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/quotaops.h>
26 #include <linux/f2fs_fs.h>
27 #include <linux/sysfs.h>
28 #include <linux/quota.h>
37 #define CREATE_TRACE_POINTS
38 #include <trace/events/f2fs.h>
40 static struct kmem_cache *f2fs_inode_cachep;
42 #ifdef CONFIG_F2FS_FAULT_INJECTION
44 char *fault_name[FAULT_MAX] = {
45 [FAULT_KMALLOC] = "kmalloc",
46 [FAULT_PAGE_ALLOC] = "page alloc",
47 [FAULT_PAGE_GET] = "page get",
48 [FAULT_ALLOC_BIO] = "alloc bio",
49 [FAULT_ALLOC_NID] = "alloc nid",
50 [FAULT_ORPHAN] = "orphan",
51 [FAULT_BLOCK] = "no more block",
52 [FAULT_DIR_DEPTH] = "too big dir depth",
53 [FAULT_EVICT_INODE] = "evict_inode fail",
54 [FAULT_TRUNCATE] = "truncate fail",
55 [FAULT_IO] = "IO error",
56 [FAULT_CHECKPOINT] = "checkpoint error",
59 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
62 struct f2fs_fault_info *ffi = &sbi->fault_info;
65 atomic_set(&ffi->inject_ops, 0);
66 ffi->inject_rate = rate;
67 ffi->inject_type = (1 << FAULT_MAX) - 1;
69 memset(ffi, 0, sizeof(struct f2fs_fault_info));
74 /* f2fs-wide shrinker description */
75 static struct shrinker f2fs_shrinker_info = {
76 .scan_objects = f2fs_shrink_scan,
77 .count_objects = f2fs_shrink_count,
78 .seeks = DEFAULT_SEEKS,
83 Opt_disable_roll_forward,
94 Opt_disable_ext_identify,
97 Opt_inline_xattr_size,
131 static match_table_t f2fs_tokens = {
132 {Opt_gc_background, "background_gc=%s"},
133 {Opt_disable_roll_forward, "disable_roll_forward"},
134 {Opt_norecovery, "norecovery"},
135 {Opt_discard, "discard"},
136 {Opt_nodiscard, "nodiscard"},
137 {Opt_noheap, "no_heap"},
139 {Opt_user_xattr, "user_xattr"},
140 {Opt_nouser_xattr, "nouser_xattr"},
142 {Opt_noacl, "noacl"},
143 {Opt_active_logs, "active_logs=%u"},
144 {Opt_disable_ext_identify, "disable_ext_identify"},
145 {Opt_inline_xattr, "inline_xattr"},
146 {Opt_noinline_xattr, "noinline_xattr"},
147 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
148 {Opt_inline_data, "inline_data"},
149 {Opt_inline_dentry, "inline_dentry"},
150 {Opt_noinline_dentry, "noinline_dentry"},
151 {Opt_flush_merge, "flush_merge"},
152 {Opt_noflush_merge, "noflush_merge"},
153 {Opt_nobarrier, "nobarrier"},
154 {Opt_fastboot, "fastboot"},
155 {Opt_extent_cache, "extent_cache"},
156 {Opt_noextent_cache, "noextent_cache"},
157 {Opt_noinline_data, "noinline_data"},
158 {Opt_data_flush, "data_flush"},
159 {Opt_mode, "mode=%s"},
160 {Opt_io_size_bits, "io_bits=%u"},
161 {Opt_fault_injection, "fault_injection=%u"},
162 {Opt_lazytime, "lazytime"},
163 {Opt_nolazytime, "nolazytime"},
164 {Opt_quota, "quota"},
165 {Opt_noquota, "noquota"},
166 {Opt_usrquota, "usrquota"},
167 {Opt_grpquota, "grpquota"},
168 {Opt_prjquota, "prjquota"},
169 {Opt_usrjquota, "usrjquota=%s"},
170 {Opt_grpjquota, "grpjquota=%s"},
171 {Opt_prjjquota, "prjjquota=%s"},
172 {Opt_offusrjquota, "usrjquota="},
173 {Opt_offgrpjquota, "grpjquota="},
174 {Opt_offprjjquota, "prjjquota="},
175 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
176 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
177 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
181 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
183 struct va_format vaf;
189 printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
193 static void init_once(void *foo)
195 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
197 inode_init_once(&fi->vfs_inode);
201 static const char * const quotatypes[] = INITQFNAMES;
202 #define QTYPE2NAME(t) (quotatypes[t])
203 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
206 struct f2fs_sb_info *sbi = F2FS_SB(sb);
210 if (sb_any_quota_loaded(sb) && !sbi->s_qf_names[qtype]) {
211 f2fs_msg(sb, KERN_ERR,
212 "Cannot change journaled "
213 "quota options when quota turned on");
216 if (f2fs_sb_has_quota_ino(sb)) {
217 f2fs_msg(sb, KERN_INFO,
218 "QUOTA feature is enabled, so ignore qf_name");
222 qname = match_strdup(args);
224 f2fs_msg(sb, KERN_ERR,
225 "Not enough memory for storing quotafile name");
228 if (sbi->s_qf_names[qtype]) {
229 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
232 f2fs_msg(sb, KERN_ERR,
233 "%s quota file already specified",
237 if (strchr(qname, '/')) {
238 f2fs_msg(sb, KERN_ERR,
239 "quotafile must be on filesystem root");
242 sbi->s_qf_names[qtype] = qname;
250 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
252 struct f2fs_sb_info *sbi = F2FS_SB(sb);
254 if (sb_any_quota_loaded(sb) && sbi->s_qf_names[qtype]) {
255 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
256 " when quota turned on");
259 kfree(sbi->s_qf_names[qtype]);
260 sbi->s_qf_names[qtype] = NULL;
264 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
267 * We do the test below only for project quotas. 'usrquota' and
268 * 'grpquota' mount options are allowed even without quota feature
269 * to support legacy quotas in quota files.
271 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
272 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
273 "Cannot enable project quota enforcement.");
276 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA] ||
277 sbi->s_qf_names[PRJQUOTA]) {
278 if (test_opt(sbi, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
279 clear_opt(sbi, USRQUOTA);
281 if (test_opt(sbi, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
282 clear_opt(sbi, GRPQUOTA);
284 if (test_opt(sbi, PRJQUOTA) && sbi->s_qf_names[PRJQUOTA])
285 clear_opt(sbi, PRJQUOTA);
287 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
288 test_opt(sbi, PRJQUOTA)) {
289 f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
294 if (!sbi->s_jquota_fmt) {
295 f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
301 if (f2fs_sb_has_quota_ino(sbi->sb) && sbi->s_jquota_fmt) {
302 f2fs_msg(sbi->sb, KERN_INFO,
303 "QUOTA feature is enabled, so ignore jquota_fmt");
304 sbi->s_jquota_fmt = 0;
306 if (f2fs_sb_has_quota_ino(sbi->sb) && sb_rdonly(sbi->sb)) {
307 f2fs_msg(sbi->sb, KERN_INFO,
308 "Filesystem with quota feature cannot be mounted RDWR "
309 "without CONFIG_QUOTA");
316 static int parse_options(struct super_block *sb, char *options)
318 struct f2fs_sb_info *sbi = F2FS_SB(sb);
319 struct request_queue *q;
320 substring_t args[MAX_OPT_ARGS];
330 while ((p = strsep(&options, ",")) != NULL) {
335 * Initialize args struct so we know whether arg was
336 * found; some options take optional arguments.
