1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
8 #include <linux/module.h>
9 #include <linux/init.h>
11 #include <linux/statfs.h>
12 #include <linux/buffer_head.h>
13 #include <linux/backing-dev.h>
14 #include <linux/kthread.h>
15 #include <linux/parser.h>
16 #include <linux/mount.h>
17 #include <linux/seq_file.h>
18 #include <linux/proc_fs.h>
19 #include <linux/random.h>
20 #include <linux/exportfs.h>
21 #include <linux/blkdev.h>
22 #include <linux/quotaops.h>
23 #include <linux/f2fs_fs.h>
24 #include <linux/sysfs.h>
25 #include <linux/quota.h>
26 #include <linux/unicode.h>
27 #include <linux/part_stat.h>
28 #include <linux/zstd.h>
29 #include <linux/lz4.h>
38 #define CREATE_TRACE_POINTS
39 #include <trace/events/f2fs.h>
41 static struct kmem_cache *f2fs_inode_cachep;
43 #ifdef CONFIG_F2FS_FAULT_INJECTION
45 const char *f2fs_fault_name[FAULT_MAX] = {
46 [FAULT_KMALLOC] = "kmalloc",
47 [FAULT_KVMALLOC] = "kvmalloc",
48 [FAULT_PAGE_ALLOC] = "page alloc",
49 [FAULT_PAGE_GET] = "page get",
50 [FAULT_ALLOC_NID] = "alloc nid",
51 [FAULT_ORPHAN] = "orphan",
52 [FAULT_BLOCK] = "no more block",
53 [FAULT_DIR_DEPTH] = "too big dir depth",
54 [FAULT_EVICT_INODE] = "evict_inode fail",
55 [FAULT_TRUNCATE] = "truncate fail",
56 [FAULT_READ_IO] = "read IO error",
57 [FAULT_CHECKPOINT] = "checkpoint error",
58 [FAULT_DISCARD] = "discard error",
59 [FAULT_WRITE_IO] = "write IO error",
60 [FAULT_SLAB_ALLOC] = "slab alloc",
61 [FAULT_DQUOT_INIT] = "dquot initialize",
64 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
67 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
70 atomic_set(&ffi->inject_ops, 0);
71 ffi->inject_rate = rate;
75 ffi->inject_type = type;
78 memset(ffi, 0, sizeof(struct f2fs_fault_info));
82 /* f2fs-wide shrinker description */
83 static struct shrinker f2fs_shrinker_info = {
84 .scan_objects = f2fs_shrink_scan,
85 .count_objects = f2fs_shrink_count,
86 .seeks = DEFAULT_SEEKS,
91 Opt_disable_roll_forward,
102 Opt_disable_ext_identify,
105 Opt_inline_xattr_size,
143 Opt_test_dummy_encryption,
145 Opt_checkpoint_disable,
146 Opt_checkpoint_disable_cap,
147 Opt_checkpoint_disable_cap_perc,
148 Opt_checkpoint_enable,
149 Opt_checkpoint_merge,
150 Opt_nocheckpoint_merge,
151 Opt_compress_algorithm,
152 Opt_compress_log_size,
153 Opt_compress_extension,
154 Opt_nocompress_extension,
165 static match_table_t f2fs_tokens = {
166 {Opt_gc_background, "background_gc=%s"},
167 {Opt_disable_roll_forward, "disable_roll_forward"},
168 {Opt_norecovery, "norecovery"},
169 {Opt_discard, "discard"},
170 {Opt_nodiscard, "nodiscard"},
171 {Opt_noheap, "no_heap"},
173 {Opt_user_xattr, "user_xattr"},
174 {Opt_nouser_xattr, "nouser_xattr"},
176 {Opt_noacl, "noacl"},
177 {Opt_active_logs, "active_logs=%u"},
178 {Opt_disable_ext_identify, "disable_ext_identify"},
179 {Opt_inline_xattr, "inline_xattr"},
180 {Opt_noinline_xattr, "noinline_xattr"},
181 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
182 {Opt_inline_data, "inline_data"},
183 {Opt_inline_dentry, "inline_dentry"},
184 {Opt_noinline_dentry, "noinline_dentry"},
185 {Opt_flush_merge, "flush_merge"},
186 {Opt_noflush_merge, "noflush_merge"},
187 {Opt_nobarrier, "nobarrier"},
188 {Opt_fastboot, "fastboot"},
189 {Opt_extent_cache, "extent_cache"},
190 {Opt_noextent_cache, "noextent_cache"},
191 {Opt_noinline_data, "noinline_data"},
192 {Opt_data_flush, "data_flush"},
193 {Opt_reserve_root, "reserve_root=%u"},
194 {Opt_resgid, "resgid=%u"},
195 {Opt_resuid, "resuid=%u"},
196 {Opt_mode, "mode=%s"},
197 {Opt_io_size_bits, "io_bits=%u"},
198 {Opt_fault_injection, "fault_injection=%u"},
199 {Opt_fault_type, "fault_type=%u"},
200 {Opt_lazytime, "lazytime"},
201 {Opt_nolazytime, "nolazytime"},
202 {Opt_quota, "quota"},
203 {Opt_noquota, "noquota"},
204 {Opt_usrquota, "usrquota"},
205 {Opt_grpquota, "grpquota"},
206 {Opt_prjquota, "prjquota"},
207 {Opt_usrjquota, "usrjquota=%s"},
208 {Opt_grpjquota, "grpjquota=%s"},
209 {Opt_prjjquota, "prjjquota=%s"},
210 {Opt_offusrjquota, "usrjquota="},
211 {Opt_offgrpjquota, "grpjquota="},
212 {Opt_offprjjquota, "prjjquota="},
213 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
214 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
215 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
216 {Opt_whint, "whint_mode=%s"},
217 {Opt_alloc, "alloc_mode=%s"},
218 {Opt_fsync, "fsync_mode=%s"},
219 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
220 {Opt_test_dummy_encryption, "test_dummy_encryption"},
221 {Opt_inlinecrypt, "inlinecrypt"},
222 {Opt_checkpoint_disable, "checkpoint=disable"},
223 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
224 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
