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 u32 f2fs_fault_rate
= 0;
46 char *fault_name
[FAULT_MAX
] = {
47 [FAULT_KMALLOC
] = "kmalloc",
48 [FAULT_PAGE_ALLOC
] = "page alloc",
52 /* f2fs-wide shrinker description */
53 static struct shrinker f2fs_shrinker_info
= {
54 .scan_objects
= f2fs_shrink_scan
,
55 .count_objects
= f2fs_shrink_count
,
56 .seeks
= DEFAULT_SEEKS
,
61 Opt_disable_roll_forward
,
70 Opt_disable_ext_identify
,
85 static match_table_t f2fs_tokens
= {
86 {Opt_gc_background
, "background_gc=%s"},
87 {Opt_disable_roll_forward
, "disable_roll_forward"},
88 {Opt_norecovery
, "norecovery"},
89 {Opt_discard
, "discard"},
90 {Opt_noheap
, "no_heap"},
91 {Opt_user_xattr
, "user_xattr"},
92 {Opt_nouser_xattr
, "nouser_xattr"},
95 {Opt_active_logs
, "active_logs=%u"},
96 {Opt_disable_ext_identify
, "disable_ext_identify"},
97 {Opt_inline_xattr
, "inline_xattr"},
98 {Opt_inline_data
, "inline_data"},
99 {Opt_inline_dentry
, "inline_dentry"},
100 {Opt_flush_merge
, "flush_merge"},
101 {Opt_nobarrier
, "nobarrier"},
102 {Opt_fastboot
, "fastboot"},
103 {Opt_extent_cache
, "extent_cache"},
104 {Opt_noextent_cache
, "noextent_cache"},
105 {Opt_noinline_data
, "noinline_data"},
106 {Opt_data_flush
, "data_flush"},
107 {Opt_fault_injection
, "fault_injection=%u"},
111 /* Sysfs support for f2fs */
113 GC_THREAD
, /* struct f2fs_gc_thread */
114 SM_INFO
, /* struct f2fs_sm_info */
115 NM_INFO
, /* struct f2fs_nm_info */
116 F2FS_SBI
, /* struct f2fs_sb_info */
120 struct attribute attr
;
121 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
122 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
123 const char *, size_t);
128 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
130 if (struct_type
== GC_THREAD
)
131 return (unsigned char *)sbi
->gc_thread
;
132 else if (struct_type
== SM_INFO
)
133 return (unsigned char *)SM_I(sbi
);
134 else if (struct_type
== NM_INFO
)
135 return (unsigned char *)NM_I(sbi
);
136 else if (struct_type
== F2FS_SBI
)
137 return (unsigned char *)sbi
;
141 static ssize_t
lifetime_write_kbytes_show(struct f2fs_attr
*a
,
142 struct f2fs_sb_info
*sbi
, char *buf
)
144 struct super_block
*sb
= sbi
->sb
;
146 if (!sb
->s_bdev
->bd_part
)
147 return snprintf(buf
, PAGE_SIZE
, "0\n");
149 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
150 (unsigned long long)(sbi
->kbytes_written
+
151 BD_PART_WRITTEN(sbi
)));
154 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
155 struct f2fs_sb_info
*sbi
, char *buf
)
157 unsigned char *ptr
= NULL
;
160 ptr
= __struct_ptr(sbi
, a
->struct_type
);
164 ui
= (unsigned int *)(ptr
+ a
->offset
);
166 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
169 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
170 struct f2fs_sb_info
*sbi
,
171 const char *buf
, size_t count
)
178 ptr
= __struct_ptr(sbi
, a
->struct_type
);
182 ui
= (unsigned int *)(ptr
+ a
->offset
);
184 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
191 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
192 struct attribute
*attr
, char *buf
)
194 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
196 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
198 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
201 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
202 const char *buf
, size_t len
)
204 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
206 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
208 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
211 static void f2fs_sb_release(struct kobject
*kobj
)
213 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
215 complete(&sbi
->s_kobj_unregister
);
218 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
219 static struct f2fs_attr f2fs_attr_##_name = { \
220 .attr = {.name = __stringify(_name), .mode = _mode }, \
223 .struct_type = _struct_type, \
227 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
228 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
229 f2fs_sbi_show, f2fs_sbi_store, \
230 offsetof(struct struct_name, elname))
232 #define F2FS_GENERAL_RO_ATTR(name) \
233 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
235 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
236 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
237 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
238 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
239 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
240 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
241 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, batched_trim_sections
, trim_sections
);
242 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
243 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
244 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
245 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
246 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ra_nid_pages
, ra_nid_pages
);
247 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, dirty_nats_ratio
, dirty_nats_ratio
);
248 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
249 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
250 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, cp_interval
, interval_time
[CP_TIME
]);
251 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, idle_interval
, interval_time
[REQ_TIME
]);
252 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes
);
254 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
255 static struct attribute
*f2fs_attrs
[] = {
256 ATTR_LIST(gc_min_sleep_time
),
257 ATTR_LIST(gc_max_sleep_time
),
258 ATTR_LIST(gc_no_gc_sleep_time
),
260 ATTR_LIST(reclaim_segments
),
261 ATTR_LIST(max_small_discards
),
262 ATTR_LIST(batched_trim_sections
),
263 ATTR_LIST(ipu_policy
),
264 ATTR_LIST(min_ipu_util
),
265 ATTR_LIST(min_fsync_blocks
),
266 ATTR_LIST(max_victim_search
),
267 ATTR_LIST(dir_level
),
268 ATTR_LIST(ram_thresh
),
269 ATTR_LIST(ra_nid_pages
),
270 ATTR_LIST(dirty_nats_ratio
),
271 ATTR_LIST(cp_interval
),
272 ATTR_LIST(idle_interval
),
273 ATTR_LIST(lifetime_write_kbytes
),
277 static const struct sysfs_ops f2fs_attr_ops
= {
278 .show
= f2fs_attr_show
,
279 .store
= f2fs_attr_store
,
282 static struct kobj_type f2fs_ktype
= {
283 .default_attrs
= f2fs_attrs
,
284 .sysfs_ops
= &f2fs_attr_ops
,
285 .release
= f2fs_sb_release
,
288 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
290 struct va_format vaf
;
296 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
300 static void init_once(void *foo
)
302 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
304 inode_init_once(&fi
->vfs_inode
);
307 static int parse_options(struct super_block
*sb
, char *options
)
309 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
310 struct request_queue
*q
;
311 substring_t args
[MAX_OPT_ARGS
];
315 #ifdef CONFIG_F2FS_FAULT_INJECTION
321 while ((p
= strsep(&options
, ",")) != NULL
) {
326 * Initialize args struct so we know whether arg was
327 * found; some options take optional arguments.
