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",
49 [FAULT_ALLOC_NID
] = "alloc nid",
50 [FAULT_ORPHAN
] = "orphan",
51 [FAULT_BLOCK
] = "no more block",
52 [FAULT_DIR_DEPTH
] = "too big dir depth",
56 /* f2fs-wide shrinker description */
57 static struct shrinker f2fs_shrinker_info
= {
58 .scan_objects
= f2fs_shrink_scan
,
59 .count_objects
= f2fs_shrink_count
,
60 .seeks
= DEFAULT_SEEKS
,
65 Opt_disable_roll_forward
,
74 Opt_disable_ext_identify
,
89 static match_table_t f2fs_tokens
= {
90 {Opt_gc_background
, "background_gc=%s"},
91 {Opt_disable_roll_forward
, "disable_roll_forward"},
92 {Opt_norecovery
, "norecovery"},
93 {Opt_discard
, "discard"},
94 {Opt_noheap
, "no_heap"},
95 {Opt_user_xattr
, "user_xattr"},
96 {Opt_nouser_xattr
, "nouser_xattr"},
99 {Opt_active_logs
, "active_logs=%u"},
100 {Opt_disable_ext_identify
, "disable_ext_identify"},
101 {Opt_inline_xattr
, "inline_xattr"},
102 {Opt_inline_data
, "inline_data"},
103 {Opt_inline_dentry
, "inline_dentry"},
104 {Opt_flush_merge
, "flush_merge"},
105 {Opt_nobarrier
, "nobarrier"},
106 {Opt_fastboot
, "fastboot"},
107 {Opt_extent_cache
, "extent_cache"},
108 {Opt_noextent_cache
, "noextent_cache"},
109 {Opt_noinline_data
, "noinline_data"},
110 {Opt_data_flush
, "data_flush"},
111 {Opt_fault_injection
, "fault_injection=%u"},
115 /* Sysfs support for f2fs */
117 GC_THREAD
, /* struct f2fs_gc_thread */
118 SM_INFO
, /* struct f2fs_sm_info */
119 NM_INFO
, /* struct f2fs_nm_info */
120 F2FS_SBI
, /* struct f2fs_sb_info */
124 struct attribute attr
;
125 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
126 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
127 const char *, size_t);
132 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
134 if (struct_type
== GC_THREAD
)
135 return (unsigned char *)sbi
->gc_thread
;
136 else if (struct_type
== SM_INFO
)
137 return (unsigned char *)SM_I(sbi
);
138 else if (struct_type
== NM_INFO
)
139 return (unsigned char *)NM_I(sbi
);
140 else if (struct_type
== F2FS_SBI
)
141 return (unsigned char *)sbi
;
145 static ssize_t
lifetime_write_kbytes_show(struct f2fs_attr
*a
,
146 struct f2fs_sb_info
*sbi
, char *buf
)
148 struct super_block
*sb
= sbi
->sb
;
150 if (!sb
->s_bdev
->bd_part
)
151 return snprintf(buf
, PAGE_SIZE
, "0\n");
153 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
154 (unsigned long long)(sbi
->kbytes_written
+
155 BD_PART_WRITTEN(sbi
)));
158 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
159 struct f2fs_sb_info
*sbi
, char *buf
)
161 unsigned char *ptr
= NULL
;
164 ptr
= __struct_ptr(sbi
, a
->struct_type
);
168 ui
= (unsigned int *)(ptr
+ a
->offset
);
170 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
173 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
174 struct f2fs_sb_info
*sbi
,
175 const char *buf
, size_t count
)
182 ptr
= __struct_ptr(sbi
, a
->struct_type
);
186 ui
= (unsigned int *)(ptr
+ a
->offset
);
188 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
195 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
196 struct attribute
*attr
, char *buf
)
198 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
200 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
202 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
205 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
206 const char *buf
, size_t len
)
208 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
210 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
212 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
215 static void f2fs_sb_release(struct kobject
*kobj
)
217 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
219 complete(&sbi
->s_kobj_unregister
);
222 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
223 static struct f2fs_attr f2fs_attr_##_name = { \
224 .attr = {.name = __stringify(_name), .mode = _mode }, \
227 .struct_type = _struct_type, \
231 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
232 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
233 f2fs_sbi_show, f2fs_sbi_store, \
234 offsetof(struct struct_name, elname))
236 #define F2FS_GENERAL_RO_ATTR(name) \
237 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL)
239 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
240 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
241 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
242 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
243 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
244 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
245 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, batched_trim_sections
, trim_sections
);
246 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
247 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
248 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
249 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
250 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ra_nid_pages
, ra_nid_pages
);
251 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, dirty_nats_ratio
, dirty_nats_ratio
);
252 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
253 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
254 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, cp_interval
, interval_time
[CP_TIME
]);
255 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, idle_interval
, interval_time
[REQ_TIME
]);
256 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes
);
258 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
259 static struct attribute
*f2fs_attrs
[] = {
260 ATTR_LIST(gc_min_sleep_time
),
261 ATTR_LIST(gc_max_sleep_time
),
262 ATTR_LIST(gc_no_gc_sleep_time
),
264 ATTR_LIST(reclaim_segments
),
265 ATTR_LIST(max_small_discards
),
266 ATTR_LIST(batched_trim_sections
),
267 ATTR_LIST(ipu_policy
),
268 ATTR_LIST(min_ipu_util
),
269 ATTR_LIST(min_fsync_blocks
),
270 ATTR_LIST(max_victim_search
),
271 ATTR_LIST(dir_level
),
272 ATTR_LIST(ram_thresh
),
273 ATTR_LIST(ra_nid_pages
),
274 ATTR_LIST(dirty_nats_ratio
),
275 ATTR_LIST(cp_interval
),
276 ATTR_LIST(idle_interval
),
277 ATTR_LIST(lifetime_write_kbytes
),
281 static const struct sysfs_ops f2fs_attr_ops
= {
282 .show
= f2fs_attr_show
,
283 .store
= f2fs_attr_store
,
286 static struct kobj_type f2fs_ktype
= {
287 .default_attrs
= f2fs_attrs
,
288 .sysfs_ops
= &f2fs_attr_ops
,
289 .release
= f2fs_sb_release
,
292 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
294 struct va_format vaf
;
300 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
304 static void init_once(void *foo
)
306 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
308 inode_init_once(&fi
->vfs_inode
);
311 static int parse_options(struct super_block
*sb
, char *options
)
313 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
314 struct request_queue
*q
;
315 substring_t args
[MAX_OPT_ARGS
];
319 #ifdef CONFIG_F2FS_FAULT_INJECTION
325 while ((p
= strsep(&options
, ",")) != NULL
) {
330 * Initialize args struct so we know whether arg was
331 * found; some options take optional arguments.