225 {Opt_checkpoint_enable, "checkpoint=enable"},
226 {Opt_checkpoint_merge, "checkpoint_merge"},
227 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
228 {Opt_compress_algorithm, "compress_algorithm=%s"},
229 {Opt_compress_log_size, "compress_log_size=%u"},
230 {Opt_compress_extension, "compress_extension=%s"},
231 {Opt_nocompress_extension, "nocompress_extension=%s"},
232 {Opt_compress_chksum, "compress_chksum"},
233 {Opt_compress_mode, "compress_mode=%s"},
234 {Opt_compress_cache, "compress_cache"},
236 {Opt_gc_merge, "gc_merge"},
237 {Opt_nogc_merge, "nogc_merge"},
238 {Opt_discard_unit, "discard_unit=%s"},
242 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
244 struct va_format vaf;
250 level = printk_get_level(fmt);
251 vaf.fmt = printk_skip_level(fmt);
253 printk("%c%cF2FS-fs (%s): %pV\n",
254 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
259 #ifdef CONFIG_UNICODE
260 static const struct f2fs_sb_encodings {
264 } f2fs_sb_encoding_map[] = {
265 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
268 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
269 const struct f2fs_sb_encodings **encoding,
272 __u16 magic = le16_to_cpu(sb->s_encoding);
275 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
276 if (magic == f2fs_sb_encoding_map[i].magic)
279 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
282 *encoding = &f2fs_sb_encoding_map[i];
283 *flags = le16_to_cpu(sb->s_encoding_flags);
288 struct kmem_cache *f2fs_cf_name_slab;
289 static int __init f2fs_create_casefold_cache(void)
291 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
293 if (!f2fs_cf_name_slab)
298 static void f2fs_destroy_casefold_cache(void)
300 kmem_cache_destroy(f2fs_cf_name_slab);
303 static int __init f2fs_create_casefold_cache(void) { return 0; }
304 static void f2fs_destroy_casefold_cache(void) { }
307 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
309 block_t limit = min((sbi->user_block_count << 1) / 1000,
310 sbi->user_block_count - sbi->reserved_blocks);
313 if (test_opt(sbi, RESERVE_ROOT) &&
314 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
315 F2FS_OPTION(sbi).root_reserved_blocks = limit;
316 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
317 F2FS_OPTION(sbi).root_reserved_blocks);
319 if (!test_opt(sbi, RESERVE_ROOT) &&
320 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
321 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
322 !gid_eq(F2FS_OPTION(sbi).s_resgid,
323 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
324 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
325 from_kuid_munged(&init_user_ns,
326 F2FS_OPTION(sbi).s_resuid),
327 from_kgid_munged(&init_user_ns,
328 F2FS_OPTION(sbi).s_resgid));
331 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
333 if (!F2FS_OPTION(sbi).unusable_cap_perc)
336 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
337 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
339 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
340 F2FS_OPTION(sbi).unusable_cap_perc;
342 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
343 F2FS_OPTION(sbi).unusable_cap,
344 F2FS_OPTION(sbi).unusable_cap_perc);
347 static void init_once(void *foo)
349 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
351 inode_init_once(&fi->vfs_inode);
355 static const char * const quotatypes[] = INITQFNAMES;
356 #define QTYPE2NAME(t) (quotatypes[t])
357 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
360 struct f2fs_sb_info *sbi = F2FS_SB(sb);
364 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
365 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
368 if (f2fs_sb_has_quota_ino(sbi)) {
369 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
373 qname = match_strdup(args);
375 f2fs_err(sbi, "Not enough memory for storing quotafile name");
378 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
379 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
382 f2fs_err(sbi, "%s quota file already specified",
386 if (strchr(qname, '/')) {
387 f2fs_err(sbi, "quotafile must be on filesystem root");
390 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
398 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
400 struct f2fs_sb_info *sbi = F2FS_SB(sb);
402 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
403 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
406 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
407 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
411 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
414 * We do the test below only for project quotas. 'usrquota' and
415 * 'grpquota' mount options are allowed even without quota feature
416 * to support legacy quotas in quota files.