338 args[0].to = args[0].from = NULL;
339 token = match_token(p, f2fs_tokens, args);
342 case Opt_gc_background:
343 name = match_strdup(&args[0]);
347 if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
349 clear_opt(sbi, FORCE_FG_GC);
350 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
351 clear_opt(sbi, BG_GC);
352 clear_opt(sbi, FORCE_FG_GC);
353 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
355 set_opt(sbi, FORCE_FG_GC);
362 case Opt_disable_roll_forward:
363 set_opt(sbi, DISABLE_ROLL_FORWARD);
366 /* this option mounts f2fs with ro */
367 set_opt(sbi, DISABLE_ROLL_FORWARD);
368 if (!f2fs_readonly(sb))
372 q = bdev_get_queue(sb->s_bdev);
373 if (blk_queue_discard(q)) {
374 set_opt(sbi, DISCARD);
375 } else if (!f2fs_sb_mounted_blkzoned(sb)) {
376 f2fs_msg(sb, KERN_WARNING,
377 "mounting with \"discard\" option, but "
378 "the device does not support discard");
382 if (f2fs_sb_mounted_blkzoned(sb)) {
383 f2fs_msg(sb, KERN_WARNING,
384 "discard is required for zoned block devices");
387 clear_opt(sbi, DISCARD);
390 set_opt(sbi, NOHEAP);
393 clear_opt(sbi, NOHEAP);
395 #ifdef CONFIG_F2FS_FS_XATTR
397 set_opt(sbi, XATTR_USER);
399 case Opt_nouser_xattr:
400 clear_opt(sbi, XATTR_USER);
402 case Opt_inline_xattr:
403 set_opt(sbi, INLINE_XATTR);
405 case Opt_noinline_xattr:
406 clear_opt(sbi, INLINE_XATTR);
408 case Opt_inline_xattr_size:
409 if (args->from && match_int(args, &arg))
411 set_opt(sbi, INLINE_XATTR_SIZE);
412 sbi->inline_xattr_size = arg;
416 f2fs_msg(sb, KERN_INFO,
417 "user_xattr options not supported");
419 case Opt_nouser_xattr:
420 f2fs_msg(sb, KERN_INFO,
421 "nouser_xattr options not supported");
423 case Opt_inline_xattr:
424 f2fs_msg(sb, KERN_INFO,
425 "inline_xattr options not supported");
427 case Opt_noinline_xattr:
428 f2fs_msg(sb, KERN_INFO,
429 "noinline_xattr options not supported");
432 #ifdef CONFIG_F2FS_FS_POSIX_ACL
434 set_opt(sbi, POSIX_ACL);
437 clear_opt(sbi, POSIX_ACL);
441 f2fs_msg(sb, KERN_INFO, "acl options not supported");
444 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
447 case Opt_active_logs:
448 if (args->from && match_int(args, &arg))
450 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
452 sbi->active_logs = arg;
454 case Opt_disable_ext_identify:
455 set_opt(sbi, DISABLE_EXT_IDENTIFY);
457 case Opt_inline_data:
458 set_opt(sbi, INLINE_DATA);
460 case Opt_inline_dentry:
461 set_opt(sbi, INLINE_DENTRY);
463 case Opt_noinline_dentry:
464 clear_opt(sbi, INLINE_DENTRY);
466 case Opt_flush_merge:
467 set_opt(sbi, FLUSH_MERGE);
469 case Opt_noflush_merge:
470 clear_opt(sbi, FLUSH_MERGE);
473 set_opt(sbi, NOBARRIER);
476 set_opt(sbi, FASTBOOT);
478 case Opt_extent_cache:
479 set_opt(sbi, EXTENT_CACHE);
481 case Opt_noextent_cache:
482 clear_opt(sbi, EXTENT_CACHE);
484 case Opt_noinline_data:
485 clear_opt(sbi, INLINE_DATA);
488 set_opt(sbi, DATA_FLUSH);
491 name = match_strdup(&args[0]);
495 if (strlen(name) == 8 &&
496 !strncmp(name, "adaptive", 8)) {
497 if (f2fs_sb_mounted_blkzoned(sb)) {
498 f2fs_msg(sb, KERN_WARNING,
499 "adaptive mode is not allowed with "
500 "zoned block device feature");
504 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
505 } else if (strlen(name) == 3 &&
506 !strncmp(name, "lfs", 3)) {
507 set_opt_mode(sbi, F2FS_MOUNT_LFS);
514 case Opt_io_size_bits:
515 if (args->from && match_int(args, &arg))
517 if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
518 f2fs_msg(sb, KERN_WARNING,
519 "Not support %d, larger than %d",
520 1 << arg, BIO_MAX_PAGES);
523 sbi->write_io_size_bits = arg;
525 case Opt_fault_injection:
526 if (args->from && match_int(args, &arg))
528 #ifdef CONFIG_F2FS_FAULT_INJECTION
529 f2fs_build_fault_attr(sbi, arg);
530 set_opt(sbi, FAULT_INJECTION);
532 f2fs_msg(sb, KERN_INFO,
533 "FAULT_INJECTION was not selected");
537 sb->s_flags |= MS_LAZYTIME;
540 sb->s_flags &= ~MS_LAZYTIME;
545 set_opt(sbi, USRQUOTA);
548 set_opt(sbi, GRPQUOTA);
551 set_opt(sbi, PRJQUOTA);
554 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
559 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
564 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
568 case Opt_offusrjquota:
569 ret = f2fs_clear_qf_name(sb, USRQUOTA);
573 case Opt_offgrpjquota:
574 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
578 case Opt_offprjjquota:
579 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
583 case Opt_jqfmt_vfsold:
584 sbi->s_jquota_fmt = QFMT_VFS_OLD;
586 case Opt_jqfmt_vfsv0:
587 sbi->s_jquota_fmt = QFMT_VFS_V0;
589 case Opt_jqfmt_vfsv1:
590 sbi->s_jquota_fmt = QFMT_VFS_V1;
593 clear_opt(sbi, QUOTA);
594 clear_opt(sbi, USRQUOTA);
595 clear_opt(sbi, GRPQUOTA);
596 clear_opt(sbi, PRJQUOTA);
606 case Opt_offusrjquota:
607 case Opt_offgrpjquota:
608 case Opt_offprjjquota:
609 case Opt_jqfmt_vfsold:
610 case Opt_jqfmt_vfsv0:
611 case Opt_jqfmt_vfsv1:
613 f2fs_msg(sb, KERN_INFO,
614 "quota operations not supported");
618 f2fs_msg(sb, KERN_ERR,
619 "Unrecognized mount option \"%s\" or missing value",
625 if (f2fs_check_quota_options(sbi))
629 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
630 f2fs_msg(sb, KERN_ERR,
631 "Should set mode=lfs with %uKB-sized IO",
632 F2FS_IO_SIZE_KB(sbi));
636 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
637 if (!test_opt(sbi, INLINE_XATTR)) {
638 f2fs_msg(sb, KERN_ERR,
639 "inline_xattr_size option should be "
640 "set with inline_xattr option");
643 if (!sbi->inline_xattr_size ||
644 sbi->inline_xattr_size >= DEF_ADDRS_PER_INODE -
645 F2FS_TOTAL_EXTRA_ATTR_SIZE -
646 DEF_INLINE_RESERVED_SIZE -
647 DEF_MIN_INLINE_SIZE) {
648 f2fs_msg(sb, KERN_ERR,
649 "inline xattr size is out of range");
656 static struct inode *f2fs_alloc_inode(struct super_block *sb)
658 struct f2fs_inode_info *fi;
660 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
664 init_once((void *) fi);
666 /* Initialize f2fs-specific inode info */
667 atomic_set(&fi->dirty_pages, 0);
668 fi->i_current_depth = 1;
670 init_rwsem(&fi->i_sem);
671 INIT_LIST_HEAD(&fi->dirty_list);
672 INIT_LIST_HEAD(&fi->gdirty_list);
673 INIT_LIST_HEAD(&fi->inmem_ilist);
674 INIT_LIST_HEAD(&fi->inmem_pages);
675 mutex_init(&fi->inmem_lock);
676 init_rwsem(&fi->dio_rwsem[READ]);
677 init_rwsem(&fi->dio_rwsem[WRITE]);
678 init_rwsem(&fi->i_mmap_sem);
679 init_rwsem(&fi->i_xattr_sem);
682 memset(&fi->i_dquot, 0, sizeof(fi->i_dquot));
683 fi->i_reserved_quota = 0;
685 /* Will be used by directory only */
686 fi->i_dir_level = F2FS_SB(sb)->dir_level;
688 return &fi->vfs_inode;
691 static int f2fs_drop_inode(struct inode *inode)
695 * This is to avoid a deadlock condition like below.
696 * writeback_single_inode(inode)
697 * - f2fs_write_data_page
698 * - f2fs_gc -> iput -> evict
699 * - inode_wait_for_writeback(inode)
701 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
702 if (!inode->i_nlink && !is_bad_inode(inode)) {
703 /* to avoid evict_inode call simultaneously */
704 atomic_inc(&inode->i_count);
705 spin_unlock(&inode->i_lock);
707 /* some remained atomic pages should discarded */
708 if (f2fs_is_atomic_file(inode))
709 drop_inmem_pages(inode);
711 /* should remain fi->extent_tree for writepage */
712 f2fs_destroy_extent_node(inode);
714 sb_start_intwrite(inode->i_sb);
715 f2fs_i_size_write(inode, 0);
717 if (F2FS_HAS_BLOCKS(inode))
718 f2fs_truncate(inode);
720 sb_end_intwrite(inode->i_sb);
722 spin_lock(&inode->i_lock);
723 atomic_dec(&inode->i_count);
725 trace_f2fs_drop_inode(inode, 0);
728 ret = generic_drop_inode(inode);
729 trace_f2fs_drop_inode(inode, ret);
733 int f2fs_inode_dirtied(struct inode *inode, bool sync)
735 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
738 spin_lock(&sbi->inode_lock[DIRTY_META]);
739 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
742 set_inode_flag(inode, FI_DIRTY_INODE);
743 stat_inc_dirty_inode(sbi, DIRTY_META);
745 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
746 list_add_tail(&F2FS_I(inode)->gdirty_list,
747 &sbi->inode_list[DIRTY_META]);
748 inc_page_count(sbi, F2FS_DIRTY_IMETA);
750 spin_unlock(&sbi->inode_lock[DIRTY_META]);
754 void f2fs_inode_synced(struct inode *inode)
756 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
758 spin_lock(&sbi->inode_lock[DIRTY_META]);
759 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
760 spin_unlock(&sbi->inode_lock[DIRTY_META]);
763 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
764 list_del_init(&F2FS_I(inode)->gdirty_list);
765 dec_page_count(sbi, F2FS_DIRTY_IMETA);
767 clear_inode_flag(inode, FI_DIRTY_INODE);
768 clear_inode_flag(inode, FI_AUTO_RECOVER);
769 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
770 spin_unlock(&sbi->inode_lock[DIRTY_META]);
774 * f2fs_dirty_inode() is called from __mark_inode_dirty()
776 * We should call set_dirty_inode to write the dirty inode through write_inode.