333 args
[0].to
= args
[0].from
= NULL
;
334 token
= match_token(p
, f2fs_tokens
, args
);
337 case Opt_gc_background
:
338 name
= match_strdup(&args
[0]);
342 if (strlen(name
) == 2 && !strncmp(name
, "on", 2)) {
344 clear_opt(sbi
, FORCE_FG_GC
);
345 } else if (strlen(name
) == 3 && !strncmp(name
, "off", 3)) {
346 clear_opt(sbi
, BG_GC
);
347 clear_opt(sbi
, FORCE_FG_GC
);
348 } else if (strlen(name
) == 4 && !strncmp(name
, "sync", 4)) {
350 set_opt(sbi
, FORCE_FG_GC
);
357 case Opt_disable_roll_forward
:
358 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
361 /* this option mounts f2fs with ro */
362 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
363 if (!f2fs_readonly(sb
))
367 q
= bdev_get_queue(sb
->s_bdev
);
368 if (blk_queue_discard(q
)) {
369 set_opt(sbi
, DISCARD
);
371 f2fs_msg(sb
, KERN_WARNING
,
372 "mounting with \"discard\" option, but "
373 "the device does not support discard");
377 set_opt(sbi
, NOHEAP
);
379 #ifdef CONFIG_F2FS_FS_XATTR
381 set_opt(sbi
, XATTR_USER
);
383 case Opt_nouser_xattr
:
384 clear_opt(sbi
, XATTR_USER
);
386 case Opt_inline_xattr
:
387 set_opt(sbi
, INLINE_XATTR
);
391 f2fs_msg(sb
, KERN_INFO
,
392 "user_xattr options not supported");
394 case Opt_nouser_xattr
:
395 f2fs_msg(sb
, KERN_INFO
,
396 "nouser_xattr options not supported");
398 case Opt_inline_xattr
:
399 f2fs_msg(sb
, KERN_INFO
,
400 "inline_xattr options not supported");
403 #ifdef CONFIG_F2FS_FS_POSIX_ACL
405 set_opt(sbi
, POSIX_ACL
);
408 clear_opt(sbi
, POSIX_ACL
);
412 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
415 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
418 case Opt_active_logs
:
419 if (args
->from
&& match_int(args
, &arg
))
421 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
423 sbi
->active_logs
= arg
;
425 case Opt_disable_ext_identify
:
426 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
428 case Opt_inline_data
:
429 set_opt(sbi
, INLINE_DATA
);
431 case Opt_inline_dentry
:
432 set_opt(sbi
, INLINE_DENTRY
);
434 case Opt_flush_merge
:
435 set_opt(sbi
, FLUSH_MERGE
);
438 set_opt(sbi
, NOBARRIER
);
441 set_opt(sbi
, FASTBOOT
);
443 case Opt_extent_cache
:
444 set_opt(sbi
, EXTENT_CACHE
);
446 case Opt_noextent_cache
:
447 clear_opt(sbi
, EXTENT_CACHE
);
449 case Opt_noinline_data
:
450 clear_opt(sbi
, INLINE_DATA
);
453 set_opt(sbi
, DATA_FLUSH
);
455 case Opt_fault_injection
:
456 if (args
->from
&& match_int(args
, &arg
))
458 #ifdef CONFIG_F2FS_FAULT_INJECTION
459 f2fs_fault_rate
= arg
;
460 atomic_set(&f2fs_ops
, 0);
462 f2fs_msg(sb
, KERN_INFO
,
463 "FAULT_INJECTION was not selected");
467 f2fs_msg(sb
, KERN_ERR
,
468 "Unrecognized mount option \"%s\" or missing value",
476 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
478 struct f2fs_inode_info
*fi
;
480 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
484 init_once((void *) fi
);
486 /* Initialize f2fs-specific inode info */
487 fi
->vfs_inode
.i_version
= 1;
488 atomic_set(&fi
->dirty_pages
, 0);
489 fi
->i_current_depth
= 1;
491 init_rwsem(&fi
->i_sem
);
492 INIT_LIST_HEAD(&fi
->dirty_list
);
493 INIT_LIST_HEAD(&fi
->inmem_pages
);
494 mutex_init(&fi
->inmem_lock
);
496 set_inode_flag(fi
, FI_NEW_INODE
);
498 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
499 set_inode_flag(fi
, FI_INLINE_XATTR
);
501 /* Will be used by directory only */
502 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
503 return &fi
->vfs_inode
;
506 static int f2fs_drop_inode(struct inode
*inode
)
509 * This is to avoid a deadlock condition like below.
510 * writeback_single_inode(inode)
511 * - f2fs_write_data_page
512 * - f2fs_gc -> iput -> evict
513 * - inode_wait_for_writeback(inode)
515 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
) {
516 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
517 /* to avoid evict_inode call simultaneously */
518 atomic_inc(&inode
->i_count
);
519 spin_unlock(&inode
->i_lock
);
521 /* some remained atomic pages should discarded */
522 if (f2fs_is_atomic_file(inode
))
523 drop_inmem_pages(inode
);
525 /* should remain fi->extent_tree for writepage */
526 f2fs_destroy_extent_node(inode
);
528 sb_start_intwrite(inode
->i_sb
);
529 i_size_write(inode
, 0);
531 if (F2FS_HAS_BLOCKS(inode
))
532 f2fs_truncate(inode
, true);
534 sb_end_intwrite(inode
->i_sb
);
536 fscrypt_put_encryption_info(inode
, NULL
);
537 spin_lock(&inode
->i_lock
);
538 atomic_dec(&inode
->i_count
);
542 return generic_drop_inode(inode
);
546 * f2fs_dirty_inode() is called from __mark_inode_dirty()
548 * We should call set_dirty_inode to write the dirty inode through write_inode.