418 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
419 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
422 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
423 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
424 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
425 if (test_opt(sbi, USRQUOTA) &&
426 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
427 clear_opt(sbi, USRQUOTA);
429 if (test_opt(sbi, GRPQUOTA) &&
430 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
431 clear_opt(sbi, GRPQUOTA);
433 if (test_opt(sbi, PRJQUOTA) &&
434 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
435 clear_opt(sbi, PRJQUOTA);
437 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
438 test_opt(sbi, PRJQUOTA)) {
439 f2fs_err(sbi, "old and new quota format mixing");
443 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
444 f2fs_err(sbi, "journaled quota format not specified");
449 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
450 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
451 F2FS_OPTION(sbi).s_jquota_fmt = 0;
457 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
459 const substring_t *arg,
462 struct f2fs_sb_info *sbi = F2FS_SB(sb);
463 #ifdef CONFIG_FS_ENCRYPTION
466 if (!f2fs_sb_has_encrypt(sbi)) {
467 f2fs_err(sbi, "Encrypt feature is off");
472 * This mount option is just for testing, and it's not worthwhile to
473 * implement the extra complexity (e.g. RCU protection) that would be
474 * needed to allow it to be set or changed during remount. We do allow
475 * it to be specified during remount, but only if there is no change.
477 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
478 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
481 err = fscrypt_set_test_dummy_encryption(
482 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
486 "Can't change test_dummy_encryption on remount");
487 else if (err == -EINVAL)
488 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
491 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
495 f2fs_warn(sbi, "Test dummy encryption mode enabled");
497 f2fs_warn(sbi, "Test dummy encryption mount option ignored");
502 #ifdef CONFIG_F2FS_FS_COMPRESSION
504 * 1. The same extension name cannot not appear in both compress and non-compress extension
506 * 2. If the compress extension specifies all files, the types specified by the non-compress
507 * extension will be treated as special cases and will not be compressed.
508 * 3. Don't allow the non-compress extension specifies all files.
510 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
512 unsigned char (*ext)[F2FS_EXTENSION_LEN];
513 unsigned char (*noext)[F2FS_EXTENSION_LEN];
514 int ext_cnt, noext_cnt, index = 0, no_index = 0;
516 ext = F2FS_OPTION(sbi).extensions;
517 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
518 noext = F2FS_OPTION(sbi).noextensions;
519 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
524 for (no_index = 0; no_index < noext_cnt; no_index++) {
525 if (!strcasecmp("*", noext[no_index])) {
526 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
529 for (index = 0; index < ext_cnt; index++) {
530 if (!strcasecmp(ext[index], noext[no_index])) {
531 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
540 #ifdef CONFIG_F2FS_FS_LZ4
541 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
543 #ifdef CONFIG_F2FS_FS_LZ4HC
547 if (strlen(str) == 3) {
548 F2FS_OPTION(sbi).compress_level = 0;
552 #ifdef CONFIG_F2FS_FS_LZ4HC
556 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
559 if (kstrtouint(str + 1, 10, &level))
562 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
563 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
567 F2FS_OPTION(sbi).compress_level = level;
570 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
576 #ifdef CONFIG_F2FS_FS_ZSTD
577 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
582 if (strlen(str) == len) {
583 F2FS_OPTION(sbi).compress_level = 0;
590 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
593 if (kstrtouint(str + 1, 10, &level))
596 if (!level || level > zstd_max_clevel()) {
597 f2fs_info(sbi, "invalid zstd compress level: %d", level);
601 F2FS_OPTION(sbi).compress_level = level;
607 static int parse_options(struct super_block *sb, char *options, bool is_remount)
609 struct f2fs_sb_info *sbi = F2FS_SB(sb);
610 substring_t args[MAX_OPT_ARGS];
611 #ifdef CONFIG_F2FS_FS_COMPRESSION
612 unsigned char (*ext)[F2FS_EXTENSION_LEN];
613 unsigned char (*noext)[F2FS_EXTENSION_LEN];
614 int ext_cnt, noext_cnt;
625 while ((p = strsep(&options, ",")) != NULL) {
631 * Initialize args struct so we know whether arg was
632 * found; some options take optional arguments.