329 args
[0].to
= args
[0].from
= NULL
;
330 token
= match_token(p
, f2fs_tokens
, args
);
333 case Opt_gc_background
:
334 name
= match_strdup(&args
[0]);
338 if (strlen(name
) == 2 && !strncmp(name
, "on", 2)) {
340 clear_opt(sbi
, FORCE_FG_GC
);
341 } else if (strlen(name
) == 3 && !strncmp(name
, "off", 3)) {
342 clear_opt(sbi
, BG_GC
);
343 clear_opt(sbi
, FORCE_FG_GC
);
344 } else if (strlen(name
) == 4 && !strncmp(name
, "sync", 4)) {
346 set_opt(sbi
, FORCE_FG_GC
);
353 case Opt_disable_roll_forward
:
354 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
357 /* this option mounts f2fs with ro */
358 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
359 if (!f2fs_readonly(sb
))
363 q
= bdev_get_queue(sb
->s_bdev
);
364 if (blk_queue_discard(q
)) {
365 set_opt(sbi
, DISCARD
);
367 f2fs_msg(sb
, KERN_WARNING
,
368 "mounting with \"discard\" option, but "
369 "the device does not support discard");
373 set_opt(sbi
, NOHEAP
);
375 #ifdef CONFIG_F2FS_FS_XATTR
377 set_opt(sbi
, XATTR_USER
);
379 case Opt_nouser_xattr
:
380 clear_opt(sbi
, XATTR_USER
);
382 case Opt_inline_xattr
:
383 set_opt(sbi
, INLINE_XATTR
);
387 f2fs_msg(sb
, KERN_INFO
,
388 "user_xattr options not supported");
390 case Opt_nouser_xattr
:
391 f2fs_msg(sb
, KERN_INFO
,
392 "nouser_xattr options not supported");
394 case Opt_inline_xattr
:
395 f2fs_msg(sb
, KERN_INFO
,
396 "inline_xattr options not supported");
399 #ifdef CONFIG_F2FS_FS_POSIX_ACL
401 set_opt(sbi
, POSIX_ACL
);
404 clear_opt(sbi
, POSIX_ACL
);
408 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
411 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
414 case Opt_active_logs
:
415 if (args
->from
&& match_int(args
, &arg
))
417 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
419 sbi
->active_logs
= arg
;
421 case Opt_disable_ext_identify
:
422 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
424 case Opt_inline_data
:
425 set_opt(sbi
, INLINE_DATA
);
427 case Opt_inline_dentry
:
428 set_opt(sbi
, INLINE_DENTRY
);
430 case Opt_flush_merge
:
431 set_opt(sbi
, FLUSH_MERGE
);
434 set_opt(sbi
, NOBARRIER
);
437 set_opt(sbi
, FASTBOOT
);
439 case Opt_extent_cache
:
440 set_opt(sbi
, EXTENT_CACHE
);
442 case Opt_noextent_cache
:
443 clear_opt(sbi
, EXTENT_CACHE
);
445 case Opt_noinline_data
:
446 clear_opt(sbi
, INLINE_DATA
);
449 set_opt(sbi
, DATA_FLUSH
);
451 case Opt_fault_injection
:
452 if (args
->from
&& match_int(args
, &arg
))
454 #ifdef CONFIG_F2FS_FAULT_INJECTION
455 f2fs_fault_rate
= arg
;
456 atomic_set(&f2fs_ops
, 0);
458 f2fs_msg(sb
, KERN_INFO
,
459 "FAULT_INJECTION was not selected");
463 f2fs_msg(sb
, KERN_ERR
,
464 "Unrecognized mount option \"%s\" or missing value",
472 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
474 struct f2fs_inode_info
*fi
;
476 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
480 init_once((void *) fi
);
482 /* Initialize f2fs-specific inode info */
483 fi
->vfs_inode
.i_version
= 1;
484 atomic_set(&fi
->dirty_pages
, 0);
485 fi
->i_current_depth
= 1;
487 init_rwsem(&fi
->i_sem
);
488 INIT_LIST_HEAD(&fi
->dirty_list
);
489 INIT_LIST_HEAD(&fi
->inmem_pages
);
490 mutex_init(&fi
->inmem_lock
);
492 set_inode_flag(fi
, FI_NEW_INODE
);
494 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
495 set_inode_flag(fi
, FI_INLINE_XATTR
);
497 /* Will be used by directory only */
498 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
499 return &fi
->vfs_inode
;
502 static int f2fs_drop_inode(struct inode
*inode
)
505 * This is to avoid a deadlock condition like below.
506 * writeback_single_inode(inode)
507 * - f2fs_write_data_page
508 * - f2fs_gc -> iput -> evict
509 * - inode_wait_for_writeback(inode)
511 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
) {
512 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
513 /* to avoid evict_inode call simultaneously */
514 atomic_inc(&inode
->i_count
);
515 spin_unlock(&inode
->i_lock
);
517 /* some remained atomic pages should discarded */
518 if (f2fs_is_atomic_file(inode
))
519 drop_inmem_pages(inode
);
521 /* should remain fi->extent_tree for writepage */
522 f2fs_destroy_extent_node(inode
);
524 sb_start_intwrite(inode
->i_sb
);
525 i_size_write(inode
, 0);
527 if (F2FS_HAS_BLOCKS(inode
))
528 f2fs_truncate(inode
, true);
530 sb_end_intwrite(inode
->i_sb
);
532 fscrypt_put_encryption_info(inode
, NULL
);
533 spin_lock(&inode
->i_lock
);
534 atomic_dec(&inode
->i_count
);
538 return generic_drop_inode(inode
);
542 * f2fs_dirty_inode() is called from __mark_inode_dirty()
544 * We should call set_dirty_inode to write the dirty inode through write_inode.
