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
;
44 Opt_disable_roll_forward
,
53 Opt_disable_ext_identify
,
63 static match_table_t f2fs_tokens
= {
64 {Opt_gc_background
, "background_gc=%s"},
65 {Opt_disable_roll_forward
, "disable_roll_forward"},
66 {Opt_norecovery
, "norecovery"},
67 {Opt_discard
, "discard"},
68 {Opt_noheap
, "no_heap"},
69 {Opt_user_xattr
, "user_xattr"},
70 {Opt_nouser_xattr
, "nouser_xattr"},
73 {Opt_active_logs
, "active_logs=%u"},
74 {Opt_disable_ext_identify
, "disable_ext_identify"},
75 {Opt_inline_xattr
, "inline_xattr"},
76 {Opt_inline_data
, "inline_data"},
77 {Opt_inline_dentry
, "inline_dentry"},
78 {Opt_flush_merge
, "flush_merge"},
79 {Opt_nobarrier
, "nobarrier"},
80 {Opt_fastboot
, "fastboot"},
84 /* Sysfs support for f2fs */
86 GC_THREAD
, /* struct f2fs_gc_thread */
87 SM_INFO
, /* struct f2fs_sm_info */
88 NM_INFO
, /* struct f2fs_nm_info */
89 F2FS_SBI
, /* struct f2fs_sb_info */
93 struct attribute attr
;
94 ssize_t (*show
)(struct f2fs_attr
*, struct f2fs_sb_info
*, char *);
95 ssize_t (*store
)(struct f2fs_attr
*, struct f2fs_sb_info
*,
96 const char *, size_t);
101 static unsigned char *__struct_ptr(struct f2fs_sb_info
*sbi
, int struct_type
)
103 if (struct_type
== GC_THREAD
)
104 return (unsigned char *)sbi
->gc_thread
;
105 else if (struct_type
== SM_INFO
)
106 return (unsigned char *)SM_I(sbi
);
107 else if (struct_type
== NM_INFO
)
108 return (unsigned char *)NM_I(sbi
);
109 else if (struct_type
== F2FS_SBI
)
110 return (unsigned char *)sbi
;
114 static ssize_t
f2fs_sbi_show(struct f2fs_attr
*a
,
115 struct f2fs_sb_info
*sbi
, char *buf
)
117 unsigned char *ptr
= NULL
;
120 ptr
= __struct_ptr(sbi
, a
->struct_type
);
124 ui
= (unsigned int *)(ptr
+ a
->offset
);
126 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
129 static ssize_t
f2fs_sbi_store(struct f2fs_attr
*a
,
130 struct f2fs_sb_info
*sbi
,
131 const char *buf
, size_t count
)
138 ptr
= __struct_ptr(sbi
, a
->struct_type
);
142 ui
= (unsigned int *)(ptr
+ a
->offset
);
144 ret
= kstrtoul(skip_spaces(buf
), 0, &t
);
151 static ssize_t
f2fs_attr_show(struct kobject
*kobj
,
152 struct attribute
*attr
, char *buf
)
154 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
156 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
158 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
161 static ssize_t
f2fs_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
162 const char *buf
, size_t len
)
164 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
166 struct f2fs_attr
*a
= container_of(attr
, struct f2fs_attr
, attr
);
168 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
171 static void f2fs_sb_release(struct kobject
*kobj
)
173 struct f2fs_sb_info
*sbi
= container_of(kobj
, struct f2fs_sb_info
,
175 complete(&sbi
->s_kobj_unregister
);
178 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
179 static struct f2fs_attr f2fs_attr_##_name = { \
180 .attr = {.name = __stringify(_name), .mode = _mode }, \
183 .struct_type = _struct_type, \
187 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
188 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
189 f2fs_sbi_show, f2fs_sbi_store, \
190 offsetof(struct struct_name, elname))
192 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_min_sleep_time
, min_sleep_time
);
193 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_max_sleep_time
, max_sleep_time
);
194 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_no_gc_sleep_time
, no_gc_sleep_time
);
195 F2FS_RW_ATTR(GC_THREAD
, f2fs_gc_kthread
, gc_idle
, gc_idle
);
196 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, reclaim_segments
, rec_prefree_segments
);
197 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, max_small_discards
, max_discards
);
198 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, batched_trim_sections
, trim_sections
);
199 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, ipu_policy
, ipu_policy
);
200 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_ipu_util
, min_ipu_util
);
201 F2FS_RW_ATTR(SM_INFO
, f2fs_sm_info
, min_fsync_blocks
, min_fsync_blocks
);
202 F2FS_RW_ATTR(NM_INFO
, f2fs_nm_info
, ram_thresh
, ram_thresh
);
203 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, max_victim_search
, max_victim_search
);
204 F2FS_RW_ATTR(F2FS_SBI
, f2fs_sb_info
, dir_level
, dir_level
);
206 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
207 static struct attribute
*f2fs_attrs
[] = {
208 ATTR_LIST(gc_min_sleep_time
),
209 ATTR_LIST(gc_max_sleep_time
),
210 ATTR_LIST(gc_no_gc_sleep_time
),
212 ATTR_LIST(reclaim_segments
),
213 ATTR_LIST(max_small_discards
),
214 ATTR_LIST(batched_trim_sections
),
215 ATTR_LIST(ipu_policy
),
216 ATTR_LIST(min_ipu_util
),
217 ATTR_LIST(min_fsync_blocks
),
218 ATTR_LIST(max_victim_search
),
219 ATTR_LIST(dir_level
),
220 ATTR_LIST(ram_thresh
),
224 static const struct sysfs_ops f2fs_attr_ops
= {
225 .show
= f2fs_attr_show
,
226 .store
= f2fs_attr_store
,
229 static struct kobj_type f2fs_ktype
= {
230 .default_attrs
= f2fs_attrs
,
231 .sysfs_ops
= &f2fs_attr_ops
,
232 .release
= f2fs_sb_release
,
235 void f2fs_msg(struct super_block
*sb
, const char *level
, const char *fmt
, ...)