550 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
552 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
555 static void f2fs_i_callback(struct rcu_head
*head
)
557 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
558 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
561 static void f2fs_destroy_inode(struct inode
*inode
)
563 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
566 static void f2fs_put_super(struct super_block
*sb
)
568 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
571 remove_proc_entry("segment_info", sbi
->s_proc
);
572 remove_proc_entry("segment_bits", sbi
->s_proc
);
573 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
575 kobject_del(&sbi
->s_kobj
);
579 /* prevent remaining shrinker jobs */
580 mutex_lock(&sbi
->umount_mutex
);
583 * We don't need to do checkpoint when superblock is clean.
584 * But, the previous checkpoint was not done by umount, it needs to do
585 * clean checkpoint again.
587 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
588 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
589 struct cp_control cpc
= {
592 write_checkpoint(sbi
, &cpc
);
595 /* write_checkpoint can update stat informaion */
596 f2fs_destroy_stats(sbi
);
599 * normally superblock is clean, so we need to release this.
600 * In addition, EIO will skip do checkpoint, we need this as well.
602 release_ino_entry(sbi
);
603 release_discard_addrs(sbi
);
605 f2fs_leave_shrinker(sbi
);
606 mutex_unlock(&sbi
->umount_mutex
);
608 /* our cp_error case, we can wait for any writeback page */
609 if (get_pages(sbi
, F2FS_WRITEBACK
))
610 f2fs_flush_merged_bios(sbi
);
612 iput(sbi
->node_inode
);
613 iput(sbi
->meta_inode
);
615 /* destroy f2fs internal modules */
616 destroy_node_manager(sbi
);
617 destroy_segment_manager(sbi
);
620 kobject_put(&sbi
->s_kobj
);
621 wait_for_completion(&sbi
->s_kobj_unregister
);
623 sb
->s_fs_info
= NULL
;
624 if (sbi
->s_chksum_driver
)
625 crypto_free_shash(sbi
->s_chksum_driver
);
626 kfree(sbi
->raw_super
);
630 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
632 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
635 trace_f2fs_sync_fs(sb
, sync
);
638 struct cp_control cpc
;
640 cpc
.reason
= __get_cp_reason(sbi
);
642 mutex_lock(&sbi
->gc_mutex
);
643 err
= write_checkpoint(sbi
, &cpc
);
644 mutex_unlock(&sbi
->gc_mutex
);
646 f2fs_trace_ios(NULL
, 1);
651 static int f2fs_freeze(struct super_block
*sb
)
655 if (f2fs_readonly(sb
))
658 err
= f2fs_sync_fs(sb
, 1);
662 static int f2fs_unfreeze(struct super_block
*sb
)
667 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
669 struct super_block
*sb
= dentry
->d_sb
;
670 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
671 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
672 block_t total_count
, user_block_count
, start_count
, ovp_count
;
674 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
675 user_block_count
= sbi
->user_block_count
;
676 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
677 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
678 buf
->f_type
= F2FS_SUPER_MAGIC
;
679 buf
->f_bsize
= sbi
->blocksize
;
681 buf
->f_blocks
= total_count
- start_count
;
682 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
683 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
685 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
686 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
688 buf
->f_namelen
= F2FS_NAME_LEN
;
689 buf
->f_fsid
.val
[0] = (u32
)id
;
690 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
695 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
697 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
699 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
)) {
700 if (test_opt(sbi
, FORCE_FG_GC
))
701 seq_printf(seq
, ",background_gc=%s", "sync");
703 seq_printf(seq
, ",background_gc=%s", "on");
705 seq_printf(seq
, ",background_gc=%s", "off");
707 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
708 seq_puts(seq
, ",disable_roll_forward");
709 if (test_opt(sbi
, DISCARD
))
710 seq_puts(seq
, ",discard");
711 if (test_opt(sbi
, NOHEAP
))
712 seq_puts(seq
, ",no_heap_alloc");
713 #ifdef CONFIG_F2FS_FS_XATTR
714 if (test_opt(sbi
, XATTR_USER
))
715 seq_puts(seq
, ",user_xattr");
717 seq_puts(seq
, ",nouser_xattr");
718 if (test_opt(sbi
, INLINE_XATTR
))
719 seq_puts(seq
, ",inline_xattr");
721 #ifdef CONFIG_F2FS_FS_POSIX_ACL
722 if (test_opt(sbi
, POSIX_ACL
))
723 seq_puts(seq
, ",acl");
725 seq_puts(seq
, ",noacl");
727 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
728 seq_puts(seq
, ",disable_ext_identify");
729 if (test_opt(sbi
, INLINE_DATA
))
730 seq_puts(seq
, ",inline_data");
732 seq_puts(seq
, ",noinline_data");
733 if (test_opt(sbi
, INLINE_DENTRY
))
734 seq_puts(seq
, ",inline_dentry");
735 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
736 seq_puts(seq
, ",flush_merge");
737 if (test_opt(sbi
, NOBARRIER
))
738 seq_puts(seq
, ",nobarrier");
739 if (test_opt(sbi
, FASTBOOT
))
740 seq_puts(seq
, ",fastboot");
741 if (test_opt(sbi
, EXTENT_CACHE
))
742 seq_puts(seq
, ",extent_cache");
744 seq_puts(seq
, ",noextent_cache");
745 if (test_opt(sbi
, DATA_FLUSH
))
746 seq_puts(seq
, ",data_flush");
747 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
752 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
754 struct super_block
*sb
= seq
->private;
755 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
756 unsigned int total_segs
=
757 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
