1977 f2fs_fault_inject
.kset
= NULL
;
1981 err
= register_shrinker(&f2fs_shrinker_info
);
1985 err
= register_filesystem(&f2fs_fs_type
);
1988 err
= f2fs_create_root_stats();
1990 goto free_filesystem
;
1991 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1995 unregister_filesystem(&f2fs_fs_type
);
1997 unregister_shrinker(&f2fs_shrinker_info
);
1999 #ifdef CONFIG_F2FS_FAULT_INJECTION
2000 if (f2fs_fault_inject
.kset
)
2001 kobject_put(&f2fs_fault_inject
);
2003 kset_unregister(f2fs_kset
);
2005 destroy_extent_cache();
2006 free_checkpoint_caches
:
2007 destroy_checkpoint_caches();
2008 free_segment_manager_caches
:
2009 destroy_segment_manager_caches();
2010 free_node_manager_caches
:
2011 destroy_node_manager_caches();
2013 destroy_inodecache();
2018 static void __exit
exit_f2fs_fs(void)
2020 remove_proc_entry("fs/f2fs", NULL
);
2021 f2fs_destroy_root_stats();
2022 unregister_filesystem(&f2fs_fs_type
);
2023 unregister_shrinker(&f2fs_shrinker_info
);
2024 #ifdef CONFIG_F2FS_FAULT_INJECTION
2025 kobject_put(&f2fs_fault_inject
);
2027 kset_unregister(f2fs_kset
);
2028 destroy_extent_cache();
2029 destroy_checkpoint_caches();
2030 destroy_segment_manager_caches();
2031 destroy_node_manager_caches();
2032 destroy_inodecache();
2033 f2fs_destroy_trace_ios();
2036 module_init(init_f2fs_fs
)
2037 module_exit(exit_f2fs_fs
)
2039 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2040 MODULE_DESCRIPTION("Flash Friendly File System");
2041 MODULE_LICENSE("GPL");