778 static void f2fs_dirty_inode(struct inode *inode, int flags)
780 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
782 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
783 inode->i_ino == F2FS_META_INO(sbi))
786 if (flags == I_DIRTY_TIME)
789 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
790 clear_inode_flag(inode, FI_AUTO_RECOVER);
792 f2fs_inode_dirtied(inode, false);
795 static void f2fs_i_callback(struct rcu_head *head)
797 struct inode *inode = container_of(head, struct inode, i_rcu);
798 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
801 static void f2fs_destroy_inode(struct inode *inode)
803 call_rcu(&inode->i_rcu, f2fs_i_callback);
806 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
808 percpu_counter_destroy(&sbi->alloc_valid_block_count);
809 percpu_counter_destroy(&sbi->total_valid_inode_count);
812 static void destroy_device_list(struct f2fs_sb_info *sbi)
816 for (i = 0; i < sbi->s_ndevs; i++) {
817 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
818 #ifdef CONFIG_BLK_DEV_ZONED
819 kfree(FDEV(i).blkz_type);
825 static void f2fs_put_super(struct super_block *sb)
827 struct f2fs_sb_info *sbi = F2FS_SB(sb);
831 f2fs_quota_off_umount(sb);
833 /* prevent remaining shrinker jobs */
834 mutex_lock(&sbi->umount_mutex);
837 * We don't need to do checkpoint when superblock is clean.
838 * But, the previous checkpoint was not done by umount, it needs to do
839 * clean checkpoint again.
841 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
842 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
843 struct cp_control cpc = {
846 write_checkpoint(sbi, &cpc);
849 /* be sure to wait for any on-going discard commands */
850 dropped = f2fs_wait_discard_bios(sbi);
852 if (f2fs_discard_en(sbi) && !sbi->discard_blks && !dropped) {
853 struct cp_control cpc = {
854 .reason = CP_UMOUNT | CP_TRIMMED,
856 write_checkpoint(sbi, &cpc);
859 /* write_checkpoint can update stat informaion */
860 f2fs_destroy_stats(sbi);
863 * normally superblock is clean, so we need to release this.
864 * In addition, EIO will skip do checkpoint, we need this as well.
866 release_ino_entry(sbi, true);
868 f2fs_leave_shrinker(sbi);
869 mutex_unlock(&sbi->umount_mutex);
871 /* our cp_error case, we can wait for any writeback page */
872 f2fs_flush_merged_writes(sbi);
874 iput(sbi->node_inode);
875 iput(sbi->meta_inode);
877 /* destroy f2fs internal modules */
878 destroy_node_manager(sbi);
879 destroy_segment_manager(sbi);
883 f2fs_unregister_sysfs(sbi);
885 sb->s_fs_info = NULL;
886 if (sbi->s_chksum_driver)
887 crypto_free_shash(sbi->s_chksum_driver);
888 kfree(sbi->raw_super);
890 destroy_device_list(sbi);
891 mempool_destroy(sbi->write_io_dummy);
893 for (i = 0; i < MAXQUOTAS; i++)
894 kfree(sbi->s_qf_names[i]);
896 destroy_percpu_info(sbi);
897 for (i = 0; i < NR_PAGE_TYPE; i++)
898 kfree(sbi->write_io[i]);
902 int f2fs_sync_fs(struct super_block *sb, int sync)
904 struct f2fs_sb_info *sbi = F2FS_SB(sb);
907 if (unlikely(f2fs_cp_error(sbi)))
910 trace_f2fs_sync_fs(sb, sync);
912 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
916 struct cp_control cpc;
918 cpc.reason = __get_cp_reason(sbi);
920 mutex_lock(&sbi->gc_mutex);
921 err = write_checkpoint(sbi, &cpc);
922 mutex_unlock(&sbi->gc_mutex);
924 f2fs_trace_ios(NULL, 1);
929 static int f2fs_freeze(struct super_block *sb)
931 if (f2fs_readonly(sb))
934 /* IO error happened before */
935 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
938 /* must be clean, since sync_filesystem() was already called */
939 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
944 static int f2fs_unfreeze(struct super_block *sb)
950 static int f2fs_statfs_project(struct super_block *sb,
951 kprojid_t projid, struct kstatfs *buf)
958 qid = make_kqid_projid(projid);
959 dquot = dqget(sb, qid);
961 return PTR_ERR(dquot);
962 spin_lock(&dq_data_lock);
964 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
965 dquot->dq_dqb.dqb_bsoftlimit :
966 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
967 if (limit && buf->f_blocks > limit) {
968 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
969 buf->f_blocks = limit;
970 buf->f_bfree = buf->f_bavail =
971 (buf->f_blocks > curblock) ?
972 (buf->f_blocks - curblock) : 0;
975 limit = dquot->dq_dqb.dqb_isoftlimit ?
976 dquot->dq_dqb.dqb_isoftlimit :
977 dquot->dq_dqb.dqb_ihardlimit;
978 if (limit && buf->f_files > limit) {
979 buf->f_files = limit;
981 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
982 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
985 spin_unlock(&dq_data_lock);
991 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
993 struct super_block *sb = dentry->d_sb;
994 struct f2fs_sb_info *sbi = F2FS_SB(sb);
995 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
996 block_t total_count, user_block_count, start_count, ovp_count;
997 u64 avail_node_count;
999 total_count = le64_to_cpu(sbi->raw_super->block_count);
1000 user_block_count = sbi->user_block_count;
1001 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1002 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
1003 buf->f_type = F2FS_SUPER_MAGIC;
1004 buf->f_bsize = sbi->blocksize;
1006 buf->f_blocks = total_count - start_count;
1007 buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
1008 buf->f_bavail = user_block_count - valid_user_blocks(sbi) -
1009 sbi->current_reserved_blocks;
1011 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1013 if (avail_node_count > user_block_count) {
1014 buf->f_files = user_block_count;
1015 buf->f_ffree = buf->f_bavail;
1017 buf->f_files = avail_node_count;
1018 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1022 buf->f_namelen = F2FS_NAME_LEN;
1023 buf->f_fsid.val[0] = (u32)id;
1024 buf->f_fsid.val[1] = (u32)(id >> 32);
1027 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1028 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1029 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1035 static inline void f2fs_show_quota_options(struct seq_file *seq,
1036 struct super_block *sb)
1039 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1041 if (sbi->s_jquota_fmt) {
1044 switch (sbi->s_jquota_fmt) {
1055 seq_printf(seq, ",jqfmt=%s", fmtname);
1058 if (sbi->s_qf_names[USRQUOTA])
1059 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1061 if (sbi->s_qf_names[GRPQUOTA])
1062 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1064 if (sbi->s_qf_names[PRJQUOTA])