760 seq_puts(seq
, "format: segment_type|valid_blocks\n"
761 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
763 for (i
= 0; i
< total_segs
; i
++) {
764 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
767 seq_printf(seq
, "%-10d", i
);
768 seq_printf(seq
, "%d|%-3u", se
->type
,
769 get_valid_blocks(sbi
, i
, 1));
770 if ((i
% 10) == 9 || i
== (total_segs
- 1))
779 static int segment_bits_seq_show(struct seq_file
*seq
, void *offset
)
781 struct super_block
*sb
= seq
->private;
782 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
783 unsigned int total_segs
=
784 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
787 seq_puts(seq
, "format: segment_type|valid_blocks|bitmaps\n"
788 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
790 for (i
= 0; i
< total_segs
; i
++) {
791 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
793 seq_printf(seq
, "%-10d", i
);
794 seq_printf(seq
, "%d|%-3u|", se
->type
,
795 get_valid_blocks(sbi
, i
, 1));
796 for (j
= 0; j
< SIT_VBLOCK_MAP_SIZE
; j
++)
797 seq_printf(seq
, "%x ", se
->cur_valid_map
[j
]);
803 #define F2FS_PROC_FILE_DEF(_name) \
804 static int _name##_open_fs(struct inode *inode, struct file *file) \
806 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \
809 static const struct file_operations f2fs_seq_##_name##_fops = { \
810 .owner = THIS_MODULE, \
811 .open = _name##_open_fs, \
813 .llseek = seq_lseek, \
814 .release = single_release, \
817 F2FS_PROC_FILE_DEF(segment_info
);
818 F2FS_PROC_FILE_DEF(segment_bits
);
820 static void default_options(struct f2fs_sb_info
*sbi
)
822 /* init some FS parameters */
823 sbi
->active_logs
= NR_CURSEG_TYPE
;
826 set_opt(sbi
, INLINE_DATA
);
827 set_opt(sbi
, EXTENT_CACHE
);
829 #ifdef CONFIG_F2FS_FS_XATTR
830 set_opt(sbi
, XATTR_USER
);
832 #ifdef CONFIG_F2FS_FS_POSIX_ACL
833 set_opt(sbi
, POSIX_ACL
);
837 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
839 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
840 struct f2fs_mount_info org_mount_opt
;
841 int err
, active_logs
;
842 bool need_restart_gc
= false;
843 bool need_stop_gc
= false;
844 bool no_extent_cache
= !test_opt(sbi
, EXTENT_CACHE
);
847 * Save the old mount options in case we
848 * need to restore them.
850 org_mount_opt
= sbi
->mount_opt
;
851 active_logs
= sbi
->active_logs
;
853 /* recover superblocks we couldn't write due to previous RO mount */
854 if (!(*flags
& MS_RDONLY
) && is_sbi_flag_set(sbi
, SBI_NEED_SB_WRITE
)) {
855 err
= f2fs_commit_super(sbi
, false);
856 f2fs_msg(sb
, KERN_INFO
,
857 "Try to recover all the superblocks, ret: %d", err
);
859 clear_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
862 sbi
->mount_opt
.opt
= 0;
863 default_options(sbi
);
865 /* parse mount options */
866 err
= parse_options(sb
, data
);
871 * Previous and new state of filesystem is RO,
872 * so skip checking GC and FLUSH_MERGE conditions.
874 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
877 /* disallow enable/disable extent_cache dynamically */
878 if (no_extent_cache
== !!test_opt(sbi
, EXTENT_CACHE
)) {
880 f2fs_msg(sbi
->sb
, KERN_WARNING
,
881 "switch extent_cache option is not allowed");
886 * We stop the GC thread if FS is mounted as RO
887 * or if background_gc = off is passed in mount
888 * option. Also sync the filesystem.
890 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
891 if (sbi
->gc_thread
) {
893 need_restart_gc
= true;
895 } else if (!sbi
->gc_thread
) {
896 err
= start_gc_thread(sbi
);
902 if (*flags
& MS_RDONLY
) {
903 writeback_inodes_sb(sb
, WB_REASON_SYNC
);
906 set_sbi_flag(sbi
, SBI_IS_DIRTY
);
907 set_sbi_flag(sbi
, SBI_IS_CLOSE
);
909 clear_sbi_flag(sbi
, SBI_IS_CLOSE
);
913 * We stop issue flush thread if FS is mounted as RO
914 * or if flush_merge is not passed in mount option.
916 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
917 destroy_flush_cmd_control(sbi
);
918 } else if (!SM_I(sbi
)->cmd_control_info
) {
919 err
= create_flush_cmd_control(sbi
);
924 /* Update the POSIXACL Flag */
925 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
926 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
930 if (need_restart_gc
) {
931 if (start_gc_thread(sbi
))
932 f2fs_msg(sbi
->sb
, KERN_WARNING
,
933 "background gc thread has stopped");
934 } else if (need_stop_gc
) {
938 sbi
->mount_opt
= org_mount_opt
;
939 sbi
->active_logs
= active_logs
;
943 static struct super_operations f2fs_sops
= {
944 .alloc_inode
= f2fs_alloc_inode
,
945 .drop_inode
= f2fs_drop_inode
,
946 .destroy_inode
= f2fs_destroy_inode
,
947 .write_inode
= f2fs_write_inode
,
948 .dirty_inode
= f2fs_dirty_inode
,
949 .show_options
= f2fs_show_options
,
950 .evict_inode
= f2fs_evict_inode
,
951 .put_super
= f2fs_put_super
,
952 .sync_fs
= f2fs_sync_fs
,
953 .freeze_fs
= f2fs_freeze
,
954 .unfreeze_fs
= f2fs_unfreeze
,
955 .statfs
= f2fs_statfs
,
956 .remount_fs
= f2fs_remount
,
959 #ifdef CONFIG_F2FS_FS_ENCRYPTION
960 static int f2fs_get_context(struct inode
*inode
, void *ctx
, size_t len
)
962 return f2fs_getxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
963 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
967 static int f2fs_set_context(struct inode
*inode
, const void *ctx
, size_t len
,
970 return f2fs_setxattr(inode
, F2FS_XATTR_INDEX_ENCRYPTION
,
971 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT
,
972 ctx
, len
, fs_data
, XATTR_CREATE
);
975 static unsigned f2fs_max_namelen(struct inode
*inode
)
977 return S_ISLNK(inode
->i_mode
) ?