237 struct va_format vaf
;
243 printk("%sF2FS-fs (%s): %pV\n", level
, sb
->s_id
, &vaf
);
247 static void init_once(void *foo
)
249 struct f2fs_inode_info
*fi
= (struct f2fs_inode_info
*) foo
;
251 inode_init_once(&fi
->vfs_inode
);
254 static int parse_options(struct super_block
*sb
, char *options
)
256 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
257 substring_t args
[MAX_OPT_ARGS
];
264 while ((p
= strsep(&options
, ",")) != NULL
) {
269 * Initialize args struct so we know whether arg was
270 * found; some options take optional arguments.
272 args
[0].to
= args
[0].from
= NULL
;
273 token
= match_token(p
, f2fs_tokens
, args
);
276 case Opt_gc_background
:
277 name
= match_strdup(&args
[0]);
281 if (strlen(name
) == 2 && !strncmp(name
, "on", 2))
283 else if (strlen(name
) == 3 && !strncmp(name
, "off", 3))
284 clear_opt(sbi
, BG_GC
);
291 case Opt_disable_roll_forward
:
292 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
295 /* this option mounts f2fs with ro */
296 set_opt(sbi
, DISABLE_ROLL_FORWARD
);
297 if (!f2fs_readonly(sb
))
301 set_opt(sbi
, DISCARD
);
304 set_opt(sbi
, NOHEAP
);
306 #ifdef CONFIG_F2FS_FS_XATTR
308 set_opt(sbi
, XATTR_USER
);
310 case Opt_nouser_xattr
:
311 clear_opt(sbi
, XATTR_USER
);
313 case Opt_inline_xattr
:
314 set_opt(sbi
, INLINE_XATTR
);
318 f2fs_msg(sb
, KERN_INFO
,
319 "user_xattr options not supported");
321 case Opt_nouser_xattr
:
322 f2fs_msg(sb
, KERN_INFO
,
323 "nouser_xattr options not supported");
325 case Opt_inline_xattr
:
326 f2fs_msg(sb
, KERN_INFO
,
327 "inline_xattr options not supported");
330 #ifdef CONFIG_F2FS_FS_POSIX_ACL
332 set_opt(sbi
, POSIX_ACL
);
335 clear_opt(sbi
, POSIX_ACL
);
339 f2fs_msg(sb
, KERN_INFO
, "acl options not supported");
342 f2fs_msg(sb
, KERN_INFO
, "noacl options not supported");
345 case Opt_active_logs
:
346 if (args
->from
&& match_int(args
, &arg
))
348 if (arg
!= 2 && arg
!= 4 && arg
!= NR_CURSEG_TYPE
)
350 sbi
->active_logs
= arg
;
352 case Opt_disable_ext_identify
:
353 set_opt(sbi
, DISABLE_EXT_IDENTIFY
);
355 case Opt_inline_data
:
356 set_opt(sbi
, INLINE_DATA
);
358 case Opt_inline_dentry
:
359 set_opt(sbi
, INLINE_DENTRY
);
361 case Opt_flush_merge
:
362 set_opt(sbi
, FLUSH_MERGE
);
365 set_opt(sbi
, NOBARRIER
);
368 set_opt(sbi
, FASTBOOT
);
371 f2fs_msg(sb
, KERN_ERR
,
372 "Unrecognized mount option \"%s\" or missing value",
380 static struct inode
*f2fs_alloc_inode(struct super_block
*sb
)
382 struct f2fs_inode_info
*fi
;
384 fi
= kmem_cache_alloc(f2fs_inode_cachep
, GFP_F2FS_ZERO
);
388 init_once((void *) fi
);
390 /* Initialize f2fs-specific inode info */
391 fi
->vfs_inode
.i_version
= 1;
392 atomic_set(&fi
->dirty_pages
, 0);
393 fi
->i_current_depth
= 1;
395 rwlock_init(&fi
->ext
.ext_lock
);
396 init_rwsem(&fi
->i_sem
);
397 INIT_RADIX_TREE(&fi
->inmem_root
, GFP_NOFS
);
398 INIT_LIST_HEAD(&fi
->inmem_pages
);
399 mutex_init(&fi
->inmem_lock
);
401 set_inode_flag(fi
, FI_NEW_INODE
);
403 if (test_opt(F2FS_SB(sb
), INLINE_XATTR
))
404 set_inode_flag(fi
, FI_INLINE_XATTR
);
406 /* Will be used by directory only */
407 fi
->i_dir_level
= F2FS_SB(sb
)->dir_level
;
409 return &fi
->vfs_inode
;
412 static int f2fs_drop_inode(struct inode
*inode
)
415 * This is to avoid a deadlock condition like below.