634 args[0].to = args[0].from = NULL;
635 token = match_token(p, f2fs_tokens, args);
638 case Opt_gc_background:
639 name = match_strdup(&args[0]);
643 if (!strcmp(name, "on")) {
644 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
645 } else if (!strcmp(name, "off")) {
646 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
647 } else if (!strcmp(name, "sync")) {
648 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
655 case Opt_disable_roll_forward:
656 set_opt(sbi, DISABLE_ROLL_FORWARD);
659 /* this option mounts f2fs with ro */
660 set_opt(sbi, NORECOVERY);
661 if (!f2fs_readonly(sb))
665 if (!f2fs_hw_support_discard(sbi)) {
666 f2fs_warn(sbi, "device does not support discard");
669 set_opt(sbi, DISCARD);
672 if (f2fs_hw_should_discard(sbi)) {
673 f2fs_warn(sbi, "discard is required for zoned block devices");
676 clear_opt(sbi, DISCARD);
679 set_opt(sbi, NOHEAP);
682 clear_opt(sbi, NOHEAP);
684 #ifdef CONFIG_F2FS_FS_XATTR
686 set_opt(sbi, XATTR_USER);
688 case Opt_nouser_xattr:
689 clear_opt(sbi, XATTR_USER);
691 case Opt_inline_xattr:
692 set_opt(sbi, INLINE_XATTR);
694 case Opt_noinline_xattr:
695 clear_opt(sbi, INLINE_XATTR);
697 case Opt_inline_xattr_size:
698 if (args->from && match_int(args, &arg))
700 set_opt(sbi, INLINE_XATTR_SIZE);
701 F2FS_OPTION(sbi).inline_xattr_size = arg;
705 f2fs_info(sbi, "user_xattr options not supported");
707 case Opt_nouser_xattr:
708 f2fs_info(sbi, "nouser_xattr options not supported");
710 case Opt_inline_xattr:
711 f2fs_info(sbi, "inline_xattr options not supported");
713 case Opt_noinline_xattr:
714 f2fs_info(sbi, "noinline_xattr options not supported");
717 #ifdef CONFIG_F2FS_FS_POSIX_ACL
719 set_opt(sbi, POSIX_ACL);
722 clear_opt(sbi, POSIX_ACL);
726 f2fs_info(sbi, "acl options not supported");
729 f2fs_info(sbi, "noacl options not supported");
732 case Opt_active_logs:
733 if (args->from && match_int(args, &arg))
735 if (arg != 2 && arg != 4 &&
736 arg != NR_CURSEG_PERSIST_TYPE)
738 F2FS_OPTION(sbi).active_logs = arg;
740 case Opt_disable_ext_identify:
741 set_opt(sbi, DISABLE_EXT_IDENTIFY);
743 case Opt_inline_data:
744 set_opt(sbi, INLINE_DATA);
746 case Opt_inline_dentry:
747 set_opt(sbi, INLINE_DENTRY);
749 case Opt_noinline_dentry:
750 clear_opt(sbi, INLINE_DENTRY);
752 case Opt_flush_merge:
753 set_opt(sbi, FLUSH_MERGE);
755 case Opt_noflush_merge:
756 clear_opt(sbi, FLUSH_MERGE);
759 set_opt(sbi, NOBARRIER);
762 set_opt(sbi, FASTBOOT);
764 case Opt_extent_cache:
765 set_opt(sbi, EXTENT_CACHE);
767 case Opt_noextent_cache:
768 clear_opt(sbi, EXTENT_CACHE);
770 case Opt_noinline_data:
771 clear_opt(sbi, INLINE_DATA);
774 set_opt(sbi, DATA_FLUSH);
776 case Opt_reserve_root:
777 if (args->from && match_int(args, &arg))
779 if (test_opt(sbi, RESERVE_ROOT)) {
780 f2fs_info(sbi, "Preserve previous reserve_root=%u",
781 F2FS_OPTION(sbi).root_reserved_blocks);
783 F2FS_OPTION(sbi).root_reserved_blocks = arg;
784 set_opt(sbi, RESERVE_ROOT);
788 if (args->from && match_int(args, &arg))
790 uid = make_kuid(current_user_ns(), arg);
791 if (!uid_valid(uid)) {
792 f2fs_err(sbi, "Invalid uid value %d", arg);
795 F2FS_OPTION(sbi).s_resuid = uid;
798 if (args->from && match_int(args, &arg))
800 gid = make_kgid(current_user_ns(), arg);
801 if (!gid_valid(gid)) {
802 f2fs_err(sbi, "Invalid gid value %d", arg);
805 F2FS_OPTION(sbi).s_resgid = gid;
808 name = match_strdup(&args[0]);
812 if (!strcmp(name, "adaptive")) {
813 if (f2fs_sb_has_blkzoned(sbi)) {
814 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
818 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
819 } else if (!strcmp(name, "lfs")) {
820 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
821 } else if (!strcmp(name, "fragment:segment")) {
822 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
823 } else if (!strcmp(name, "fragment:block")) {
824 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
831 case Opt_io_size_bits:
832 if (args->from && match_int(args, &arg))
834 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
835 f2fs_warn(sbi, "Not support %d, larger than %d",
836 1 << arg, BIO_MAX_VECS);
839 F2FS_OPTION(sbi).write_io_size_bits = arg;
841 #ifdef CONFIG_F2FS_FAULT_INJECTION
842 case Opt_fault_injection:
843 if (args->from && match_int(args, &arg))
845 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
846 set_opt(sbi, FAULT_INJECTION);
850 if (args->from && match_int(args, &arg))
852 f2fs_build_fault_attr(sbi, 0, arg);
853 set_opt(sbi, FAULT_INJECTION);
856 case Opt_fault_injection:
857 f2fs_info(sbi, "fault_injection options not supported");
861 f2fs_info(sbi, "fault_type options not supported");
865 sb->s_flags |= SB_LAZYTIME;
868 sb->s_flags &= ~SB_LAZYTIME;
873 set_opt(sbi, USRQUOTA);
876 set_opt(sbi, GRPQUOTA);
879 set_opt(sbi, PRJQUOTA);
882 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
887 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
892 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
896 case Opt_offusrjquota:
897 ret = f2fs_clear_qf_name(sb, USRQUOTA);
901 case Opt_offgrpjquota:
902 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
906 case Opt_offprjjquota:
907 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
911 case Opt_jqfmt_vfsold:
912 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
914 case Opt_jqfmt_vfsv0:
915 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
917 case Opt_jqfmt_vfsv1:
918 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
921 clear_opt(sbi, QUOTA);
922 clear_opt(sbi, USRQUOTA);
923 clear_opt(sbi, GRPQUOTA);
924 clear_opt(sbi, PRJQUOTA);
934 case Opt_offusrjquota:
935 case Opt_offgrpjquota:
936 case Opt_offprjjquota:
937 case Opt_jqfmt_vfsold:
938 case Opt_jqfmt_vfsv0:
939 case Opt_jqfmt_vfsv1:
941 f2fs_info(sbi, "quota operations not supported");
945 name = match_strdup(&args[0]);
948 if (!strcmp(name, "user-based")) {
949 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
950 } else if (!strcmp(name, "off")) {
951 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
952 } else if (!strcmp(name, "fs-based")) {
953 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
961 name = match_strdup(&args[0]);
965 if (!strcmp(name, "default")) {
966 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
967 } else if (!strcmp(name, "reuse")) {
968 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
976 name = match_strdup(&args[0]);
979 if (!strcmp(name, "posix")) {
980 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
981 } else if (!strcmp(name, "strict")) {
982 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
983 } else if (!strcmp(name, "nobarrier")) {
984 F2FS_OPTION(sbi).fsync_mode =
985 FSYNC_MODE_NOBARRIER;
992 case Opt_test_dummy_encryption:
993 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
998 case Opt_inlinecrypt:
999 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1000 sb->s_flags |= SB_INLINECRYPT;
1002 f2fs_info(sbi, "inline encryption not supported");
1005 case Opt_checkpoint_disable_cap_perc:
1006 if (args->from && match_int(args, &arg))
1008 if (arg < 0 || arg > 100)
1010 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1011 set_opt(sbi, DISABLE_CHECKPOINT);
1013 case Opt_checkpoint_disable_cap:
1014 if (args->from && match_int(args, &arg))
1016 F2FS_OPTION(sbi).unusable_cap = arg;
1017 set_opt(sbi, DISABLE_CHECKPOINT);
1019 case Opt_checkpoint_disable:
1020 set_opt(sbi, DISABLE_CHECKPOINT);
1022 case Opt_checkpoint_enable:
1023 clear_opt(sbi, DISABLE_CHECKPOINT);
1025 case Opt_checkpoint_merge:
1026 set_opt(sbi, MERGE_CHECKPOINT);
1028 case Opt_nocheckpoint_merge:
1029 clear_opt(sbi, MERGE_CHECKPOINT);
1031 #ifdef CONFIG_F2FS_FS_COMPRESSION
1032 case Opt_compress_algorithm:
1033 if (!f2fs_sb_has_compression(sbi)) {
1034 f2fs_info(sbi, "Image doesn't support compression");
1037 name = match_strdup(&args[0]);
1040 if (!strcmp(name, "lzo")) {
1041 #ifdef CONFIG_F2FS_FS_LZO