978 inode
->i_sb
->s_blocksize
: F2FS_NAME_LEN
;
981 static struct fscrypt_operations f2fs_cryptops
= {
982 .get_context
= f2fs_get_context
,
983 .set_context
= f2fs_set_context
,
984 .is_encrypted
= f2fs_encrypted_inode
,
985 .empty_dir
= f2fs_empty_dir
,
986 .max_namelen
= f2fs_max_namelen
,
989 static struct fscrypt_operations f2fs_cryptops
= {
990 .is_encrypted
= f2fs_encrypted_inode
,
994 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
995 u64 ino
, u32 generation
)
997 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
1000 if (check_nid_range(sbi
, ino
))
1001 return ERR_PTR(-ESTALE
);
1004 * f2fs_iget isn't quite right if the inode is currently unallocated!
1005 * However f2fs_iget currently does appropriate checks to handle stale
1006 * inodes so everything is OK.
1008 inode
= f2fs_iget(sb
, ino
);
1010 return ERR_CAST(inode
);
1011 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
1012 /* we didn't find the right inode.. */
1014 return ERR_PTR(-ESTALE
);
1019 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1020 int fh_len
, int fh_type
)
1022 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1023 f2fs_nfs_get_inode
);
1026 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1027 int fh_len
, int fh_type
)
1029 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1030 f2fs_nfs_get_inode
);
1033 static const struct export_operations f2fs_export_ops
= {
1034 .fh_to_dentry
= f2fs_fh_to_dentry
,
1035 .fh_to_parent
= f2fs_fh_to_parent
,
1036 .get_parent
= f2fs_get_parent
,
1039 static loff_t
max_file_blocks(void)
1041 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
1042 loff_t leaf_count
= ADDRS_PER_BLOCK
;
1044 /* two direct node blocks */
1045 result
+= (leaf_count
* 2);
1047 /* two indirect node blocks */
1048 leaf_count
*= NIDS_PER_BLOCK
;
1049 result
+= (leaf_count
* 2);
1051 /* one double indirect node block */
1052 leaf_count
*= NIDS_PER_BLOCK
;
1053 result
+= leaf_count
;
1058 static int __f2fs_commit_super(struct buffer_head
*bh
,
1059 struct f2fs_super_block
*super
)
1063 memcpy(bh
->b_data
+ F2FS_SUPER_OFFSET
, super
, sizeof(*super
));
1064 set_buffer_uptodate(bh
);
1065 set_buffer_dirty(bh
);
1068 /* it's rare case, we can do fua all the time */
1069 return __sync_dirty_buffer(bh
, WRITE_FLUSH_FUA
);
1072 static inline bool sanity_check_area_boundary(struct f2fs_sb_info
*sbi
,
1073 struct buffer_head
*bh
)
1075 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1076 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1077 struct super_block
*sb
= sbi
->sb
;
1078 u32 segment0_blkaddr
= le32_to_cpu(raw_super
->segment0_blkaddr
);
1079 u32 cp_blkaddr
= le32_to_cpu(raw_super
->cp_blkaddr
);
1080 u32 sit_blkaddr
= le32_to_cpu(raw_super
->sit_blkaddr
);
1081 u32 nat_blkaddr
= le32_to_cpu(raw_super
->nat_blkaddr
);
1082 u32 ssa_blkaddr
= le32_to_cpu(raw_super
->ssa_blkaddr
);
1083 u32 main_blkaddr
= le32_to_cpu(raw_super
->main_blkaddr
);
1084 u32 segment_count_ckpt
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1085 u32 segment_count_sit
= le32_to_cpu(raw_super
->segment_count_sit
);
1086 u32 segment_count_nat
= le32_to_cpu(raw_super
->segment_count_nat
);
1087 u32 segment_count_ssa
= le32_to_cpu(raw_super
->segment_count_ssa
);
1088 u32 segment_count_main
= le32_to_cpu(raw_super
->segment_count_main
);
1089 u32 segment_count
= le32_to_cpu(raw_super
->segment_count
);
1090 u32 log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1091 u64 main_end_blkaddr
= main_blkaddr
+
1092 (segment_count_main
<< log_blocks_per_seg
);
1093 u64 seg_end_blkaddr
= segment0_blkaddr
+
1094 (segment_count
<< log_blocks_per_seg
);
1096 if (segment0_blkaddr
!= cp_blkaddr
) {
1097 f2fs_msg(sb
, KERN_INFO
,
1098 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1099 segment0_blkaddr
, cp_blkaddr
);
1103 if (cp_blkaddr
+ (segment_count_ckpt
<< log_blocks_per_seg
) !=
1105 f2fs_msg(sb
, KERN_INFO
,
1106 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1107 cp_blkaddr
, sit_blkaddr
,
1108 segment_count_ckpt
<< log_blocks_per_seg
);
1112 if (sit_blkaddr
+ (segment_count_sit
<< log_blocks_per_seg
) !=
1114 f2fs_msg(sb
, KERN_INFO
,
1115 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1116 sit_blkaddr
, nat_blkaddr
,
1117 segment_count_sit
<< log_blocks_per_seg
);
1121 if (nat_blkaddr
+ (segment_count_nat
<< log_blocks_per_seg
) !=
1123 f2fs_msg(sb
, KERN_INFO
,
1124 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1125 nat_blkaddr
, ssa_blkaddr
,
1126 segment_count_nat
<< log_blocks_per_seg
);
1130 if (ssa_blkaddr
+ (segment_count_ssa
<< log_blocks_per_seg
) !=
1132 f2fs_msg(sb
, KERN_INFO
,
1133 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1134 ssa_blkaddr
, main_blkaddr
,
1135 segment_count_ssa
<< log_blocks_per_seg
);
1139 if (main_end_blkaddr
> seg_end_blkaddr
) {
1140 f2fs_msg(sb
, KERN_INFO
,
1141 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1144 (segment_count
<< log_blocks_per_seg
),
1145 segment_count_main
<< log_blocks_per_seg
);
1147 } else if (main_end_blkaddr
< seg_end_blkaddr
) {
1151 /* fix in-memory information all the time */
1152 raw_super
->segment_count