416 * writeback_single_inode(inode)
417 * - f2fs_write_data_page
418 * - f2fs_gc -> iput -> evict
419 * - inode_wait_for_writeback(inode)
421 if (!inode_unhashed(inode
) && inode
->i_state
& I_SYNC
)
423 return generic_drop_inode(inode
);
427 * f2fs_dirty_inode() is called from __mark_inode_dirty()
429 * We should call set_dirty_inode to write the dirty inode through write_inode.
431 static void f2fs_dirty_inode(struct inode
*inode
, int flags
)
433 set_inode_flag(F2FS_I(inode
), FI_DIRTY_INODE
);
436 static void f2fs_i_callback(struct rcu_head
*head
)
438 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
439 kmem_cache_free(f2fs_inode_cachep
, F2FS_I(inode
));
442 static void f2fs_destroy_inode(struct inode
*inode
)
444 call_rcu(&inode
->i_rcu
, f2fs_i_callback
);
447 static void f2fs_put_super(struct super_block
*sb
)
449 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
452 remove_proc_entry("segment_info", sbi
->s_proc
);
453 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
455 kobject_del(&sbi
->s_kobj
);
457 f2fs_destroy_stats(sbi
);
461 * We don't need to do checkpoint when superblock is clean.
462 * But, the previous checkpoint was not done by umount, it needs to do
463 * clean checkpoint again.
465 if (is_sbi_flag_set(sbi
, SBI_IS_DIRTY
) ||
466 !is_set_ckpt_flags(F2FS_CKPT(sbi
), CP_UMOUNT_FLAG
)) {
467 struct cp_control cpc
= {
470 write_checkpoint(sbi
, &cpc
);
474 * normally superblock is clean, so we need to release this.
475 * In addition, EIO will skip do checkpoint, we need this as well.
477 release_dirty_inode(sbi
);
478 release_discard_addrs(sbi
);
480 iput(sbi
->node_inode
);
481 iput(sbi
->meta_inode
);
483 /* destroy f2fs internal modules */
484 destroy_node_manager(sbi
);
485 destroy_segment_manager(sbi
);
488 kobject_put(&sbi
->s_kobj
);
489 wait_for_completion(&sbi
->s_kobj_unregister
);
491 sb
->s_fs_info
= NULL
;
492 brelse(sbi
->raw_super_buf
);
496 int f2fs_sync_fs(struct super_block
*sb
, int sync
)
498 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
500 trace_f2fs_sync_fs(sb
, sync
);
503 struct cp_control cpc
;
505 cpc
.reason
= __get_cp_reason(sbi
);
507 mutex_lock(&sbi
->gc_mutex
);
508 write_checkpoint(sbi
, &cpc
);
509 mutex_unlock(&sbi
->gc_mutex
);
511 f2fs_balance_fs(sbi
);
513 f2fs_trace_ios(NULL
, NULL
, 1);
518 static int f2fs_freeze(struct super_block
*sb
)
522 if (f2fs_readonly(sb
))
525 err
= f2fs_sync_fs(sb
, 1);
529 static int f2fs_unfreeze(struct super_block
*sb
)
534 static int f2fs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
536 struct super_block
*sb
= dentry
->d_sb
;
537 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
538 u64 id
= huge_encode_dev(sb
->s_bdev
->bd_dev
);
539 block_t total_count
, user_block_count
, start_count
, ovp_count
;
541 total_count
= le64_to_cpu(sbi
->raw_super
->block_count
);
542 user_block_count
= sbi
->user_block_count
;
543 start_count
= le32_to_cpu(sbi
->raw_super
->segment0_blkaddr
);
544 ovp_count
= SM_I(sbi
)->ovp_segments
<< sbi
->log_blocks_per_seg
;
545 buf
->f_type
= F2FS_SUPER_MAGIC
;
546 buf
->f_bsize
= sbi
->blocksize
;
548 buf
->f_blocks
= total_count
- start_count
;
549 buf
->f_bfree
= buf
->f_blocks
- valid_user_blocks(sbi
) - ovp_count
;
550 buf
->f_bavail
= user_block_count
- valid_user_blocks(sbi
);
552 buf
->f_files
= sbi
->total_node_count
- F2FS_RESERVED_NODE_NUM
;
553 buf
->f_ffree
= buf
->f_files
- valid_inode_count(sbi
);
555 buf
->f_namelen
= F2FS_NAME_LEN
;
556 buf
->f_fsid
.val
[0] = (u32
)id
;
557 buf
->f_fsid
.val
[1] = (u32
)(id