546 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
548 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
551 static void f2fs_i_callback(struct rcu_head
*head
)
553 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
554 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
557 static void f2fs_destroy_inode(struct inode
*inode
)
559 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
562 static void f2fs_put_super(struct super_block
*sb
)
564 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
567 remove_proc_entry("segment_info", sbi
->s_proc
);
568 remove_proc_entry("segment_bits", sbi
->s_proc
);
569 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
571 kobject_del(&sbi
->s_kobj
);
575 /* prevent remaining shrinker jobs */
576 mutex_lock(&sbi
->umount_mutex
);
579 * We don't need to do checkpoint when superblock is clean.
580 * But, the previous checkpoint was not done by umount, it needs to do
581 * clean checkpoint again.
583 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
584 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
585 struct cp_control cpc
= {
588 write_checkpoint(sbi
, &cpc
);
591 /* write_checkpoint can update stat informaion */
592 f2fs_destroy_stats(sbi
);
595 * normally superblock is clean, so we need to release this.
596 * In addition, EIO will skip do checkpoint, we need this as well.
598 release_ino_entry(sbi
);
599 release_discard_addrs(sbi
);
601 f2fs_leave_shrinker(sbi
);
602 mutex_unlock(&sbi
->umount_mutex
);
604 /* our cp_error case, we can wait for any writeback page */
605 if (get_pages(sbi
, F2FS_WRITEBACK
))
606 f2fs_flush_merged_bios(sbi
);
608 iput(sbi
->node_inode
);
609 iput(sbi
->meta_inode
);
611 /* destroy f2fs internal modules */
612 destroy_node_manager(sbi
);
613 destroy_segment_manager(sbi
);
616 kobject_put(&sbi
->s_kobj
);
617 wait_for_completion(&sbi
->s_kobj_unregister
);
619 sb
->s_fs_info
= NULL
;
620 if (sbi
->s_chksum_driver
)
621 crypto_free_shash(sbi
->s_chksum_driver
);
622 kfree(sbi
->raw_super
);
626 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
628 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
631 trace_f2fs_sync_fs(sb
, sync
);
634 struct cp_control cpc
;
636 cpc
.reason
= __get_cp_reason(sbi
);
638 mutex_lock(&sbi
->gc_mutex
);
639 err
= write_checkpoint(sbi
, &cpc
);
640 mutex_unlock(&sbi
->gc_mutex
);
642 f2fs_trace_ios(NULL
, 1);
647 static int f2fs_freeze(struct super_block
*sb
)
651 if (f2fs_readonly(sb
))
654 err
= f2fs_sync_fs(sb
, 1);
658 static int f2fs_unfreeze(struct super_block
*sb
)
663 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
665 struct super_block
*sb
= dentry
->d_sb
;
666 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
667 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
668 block_t total_count
, user_block_count
, start_count
, ovp_count
;
670 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
671 user_block_count
= sbi
->user_block_count
;
672 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
673 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
674 buf
->f_type
= F2FS_SUPER_MAGIC
;
675 buf
->f_bsize
= sbi
->blocksize
;
677 buf
->f_blocks
= total_count
- start_count
;
678 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
679 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
681 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
682 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
684 buf
->f_namelen
= F2FS_NAME_LEN
;
685 buf
->f_fsid
.val
[0] = (u32
)id
;
686 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
691 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
693 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
695 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
)) {
696 if (test_opt(sbi
, FORCE_FG_GC
))
697 seq_printf(seq
, ",background_gc=%s", "sync");
699 seq_printf(seq
, ",background_gc=%s", "on");
701 seq_printf(seq
, ",background_gc=%s", "off");
703 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
704 seq_puts(seq
, ",disable_roll_forward");
705 if (test_opt(sbi
, DISCARD
))
706 seq_puts(seq
, ",discard");
707 if (test_opt(sbi
, NOHEAP
))
708 seq_puts(seq
, ",no_heap_alloc");
709 #ifdef CONFIG_F2FS_FS_XATTR
710 if (test_opt(sbi
, XATTR_USER
))
711 seq_puts(seq
, ",user_xattr");
713 seq_puts(seq
, ",nouser_xattr");
714 if (test_opt(sbi
, INLINE_XATTR
))
715 seq_puts(seq
, ",inline_xattr");
717 #ifdef CONFIG_F2FS_FS_POSIX_ACL
718 if (test_opt(sbi
, POSIX_ACL
))
719 seq_puts(seq
, ",acl");
721 seq_puts(seq
, ",noacl");
723 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
724 seq_puts(seq
, ",disable_ext_identify");
725 if (test_opt(sbi
, INLINE_DATA
))
726 seq_puts(seq
, ",inline_data");
728 seq_puts(seq
, ",noinline_data");
729 if (test_opt(sbi
, INLINE_DENTRY
))
730 seq_puts(seq
, ",inline_dentry");
731 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
732 seq_puts(seq
, ",flush_merge");
733 if (test_opt(sbi
, NOBARRIER
))
734 seq_puts(seq
, ",nobarrier");
735 if (test_opt(sbi
, FASTBOOT
))
736 seq_puts(seq
, ",fastboot");
737 if (test_opt(sbi
, EXTENT_CACHE
))
738 seq_puts(seq
, ",extent_cache");
740 seq_puts(seq
, ",noextent_cache");
741 if (test_opt(sbi
, DATA_FLUSH
))
742 seq_puts(seq
, ",data_flush");
743 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
748 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
750 struct super_block
*sb
= seq
->private;
751 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
752 unsigned int total_segs
=
753 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