1042 F2FS_OPTION(sbi).compress_level = 0;
1043 F2FS_OPTION(sbi).compress_algorithm =
1046 f2fs_info(sbi, "kernel doesn't support lzo compression");
1048 } else if (!strncmp(name, "lz4", 3)) {
1049 #ifdef CONFIG_F2FS_FS_LZ4
1050 ret = f2fs_set_lz4hc_level(sbi, name);
1055 F2FS_OPTION(sbi).compress_algorithm =
1058 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1060 } else if (!strncmp(name, "zstd", 4)) {
1061 #ifdef CONFIG_F2FS_FS_ZSTD
1062 ret = f2fs_set_zstd_level(sbi, name);
1067 F2FS_OPTION(sbi).compress_algorithm =
1070 f2fs_info(sbi, "kernel doesn't support zstd compression");
1072 } else if (!strcmp(name, "lzo-rle")) {
1073 #ifdef CONFIG_F2FS_FS_LZORLE
1074 F2FS_OPTION(sbi).compress_level = 0;
1075 F2FS_OPTION(sbi).compress_algorithm =
1078 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1086 case Opt_compress_log_size:
1087 if (!f2fs_sb_has_compression(sbi)) {
1088 f2fs_info(sbi, "Image doesn't support compression");
1091 if (args->from && match_int(args, &arg))
1093 if (arg < MIN_COMPRESS_LOG_SIZE ||
1094 arg > MAX_COMPRESS_LOG_SIZE) {
1096 "Compress cluster log size is out of range");
1099 F2FS_OPTION(sbi).compress_log_size = arg;
1101 case Opt_compress_extension:
1102 if (!f2fs_sb_has_compression(sbi)) {
1103 f2fs_info(sbi, "Image doesn't support compression");
1106 name = match_strdup(&args[0]);
1110 ext = F2FS_OPTION(sbi).extensions;
1111 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1113 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1114 ext_cnt >= COMPRESS_EXT_NUM) {
1116 "invalid extension length/number");
1121 strcpy(ext[ext_cnt], name);
1122 F2FS_OPTION(sbi).compress_ext_cnt++;
1125 case Opt_nocompress_extension:
1126 if (!f2fs_sb_has_compression(sbi)) {
1127 f2fs_info(sbi, "Image doesn't support compression");
1130 name = match_strdup(&args[0]);
1134 noext = F2FS_OPTION(sbi).noextensions;
1135 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1137 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1138 noext_cnt >= COMPRESS_EXT_NUM) {
1140 "invalid extension length/number");
1145 strcpy(noext[noext_cnt], name);
1146 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1149 case Opt_compress_chksum:
1150 F2FS_OPTION(sbi).compress_chksum = true;
1152 case Opt_compress_mode:
1153 name = match_strdup(&args[0]);
1156 if (!strcmp(name, "fs")) {
1157 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1158 } else if (!strcmp(name, "user")) {
1159 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1166 case Opt_compress_cache:
1167 set_opt(sbi, COMPRESS_CACHE);
1170 case Opt_compress_algorithm:
1171 case Opt_compress_log_size:
1172 case Opt_compress_extension:
1173 case Opt_nocompress_extension:
1174 case Opt_compress_chksum:
1175 case Opt_compress_mode:
1176 case Opt_compress_cache:
1177 f2fs_info(sbi, "compression options not supported");
1184 set_opt(sbi, GC_MERGE);
1186 case Opt_nogc_merge:
1187 clear_opt(sbi, GC_MERGE);
1189 case Opt_discard_unit:
1190 name = match_strdup(&args[0]);
1193 if (!strcmp(name, "block")) {
1194 F2FS_OPTION(sbi).discard_unit =
1196 } else if (!strcmp(name, "segment")) {
1197 F2FS_OPTION(sbi).discard_unit =
1198 DISCARD_UNIT_SEGMENT;
1199 } else if (!strcmp(name, "section")) {
1200 F2FS_OPTION(sbi).discard_unit =
1201 DISCARD_UNIT_SECTION;
1209 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1216 if (f2fs_check_quota_options(sbi))
1219 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1220 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1223 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1224 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1228 #ifndef CONFIG_UNICODE
1229 if (f2fs_sb_has_casefold(sbi)) {
1231 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1236 * The BLKZONED feature indicates that the drive was formatted with
1237 * zone alignment optimization. This is optional for host-aware
1238 * devices, but mandatory for host-managed zoned block devices.