= cpu_to_le32((main_end_blkaddr
-
1153 segment0_blkaddr
) >> log_blocks_per_seg
);
1155 if (f2fs_readonly(sb
) || bdev_read_only(sb
->s_bdev
)) {
1156 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1159 err
= __f2fs_commit_super(bh
, NULL
);
1160 res
= err
? "failed" : "done";
1162 f2fs_msg(sb
, KERN_INFO
,
1163 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1166 (segment_count
<< log_blocks_per_seg
),
1167 segment_count_main
<< log_blocks_per_seg
);
1174 static int sanity_check_raw_super(struct f2fs_sb_info
*sbi
,
1175 struct buffer_head
*bh
)
1177 struct f2fs_super_block
*raw_super
= (struct f2fs_super_block
*)
1178 (bh
->b_data
+ F2FS_SUPER_OFFSET
);
1179 struct super_block
*sb
= sbi
->sb
;
1180 unsigned int blocksize
;
1182 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
1183 f2fs_msg(sb
, KERN_INFO
,
1184 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1185 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
1189 /* Currently, support only 4KB page cache size */
1190 if (F2FS_BLKSIZE
!= PAGE_SIZE
) {
1191 f2fs_msg(sb
, KERN_INFO
,
1192 "Invalid page_cache_size (%lu), supports only 4KB\n",
1197 /* Currently, support only 4KB block size */
1198 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
1199 if (blocksize
!= F2FS_BLKSIZE
) {
1200 f2fs_msg(sb
, KERN_INFO
,
1201 "Invalid blocksize (%u), supports only 4KB\n",
1206 /* check log blocks per segment */
1207 if (le32_to_cpu(raw_super
->log_blocks_per_seg
) != 9) {
1208 f2fs_msg(sb
, KERN_INFO
,
1209 "Invalid log blocks per segment (%u)\n",
1210 le32_to_cpu(raw_super
->log_blocks_per_seg
));
1214 /* Currently, support 512/1024/2048/4096 bytes sector size */
1215 if (le32_to_cpu(raw_super
->log_sectorsize
) >
1216 F2FS_MAX_LOG_SECTOR_SIZE
||
1217 le32_to_cpu(raw_super
->log_sectorsize
) <
1218 F2FS_MIN_LOG_SECTOR_SIZE
) {
1219 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
1220 le32_to_cpu(raw_super
->log_sectorsize
));
1223 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
1224 le32_to_cpu(raw_super
->log_sectorsize
) !=
1225 F2FS_MAX_LOG_SECTOR_SIZE
) {
1226 f2fs_msg(sb
, KERN_INFO
,
1227 "Invalid log sectors per block(%u) log sectorsize(%u)",
1228 le32_to_cpu(raw_super
->log_sectors_per_block
),
1229 le32_to_cpu(raw_super
->log_sectorsize
));
1233 /* check reserved ino info */
1234 if (le32_to_cpu(raw_super
->node_ino
) != 1 ||
1235 le32_to_cpu(raw_super
->meta_ino
) != 2 ||
1236 le32_to_cpu(raw_super
->root_ino
) != 3) {
1237 f2fs_msg(sb
, KERN_INFO
,
1238 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1239 le32_to_cpu(raw_super
->node_ino
),
1240 le32_to_cpu(raw_super
->meta_ino
),
1241 le32_to_cpu(raw_super
->root_ino
));
1245 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1246 if (sanity_check_area_boundary(sbi
, bh
))
1252 int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
1254 unsigned int total
, fsmeta
;
1255 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
1256 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
1258 total
= le32_to_cpu(raw_super
->segment_count
);
1259 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
1260 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
1261 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
1262 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
1263 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
1265 if (unlikely(fsmeta
>= total
))
1268 if (unlikely(f2fs_cp_error(sbi
))) {
1269 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
1275 static void init_sb_info(struct f2fs_sb_info
*sbi
)
1277 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
1280 sbi
->log_sectors_per_block
=
1281 le32_to_cpu(raw_super
->log_sectors_per_block
);
1282 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
1283 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
1284 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
1285 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
1286 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
1287 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
1288 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
1289 sbi
->total_node_count
=
1290 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
1291 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
1292 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
1293 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
1294 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
1295 sbi
->cur_victim_sec
= NULL_SECNO
;
1296 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
1298 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
1299 atomic_set(&sbi
->nr_pages
[i
], 0);
1301 sbi
->dir_level
= DEF_DIR_LEVEL
;
1302 sbi
->interval_time
[CP_TIME
] = DEF_CP_INTERVAL
;
1303 sbi
->interval_time
[REQ_TIME
] = DEF_IDLE_INTERVAL
;
1304 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
1306 INIT_LIST_HEAD(&sbi
->s_list
);
1307 mutex_init(&sbi
->umount_mutex
);
1311 * Read f2fs raw super block.
1312 * Because we have two copies of super block, so read both of them
1313 * to get the first valid one. If any one of them is broken, we pass
1314 * them recovery flag back to the caller.