756 seq_puts(seq
, "format: segment_type|valid_blocks\n"
757 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
759 for (i
= 0; i
< total_segs
; i
++) {
760 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
763 seq_printf(seq
, "%-10d", i
);
764 seq_printf(seq
, "%d|%-3u", se
->type
,
765 get_valid_blocks(sbi
, i
, 1));
766 if ((i
% 10) == 9 || i
== (total_segs
- 1))
775 static int segment_bits_seq_show(struct seq_file
*seq
, void *offset
)
777 struct super_block
*sb
= seq
->private;
778 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
779 unsigned int total_segs
=
780 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
783 seq_puts(seq
, "format: segment_type|valid_blocks|bitmaps\n"
784 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
786 for (i
= 0; i
< total_segs
; i
++) {
787 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
789 seq_printf(seq
, "%-10d", i
);
790 seq_printf(seq
, "%d|%-3u|", se
->type
,
791 get_valid_blocks(sbi
, i
, 1));
792 for (j
= 0; j
< SIT_VBLOCK_MAP_SIZE
; j
++)
793 seq_printf(seq
, "%x ", se
->cur_valid_map
[j
]);
799 #define F2FS_PROC_FILE_DEF(_name) \
800 static int _name##_open_fs(struct inode *inode, struct file *file) \
802 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
805 static const struct file_operations f2fs_seq_##_name##_fops = { \
806 .owner = THIS_MODULE, \
807 .open = _name##_open_fs, \
809 .llseek = seq_lseek, \
810 .release = single_release, \
813 F2FS_PROC_FILE_DEF(segment_info
);
814 F2FS_PROC_FILE_DEF(segment_bits
);
816 static void default_options(struct f2fs_sb_info
*sbi
)
818 /* init some FS parameters */
819 sbi
->active_logs
= NR_CURSEG_TYPE
;
822 set_opt(sbi
, INLINE_DATA
);
823 set_opt(sbi
, EXTENT_CACHE
);
825 #ifdef CONFIG_F2FS_FS_XATTR
826 set_opt(sbi
, XATTR_USER
);
828 #ifdef CONFIG_F2FS_FS_POSIX_ACL
829 set_opt(sbi
, POSIX_ACL
);
833 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
835 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
836 struct f2fs_mount_info org_mount_opt
;
837 int err
, active_logs
;
838 bool need_restart_gc
= false;
839 bool need_stop_gc
= false;
840 bool no_extent_cache
= !test_opt(sbi
, EXTENT_CACHE
);
843 * Save the old mount options in case we
844 * need to restore them.
846 org_mount_opt
= sbi
->mount_opt
;
847 active_logs
= sbi
->active_logs
;
849 /* recover superblocks we couldn't write due to previous RO mount */
850 if (!(*flags
& MS_RDONLY
) && is_sbi_flag_set(sbi
, SBI_NEED_SB_WRITE
)) {
851 err
= f2fs_commit_super(sbi
, false);
852 f2fs_msg(sb
, KERN_INFO
,
853 "Try to recover all the superblocks, ret: %d", err
);
855 clear_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
858 sbi
->mount_opt
.opt
= 0;
859 default_options(sbi
);
861 /* parse mount options */
862 err
= parse_options(sb
, data
);
867 * Previous and new state of filesystem is RO,
868 * so skip checking GC and FLUSH_MERGE conditions.
870 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
873 /* disallow enable/disable extent_cache dynamically */
874 if (no_extent_cache
== !!test_opt(sbi
, EXTENT_CACHE
)) {
876 f2fs_msg(sbi
->sb
, KERN_WARNING
,
877 "switch extent_cache option is not allowed");
882 * We stop the GC thread if FS is mounted as RO
883 * or if background_gc = off is passed in mount
884 * option. Also sync the filesystem.
886 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
887 if (sbi
->gc_thread
) {
889 need_restart_gc
= true;
891 } else if (!sbi
->gc_thread
) {
892 err
= start_gc_thread(sbi
);
898 if (*flags
& MS_RDONLY
) {
899 writeback_inodes_sb(sb
, WB_REASON_SYNC
);
902 set_sbi_flag(sbi
, SBI_IS_DIRTY
);
903 set_sbi_flag(sbi
, SBI_IS_CLOSE
);
905 clear_sbi_flag(sbi
, SBI_IS_CLOSE
);
909 * We stop issue flush thread if FS is mounted as RO
910 * or if flush_merge is not passed in mount option.
912 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
913 destroy_flush_cmd_control(sbi
);
914 } else if (!SM_I(sbi
)->cmd_control_info
) {
915 err
= create_flush_cmd_control(sbi
);
920 /* Update the POSIXACL Flag */
921 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
922 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
926 if (need_restart_gc
) {
927 if (start_gc_thread(sbi
))
928 f2fs_msg(sbi
->sb
, KERN_WARNING
,
929 "background gc thread has stopped");
930 } else if (need_stop_gc
) {
934 sbi
->mount_opt
= org_mount_opt
;
935 sbi
->active_logs
= active_logs
;
939 static struct super_operations f2fs_sops
= {
940 .alloc_inode
= f2fs_alloc_inode
,
941 .drop_inode
= f2fs_drop_inode
,
942 .destroy_inode
= f2fs_destroy_inode
,
943 .write_inode
= f2fs_write_inode
,
944 .dirty_inode
= f2fs_dirty_inode
,
945 .show_options
= f2fs_show_options
,
946 .evict_inode
= f2fs_evict_inode
,
947 .put_super
= f2fs_put_super
,
948 .sync_fs
= f2fs_sync_fs
,
949 .freeze_fs
= f2fs_freeze
,
950 .unfreeze_fs
= f2fs_unfreeze
,
951 .statfs
= f2fs_statfs
,
952 .remount_fs
= f2fs_remount
,
955 #ifdef CONFIG_F2FS_FS_ENCRYPTION
956 static int f2fs_get_context(struct inode
*inode
, void *ctx
, size_t len
)
958 return f2fs_getxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
959 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
963 static int f2fs_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
966 return f2fs_setxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
967 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
968 ctx
, len
, fs_data
, XATTR_CREATE
);
971 static unsigned f2fs_max_namelen(struct inode
*inode
)
973 return S_ISLNK(inode
->i_mode
) ?