>> 32);
562 static int f2fs_show_options(struct seq_file
*seq
, struct dentry
*root
)
564 struct f2fs_sb_info
*sbi
= F2FS_SB(root
->d_sb
);
566 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, BG_GC
))
567 seq_printf(seq
, ",background_gc=%s", "on");
569 seq_printf(seq
, ",background_gc=%s", "off");
570 if (test_opt(sbi
, DISABLE_ROLL_FORWARD
))
571 seq_puts(seq
, ",disable_roll_forward");
572 if (test_opt(sbi
, DISCARD
))
573 seq_puts(seq
, ",discard");
574 if (test_opt(sbi
, NOHEAP
))
575 seq_puts(seq
, ",no_heap_alloc");
576 #ifdef CONFIG_F2FS_FS_XATTR
577 if (test_opt(sbi
, XATTR_USER
))
578 seq_puts(seq
, ",user_xattr");
580 seq_puts(seq
, ",nouser_xattr");
581 if (test_opt(sbi
, INLINE_XATTR
))
582 seq_puts(seq
, ",inline_xattr");
584 #ifdef CONFIG_F2FS_FS_POSIX_ACL
585 if (test_opt(sbi
, POSIX_ACL
))
586 seq_puts(seq
, ",acl");
588 seq_puts(seq
, ",noacl");
590 if (test_opt(sbi
, DISABLE_EXT_IDENTIFY
))
591 seq_puts(seq
, ",disable_ext_identify");
592 if (test_opt(sbi
, INLINE_DATA
))
593 seq_puts(seq
, ",inline_data");
594 if (test_opt(sbi
, INLINE_DENTRY
))
595 seq_puts(seq
, ",inline_dentry");
596 if (!f2fs_readonly(sbi
->sb
) && test_opt(sbi
, FLUSH_MERGE
))
597 seq_puts(seq
, ",flush_merge");
598 if (test_opt(sbi
, NOBARRIER
))
599 seq_puts(seq
, ",nobarrier");
600 if (test_opt(sbi
, FASTBOOT
))
601 seq_puts(seq
, ",fastboot");
602 seq_printf(seq
, ",active_logs=%u", sbi
->active_logs
);
607 static int segment_info_seq_show(struct seq_file
*seq
, void *offset
)
609 struct super_block
*sb
= seq
->private;
610 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
611 unsigned int total_segs
=
612 le32_to_cpu(sbi
->raw_super
->segment_count_main
);
615 seq_puts(seq
, "format: segment_type|valid_blocks\n"
616 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
618 for (i
= 0; i
< total_segs
; i
++) {
619 struct seg_entry
*se
= get_seg_entry(sbi
, i
);
622 seq_printf(seq
, "%-5d", i
);
623 seq_printf(seq
, "%d|%-3u", se
->type
,
624 get_valid_blocks(sbi
, i
, 1));
625 if ((i
% 10) == 9 || i
== (total_segs
- 1))
634 static int segment_info_open_fs(struct inode
*inode
, struct file
*file
)
636 return single_open(file
, segment_info_seq_show
, PDE_DATA(inode
));
639 static const struct file_operations f2fs_seq_segment_info_fops
= {
640 .owner
= THIS_MODULE
,
641 .open
= segment_info_open_fs
,
644 .release
= single_release
,
647 static int f2fs_remount(struct super_block
*sb
, int *flags
, char *data
)
649 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
650 struct f2fs_mount_info org_mount_opt
;
651 int err
, active_logs
;
652 bool need_restart_gc
= false;
653 bool need_stop_gc
= false;
658 * Save the old mount options in case we
659 * need to restore them.
661 org_mount_opt
= sbi
->mount_opt
;
662 active_logs
= sbi
->active_logs
;
664 sbi
->mount_opt
.opt
= 0;
665 sbi
->active_logs
= NR_CURSEG_TYPE
;
667 /* parse mount options */
668 err
= parse_options(sb
, data
);
673 * Previous and new state of filesystem is RO,
674 * so skip checking GC and FLUSH_MERGE conditions.
676 if (f2fs_readonly(sb
) && (*flags
& MS_RDONLY
))
680 * We stop the GC thread if FS is mounted as RO
681 * or if background_gc = off is passed in mount
682 * option. Also sync the filesystem.
684 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, BG_GC
)) {
685 if (sbi
->gc_thread
) {
688 need_restart_gc
= true;
690 } else if (!sbi
->gc_thread
) {
691 err
= start_gc_thread(sbi
);
698 * We stop issue flush thread if FS is mounted as RO
699 * or if flush_merge is not passed in mount option.