1316 static int read_raw_super_block(struct f2fs_sb_info
*sbi
,
1317 struct f2fs_super_block
**raw_super
,
1318 int *valid_super_block
, int *recovery
)
1320 struct super_block
*sb
= sbi
->sb
;
1322 struct buffer_head
*bh
;
1323 struct f2fs_super_block
*super
;
1326 super
= kzalloc(sizeof(struct f2fs_super_block
), GFP_KERNEL
);
1330 for (block
= 0; block
< 2; block
++) {
1331 bh
= sb_bread(sb
, block
);
1333 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
1339 /* sanity checking of raw super */
1340 if (sanity_check_raw_super(sbi
, bh
)) {
1341 f2fs_msg(sb
, KERN_ERR
,
1342 "Can't find valid F2FS filesystem in %dth superblock",
1350 memcpy(super
, bh
->b_data
+ F2FS_SUPER_OFFSET
,
1352 *valid_super_block
= block
;
1358 /* Fail to read any one of the superblocks*/
1362 /* No valid superblock */
1371 int f2fs_commit_super(struct f2fs_sb_info
*sbi
, bool recover
)
1373 struct buffer_head
*bh
;
1376 if ((recover
&& f2fs_readonly(sbi
->sb
)) ||
1377 bdev_read_only(sbi
->sb
->s_bdev
)) {
1378 set_sbi_flag(sbi
, SBI_NEED_SB_WRITE
);
1382 /* write back-up superblock first */
1383 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
? 0: 1);
1386 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1389 /* if we are in recovery path, skip writing valid superblock */
1393 /* write current valid superblock */
1394 bh
= sb_getblk(sbi
->sb
, sbi
->valid_super_block
);
1397 err
= __f2fs_commit_super(bh
, F2FS_RAW_SUPER(sbi
));
1402 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
1404 struct f2fs_sb_info
*sbi
;
1405 struct f2fs_super_block
*raw_super
;
1408 bool retry
= true, need_fsck
= false;
1409 char *options
= NULL
;
1410 int recovery
, i
, valid_super_block
;
1411 struct curseg_info
*seg_i
;
1416 valid_super_block
= -1;
1419 /* allocate memory for f2fs-specific super block info */
1420 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
1426 /* Load the checksum driver */
1427 sbi
->s_chksum_driver
= crypto_alloc_shash("crc32", 0, 0);
1428 if (IS_ERR(sbi
->s_chksum_driver
)) {
1429 f2fs_msg(sb
, KERN_ERR
, "Cannot load crc32 driver.");
1430 err
= PTR_ERR(sbi
->s_chksum_driver
);
1431 sbi
->s_chksum_driver
= NULL
;
1435 /* set a block size */
1436 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
1437 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
1441 err
= read_raw_super_block(sbi
, &raw_super
, &valid_super_block
,
1446 sb
->s_fs_info
= sbi
;
1447 default_options(sbi
);
1448 /* parse mount options */
1449 options
= kstrdup((const char *)data
, GFP_KERNEL
);
1450 if (data
&& !options
) {
1455 err
= parse_options(sb
, options
);
1459 sbi
->max_file_blocks
= max_file_blocks();
1460 sb
->s_maxbytes
= sbi
->max_file_blocks
<<
1461 le32_to_cpu(raw_super
->log_blocksize
);
1462 sb
->s_max_links
= F2FS_LINK_MAX
;
1463 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1465 sb
->s_op
= &f2fs_sops
;
1466 sb
->s_cop
= &f2fs_cryptops
;
1467 sb
->s_xattr
= f2fs_xattr_handlers
;
1468 sb
->s_export_op
= &f2fs_export_ops
;
1469 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1470 sb
->s_time_gran
= 1;
1471 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1472 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1473 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1475 /* init f2fs-specific super block info */
1476 sbi
->raw_super
= raw_super
;
1477 sbi
->valid_super_block
= valid_super_block
;
1478 mutex_init(&sbi
->gc_mutex
);
1479 mutex_init(&sbi
->writepages
);
1480 mutex_init(&sbi
->cp_mutex
);
1481 init_rwsem(&sbi
->node_write
);
1483 /* disallow all the data/node/meta page writes */
1484 set_sbi_flag(sbi
, SBI_POR_DOING
);
1485 spin_lock_init(&sbi
->stat_lock
);
1487 init_rwsem(&sbi
->read_io
.io_rwsem
);
1488 sbi
->read_io
.sbi
= sbi
;
1489 sbi
->read_io
.bio
= NULL
;
1490 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1491 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1492 sbi
->write_io
[i
].sbi
= sbi
;
1493 sbi
->write_io
[i
].bio
= NULL
;
1496 init_rwsem(&sbi
->cp_rwsem
);
1497 init_waitqueue_head(&sbi
->cp_wait
);
1500 /* get an inode for meta space */
1501 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1502 if (IS_ERR(sbi
->meta_inode
)) {
1503 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1504 err
= PTR_ERR(sbi
->meta_inode
);
1508 err
= get_valid_checkpoint(sbi
);
1510 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1511 goto free_meta_inode
;
1514 sbi
->total_valid_node_count
=
1515 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1516 sbi
->total_valid_inode_count
=
1517 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1518 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1519 sbi
->total_valid_block_count
=
1520 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1521 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1522 sbi
->alloc_valid_block_count
= 0;
1523 for (i
= 0; i
< NR_INODE_TYPE
; i
++) {
1524 INIT_LIST_HEAD(&sbi
->inode_list
[i
]);
1525 spin_lock_init(&sbi
->inode_lock
[i
]);
1528 init_extent_cache_info(sbi
);
1530 init_ino_entry_info(sbi
);
1532 /* setup f2fs internal modules */
1533 err
= build_segment_manager(sbi
);
1535 f2fs_msg(sb
, KERN_ERR
,
1536 "Failed to initialize F2FS segment manager");
1539 err
= build_node_manager(sbi
);
1541 f2fs_msg(sb
, KERN_ERR
,
1542 "Failed to initialize F2FS node manager");
1546 /* For write statistics */
1547 if (sb
->s_bdev
->bd_part
)
1548 sbi
->sectors_written_start
=
1549 (u64
)part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]);
1551 /* Read accumulated write IO statistics if exists */
1552 seg_i
= CURSEG_I(sbi
, CURSEG_HOT_NODE
);
1553 if (__exist_node_summaries(sbi
))
1554 sbi
->kbytes_written
=
1555 le64_to_cpu(seg_i
->journal
->info
.kbytes_written
);
1557 build_gc_manager(sbi
);
1559 /* get an inode for node space */
1560 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1561 if (IS_ERR(sbi
->node_inode
)) {
1562 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1563 err
= PTR_ERR(sbi
->node_inode
);
1567 f2fs_join_shrinker(sbi
);
1569 /* if there are nt orphan nodes free them */
1570 err
= recover_orphan_inodes(sbi
);
1572 goto free_node_inode
;
1574 /* read root inode and dentry */
1575 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1577 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1578 err
= PTR_ERR(root
);
1579 goto free_node_inode
;
1581 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1584 goto free_node_inode
;
1587 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1590 goto free_root_inode
;
1593 err
= f2fs_build_stats(sbi
);
1595 goto free_root_inode
;
1598 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1601 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1602 &f2fs_seq_segment_info_fops
, sb
);
1603 proc_create_data("segment_bits", S_IRUGO
, sbi
->s_proc
,
1604 &f2fs_seq_segment_bits_fops
, sb
);
1607 sbi
->s_kobj
.kset
= f2fs_kset
;
1608 init_completion(&sbi
->s_kobj_unregister
);
1609 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1614 /* recover fsynced data */
1615 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1617 * mount should be failed, when device has readonly mode, and
1618 * previous checkpoint was not done by clean system shutdown.