1065 seq_show_option(seq, "prjjquota", sbi->s_qf_names[PRJQUOTA]);
1069 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1071 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1073 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1074 if (test_opt(sbi, FORCE_FG_GC))
1075 seq_printf(seq, ",background_gc=%s", "sync");
1077 seq_printf(seq, ",background_gc=%s", "on");
1079 seq_printf(seq, ",background_gc=%s", "off");
1081 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1082 seq_puts(seq, ",disable_roll_forward");
1083 if (test_opt(sbi, DISCARD))
1084 seq_puts(seq, ",discard");
1085 if (test_opt(sbi, NOHEAP))
1086 seq_puts(seq, ",no_heap");
1088 seq_puts(seq, ",heap");
1089 #ifdef CONFIG_F2FS_FS_XATTR
1090 if (test_opt(sbi, XATTR_USER))
1091 seq_puts(seq, ",user_xattr");
1093 seq_puts(seq, ",nouser_xattr");
1094 if (test_opt(sbi, INLINE_XATTR))
1095 seq_puts(seq, ",inline_xattr");
1097 seq_puts(seq, ",noinline_xattr");
1098 if (test_opt(sbi, INLINE_XATTR_SIZE))
1099 seq_printf(seq, ",inline_xattr_size=%u",
1100 sbi->inline_xattr_size);
1102 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1103 if (test_opt(sbi, POSIX_ACL))
1104 seq_puts(seq, ",acl");
1106 seq_puts(seq, ",noacl");
1108 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1109 seq_puts(seq, ",disable_ext_identify");
1110 if (test_opt(sbi, INLINE_DATA))
1111 seq_puts(seq, ",inline_data");
1113 seq_puts(seq, ",noinline_data");
1114 if (test_opt(sbi, INLINE_DENTRY))
1115 seq_puts(seq, ",inline_dentry");
1117 seq_puts(seq, ",noinline_dentry");
1118 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1119 seq_puts(seq, ",flush_merge");
1120 if (test_opt(sbi, NOBARRIER))
1121 seq_puts(seq, ",nobarrier");
1122 if (test_opt(sbi, FASTBOOT))
1123 seq_puts(seq, ",fastboot");
1124 if (test_opt(sbi, EXTENT_CACHE))
1125 seq_puts(seq, ",extent_cache");
1127 seq_puts(seq, ",noextent_cache");
1128 if (test_opt(sbi, DATA_FLUSH))
1129 seq_puts(seq, ",data_flush");
1131 seq_puts(seq, ",mode=");
1132 if (test_opt(sbi, ADAPTIVE))
1133 seq_puts(seq, "adaptive");
1134 else if (test_opt(sbi, LFS))
1135 seq_puts(seq, "lfs");
1136 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
1137 if (F2FS_IO_SIZE_BITS(sbi))
1138 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1139 #ifdef CONFIG_F2FS_FAULT_INJECTION
1140 if (test_opt(sbi, FAULT_INJECTION))
1141 seq_printf(seq, ",fault_injection=%u",
1142 sbi->fault_info.inject_rate);
1145 if (test_opt(sbi, QUOTA))
1146 seq_puts(seq, ",quota");
1147 if (test_opt(sbi, USRQUOTA))
1148 seq_puts(seq, ",usrquota");
1149 if (test_opt(sbi, GRPQUOTA))
1150 seq_puts(seq, ",grpquota");
1151 if (test_opt(sbi, PRJQUOTA))
1152 seq_puts(seq, ",prjquota");
1154 f2fs_show_quota_options(seq, sbi->sb);
1159 static void default_options(struct f2fs_sb_info *sbi)
1161 /* init some FS parameters */
1162 sbi->active_logs = NR_CURSEG_TYPE;
1163 sbi->inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1165 set_opt(sbi, BG_GC);
1166 set_opt(sbi, INLINE_XATTR);
1167 set_opt(sbi, INLINE_DATA);
1168 set_opt(sbi, INLINE_DENTRY);
1169 set_opt(sbi, EXTENT_CACHE);
1170 set_opt(sbi, NOHEAP);
1171 sbi->sb->s_flags |= MS_LAZYTIME;
1172 set_opt(sbi, FLUSH_MERGE);
1173 if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1174 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1175 set_opt(sbi, DISCARD);
1177 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1180 #ifdef CONFIG_F2FS_FS_XATTR
1181 set_opt(sbi, XATTR_USER);
1183 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1184 set_opt(sbi, POSIX_ACL);
1187 #ifdef CONFIG_F2FS_FAULT_INJECTION
1188 f2fs_build_fault_attr(sbi, 0);
1193 static int f2fs_enable_quotas(struct super_block *sb);
1195 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1197 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1198 struct f2fs_mount_info org_mount_opt;
1199 unsigned long old_sb_flags;
1200 int err, active_logs;
1201 bool need_restart_gc = false;
1202 bool need_stop_gc = false;
1203 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1204 #ifdef CONFIG_F2FS_FAULT_INJECTION
1205 struct f2fs_fault_info ffi = sbi->fault_info;
1209 char *s_qf_names[MAXQUOTAS];
1214 * Save the old mount options in case we
1215 * need to restore them.
1217 org_mount_opt = sbi->mount_opt;
1218 old_sb_flags = sb->s_flags;
1219 active_logs = sbi->active_logs;
1222 s_jquota_fmt = sbi->s_jquota_fmt;
1223 for (i = 0; i < MAXQUOTAS; i++) {
1224 if (sbi->s_qf_names[i]) {
1225 s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
1227 if (!s_qf_names[i]) {
1228 for (j = 0; j < i; j++)
1229 kfree(s_qf_names[j]);
1233 s_qf_names[i] = NULL;
1238 /* recover superblocks we couldn't write due to previous RO mount */
1239 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1240 err = f2fs_commit_super(sbi, false);
1241 f2fs_msg(sb, KERN_INFO,
1242 "Try to recover all the superblocks, ret: %d", err);
1244 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1247 default_options(sbi);
1249 /* parse mount options */
1250 err = parse_options(sb, data);
1255 * Previous and new state of filesystem is RO,
1256 * so skip checking GC and FLUSH_MERGE conditions.
1258 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1262 if (!f2fs_readonly(sb) && (*flags & MS_RDONLY)) {
1263 err = dquot_suspend(sb, -1);
1267 /* dquot_resume needs RW */
1268 sb->s_flags &= ~MS_RDONLY;
1269 if (sb_any_quota_suspended(sb)) {
1270 dquot_resume(sb, -1);
1271 } else if (f2fs_sb_has_quota_ino(sb)) {
1272 err = f2fs_enable_quotas(sb);
1278 /* disallow enable/disable extent_cache dynamically */
1279 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1281 f2fs_msg(sbi->sb, KERN_WARNING,
1282 "switch extent_cache option is not allowed");
1287 * We stop the GC thread if FS is mounted as RO
1288 * or if background_gc = off is passed in mount
1289 * option. Also sync the filesystem.
1291 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1292 if (sbi->gc_thread) {
1293 stop_gc_thread(sbi);
1294 need_restart_gc = true;
1296 } else if (!sbi->gc_thread) {
1297 err = start_gc_thread(sbi);
1300 need_stop_gc = true;
1303 if (*flags & MS_RDONLY) {
1304 writeback_inodes_sb(sb, WB_REASON_SYNC);
1307 set_sbi_flag(sbi, SBI_IS_DIRTY);
1308 set_sbi_flag(sbi, SBI_IS_CLOSE);
1309 f2fs_sync_fs(sb, 1);
1310 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1314 * We stop issue flush thread if FS is mounted as RO
1315 * or if flush_merge is not passed in mount option.