701 if ((*flags
& MS_RDONLY
) || !test_opt(sbi
, FLUSH_MERGE
)) {
702 destroy_flush_cmd_control(sbi
);
703 } else if (!SM_I(sbi
)->cmd_control_info
) {
704 err
= create_flush_cmd_control(sbi
);
709 /* Update the POSIXACL Flag */
710 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
711 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
714 if (need_restart_gc
) {
715 if (start_gc_thread(sbi
))
716 f2fs_msg(sbi
->sb
, KERN_WARNING
,
717 "background gc thread has stopped");
718 } else if (need_stop_gc
) {
722 sbi
->mount_opt
= org_mount_opt
;
723 sbi
->active_logs
= active_logs
;
727 static struct super_operations f2fs_sops
= {
728 .alloc_inode
= f2fs_alloc_inode
,
729 .drop_inode
= f2fs_drop_inode
,
730 .destroy_inode
= f2fs_destroy_inode
,
731 .write_inode
= f2fs_write_inode
,
732 .dirty_inode
= f2fs_dirty_inode
,
733 .show_options
= f2fs_show_options
,
734 .evict_inode
= f2fs_evict_inode
,
735 .put_super
= f2fs_put_super
,
736 .sync_fs
= f2fs_sync_fs
,
737 .freeze_fs
= f2fs_freeze
,
738 .unfreeze_fs
= f2fs_unfreeze
,
739 .statfs
= f2fs_statfs
,
740 .remount_fs
= f2fs_remount
,
743 static struct inode
*f2fs_nfs_get_inode(struct super_block
*sb
,
744 u64 ino
, u32 generation
)
746 struct f2fs_sb_info
*sbi
= F2FS_SB(sb
);
749 if (check_nid_range(sbi
, ino
))
750 return ERR_PTR(-ESTALE
);
753 * f2fs_iget isn't quite right if the inode is currently unallocated!
754 * However f2fs_iget currently does appropriate checks to handle stale
755 * inodes so everything is OK.
757 inode
= f2fs_iget(sb
, ino
);
759 return ERR_CAST(inode
);
760 if (unlikely(generation
&& inode
->i_generation
!= generation
)) {
761 /* we didn't find the right inode.. */
763 return ERR_PTR(-ESTALE
);
768 static struct dentry
*f2fs_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
769 int fh_len
, int fh_type
)
771 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
775 static struct dentry
*f2fs_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
776 int fh_len
, int fh_type
)
778 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
782 static const struct export_operations f2fs_export_ops
= {
783 .fh_to_dentry
= f2fs_fh_to_dentry
,
784 .fh_to_parent
= f2fs_fh_to_parent
,
785 .get_parent
= f2fs_get_parent
,
788 static loff_t
max_file_size(unsigned bits
)
790 loff_t result
= (DEF_ADDRS_PER_INODE
- F2FS_INLINE_XATTR_ADDRS
);
791 loff_t leaf_count
= ADDRS_PER_BLOCK
;
793 /* two direct node blocks */
794 result
+= (leaf_count
* 2);
796 /* two indirect node blocks */
797 leaf_count
*= NIDS_PER_BLOCK
;
798 result
+= (leaf_count
* 2);
800 /* one double indirect node block */
801 leaf_count
*= NIDS_PER_BLOCK
;
802 result
+= leaf_count
;
808 static int sanity_check_raw_super(struct super_block
*sb
,
809 struct f2fs_super_block
*raw_super
)
811 unsigned int blocksize
;
813 if (F2FS_SUPER_MAGIC
!= le32_to_cpu(raw_super
->magic
)) {
814 f2fs_msg(sb
, KERN_INFO
,
815 "Magic Mismatch, valid(0x%x) - read(0x%x)",
816 F2FS_SUPER_MAGIC
, le32_to_cpu(raw_super
->magic
));
820 /* Currently, support only 4KB page cache size */
821 if (F2FS_BLKSIZE
!= PAGE_CACHE_SIZE
) {
822 f2fs_msg(sb
, KERN_INFO
,
823 "Invalid page_cache_size (%lu), supports only 4KB\n",
828 /* Currently, support only 4KB block size */
829 blocksize
= 1 << le32_to_cpu(raw_super
->log_blocksize
);
830 if (blocksize
!= F2FS_BLKSIZE
) {
831 f2fs_msg(sb
, KERN_INFO
,
832 "Invalid blocksize (%u), supports only 4KB\n",
837 /* Currently, support 512/1024/2048/4096 bytes sector size */
838 if (le32_to_cpu(raw_super
->log_sectorsize
) >
839 F2FS_MAX_LOG_SECTOR_SIZE
||
840 le32_to_cpu(raw_super
->log_sectorsize
) <
841 F2FS_MIN_LOG_SECTOR_SIZE
) {
842 f2fs_msg(sb
, KERN_INFO
, "Invalid log sectorsize (%u)",
843 le32_to_cpu(raw_super
->log_sectorsize
));
846 if (le32_to_cpu(raw_super
->log_sectors_per_block
) +
847 le32_to_cpu(raw_super
->log_sectorsize
) !=
848 F2FS_MAX_LOG_SECTOR_SIZE
) {
849 f2fs_msg(sb
, KERN_INFO
,
850 "Invalid log sectors per block(%u) log sectorsize(%u)",