1240 #ifndef CONFIG_BLK_DEV_ZONED
1241 if (f2fs_sb_has_blkzoned(sbi)) {
1242 f2fs_err(sbi, "Zoned block device support is not enabled");
1246 if (f2fs_sb_has_blkzoned(sbi)) {
1247 if (F2FS_OPTION(sbi).discard_unit !=
1248 DISCARD_UNIT_SECTION) {
1249 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1250 F2FS_OPTION(sbi).discard_unit =
1251 DISCARD_UNIT_SECTION;
1255 #ifdef CONFIG_F2FS_FS_COMPRESSION
1256 if (f2fs_test_compress_extension(sbi)) {
1257 f2fs_err(sbi, "invalid compress or nocompress extension");
1262 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1263 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1264 F2FS_IO_SIZE_KB(sbi));
1268 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1269 int min_size, max_size;
1271 if (!f2fs_sb_has_extra_attr(sbi) ||
1272 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1273 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1276 if (!test_opt(sbi, INLINE_XATTR)) {
1277 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1281 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1282 max_size = MAX_INLINE_XATTR_SIZE;
1284 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1285 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1286 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1287 min_size, max_size);
1292 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1293 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1297 /* Not pass down write hints if the number of active logs is lesser
1298 * than NR_CURSEG_PERSIST_TYPE.
1300 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE)
1301 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1303 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1304 f2fs_err(sbi, "Allow to mount readonly mode only");
1310 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1312 struct f2fs_inode_info *fi;
1314 fi = f2fs_kmem_cache_alloc(f2fs_inode_cachep,
1315 GFP_F2FS_ZERO, false, F2FS_SB(sb));
1319 init_once((void *) fi);
1321 /* Initialize f2fs-specific inode info */
1322 atomic_set(&fi->dirty_pages, 0);
1323 atomic_set(&fi->i_compr_blocks, 0);
1324 init_rwsem(&fi->i_sem);
1325 spin_lock_init(&fi->i_size_lock);
1326 INIT_LIST_HEAD(&fi->dirty_list);
1327 INIT_LIST_HEAD(&fi->gdirty_list);
1328 INIT_LIST_HEAD(&fi->inmem_ilist);
1329 INIT_LIST_HEAD(&fi->inmem_pages);
1330 mutex_init(&fi->inmem_lock);
1331 init_rwsem(&fi->i_gc_rwsem[READ]);
1332 init_rwsem(&fi->i_gc_rwsem[WRITE]);
1333 init_rwsem(&fi->i_xattr_sem);
1335 /* Will be used by directory only */
1336 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1338 return &fi->vfs_inode;
1341 static int f2fs_drop_inode(struct inode *inode)
1343 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1347 * during filesystem shutdown, if checkpoint is disabled,
1348 * drop useless meta/node dirty pages.
1350 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1351 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1352 inode->i_ino == F2FS_META_INO(sbi)) {
1353 trace_f2fs_drop_inode(inode, 1);
1359 * This is to avoid a deadlock condition like below.