1317 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1318 clear_opt(sbi, FLUSH_MERGE);
1319 destroy_flush_cmd_control(sbi, false);
1321 err = create_flush_cmd_control(sbi);
1327 /* Release old quota file names */
1328 for (i = 0; i < MAXQUOTAS; i++)
1329 kfree(s_qf_names[i]);
1331 /* Update the POSIXACL Flag */
1332 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1333 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1337 if (need_restart_gc) {
1338 if (start_gc_thread(sbi))
1339 f2fs_msg(sbi->sb, KERN_WARNING,
1340 "background gc thread has stopped");
1341 } else if (need_stop_gc) {
1342 stop_gc_thread(sbi);
1346 sbi->s_jquota_fmt = s_jquota_fmt;
1347 for (i = 0; i < MAXQUOTAS; i++) {
1348 kfree(sbi->s_qf_names[i]);
1349 sbi->s_qf_names[i] = s_qf_names[i];
1352 sbi->mount_opt = org_mount_opt;
1353 sbi->active_logs = active_logs;
1354 sb->s_flags = old_sb_flags;
1355 #ifdef CONFIG_F2FS_FAULT_INJECTION
1356 sbi->fault_info = ffi;
1362 /* Read data from quotafile */
1363 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1364 size_t len, loff_t off)
1366 struct inode *inode = sb_dqopt(sb)->files[type];
1367 struct address_space *mapping = inode->i_mapping;
1368 block_t blkidx = F2FS_BYTES_TO_BLK(off);
1369 int offset = off & (sb->s_blocksize - 1);
1372 loff_t i_size = i_size_read(inode);
1379 if (off + len > i_size)
1382 while (toread > 0) {
1383 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1385 page = read_mapping_page(mapping, blkidx, NULL);
1387 if (PTR_ERR(page) == -ENOMEM) {
1388 congestion_wait(BLK_RW_ASYNC, HZ/50);
1391 return PTR_ERR(page);
1396 if (unlikely(page->mapping != mapping)) {
1397 f2fs_put_page(page, 1);
1400 if (unlikely(!PageUptodate(page))) {
1401 f2fs_put_page(page, 1);
1405 kaddr = kmap_atomic(page);
1406 memcpy(data, kaddr + offset, tocopy);
1407 kunmap_atomic(kaddr);
1408 f2fs_put_page(page, 1);
1418 /* Write to quotafile */
1419 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1420 const char *data, size_t len, loff_t off)
1422 struct inode *inode = sb_dqopt(sb)->files[type];
1423 struct address_space *mapping = inode->i_mapping;
1424 const struct address_space_operations *a_ops = mapping->a_ops;
1425 int offset = off & (sb->s_blocksize - 1);
1426 size_t towrite = len;
1432 while (towrite > 0) {
1433 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1436 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1438 if (unlikely(err)) {
1439 if (err == -ENOMEM) {
1440 congestion_wait(BLK_RW_ASYNC, HZ/50);
1446 kaddr = kmap_atomic(page);
1447 memcpy(kaddr + offset, data, tocopy);
1448 kunmap_atomic(kaddr);
1449 flush_dcache_page(page);
1451 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1462 inode->i_mtime = inode->i_ctime = current_time(inode);
1463 f2fs_mark_inode_dirty_sync(inode, false);
1464 return len - towrite;
1467 static struct dquot **f2fs_get_dquots(struct inode *inode)
1469 return F2FS_I(inode)->i_dquot;
1472 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1474 return &F2FS_I(inode)->i_reserved_quota;
1477 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1479 return dquot_quota_on_mount(sbi->sb, sbi->s_qf_names[type],
1480 sbi->s_jquota_fmt, type);
1483 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1488 if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) {
1489 err = f2fs_enable_quotas(sbi->sb);
1491 f2fs_msg(sbi->sb, KERN_ERR,
1492 "Cannot turn on quota_ino: %d", err);
1498 for (i = 0; i < MAXQUOTAS; i++) {
1499 if (sbi->s_qf_names[i]) {
1500 err = f2fs_quota_on_mount(sbi, i);
1505 f2fs_msg(sbi->sb, KERN_ERR,
1506 "Cannot turn on quotas: %d on %d", err, i);
1512 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1515 struct inode *qf_inode;
1516 unsigned long qf_inum;
1519 BUG_ON(!f2fs_sb_has_quota_ino(sb));
1521 qf_inum = f2fs_qf_ino(sb, type);
1525 qf_inode = f2fs_iget(sb, qf_inum);
1526 if (IS_ERR(qf_inode)) {
1527 f2fs_msg(sb, KERN_ERR,
1528 "Bad quota inode %u:%lu", type, qf_inum);
1529 return PTR_ERR(qf_inode);
1532 /* Don't account quota for quota files to avoid recursion */
1533 qf_inode->i_flags |= S_NOQUOTA;
1534 err = dquot_enable(qf_inode, type, format_id, flags);
1539 static int f2fs_enable_quotas(struct super_block *sb)
1542 unsigned long qf_inum;
1543 bool quota_mopt[MAXQUOTAS] = {
1544 test_opt(F2FS_SB(sb), USRQUOTA),
1545 test_opt(F2FS_SB(sb), GRPQUOTA),
1546 test_opt(F2FS_SB(sb), PRJQUOTA),
1549 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
1550 for (type = 0; type < MAXQUOTAS; type++) {
1551 qf_inum = f2fs_qf_ino(sb, type);
1553 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1554 DQUOT_USAGE_ENABLED |
1555 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1557 f2fs_msg(sb, KERN_ERR,
1558 "Failed to enable quota tracking "
1559 "(type=%d, err=%d). Please run "
1560 "fsck to fix.", type, err);
1561 for (type--; type >= 0; type--)
1562 dquot_quota_off(sb, type);
1570 static int f2fs_quota_sync(struct super_block *sb, int type)
1572 struct quota_info *dqopt = sb_dqopt(sb);
1576 ret = dquot_writeback_dquots(sb, type);
1581 * Now when everything is written we can discard the pagecache so
1582 * that userspace sees the changes.
1584 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1585 if (type != -1 && cnt != type)
1587 if (!sb_has_quota_active(sb, cnt))
1590 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1594 inode_lock(dqopt->files[cnt]);
1595 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1596 inode_unlock(dqopt->files[cnt]);
1601 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1602 const struct path *path)
1604 struct inode *inode;
1607 err = f2fs_quota_sync(sb, type);
1611 err = dquot_quota_on(sb, type, format_id, path);
1615 inode = d_inode(path->dentry);
1618 F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL;
1619 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1620 S_NOATIME | S_IMMUTABLE);
1621 inode_unlock(inode);
1622 f2fs_mark_inode_dirty_sync(inode, false);
1627 static int f2fs_quota_off(struct super_block *sb, int type)
1629 struct inode *inode = sb_dqopt(sb)->files[type];
1632 if (!inode || !igrab(inode))
1633 return dquot_quota_off(sb, type);
1635 f2fs_quota_sync(sb, type);
1637 err = dquot_quota_off(sb, type);
1638 if (err || f2fs_sb_has_quota_ino(sb))
1642 F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL);
1643 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1644 inode_unlock(inode);
1645 f2fs_mark_inode_dirty_sync(inode, false);
1651 void f2fs_quota_off_umount(struct super_block *sb)
1655 for (type = 0; type < MAXQUOTAS; type++)
1656 f2fs_quota_off(sb, type);
1659 int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1661 *projid = F2FS_I(inode)->i_projid;
1665 static const struct dquot_operations f2fs_quota_operations = {
1666 .get_reserved_space = f2fs_get_reserved_space,
1667 .write_dquot = dquot_commit,
1668 .acquire_dquot = dquot_acquire,
1669 .release_dquot = dquot_release,
1670 .mark_dirty = dquot_mark_dquot_dirty,
1671 .write_info = dquot_commit_info,
1672 .alloc_dquot = dquot_alloc,
1673 .destroy_dquot = dquot_destroy,
1674 .get_projid = f2fs_get_projid,
1675 .get_next_id = dquot_get_next_id,
1678 static const struct quotactl_ops f2fs_quotactl_ops = {
1679 .quota_on = f2fs_quota_on,
1680 .quota_off = f2fs_quota_off,
1681 .quota_sync = f2fs_quota_sync,
1682 .get_state = dquot_get_state,
1683 .set_info = dquot_set_dqinfo,
1684 .get_dqblk = dquot_get_dqblk,
1685 .set_dqblk = dquot_set_dqblk,
1686 .get_nextdqblk = dquot_get_next_dqblk,
1689 void f2fs_quota_off_umount(struct super_block *sb)
1694 static const struct super_operations f2fs_sops = {
1695 .alloc_inode = f2fs_alloc_inode,
1696 .drop_inode = f2fs_drop_inode,
1697 .destroy_inode = f2fs_destroy_inode,
1698 .write_inode = f2fs_write_inode,
1699 .dirty_inode = f2fs_dirty_inode,
1700 .show_options = f2fs_show_options,
1702 .quota_read = f2fs_quota_read,
1703 .quota_write = f2fs_quota_write,
1704 .get_dquots = f2fs_get_dquots,
1706 .evict_inode = f2fs_evict_inode,
1707 .put_super = f2fs_put_super,
1708 .sync_fs = f2fs_sync_fs,
1709 .freeze_fs = f2fs_freeze,
1710 .unfreeze_fs = f2fs_unfreeze,
1711 .statfs = f2fs_statfs,
1712 .remount_fs = f2fs_remount,
1715 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1716 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1718 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1719 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1723 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1726 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1727 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1728 ctx, len, fs_data, XATTR_CREATE);
1731 static unsigned f2fs_max_namelen(struct inode *inode)
1733 return S_ISLNK(inode->i_mode) ?
1734 inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1737 static const struct fscrypt_operations f2fs_cryptops = {
1738 .key_prefix = "f2fs:",
1739 .get_context = f2fs_get_context,
1740 .set_context = f2fs_set_context,
1741 .empty_dir = f2fs_empty_dir,
1742 .max_namelen = f2fs_max_namelen,
1746 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1747 u64 ino, u32 generation)
1749 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1750 struct inode *inode;
1752 if (check_nid_range(sbi, ino))
1753 return ERR_PTR(-ESTALE);
1756 * f2fs_iget isn't quite right if the inode is currently unallocated!
1757 * However f2fs_iget currently does appropriate checks to handle stale
1758 * inodes so everything is OK.