851 le32_to_cpu(raw_super
->log_sectors_per_block
),
852 le32_to_cpu(raw_super
->log_sectorsize
));
858 static int sanity_check_ckpt(struct f2fs_sb_info
*sbi
)
860 unsigned int total
, fsmeta
;
861 struct f2fs_super_block
*raw_super
= F2FS_RAW_SUPER(sbi
);
862 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(sbi
);
864 total
= le32_to_cpu(raw_super
->segment_count
);
865 fsmeta
= le32_to_cpu(raw_super
->segment_count_ckpt
);
866 fsmeta
+= le32_to_cpu(raw_super
->segment_count_sit
);
867 fsmeta
+= le32_to_cpu(raw_super
->segment_count_nat
);
868 fsmeta
+= le32_to_cpu(ckpt
->rsvd_segment_count
);
869 fsmeta
+= le32_to_cpu(raw_super
->segment_count_ssa
);
871 if (unlikely(fsmeta
>= total
))
874 if (unlikely(f2fs_cp_error(sbi
))) {
875 f2fs_msg(sbi
->sb
, KERN_ERR
, "A bug case: need to run fsck");
881 static void init_sb_info(struct f2fs_sb_info
*sbi
)
883 struct f2fs_super_block
*raw_super
= sbi
->raw_super
;
886 sbi
->log_sectors_per_block
=
887 le32_to_cpu(raw_super
->log_sectors_per_block
);
888 sbi
->log_blocksize
= le32_to_cpu(raw_super
->log_blocksize
);
889 sbi
->blocksize
= 1 << sbi
->log_blocksize
;
890 sbi
->log_blocks_per_seg
= le32_to_cpu(raw_super
->log_blocks_per_seg
);
891 sbi
->blocks_per_seg
= 1 << sbi
->log_blocks_per_seg
;
892 sbi
->segs_per_sec
= le32_to_cpu(raw_super
->segs_per_sec
);
893 sbi
->secs_per_zone
= le32_to_cpu(raw_super
->secs_per_zone
);
894 sbi
->total_sections
= le32_to_cpu(raw_super
->section_count
);
895 sbi
->total_node_count
=
896 (le32_to_cpu(raw_super
->segment_count_nat
) / 2)
897 * sbi
->blocks_per_seg
* NAT_ENTRY_PER_BLOCK
;
898 sbi
->root_ino_num
= le32_to_cpu(raw_super
->root_ino
);
899 sbi
->node_ino_num
= le32_to_cpu(raw_super
->node_ino
);
900 sbi
->meta_ino_num
= le32_to_cpu(raw_super
->meta_ino
);
901 sbi
->cur_victim_sec
= NULL_SECNO
;
902 sbi
->max_victim_search
= DEF_MAX_VICTIM_SEARCH
;
904 for (i
= 0; i
< NR_COUNT_TYPE
; i
++)
905 atomic_set(&sbi
->nr_pages
[i
], 0);
907 sbi
->dir_level
= DEF_DIR_LEVEL
;
908 clear_sbi_flag(sbi
, SBI_NEED_FSCK
);
912 * Read f2fs raw super block.
913 * Because we have two copies of super block, so read the first one at first,
914 * if the first one is invalid, move to read the second one.
916 static int read_raw_super_block(struct super_block
*sb
,
917 struct f2fs_super_block
**raw_super
,
918 struct buffer_head
**raw_super_buf
)
923 *raw_super_buf
= sb_bread(sb
, block
);
924 if (!*raw_super_buf
) {
925 f2fs_msg(sb
, KERN_ERR
, "Unable to read %dth superblock",
935 *raw_super
= (struct f2fs_super_block
*)
936 ((char *)(*raw_super_buf
)->b_data
+ F2FS_SUPER_OFFSET
);
938 /* sanity checking of raw super */
939 if (sanity_check_raw_super(sb
, *raw_super
)) {
940 brelse(*raw_super_buf
);
941 f2fs_msg(sb
, KERN_ERR
,
942 "Can't find valid F2FS filesystem in %dth superblock",
955 static int f2fs_fill_super(struct super_block
*sb
, void *data
, int silent
)
957 struct f2fs_sb_info
*sbi
;
958 struct f2fs_super_block
*raw_super
= NULL
;
959 struct buffer_head
*raw_super_buf
;
963 char *options
= NULL
;
967 /* allocate memory for f2fs-specific super block info */
968 sbi
= kzalloc(sizeof(struct f2fs_sb_info
), GFP_KERNEL
);
972 /* set a block size */
973 if (unlikely(!sb_set_blocksize(sb
, F2FS_BLKSIZE
))) {
974 f2fs_msg(sb
, KERN_ERR
, "unable to set blocksize");
978 err
= read_raw_super_block(sb
, &raw_super
, &raw_super_buf
);
983 /* init some FS parameters */
984 sbi
->active_logs
= NR_CURSEG_TYPE
;
988 #ifdef CONFIG_F2FS_FS_XATTR
989 set_opt(sbi
, XATTR_USER
);
991 #ifdef CONFIG_F2FS_FS_POSIX_ACL
992 set_opt(sbi
, POSIX_ACL
);
994 /* parse mount options */
995 options
= kstrdup((const char *)data
, GFP_KERNEL
);
996 if (data
&& !options
) {
1001 err
= parse_options(sb
, options
);
1005 sb
->s_maxbytes
= max_file_size(le32_to_cpu(raw_super
->log_blocksize
));
1006 sb
->s_max_links
= F2FS_LINK_MAX
;
1007 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
1009 sb
->s_op
= &f2fs_sops
;
1010 sb
->s_xattr
= f2fs_xattr_handlers
;
1011 sb