1620 if (bdev_read_only(sb
->s_bdev
) &&
1621 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1627 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1629 err
= recover_fsync_data(sbi
, false);
1632 f2fs_msg(sb
, KERN_ERR
,
1633 "Cannot recover all fsync data errno=%ld", err
);
1637 err
= recover_fsync_data(sbi
, true);
1639 if (!f2fs_readonly(sb
) && err
> 0) {
1641 f2fs_msg(sb
, KERN_ERR
,
1642 "Need to recover fsync data");
1647 /* recover_fsync_data() cleared this already */
1648 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1651 * If filesystem is not mounted as read-only then
1652 * do start the gc_thread.
1654 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1655 /* After POR, we can run background GC thread.*/
1656 err
= start_gc_thread(sbi
);
1662 /* recover broken superblock */
1664 err
= f2fs_commit_super(sbi
, true);
1665 f2fs_msg(sb
, KERN_INFO
,
1666 "Try to recover %dth superblock, ret: %ld",
1667 sbi
->valid_super_block
? 1 : 2, err
);
1670 f2fs_update_time(sbi
, CP_TIME
);
1671 f2fs_update_time(sbi
, REQ_TIME
);
1675 kobject_del(&sbi
->s_kobj
);
1676 kobject_put(&sbi
->s_kobj
);
1677 wait_for_completion(&sbi
->s_kobj_unregister
);
1680 remove_proc_entry("segment_info", sbi
->s_proc
);
1681 remove_proc_entry("segment_bits", sbi
->s_proc
);
1682 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1684 f2fs_destroy_stats(sbi
);
1689 mutex_lock(&sbi
->umount_mutex
);
1690 f2fs_leave_shrinker(sbi
);
1691 iput(sbi
->node_inode
);
1692 mutex_unlock(&sbi
->umount_mutex
);
1694 destroy_node_manager(sbi
);
1696 destroy_segment_manager(sbi
);
1699 make_bad_inode(sbi
->meta_inode
);
1700 iput(sbi
->meta_inode
);
1706 if (sbi
->s_chksum_driver
)
1707 crypto_free_shash(sbi
->s_chksum_driver
);
1710 /* give only one another chance */
1713 shrink_dcache_sb(sb
);
1719 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1720 const char *dev_name
, void *data
)
1722 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1725 static void kill_f2fs_super(struct super_block
*sb
)
1728 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1729 kill_block_super(sb
);
1732 static struct file_system_type f2fs_fs_type
= {
1733 .owner
= THIS_MODULE
,
1735 .mount
= f2fs_mount
,
1736 .kill_sb
= kill_f2fs_super
,
1737 .fs_flags
= FS_REQUIRES_DEV
,
1739 MODULE_ALIAS_FS("f2fs");
1741 static int __init
init_inodecache(void)
1743 f2fs_inode_cachep
= kmem_cache_create("f2fs_inode_cache",
1744 sizeof(struct f2fs_inode_info
), 0,
1745 SLAB_RECLAIM_ACCOUNT
|SLAB_ACCOUNT
, NULL
);
1746 if (!f2fs_inode_cachep
)
1751 static void destroy_inodecache(void)
1754 * Make sure all delayed rcu free inodes are flushed before we
1758 kmem_cache_destroy(f2fs_inode_cachep
);
1761 static int __init
init_f2fs_fs(void)
1765 f2fs_build_trace_ios();
1767 err
= init_inodecache();
1770 err
= create_node_manager_caches();
1772 goto free_inodecache
;
1773 err
= create_segment_manager_caches();
1775 goto free_node_manager_caches
;
1776 err
= create_checkpoint_caches();
1778 goto free_segment_manager_caches
;
1779 err
= create_extent_cache();
1781 goto free_checkpoint_caches
;
1782 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1785 goto free_extent_cache
;
1787 err
= register_shrinker(&f2fs_shrinker_info
);
1791 err
= register_filesystem(&f2fs_fs_type
);
1794 err
= f2fs_create_root_stats();
1796 goto free_filesystem
;
1797 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1801 unregister_filesystem(&f2fs_fs_type
);
1803 unregister_shrinker(&f2fs_shrinker_info
);
1805 kset_unregister(f2fs_kset
);
1807 destroy_extent_cache();
1808 free_checkpoint_caches
:
1809 destroy_checkpoint_caches();
1810 free_segment_manager_caches
:
1811 destroy_segment_manager_caches();
1812 free_node_manager_caches
:
1813 destroy_node_manager_caches();
1815 destroy_inodecache();
1820 static void __exit
exit_f2fs_fs(void)
1822 remove_proc_entry("fs/f2fs", NULL
);
1823 f2fs_destroy_root_stats();
1824 unregister_shrinker(&f2fs_shrinker_info
);
1825 unregister_filesystem(&f2fs_fs_type
);
1826 destroy_extent_cache();
1827 destroy_checkpoint_caches();
1828 destroy_segment_manager_caches();
1829 destroy_node_manager_caches();
1830 destroy_inodecache();
1831 kset_unregister(f2fs_kset
);
1832 f2fs_destroy_trace_ios();
1835 module_init(init_f2fs_fs
)
1836 module_exit(exit_f2fs_fs
)
1838 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1839 MODULE_DESCRIPTION("Flash Friendly File System");
1840 MODULE_LICENSE("GPL");