1760 inode = f2fs_iget(sb, ino);
1762 return ERR_CAST(inode);
1763 if (unlikely(generation && inode->i_generation != generation)) {
1764 /* we didn't find the right inode.. */
1766 return ERR_PTR(-ESTALE);
1771 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1772 int fh_len, int fh_type)
1774 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1775 f2fs_nfs_get_inode);
1778 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1779 int fh_len, int fh_type)
1781 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1782 f2fs_nfs_get_inode);
1785 static const struct export_operations f2fs_export_ops = {
1786 .fh_to_dentry = f2fs_fh_to_dentry,
1787 .fh_to_parent = f2fs_fh_to_parent,
1788 .get_parent = f2fs_get_parent,
1791 static loff_t max_file_blocks(void)
1794 loff_t leaf_count = ADDRS_PER_BLOCK;
1797 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1798 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1799 * space in inode.i_addr, it will be more safe to reassign
1803 /* two direct node blocks */
1804 result += (leaf_count * 2);
1806 /* two indirect node blocks */
1807 leaf_count *= NIDS_PER_BLOCK;
1808 result += (leaf_count * 2);
1810 /* one double indirect node block */
1811 leaf_count *= NIDS_PER_BLOCK;
1812 result += leaf_count;
1817 static int __f2fs_commit_super(struct buffer_head *bh,
1818 struct f2fs_super_block *super)
1822 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1823 set_buffer_uptodate(bh);
1824 set_buffer_dirty(bh);
1827 /* it's rare case, we can do fua all the time */
1828 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
1831 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1832 struct buffer_head *bh)
1834 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1835 (bh->b_data + F2FS_SUPER_OFFSET);
1836 struct super_block *sb = sbi->sb;
1837 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1838 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1839 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1840 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1841 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1842 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1843 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1844 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1845 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1846 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1847 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1848 u32 segment_count = le32_to_cpu(raw_super->segment_count);
1849 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1850 u64 main_end_blkaddr = main_blkaddr +
1851 (segment_count_main << log_blocks_per_seg);
1852 u64 seg_end_blkaddr = segment0_blkaddr +
1853 (segment_count << log_blocks_per_seg);
1855 if (segment0_blkaddr != cp_blkaddr) {
1856 f2fs_msg(sb, KERN_INFO,
1857 "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1858 segment0_blkaddr, cp_blkaddr);
1862 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1864 f2fs_msg(sb, KERN_INFO,
1865 "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1866 cp_blkaddr, sit_blkaddr,
1867 segment_count_ckpt << log_blocks_per_seg);
1871 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1873 f2fs_msg(sb, KERN_INFO,
1874 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1875 sit_blkaddr, nat_blkaddr,
1876 segment_count_sit << log_blocks_per_seg);
1880 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1882 f2fs_msg(sb, KERN_INFO,
1883 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1884 nat_blkaddr, ssa_blkaddr,
1885 segment_count_nat << log_blocks_per_seg);
1889 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1891 f2fs_msg(sb, KERN_INFO,
1892 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1893 ssa_blkaddr, main_blkaddr,
1894 segment_count_ssa << log_blocks_per_seg);
1898 if (main_end_blkaddr > seg_end_blkaddr) {
1899 f2fs_msg(sb, KERN_INFO,
1900 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1903 (segment_count << log_blocks_per_seg),
1904 segment_count_main << log_blocks_per_seg);
1906 } else if (main_end_blkaddr < seg_end_blkaddr) {
1910 /* fix in-memory information all the time */
1911 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1912 segment0_blkaddr) >> log_blocks_per_seg);
1914 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1915 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1918 err = __f2fs_commit_super(bh, NULL);
1919 res = err ? "failed" : "done";
1921 f2fs_msg(sb, KERN_INFO,
1922 "Fix alignment : %s, start(%u) end(%u) block(%u)",
1925 (segment_count << log_blocks_per_seg),
1926 segment_count_main << log_blocks_per_seg);
1933 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1934 struct buffer_head *bh)
1936 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1937 (bh->b_data + F2FS_SUPER_OFFSET);
1938 struct super_block *sb = sbi->sb;
1939 unsigned int blocksize;
1941 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
1942 f2fs_msg(sb, KERN_INFO,
1943 "Magic Mismatch, valid(0x%x) - read(0x%x)",
1944 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1948 /* Currently, support only 4KB page cache size */
1949 if (F2FS_BLKSIZE != PAGE_SIZE) {
1950 f2fs_msg(sb, KERN_INFO,
1951 "Invalid page_cache_size (%lu), supports only 4KB\n",
1956 /* Currently, support only 4KB block size */
1957 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1958 if (blocksize != F2FS_BLKSIZE) {
1959 f2fs_msg(sb, KERN_INFO,
1960 "Invalid blocksize (%u), supports only 4KB\n",
1965 /* check log blocks per segment */
1966 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1967 f2fs_msg(sb, KERN_INFO,
1968 "Invalid log blocks per segment (%u)\n",
1969 le32_to_cpu(raw_super->log_blocks_per_seg));
1973 /* Currently, support 512/1024/2048/4096 bytes sector size */
1974 if (le32_to_cpu(raw_super->log_sectorsize) >
1975 F2FS_MAX_LOG_SECTOR_SIZE ||
1976 le32_to_cpu(raw_super->log_sectorsize) <
1977 F2FS_MIN_LOG_SECTOR_SIZE) {
1978 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1979 le32_to_cpu(raw_super->log_sectorsize));
1982 if (le32_to_cpu(raw_super->log_sectors_per_block) +
1983 le32_to_cpu(raw_super->log_sectorsize) !=
1984 F2FS_MAX_LOG_SECTOR_SIZE) {
1985 f2fs_msg(sb, KERN_INFO,
1986 "Invalid log sectors per block(%u) log sectorsize(%u)",
1987 le32_to_cpu(raw_super->log_sectors_per_block),
1988 le32_to_cpu(raw_super->log_sectorsize));
1992 /* check reserved ino info */
1993 if (le32_to_cpu(raw_super->node_ino) != 1 ||
1994 le32_to_cpu(raw_super->meta_ino) != 2 ||
1995 le32_to_cpu(raw_super->root_ino) != 3) {
1996 f2fs_msg(sb, KERN_INFO,
1997 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1998 le32_to_cpu(raw_super->node_ino),
1999 le32_to_cpu(raw_super->meta_ino),
2000 le32_to_cpu(raw_super->root_ino));
2004 if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) {
2005 f2fs_msg(sb, KERN_INFO,
2006 "Invalid segment count (%u)",
2007 le32_to_cpu(raw_super->segment_count));
2011 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2012 if (sanity_check_area_boundary(sbi, bh))
2018 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
2020 unsigned int total, fsmeta;
2021 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2022 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2023 unsigned int ovp_segments, reserved_segments;
2024 unsigned int main_segs, blocks_per_seg;
2027 total = le32_to_cpu(raw_super->segment_count);
2028 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2029 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
2030 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
2031 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2032 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2034 if (unlikely(fsmeta >= total))
2037 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2038 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2040 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2041 ovp_segments == 0 || reserved_segments == 0)) {
2042 f2fs_msg(sbi->sb, KERN_ERR,
2043 "Wrong layout: check mkfs.f2fs version");
2047 main_segs = le32_to_cpu(raw_super->segment_count_main);
2048 blocks_per_seg = sbi->blocks_per_seg;
2050 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2051 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2052 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2055 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2056 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2057 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2061 if (unlikely(f2fs_cp_error(sbi))) {
2062 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2068 static void init_sb_info(struct f2fs_sb_info *sbi)
2070 struct f2fs_super_block *raw_super = sbi->raw_super;
2073 sbi->log_sectors_per_block =
2074 le32_to_cpu(raw_super->log_sectors_per_block);
2075 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2076 sbi->blocksize = 1 << sbi->log_blocksize;
2077 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2078 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2079 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2080 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2081 sbi->total_sections = le32_to_cpu(raw_super->section_count);
2082 sbi->total_node_count =
2083 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2084 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2085 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2086 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2087 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2088 sbi->cur_victim_sec = NULL_SECNO;
2089 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2091 sbi->dir_level = DEF_DIR_LEVEL;
2092 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2093 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2094 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2096 for (i = 0; i < NR_COUNT_TYPE; i++)
2097 atomic_set(&sbi->nr_pages[i], 0);
2099 atomic_set(&sbi->wb_sync_req, 0);
2101 INIT_LIST_HEAD(&sbi->s_list);
2102 mutex_init(&sbi->umount_mutex);
2103 for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2104 for (j = HOT; j < NR_TEMP_TYPE; j++)
2105 mutex_init(&sbi->wio_mutex[i][j]);
2106 spin_lock_init(&sbi->cp_lock);
2108 sbi->dirty_device = 0;
2109 spin_lock_init(&sbi->dev_lock);
2112 static int init_percpu_info(struct f2fs_sb_info *sbi)
2116 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2120 return percpu_counter_init(&sbi->total_valid_inode_count, 0,
2124 #ifdef CONFIG_BLK_DEV_ZONED
2125 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2127 struct block_device *bdev = FDEV(devi).bdev;
2128 sector_t nr_sectors = bdev->bd_part->nr_sects;
2129 sector_t sector = 0;
2130 struct blk_zone *zones;
2131 unsigned int i, nr_zones;
2135 if (!f2fs_sb_mounted_blkzoned(sbi->sb))
2138 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2139 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2141 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2142 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2143 __ilog2_u32(sbi->blocks_per_blkz))
2145 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2146 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2147 sbi->log_blocks_per_blkz;
2148 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2149 FDEV(devi).nr_blkz++;
2151 FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
2152 if (!FDEV(devi).blkz_type)
2155 #define F2FS_REPORT_NR_ZONES 4096
2157 zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
2162 /* Get block zones type */
2163 while (zones && sector < nr_sectors) {
2165 nr_zones = F2FS_REPORT_NR_ZONES;
2166 err = blkdev_report_zones(bdev, sector,
2176 for (i = 0; i < nr_zones; i++) {
2177 FDEV(devi).blkz_type[n] = zones[i].type;
2178 sector += zones[i].len;
2190 * Read f2fs raw super block.