->s_export_op
= &f2fs_export_ops
;
1012 sb
->s_magic
= F2FS_SUPER_MAGIC
;
1013 sb
->s_time_gran
= 1;
1014 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
1015 (test_opt(sbi
, POSIX_ACL
) ? MS_POSIXACL
: 0);
1016 memcpy(sb
->s_uuid
, raw_super
->uuid
, sizeof(raw_super
->uuid
));
1018 /* init f2fs-specific super block info */
1020 sbi
->raw_super
= raw_super
;
1021 sbi
->raw_super_buf
= raw_super_buf
;
1022 mutex_init(&sbi
->gc_mutex
);
1023 mutex_init(&sbi
->writepages
);
1024 mutex_init(&sbi
->cp_mutex
);
1025 init_rwsem(&sbi
->node_write
);
1026 clear_sbi_flag(sbi
, SBI_POR_DOING
);
1027 spin_lock_init(&sbi
->stat_lock
);
1029 init_rwsem(&sbi
->read_io
.io_rwsem
);
1030 sbi
->read_io
.sbi
= sbi
;
1031 sbi
->read_io
.bio
= NULL
;
1032 for (i
= 0; i
< NR_PAGE_TYPE
; i
++) {
1033 init_rwsem(&sbi
->write_io
[i
].io_rwsem
);
1034 sbi
->write_io
[i
].sbi
= sbi
;
1035 sbi
->write_io
[i
].bio
= NULL
;
1038 init_rwsem(&sbi
->cp_rwsem
);
1039 init_waitqueue_head(&sbi
->cp_wait
);
1042 /* get an inode for meta space */
1043 sbi
->meta_inode
= f2fs_iget(sb
, F2FS_META_INO(sbi
));
1044 if (IS_ERR(sbi
->meta_inode
)) {
1045 f2fs_msg(sb
, KERN_ERR
, "Failed to read F2FS meta data inode");
1046 err
= PTR_ERR(sbi
->meta_inode
);
1050 err
= get_valid_checkpoint(sbi
);
1052 f2fs_msg(sb
, KERN_ERR
, "Failed to get valid F2FS checkpoint");
1053 goto free_meta_inode
;
1056 /* sanity checking of checkpoint */
1058 if (sanity_check_ckpt(sbi
)) {
1059 f2fs_msg(sb
, KERN_ERR
, "Invalid F2FS checkpoint");
1063 sbi
->total_valid_node_count
=
1064 le32_to_cpu(sbi
->ckpt
->valid_node_count
);
1065 sbi
->total_valid_inode_count
=
1066 le32_to_cpu(sbi
->ckpt
->valid_inode_count
);
1067 sbi
->user_block_count
= le64_to_cpu(sbi
->ckpt
->user_block_count
);
1068 sbi
->total_valid_block_count
=
1069 le64_to_cpu(sbi
->ckpt
->valid_block_count
);
1070 sbi
->last_valid_block_count
= sbi
->total_valid_block_count
;
1071 sbi
->alloc_valid_block_count
= 0;
1072 INIT_LIST_HEAD(&sbi
->dir_inode_list
);
1073 spin_lock_init(&sbi
->dir_inode_lock
);
1075 init_ino_entry_info(sbi
);
1077 /* setup f2fs internal modules */
1078 err
= build_segment_manager(sbi
);
1080 f2fs_msg(sb
, KERN_ERR
,
1081 "Failed to initialize F2FS segment manager");
1084 err
= build_node_manager(sbi
);
1086 f2fs_msg(sb
, KERN_ERR
,
1087 "Failed to initialize F2FS node manager");
1091 build_gc_manager(sbi
);
1093 /* get an inode for node space */
1094 sbi
->node_inode
= f2fs_iget(sb
, F2FS_NODE_INO(sbi
));
1095 if (IS_ERR(sbi
->node_inode
)) {
1096 f2fs_msg(sb
, KERN_ERR
, "Failed to read node inode");
1097 err
= PTR_ERR(sbi
->node_inode
);
1101 /* if there are nt orphan nodes free them */
1102 recover_orphan_inodes(sbi
);
1104 /* read root inode and dentry */
1105 root
= f2fs_iget(sb
, F2FS_ROOT_INO(sbi
));
1107 f2fs_msg(sb
, KERN_ERR
, "Failed to read root inode");
1108 err
= PTR_ERR(root
);
1109 goto free_node_inode
;
1111 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
1114 goto free_node_inode
;
1117 sb
->s_root
= d_make_root(root
); /* allocate root dentry */
1120 goto free_root_inode
;
1123 err
= f2fs_build_stats(sbi
);
1125 goto free_root_inode
;
1128 sbi
->s_proc
= proc_mkdir(sb
->s_id
, f2fs_proc_root
);
1131 proc_create_data("segment_info", S_IRUGO
, sbi
->s_proc
,
1132 &f2fs_seq_segment_info_fops
, sb
);
1134 if (test_opt(sbi
, DISCARD
)) {
1135 struct request_queue
*q
= bdev_get_queue(sb
->s_bdev
);
1136 if (!blk_queue_discard(q
))
1137 f2fs_msg(sb
, KERN_WARNING
,
1138 "mounting with \"discard\" option, but "
1139 "the device does not support discard");
1142 sbi
->s_kobj
.kset
= f2fs_kset
;
1143 init_completion(&sbi
->s_kobj_unregister
);
1144 err
= kobject_init_and_add(&sbi
->s_kobj
, &f2fs_ktype
, NULL
,
1150 set_sbi_flag(sbi
, SBI_NEED_FSCK
);
1152 /* recover fsynced data */
1153 if (!test_opt(sbi
, DISABLE_ROLL_FORWARD
)) {
1155 * mount should be failed, when device has readonly mode, and
1156 * previous checkpoint was not done by clean system shutdown.