974 inode
->i_sb
->s_blocksize
: F2FS_NAME_LEN
;
977 static struct fscrypt_operations f2fs_cryptops
= {
978 .get_context
= f2fs_get_context
,
979 .set_context
= f2fs_set_context
,
980 .is_encrypted
= f2fs_encrypted_inode
,
981 .empty_dir
= f2fs_empty_dir
,
982 .max_namelen
= f2fs_max_namelen
,
985 static struct fscrypt_operations f2fs_cryptops
= {
986 .is_encrypted
= f2fs_encrypted_inode
,
990 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
991 u64 ino
, u32 generation
)
993 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
996 if (check_nid_range(sbi
, ino
))
997 return ERR_PTR(-ESTALE
);
1000 * f2fs_iget isn't quite right if the inode is currently unallocated!
1001 * However f2fs_iget currently does appropriate checks to handle stale
1002 * inodes so everything is OK.
1004 inode
= f2fs_iget(sb
, ino
);
1006 return ERR_CAST(inode
);
1007 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
1008 /* we didn't find the right inode.. */
1010 return ERR_PTR(-ESTALE
);
1015 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1016 int fh_len
, int fh_type
)
1018 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1019 f2fs_nfs_get_inode
);
1022 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1023 int fh_len
, int fh_type
)
1025 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1026 f2fs_nfs_get_inode
);
1029 static const struct export_operations f2fs_export_ops
= {
1030 .fh_to_dentry
= f2fs_fh_to_dentry
,
1031 .fh_to_parent
= f2fs_fh_to_parent
,
1032 .get_parent
= f2fs_get_parent
,
1035 static loff_t
max_file_blocks(void)
1037 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
1038 loff_t leaf_count
= ADDRS_PER_BLOCK
;
1040 /* two direct node blocks */
1041 result
+= (leaf_count
* 2);
1043 /* two indirect node blocks */
1044 leaf_count
*= NIDS_PER_BLOCK
;
1045 result
+= (leaf_count
* 2);
1047 /* one double indirect node block */
1048 leaf_count
*= NIDS_PER_BLOCK
;
1049 result
+= leaf_count
;
1054 static int __f2fs_commit_super(struct buffer_head
*bh
,
1055 struct f2fs_super_block
*super
)
1059 memcpy(bh
->b_data
+ F2FS_SUPER_OFFSET
, super
, sizeof(*super
));
1060 set_buffer_uptodate(bh
);
1061 set_buffer_dirty(bh
);
1064 /* it's rare case, we can do fua all the time */
1065 return __sync_dirty_buffer(bh
, WRITE_FLUSH_FUA
);
1068 static inline bool sanity_check_area_boundary(struct f2fs_sb_info
*sbi
,
1069 struct buffer_head
*bh
)
1071 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1072 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1073 struct super_block
*sb
= sbi
->sb
;
1074 u32 segment0_blkaddr
= le32_to_cpu(raw_super
->segment0_blkaddr
);
1075 u32 cp_blkaddr
= le32_to_cpu(raw_super
->cp_blkaddr
);
1076 u32 sit_blkaddr
= le32_to_cpu(raw_super
->sit_blkaddr
);
1077 u32 nat_blkaddr
= le32_to_cpu(raw_super
->nat_blkaddr
);
1078 u32 ssa_blkaddr
= le32_to_cpu(raw_super
->ssa_blkaddr
);
1079 u32 main_blkaddr
= le32_to_cpu(raw_super
->main_blkaddr
);
1080 u32 segment_count_ckpt
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1081 u32 segment_count_sit
= le32_to_cpu(raw_super
->segment_count_sit
);
1082 u32 segment_count_nat
= le32_to_cpu(raw_super
->segment_count_nat
);
1083 u32 segment_count_ssa
= le32_to_cpu(raw_super
->segment_count_ssa
);
1084 u32 segment_count_main
= le32_to_cpu(raw_super
->segment_count_main
);
1085 u32 segment_count
= le32_to_cpu(raw_super
->segment_count
);
1086 u32 log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1087 u64 main_end_blkaddr
= main_blkaddr
+
1088 (segment_count_main
<< log_blocks_per_seg
);
1089 u64 seg_end_blkaddr
= segment0_blkaddr
+
1090 (segment_count
<< log_blocks_per_seg
);
1092 if (segment0_blkaddr
!= cp_blkaddr
) {
1093 f2fs_msg(sb
, KERN_INFO
,
1094 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1095 segment0_blkaddr
, cp_blkaddr
);
1099 if (cp_blkaddr
+ (segment_count_ckpt
<< log_blocks_per_seg
) !=
1101 f2fs_msg(sb
, KERN_INFO
,
1102 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1103 cp_blkaddr
, sit_blkaddr
,
1104 segment_count_ckpt
<< log_blocks_per_seg
);
1108 if (sit_blkaddr
+ (segment_count_sit
<< log_blocks_per_seg
) !=
1110 f2fs_msg(sb
, KERN_INFO
,
1111 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1112 sit_blkaddr
, nat_blkaddr
,
1113 segment_count_sit
<< log_blocks_per_seg
);
1117 if (nat_blkaddr
+ (segment_count_nat
<< log_blocks_per_seg
) !=
1119 f2fs_msg(sb
, KERN_INFO
,
1120 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1121 nat_blkaddr
, ssa_blkaddr
,
1122 segment_count_nat
<< log_blocks_per_seg
);
1126 if (ssa_blkaddr
+ (segment_count_ssa
<< log_blocks_per_seg
) !=
1128 f2fs_msg(sb
, KERN_INFO
,
1129 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1130 ssa_blkaddr
, main_blkaddr
,
1131 segment_count_ssa
<< log_blocks_per_seg
);
1135 if (main_end_blkaddr
> seg_end_blkaddr
) {
1136 f2fs_msg(sb
, KERN_INFO
,
1137 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1140 (segment_count
<< log_blocks_per_seg
),
1141 segment_count_main
<< log_blocks_per_seg
);
1143 } else if (main_end_blkaddr
< seg_end_blkaddr
) {
1147 /* fix in-memory information all the time */
1148 raw_super
->segment_count