2191 * Because we have two copies of super block, so read both of them
2192 * to get the first valid one. If any one of them is broken, we pass
2193 * them recovery flag back to the caller.
2195 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2196 struct f2fs_super_block **raw_super,
2197 int *valid_super_block, int *recovery)
2199 struct super_block *sb = sbi->sb;
2201 struct buffer_head *bh;
2202 struct f2fs_super_block *super;
2205 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2209 for (block = 0; block < 2; block++) {
2210 bh = sb_bread(sb, block);
2212 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2218 /* sanity checking of raw super */
2219 if (sanity_check_raw_super(sbi, bh)) {
2220 f2fs_msg(sb, KERN_ERR,
2221 "Can't find valid F2FS filesystem in %dth superblock",
2229 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2231 *valid_super_block = block;
2237 /* Fail to read any one of the superblocks*/
2241 /* No valid superblock */
2250 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2252 struct buffer_head *bh;
2255 if ((recover && f2fs_readonly(sbi->sb)) ||
2256 bdev_read_only(sbi->sb->s_bdev)) {
2257 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2261 /* write back-up superblock first */
2262 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
2265 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2268 /* if we are in recovery path, skip writing valid superblock */
2272 /* write current valid superblock */
2273 bh = sb_getblk(sbi->sb, sbi->valid_super_block);
2276 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2281 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2283 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2284 unsigned int max_devices = MAX_DEVICES;
2287 /* Initialize single device information */
2288 if (!RDEV(0).path[0]) {
2289 if (!bdev_is_zoned(sbi->sb->s_bdev))
2295 * Initialize multiple devices information, or single
2296 * zoned block device information.
2298 sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info),
2303 for (i = 0; i < max_devices; i++) {
2305 if (i > 0 && !RDEV(i).path[0])
2308 if (max_devices == 1) {
2309 /* Single zoned block device mount */
2311 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2312 sbi->sb->s_mode, sbi->sb->s_type);
2314 /* Multi-device mount */
2315 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2316 FDEV(i).total_segments =
2317 le32_to_cpu(RDEV(i).total_segments);
2319 FDEV(i).start_blk = 0;
2320 FDEV(i).end_blk = FDEV(i).start_blk +
2321 (FDEV(i).total_segments <<
2322 sbi->log_blocks_per_seg) - 1 +
2323 le32_to_cpu(raw_super->segment0_blkaddr);
2325 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2326 FDEV(i).end_blk = FDEV(i).start_blk +
2327 (FDEV(i).total_segments <<
2328 sbi->log_blocks_per_seg) - 1;
2330 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2331 sbi->sb->s_mode, sbi->sb->s_type);
2333 if (IS_ERR(FDEV(i).bdev))
2334 return PTR_ERR(FDEV(i).bdev);
2336 /* to release errored devices */
2337 sbi->s_ndevs = i + 1;
2339 #ifdef CONFIG_BLK_DEV_ZONED
2340 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2341 !f2fs_sb_mounted_blkzoned(sbi->sb)) {
2342 f2fs_msg(sbi->sb, KERN_ERR,
2343 "Zoned block device feature not enabled\n");
2346 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2347 if (init_blkz_info(sbi, i)) {
2348 f2fs_msg(sbi->sb, KERN_ERR,
2349 "Failed to initialize F2FS blkzone information");
2352 if (max_devices == 1)
2354 f2fs_msg(sbi->sb, KERN_INFO,
2355 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2357 FDEV(i).total_segments,
2358 FDEV(i).start_blk, FDEV(i).end_blk,
2359 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2360 "Host-aware" : "Host-managed");
2364 f2fs_msg(sbi->sb, KERN_INFO,
2365 "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2367 FDEV(i).total_segments,
2368 FDEV(i).start_blk, FDEV(i).end_blk);
2370 f2fs_msg(sbi->sb, KERN_INFO,
2371 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2375 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2377 struct f2fs_sb_info *sbi;
2378 struct f2fs_super_block *raw_super;
2381 bool retry = true, need_fsck = false;
2382 char *options = NULL;
2383 int recovery, i, valid_super_block;
2384 struct curseg_info *seg_i;
2389 valid_super_block = -1;
2392 /* allocate memory for f2fs-specific super block info */
2393 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2399 /* Load the checksum driver */
2400 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2401 if (IS_ERR(sbi->s_chksum_driver)) {
2402 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2403 err = PTR_ERR(sbi->s_chksum_driver);
2404 sbi->s_chksum_driver = NULL;
2408 /* set a block size */
2409 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2410 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2414 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2419 sb->s_fs_info = sbi;
2420 sbi->raw_super = raw_super;
2422 /* precompute checksum seed for metadata */
2423 if (f2fs_sb_has_inode_chksum(sb))
2424 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2425 sizeof(raw_super->uuid));
2428 * The BLKZONED feature indicates that the drive was formatted with
2429 * zone alignment optimization. This is optional for host-aware
2430 * devices, but mandatory for host-managed zoned block devices.
2432 #ifndef CONFIG_BLK_DEV_ZONED
2433 if (f2fs_sb_mounted_blkzoned(sb)) {
2434 f2fs_msg(sb, KERN_ERR,
2435 "Zoned block device support is not enabled\n");
2440 default_options(sbi);
2441 /* parse mount options */
2442 options = kstrdup((const char *)data, GFP_KERNEL);
2443 if (data && !options) {
2448 err = parse_options(sb, options);
2452 sbi->max_file_blocks = max_file_blocks();
2453 sb->s_maxbytes = sbi->max_file_blocks <<
2454 le32_to_cpu(raw_super->log_blocksize);
2455 sb->s_max_links = F2FS_LINK_MAX;
2456 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2459 sb->dq_op = &f2fs_quota_operations;
2460 if (f2fs_sb_has_quota_ino(sb))
2461 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2463 sb->s_qcop = &f2fs_quotactl_ops;
2464 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2467 sb->s_op = &f2fs_sops;
2468 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2469 sb->s_cop = &f2fs_cryptops;
2471 sb->s_xattr = f2fs_xattr_handlers;
2472 sb->s_export_op = &f2fs_export_ops;
2473 sb->s_magic = F2FS_SUPER_MAGIC;
2474 sb->s_time_gran = 1;
2475 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2476 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
2477 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2479 /* init f2fs-specific super block info */
2480 sbi->valid_super_block = valid_super_block;
2481 mutex_init(&sbi->gc_mutex);
2482 mutex_init(&sbi->cp_mutex);
2483 init_rwsem(&sbi->node_write);
2484 init_rwsem(&sbi->node_change);
2486 /* disallow all the data/node/meta page writes */
2487 set_sbi_flag(sbi, SBI_POR_DOING);
2488 spin_lock_init(&sbi->stat_lock);
2490 /* init iostat info */
2491 spin_lock_init(&sbi->iostat_lock);
2492 sbi->iostat_enable = false;
2494 for (i = 0; i < NR_PAGE_TYPE; i++) {
2495 int n = (i == META) ? 1: NR_TEMP_TYPE;
2498 sbi->write_io[i] = kmalloc(n * sizeof(struct f2fs_bio_info),
2500 if (!sbi->write_io[i]) {
2505 for (j = HOT; j < n; j++) {
2506 init_rwsem(&sbi->write_io[i][j].io_rwsem);
2507 sbi->write_io[i][j].sbi = sbi;
2508 sbi->write_io[i][j].bio = NULL;
2509 spin_lock_init(&sbi->write_io[i][j].io_lock);
2510 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2514 init_rwsem(&sbi->cp_rwsem);
2515 init_waitqueue_head(&sbi->cp_wait);
2518 err = init_percpu_info(sbi);
2522 if (F2FS_IO_SIZE(sbi) > 1) {
2523 sbi->write_io_dummy =
2524 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2525 if (!sbi->write_io_dummy) {
2531 /* get an inode for meta space */
2532 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2533 if (IS_ERR(sbi->meta_inode)) {
2534 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2535 err = PTR_ERR(sbi->meta_inode);
2539 err = get_valid_checkpoint(sbi);
2541 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2542 goto free_meta_inode;
2545 /* Initialize device list */
2546 err = f2fs_scan_devices(sbi);
2548 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2552 sbi->total_valid_node_count =
2553 le32_to_cpu(sbi->ckpt->valid_node_count);
2554 percpu_counter_set(&sbi->total_valid_inode_count,
2555 le32_to_cpu(sbi->ckpt->valid_inode_count));
2556 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2557 sbi->total_valid_block_count =
2558 le64_to_cpu(sbi->ckpt->valid_block_count);
2559 sbi->last_valid_block_count = sbi->total_valid_block_count;
2560 sbi->reserved_blocks = 0;
2561 sbi->current_reserved_blocks = 0;
2563 for (i = 0; i < NR_INODE_TYPE; i++) {
2564 INIT_LIST_HEAD(&sbi->inode_list[i]);
2565 spin_lock_init(&sbi->inode_lock[i]);
2568 init_extent_cache_info(sbi);
2570 init_ino_entry_info(sbi);
2572 /* setup f2fs internal modules */
2573 err = build_segment_manager(sbi);
2575 f2fs_msg(sb, KERN_ERR,
2576 "Failed to initialize F2FS segment manager");
2579 err = build_node_manager(sbi);
2581 f2fs_msg(sb, KERN_ERR,
2582 "Failed to initialize F2FS node manager");
2586 /* For write statistics */
2587 if (sb->s_bdev->bd_part)
2588 sbi->sectors_written_start =
2589 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2591 /* Read accumulated write IO statistics if exists */
2592 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2593 if (__exist_node_summaries(sbi))
2594 sbi->kbytes_written =
2595 le64_to_cpu(seg_i->journal->info.kbytes_written);
2597 build_gc_manager(sbi);
2599 /* get an inode for node space */
2600 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2601 if (IS_ERR(sbi->node_inode)) {
2602 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2603 err = PTR_ERR(sbi->node_inode);
2607 f2fs_join_shrinker(sbi);
2609 err = f2fs_build_stats(sbi);
2613 /* read root inode and dentry */
2614 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2616 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2617 err = PTR_ERR(root);
2618 goto free_node_inode;
2620 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2623 goto free_node_inode;
2626 sb->s_root = d_make_root(root); /* allocate root dentry */
2629 goto free_root_inode;
2632 err = f2fs_register_sysfs(sbi);
2634 goto free_root_inode;
2638 * Turn on quotas which were not enabled for read-only mounts if
2639 * filesystem has quota feature, so that they are updated correctly.