1158 if (bdev_read_only(sb
->s_bdev
) &&
1159 !is_set_ckpt_flags(sbi
->ckpt
, CP_UMOUNT_FLAG
)) {
1163 err
= recover_fsync_data(sbi
);
1165 f2fs_msg(sb
, KERN_ERR
,
1166 "Cannot recover all fsync data errno=%ld", err
);
1172 * If filesystem is not mounted as read-only then
1173 * do start the gc_thread.
1175 if (test_opt(sbi
, BG_GC
) && !f2fs_readonly(sb
)) {
1176 /* After POR, we can run background GC thread.*/
1177 err
= start_gc_thread(sbi
);
1185 kobject_del(&sbi
->s_kobj
);
1188 remove_proc_entry("segment_info", sbi
->s_proc
);
1189 remove_proc_entry(sb
->s_id
, f2fs_proc_root
);
1191 f2fs_destroy_stats(sbi
);
1196 iput(sbi
->node_inode
);
1198 destroy_node_manager(sbi
);
1200 destroy_segment_manager(sbi
);
1204 make_bad_inode(sbi
->meta_inode
);
1205 iput(sbi
->meta_inode
);
1209 brelse(raw_super_buf
);
1213 /* give only one another chance */
1216 shrink_dcache_sb(sb
);
1222 static struct dentry
*f2fs_mount(struct file_system_type
*fs_type
, int flags
,
1223 const char *dev_name
, void *data
)
1225 return mount_bdev(fs_type
, flags
, dev_name
, data
, f2fs_fill_super
);
1228 static void kill_f2fs_super(struct super_block
*sb
)
1231 set_sbi_flag(F2FS_SB(sb
), SBI_IS_CLOSE
);
1232 kill_block_super(sb
);
1235 static struct file_system_type f2fs_fs_type
= {
1236 .owner
= THIS_MODULE
,
1238 .mount
= f2fs_mount
,
1239 .kill_sb
= kill_f2fs_super
,
1240 .fs_flags
= FS_REQUIRES_DEV
,
1242 MODULE_ALIAS_FS("f2fs");
1244 static int __init
init_inodecache(void)
1246 f2fs_inode_cachep
= f2fs_kmem_cache_create("f2fs_inode_cache",
1247 sizeof(struct f2fs_inode_info
));
1248 if (!f2fs_inode_cachep
)
1253 static void destroy_inodecache(void)
1256 * Make sure all delayed rcu free inodes are flushed before we
1260 kmem_cache_destroy(f2fs_inode_cachep
);
1263 static int __init
init_f2fs_fs(void)
1267 f2fs_build_trace_ios();
1269 err
= init_inodecache();
1272 err
= create_node_manager_caches();
1274 goto free_inodecache
;
1275 err
= create_segment_manager_caches();
1277 goto free_node_manager_caches
;
1278 err
= create_checkpoint_caches();
1280 goto free_segment_manager_caches
;
1281 f2fs_kset
= kset_create_and_add("f2fs", NULL
, fs_kobj
);
1284 goto free_checkpoint_caches
;
1286 err
= register_filesystem(&f2fs_fs_type
);
1289 f2fs_create_root_stats();
1290 f2fs_proc_root
= proc_mkdir("fs/f2fs", NULL
);
1294 kset_unregister(f2fs_kset
);
1295 free_checkpoint_caches
:
1296 destroy_checkpoint_caches();
1297 free_segment_manager_caches
:
1298 destroy_segment_manager_caches();
1299 free_node_manager_caches
:
1300 destroy_node_manager_caches();
1302 destroy_inodecache();
1307 static void __exit
exit_f2fs_fs(void)
1309 remove_proc_entry("fs/f2fs", NULL
);
1310 f2fs_destroy_root_stats();
1311 unregister_filesystem(&f2fs_fs_type
);
1312 destroy_checkpoint_caches();
1313 destroy_segment_manager_caches();
1314 destroy_node_manager_caches();
1315 destroy_inodecache();
1316 kset_unregister(f2fs_kset
);
1317 f2fs_destroy_trace_ios();
1320 module_init(init_f2fs_fs
)
1321 module_exit(exit_f2fs_fs
)
1323 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1324 MODULE_DESCRIPTION("Flash Friendly File System");
1325 MODULE_LICENSE("GPL");