= cpu_to_le32((main_end_blkaddr
-
1149 segment0_blkaddr
) >> log_blocks_per_seg
);
1151 if (f2fs_readonly(sb
) || bdev_read_only(sb
->s_bdev
)) {
1152 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1155 err
= __f2fs_commit_super(bh
, NULL
);
1156 res
= err
? "failed" : "done";
1158 f2fs_msg(sb
, KERN_INFO
,
1159 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1162 (segment_count
<< log_blocks_per_seg
),
1163 segment_count_main
<< log_blocks_per_seg
);
1170 static int sanity_check_raw_super(struct f2fs_sb_info
*sbi
,
1171 struct buffer_head
*bh
)
1173 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1174 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1175 struct super_block
*sb
= sbi
->sb
;
1176 unsigned int blocksize
;
1178 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
1179 f2fs_msg(sb
, KERN_INFO
,
1180 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1181 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
1185 /* Currently, support only 4KB page cache size */
1186 if (F2FS_BLKSIZE
!= PAGE_SIZE
) {
1187 f2fs_msg(sb
, KERN_INFO
,
1188 "Invalid page_cache_size (%lu), supports only 4KB\n",
1193 /* Currently, support only 4KB block size */
1194 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
1195 if (blocksize
!= F2FS_BLKSIZE
) {
1196 f2fs_msg(sb
, KERN_INFO
,
1197 "Invalid blocksize (%u), supports only 4KB\n",
1202 /* check log blocks per segment */
1203 if (le32_to_cpu(raw_super
->log_blocks_per_seg
) != 9) {
1204 f2fs_msg(sb
, KERN_INFO
,
1205 "Invalid log blocks per segment (%u)\n",
1206 le32_to_cpu(raw_super
->log_blocks_per_seg
));
1210 /* Currently, support 512/1024/2048/4096 bytes sector size */
1211 if (le32_to_cpu(raw_super
->log_sectorsize
) >
1212 F2FS_MAX_LOG_SECTOR_SIZE
||
1213 le32_to_cpu(raw_super
->log_sectorsize
) <
1214 F2FS_MIN_LOG_SECTOR_SIZE
) {
1215 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
1216 le32_to_cpu(raw_super
->log_sectorsize
));
1219 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
1220 le32_to_cpu(raw_super
->log_sectorsize
) !=
1221 F2FS_MAX_LOG_SECTOR_SIZE
) {
1222 f2fs_msg(sb
, KERN_INFO
,
1223 "Invalid log sectors per block(%u) log sectorsize(%u)",
1224 le32_to_cpu(raw_super
->log_sectors_per_block
),
1225 le32_to_cpu(raw_super
->log_sectorsize
));
1229 /* check reserved ino info */
1230 if (le32_to_cpu(raw_super
->node_ino
) != 1 ||
1231 le32_to_cpu(raw_super
->meta_ino
) != 2 ||
1232 le32_to_cpu(raw_super
->root_ino
) != 3) {
1233 f2fs_msg(sb
, KERN_INFO
,
1234 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1235 le32_to_cpu(raw_super
->node_ino
),
1236 le32_to_cpu(raw_super
->meta_ino
),
1237 le32_to_cpu(raw_super
->root_ino
));
1241 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1242 if (sanity_check_area_boundary(sbi
, bh
))
1248 int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
1250 unsigned int total
, fsmeta
;
1251 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
1252 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
1254 total
= le32_to_cpu(raw_super
->segment_count
);
1255 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1256 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
1257 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
1258 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
1259 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
1261 if (unlikely(fsmeta
>= total
))
1264 if (unlikely(f2fs_cp_error(sbi
))) {
1265 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
1271 static void init_sb_info(struct f2fs_sb_info
*sbi
)
1273 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
1276 sbi
->log_sectors_per_block
=
1277 le32_to_cpu(raw_super
->log_sectors_per_block
);
1278 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
1279 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
1280 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1281 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
1282 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
1283 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
1284 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
1285 sbi
->total_node_count
=
1286 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
1287 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
1288 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
1289 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
1290 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
1291 sbi
->cur_victim_sec
= NULL_SECNO
;
1292 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
1294 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
1295 atomic_set(&sbi
->nr_pages
[i
], 0);
1297 sbi
->dir_level
= DEF_DIR_LEVEL
;
1298 sbi
->interval_time
[CP_TIME
] = DEF_CP_INTERVAL
;
1299 sbi
->interval_time
[REQ_TIME
] = DEF_IDLE_INTERVAL
;
1300 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
1302 INIT_LIST_HEAD(&sbi
->s_list
);
1303 mutex_init(&sbi
->umount_mutex
);
1307 * Read f2fs raw super block.
1308 * Because we have two copies of super block, so read both of them
1309 * to get the first valid one. If any one of them is broken, we pass
1310 * them recovery flag back to the caller.