2641 if (f2fs_sb_has_quota_ino(sb) && !sb_rdonly(sb)) {
2642 err = f2fs_enable_quotas(sb);
2644 f2fs_msg(sb, KERN_ERR,
2645 "Cannot turn on quotas: error %d", err);
2650 /* if there are nt orphan nodes free them */
2651 err = recover_orphan_inodes(sbi);
2655 /* recover fsynced data */
2656 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2658 * mount should be failed, when device has readonly mode, and
2659 * previous checkpoint was not done by clean system shutdown.
2661 if (bdev_read_only(sb->s_bdev) &&
2662 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2668 set_sbi_flag(sbi, SBI_NEED_FSCK);
2673 err = recover_fsync_data(sbi, false);
2676 f2fs_msg(sb, KERN_ERR,
2677 "Cannot recover all fsync data errno=%d", err);
2681 err = recover_fsync_data(sbi, true);
2683 if (!f2fs_readonly(sb) && err > 0) {
2685 f2fs_msg(sb, KERN_ERR,
2686 "Need to recover fsync data");
2691 /* recover_fsync_data() cleared this already */
2692 clear_sbi_flag(sbi, SBI_POR_DOING);
2695 * If filesystem is not mounted as read-only then
2696 * do start the gc_thread.
2698 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2699 /* After POR, we can run background GC thread.*/
2700 err = start_gc_thread(sbi);
2706 /* recover broken superblock */
2708 err = f2fs_commit_super(sbi, true);
2709 f2fs_msg(sb, KERN_INFO,
2710 "Try to recover %dth superblock, ret: %d",
2711 sbi->valid_super_block ? 1 : 2, err);
2714 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2715 cur_cp_version(F2FS_CKPT(sbi)));
2716 f2fs_update_time(sbi, CP_TIME);
2717 f2fs_update_time(sbi, REQ_TIME);
2722 if (f2fs_sb_has_quota_ino(sb) && !sb_rdonly(sb))
2723 f2fs_quota_off_umount(sbi->sb);
2725 f2fs_sync_inode_meta(sbi);
2727 * Some dirty meta pages can be produced by recover_orphan_inodes()
2728 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2729 * followed by write_checkpoint() through f2fs_write_node_pages(), which
2730 * falls into an infinite loop in sync_meta_pages().
2732 truncate_inode_pages_final(META_MAPPING(sbi));
2736 f2fs_unregister_sysfs(sbi);
2741 truncate_inode_pages_final(NODE_MAPPING(sbi));
2742 mutex_lock(&sbi->umount_mutex);
2743 release_ino_entry(sbi, true);
2744 f2fs_leave_shrinker(sbi);
2745 iput(sbi->node_inode);
2746 mutex_unlock(&sbi->umount_mutex);
2747 f2fs_destroy_stats(sbi);
2749 destroy_node_manager(sbi);
2751 destroy_segment_manager(sbi);
2753 destroy_device_list(sbi);
2756 make_bad_inode(sbi->meta_inode);
2757 iput(sbi->meta_inode);
2759 mempool_destroy(sbi->write_io_dummy);
2761 for (i = 0; i < NR_PAGE_TYPE; i++)
2762 kfree(sbi->write_io[i]);
2763 destroy_percpu_info(sbi);
2765 for (i = 0; i < MAXQUOTAS; i++)
2766 kfree(sbi->s_qf_names[i]);
2772 if (sbi->s_chksum_driver)
2773 crypto_free_shash(sbi->s_chksum_driver);
2776 /* give only one another chance */
2779 shrink_dcache_sb(sb);
2785 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2786 const char *dev_name, void *data)
2788 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2791 static void kill_f2fs_super(struct super_block *sb)
2794 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2795 stop_gc_thread(F2FS_SB(sb));
2796 stop_discard_thread(F2FS_SB(sb));
2798 kill_block_super(sb);
2801 static struct file_system_type f2fs_fs_type = {
2802 .owner = THIS_MODULE,
2804 .mount = f2fs_mount,
2805 .kill_sb = kill_f2fs_super,
2806 .fs_flags = FS_REQUIRES_DEV,
2808 MODULE_ALIAS_FS("f2fs");
2810 static int __init init_inodecache(void)
2812 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2813 sizeof(struct f2fs_inode_info), 0,
2814 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2815 if (!f2fs_inode_cachep)
2820 static void destroy_inodecache(void)
2823 * Make sure all delayed rcu free inodes are flushed before we
2827 kmem_cache_destroy(f2fs_inode_cachep);
2830 static int __init init_f2fs_fs(void)
2834 f2fs_build_trace_ios();
2836 err = init_inodecache();
2839 err = create_node_manager_caches();
2841 goto free_inodecache;
2842 err = create_segment_manager_caches();
2844 goto free_node_manager_caches;
2845 err = create_checkpoint_caches();
2847 goto free_segment_manager_caches;
2848 err = create_extent_cache();
2850 goto free_checkpoint_caches;
2851 err = f2fs_init_sysfs();
2853 goto free_extent_cache;
2854 err = register_shrinker(&f2fs_shrinker_info);
2857 err = register_filesystem(&f2fs_fs_type);
2860 err = f2fs_create_root_stats();
2862 goto free_filesystem;
2866 unregister_filesystem(&f2fs_fs_type);
2868 unregister_shrinker(&f2fs_shrinker_info);
2872 destroy_extent_cache();
2873 free_checkpoint_caches:
2874 destroy_checkpoint_caches();
2875 free_segment_manager_caches:
2876 destroy_segment_manager_caches();
2877 free_node_manager_caches:
2878 destroy_node_manager_caches();
2880 destroy_inodecache();
2885 static void __exit exit_f2fs_fs(void)
2887 f2fs_destroy_root_stats();
2888 unregister_filesystem(&f2fs_fs_type);
2889 unregister_shrinker(&f2fs_shrinker_info);
2891 destroy_extent_cache();
2892 destroy_checkpoint_caches();
2893 destroy_segment_manager_caches();
2894 destroy_node_manager_caches();
2895 destroy_inodecache();
2896 f2fs_destroy_trace_ios();
2899 module_init(init_f2fs_fs)
2900 module_exit(exit_f2fs_fs)
2902 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2903 MODULE_DESCRIPTION("Flash Friendly File System");
2904 MODULE_LICENSE("GPL");