1312 static int read_raw_super_block(struct f2fs_sb_info
*sbi
,
1313 struct f2fs_super_block
**raw_super
,
1314 int *valid_super_block
, int *recovery
)
1316 struct super_block
*sb
= sbi
->sb
;
1318 struct buffer_head
*bh
;
1319 struct f2fs_super_block
*super
;
1322 super
= kzalloc(sizeof(struct f2fs_super_block
), GFP_KERNEL
);
1326 for (block
= 0; block
< 2; block
++) {
1327 bh
= sb_bread(sb
, block
);
1329 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1335 /* sanity checking of raw super */
1336 if (sanity_check_raw_super(sbi
, bh
)) {
1337 f2fs_msg(sb
, KERN_ERR
,
1338 "Can't find valid F2FS filesystem in %dth superblock",
1346 memcpy(super
, bh
->b_data
+ F2FS_SUPER_OFFSET
,
1348 *valid_super_block
= block
;
1354 /* Fail to read any one of the superblocks*/
1358 /* No valid superblock */
1367 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1369 struct buffer_head
*bh
;
1372 if ((recover
&& f2fs_readonly(sbi
->sb
)) ||
1373 bdev_read_only(sbi
->sb
->s_bdev
)) {
1374 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1378 /* write back-up superblock first */
1379 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
? 0: 1);
1382 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1385 /* if we are in recovery path, skip writing valid superblock */
1389 /* write current valid superblock */
1390 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
);
1393 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1398 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1400 struct f2fs_sb_info
*sbi
;
1401 struct f2fs_super_block
*raw_super
;
1404 bool retry
= true, need_fsck
= false;
1405 char *options
= NULL
;
1406 int recovery
, i
, valid_super_block
;
1407 struct curseg_info
*seg_i
;
1412 valid_super_block
= -1;
1415 /* allocate memory for f2fs-specific super block info */
1416 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1422 /* Load the checksum driver */
1423 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32", 0, 0);
1424 if (IS_ERR(sbi
->s_chksum_driver
)) {
1425 f2fs_msg(sb
, KERN_ERR
, "Cannot load crc32 driver.");
1426 err
= PTR_ERR(sbi
->s_chksum_driver
);
1427 sbi
->s_chksum_driver
= NULL
;
1431 /* set a block size */
1432 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1433 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1437 err
= read_raw_super_block(sbi
, &raw_super
, &valid_super_block
,
1442 sb
->s_fs_info
= sbi
;
1443 default_options(sbi
);
1444 /* parse mount options */
1445 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1446 if (data
&& !options
) {
1451 err
= parse_options(sb
, options
);
1455 sbi
->max_file_blocks
= max_file_blocks();
1456 sb
->s_maxbytes
= sbi
->max_file_blocks
<<
1457 le32_to_cpu(raw_super
->log_blocksize
);
1458 sb
->s_max_links
= F2FS_LINK_MAX
;
1459 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1461 sb
->s_op
= &f2fs_sops
;
1462 sb
->s_cop
= &f2fs_cryptops
;
1463 sb
->s_xattr
= f2fs_xattr_handlers
;
1464 sb
->s_export_op
= &f2fs_export_ops
;
1465 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1466 sb
->s_time_gran
= 1;
1467 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1468 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1469 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1471 /* init f2fs-specific super block info */
1472 sbi
->raw_super
= raw_super
;
1473 sbi
->valid_super_block
= valid_super_block
;
1474 mutex_init(&sbi
->gc_mutex
);
1475 mutex_init(&sbi
->writepages
);
1476 mutex_init(&sbi
->cp_mutex
);
1477 init_rwsem(&sbi
->node_write
);
1479 /* disallow all the data/node/meta page writes */
1480 set_sbi_flag(sbi
, SBI_POR_DOING
);
1481 spin_lock_init(&sbi
->stat_lock
);
1483 init_rwsem(&sbi
->read_io
.io_rwsem
);
1484 sbi
->read_io
.sbi
= sbi
;
1485 sbi
->read_io
.bio
= NULL
;
1486 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1487 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1488 sbi
->write_io
[i
].sbi
= sbi
;
1489 sbi
->write_io
[i
].bio
= NULL
;
1492 init_rwsem(&sbi
->cp_rwsem
);
1493 init_waitqueue_head(&sbi
->cp_wait
);
1496 /* get an inode for meta space */
1497 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1498 if (IS_ERR(sbi
->meta_inode
)) {
1499 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1500 err
= PTR_ERR(sbi
->meta_inode
);
1504 err
= get_valid_checkpoint(sbi
);
1506 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1507 goto free_meta_inode
;
1510 sbi
->total_valid_node_count
=
1511 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1512 sbi
->total_valid_inode_count
=
1513 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1514 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1515 sbi
->total_valid_block_count
=
1516 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1517 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1518 sbi
->alloc_valid_block_count
= 0;
1519 for (i
= 0; i
< NR_INODE_TYPE
; i
++) {
1520 INIT_LIST_HEAD(&sbi
->inode_list
[i
]);
1521 spin_lock_init(&sbi
->inode_lock
[i
]);
1524 init_extent_cache_info(sbi
);
1526 init_ino_entry_info(sbi
);
1528 /* setup f2fs internal modules */
1529 err
= build_segment_manager(sbi
);
1531 f2fs_msg(sb
, KERN_ERR
,
1532 "Failed to initialize F2FS segment manager");
1535 err
= build_node_manager(sbi
);
1537 f2fs_msg(sb
, KERN_ERR
,
1538 "Failed to initialize F2FS node manager");
1542 /* For write statistics */
1543 if (sb
->s_bdev
->bd_part
)
1544 sbi
->sectors_written_start
=
1545 (u64
)part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
1547 /* Read accumulated write IO statistics if exists */
1548 seg_i
= CURSEG_I(sbi
, CURSEG_HOT_NODE
);
1549 if (__exist_node_summaries(sbi
))
1550 sbi
->kbytes_written
=
1551 le64_to_cpu(seg_i
->journal
->info
.kbytes_written
);
1553 build_gc_manager(sbi
);
1555 /* get an inode for node space */
1556 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1557 if (IS_ERR(sbi
->node_inode
)) {
1558 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1559 err
= PTR_ERR(sbi
->node_inode
);
1563 f2fs_join_shrinker(sbi
);
1565 /* if there are nt orphan nodes free them */
1566 err
= recover_orphan_inodes(sbi
);
1568 goto free_node_inode
;
1570 /* read root inode and dentry */
1571 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1573 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1574 err
= PTR_ERR(root
);
1575 goto free_node_inode
;
1577 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1580 goto free_node_inode
;
1583 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1586 goto free_root_inode
;
1589 err
= f2fs_build_stats(sbi
);
1591 goto free_root_inode
;
1594 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1597 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1598 &f2fs_seq_segment_info_fops
, sb
);
1599 proc_create_data("segment_bits", S_IRUGO
, sbi
->s_proc
,
1600 &f2fs_seq_segment_bits_fops
, sb
);
1603 sbi
->s_kobj
.kset
= f2fs_kset
;
1604 init_completion(&sbi
->s_kobj_unregister
);
1605 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1610 /* recover fsynced data */
1611 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1613 * mount should be failed, when device has readonly mode, and
1614 * previous checkpoint was not done by clean system shutdown.
1616 if (bdev_read_only(sb
->s_bdev
) &&
1617 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1623 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1625 err
= recover_fsync_data(sbi
, false);
1628 f2fs_msg(sb
, KERN_ERR
,
1629 "Cannot recover all fsync data errno=%ld", err
);
1633 err
= recover_fsync_data(sbi
, true);
1635 if (!f2fs_readonly(sb
) && err
> 0) {
1637 f2fs_msg(sb
, KERN_ERR
,
1638 "Need to recover fsync data");
1643 /* recover_fsync_data() cleared this already */
1644 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1647 * If filesystem is not mounted as read-only then
1648 * do start the gc_thread.
1650 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1651 /* After POR, we can run background GC thread.*/
1652 err
= start_gc_thread(sbi
);
1658 /* recover broken superblock */
1660 err
= f2fs_commit_super(sbi
, true);
1661 f2fs_msg(sb
, KERN_INFO
,
1662 "Try to recover %dth superblock, ret: %ld",
1663 sbi
->valid_super_block
? 1 : 2, err
);
1666 f2fs_update_time(sbi
, CP_TIME
);
1667 f2fs_update_time(sbi
, REQ_TIME
);
1671 kobject_del(&sbi
->s_kobj
);
1672 kobject_put(&sbi
->s_kobj
);
1673 wait_for_completion(&sbi
->s_kobj_unregister
);
1676 remove_proc_entry("segment_info", sbi
->s_proc
);
1677 remove_proc_entry("segment_bits", sbi
->s_proc
);
1678 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1680 f2fs_destroy_stats(sbi
);
1685 mutex_lock(&sbi
->umount_mutex
);
1686 f2fs_leave_shrinker(sbi
);
1687 iput(sbi
->node_inode
);
1688 mutex_unlock(&sbi
->umount_mutex
);
1690 destroy_node_manager(sbi
);
1692 destroy_segment_manager(sbi
);
1695 make_bad_inode(sbi
->meta_inode
);
1696 iput(sbi
->meta_inode
);
1702 if (sbi
->s_chksum_driver
)
1703 crypto_free_shash(sbi
->s_chksum_driver
);
1706 /* give only one another chance */
1709 shrink_dcache_sb(sb
);
1715 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1716 const char *dev_name
, void *data
)
1718 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1721 static void kill_f2fs_super(struct super_block
*sb
)
1724 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1725 kill_block_super(sb
);
1728 static struct file_system_type f2fs_fs_type
= {
1729 .owner
= THIS_MODULE
,
1731 .mount
= f2fs_mount
,
1732 .kill_sb
= kill_f2fs_super
,
1733 .fs_flags
= FS_REQUIRES_DEV
,
1735 MODULE_ALIAS_FS("f2fs");
1737 static int __init
init_inodecache(void)
1739 f2fs_inode_cachep
= kmem_cache_create("f2fs_inode_cache",
1740 sizeof(struct f2fs_inode_info
), 0,
1741 SLAB_RECLAIM_ACCOUNT
|SLAB_ACCOUNT
, NULL
);
1742 if (!f2fs_inode_cachep
)
1747 static void destroy_inodecache(void)
1750 * Make sure all delayed rcu free inodes are flushed before we
1754 kmem_cache_destroy(f2fs_inode_cachep
);
1757 static int __init
init_f2fs_fs(void)
1761 f2fs_build_trace_ios();
1763 err
= init_inodecache();
1766 err
= create_node_manager_caches();
1768 goto free_inodecache
;
1769 err
= create_segment_manager_caches();
1771 goto free_node_manager_caches
;
1772 err
= create_checkpoint_caches();
1774 goto free_segment_manager_caches
;
1775 err
= create_extent_cache();
1777 goto free_checkpoint_caches
;
1778 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1781 goto free_extent_cache
;
1783 err
= register_shrinker(&f2fs_shrinker_info
);
1787 err
= register_filesystem(&f2fs_fs_type
);
1790 err
= f2fs_create_root_stats();
1792 goto free_filesystem
;
1793 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1797 unregister_filesystem(&f2fs_fs_type
);
1799 unregister_shrinker(&f2fs_shrinker_info
);
1801 kset_unregister(f2fs_kset
);
1803 destroy_extent_cache();
1804 free_checkpoint_caches
:
1805 destroy_checkpoint_caches();
1806 free_segment_manager_caches
:
1807 destroy_segment_manager_caches();
1808 free_node_manager_caches
:
1809 destroy_node_manager_caches();
1811 destroy_inodecache();
1816 static void __exit
exit_f2fs_fs(void)
1818 remove_proc_entry("fs/f2fs", NULL
);
1819 f2fs_destroy_root_stats();
1820 unregister_shrinker(&f2fs_shrinker_info
);
1821 unregister_filesystem(&f2fs_fs_type
);
1822 destroy_extent_cache();
1823 destroy_checkpoint_caches();
1824 destroy_segment_manager_caches();
1825 destroy_node_manager_caches();
1826 destroy_inodecache();
1827 kset_unregister(f2fs_kset
);
1828 f2fs_destroy_trace_ios();
1831 module_init(init_f2fs_fs
)
1832 module_exit(exit_f2fs_fs
)
1834 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1835 MODULE_DESCRIPTION("Flash Friendly File System");
1836 MODULE_LICENSE("GPL");