1360 * writeback_single_inode(inode)
1361 * - f2fs_write_data_page
1362 * - f2fs_gc -> iput -> evict
1363 * - inode_wait_for_writeback(inode)
1365 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1366 if (!inode->i_nlink && !is_bad_inode(inode)) {
1367 /* to avoid evict_inode call simultaneously */
1368 atomic_inc(&inode->i_count);
1369 spin_unlock(&inode->i_lock);
1371 /* some remained atomic pages should discarded */
1372 if (f2fs_is_atomic_file(inode))
1373 f2fs_drop_inmem_pages(inode);
1375 /* should remain fi->extent_tree for writepage */
1376 f2fs_destroy_extent_node(inode);
1378 sb_start_intwrite(inode->i_sb);
1379 f2fs_i_size_write(inode, 0);
1381 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1382 inode, NULL, 0, DATA);
1383 truncate_inode_pages_final(inode->i_mapping);
1385 if (F2FS_HAS_BLOCKS(inode))
1386 f2fs_truncate(inode);
1388 sb_end_intwrite(inode->i_sb);
1390 spin_lock(&inode->i_lock);
1391 atomic_dec(&inode->i_count);
1393 trace_f2fs_drop_inode(inode, 0);
1396 ret = generic_drop_inode(inode);
1398 ret = fscrypt_drop_inode(inode);
1399 trace_f2fs_drop_inode(inode, ret);
1403 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1405 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1408 spin_lock(&sbi->inode_lock[DIRTY_META]);
1409 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1412 set_inode_flag(inode, FI_DIRTY_INODE);
1413 stat_inc_dirty_inode(sbi, DIRTY_META);
1415 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1416 list_add_tail(&F2FS_I(inode)->gdirty_list,
1417 &sbi->inode_list[DIRTY_META]);
1418 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1420 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1424 void f2fs_inode_synced(struct inode *inode)
1426 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1428 spin_lock(&sbi->inode_lock[DIRTY_META]);
1429 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1430 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1433 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1434 list_del_init(&F2FS_I(inode)->gdirty_list);
1435 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1437 clear_inode_flag(inode, FI_DIRTY_INODE);
1438 clear_inode_flag(inode, FI_AUTO_RECOVER);
1439 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1440 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1444 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1446 * We should call set_dirty_inode to write the dirty inode through write_inode.
1448 static void f2fs_dirty_inode(struct inode *inode, int flags)
1450 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1452 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1453 inode->i_ino == F2FS_META_INO(sbi))
1456 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1457 clear_inode_flag(inode, FI_AUTO_RECOVER);
1459 f2fs_inode_dirtied(inode, false);
1462 static void f2fs_free_inode(struct inode *inode)
1464 fscrypt_free_inode(inode);
1465 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1468 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1470 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1471 percpu_counter_destroy(&sbi->total_valid_inode_count);
1474 static void destroy_device_list(struct f2fs_sb_info *sbi)
1478 for (i = 0; i < sbi->s_ndevs; i++) {
1479 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1480 #ifdef CONFIG_BLK_DEV_ZONED
1481 kvfree(FDEV(i).blkz_seq);
1482 kfree(FDEV(i).zone_capacity_blocks);
1488 static void f2fs_put_super(struct super_block *sb)
1490 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1494 /* unregister procfs/sysfs entries in advance to avoid race case */
1495 f2fs_unregister_sysfs(sbi);
1497 f2fs_quota_off_umount(sb);
1499 /* prevent remaining shrinker jobs */
1500 mutex_lock(&sbi->umount_mutex);
1503 * flush all issued checkpoints and stop checkpoint issue thread.
1504 * after then, all checkpoints should be done by each process context.
1506 f2fs_stop_ckpt_thread(sbi);
1509 * We don't need to do checkpoint when superblock is clean.
1510 * But, the previous checkpoint was not done by umount, it needs to do
1511 * clean checkpoint again.
1513 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1514 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1515 struct cp_control cpc = {
1516 .reason = CP_UMOUNT,
1518 f2fs_write_checkpoint(sbi, &cpc);
1521 /* be sure to wait for any on-going discard commands */
1522 dropped = f2fs_issue_discard_timeout(sbi);
1524 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1525 !sbi->discard_blks && !dropped) {
1526 struct cp_control cpc = {
1527 .reason = CP_UMOUNT | CP_TRIMMED,
1529 f2fs_write_checkpoint(sbi, &cpc);
1533 * normally superblock is clean, so we need to release this.
1534 * In addition, EIO will skip do checkpoint, we need this as well.
1536 f2fs_release_ino_entry(sbi, true);
1538 f2fs_leave_shrinker(sbi);
1539 mutex_unlock(&sbi->umount_mutex);
1541 /* our cp_error case, we can wait for any writeback page */
1542 f2fs_flush_merged_writes(sbi);
1544 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1546 f2fs_bug_on(sbi, sbi->fsync_node_num);
1548 f2fs_destroy_compress_inode(sbi);
1550 iput(sbi->node_inode);
1551 sbi->node_inode = NULL;
1553 iput(sbi->meta_inode);
1554 sbi->meta_inode = NULL;
1557 * iput() can update stat information, if f2fs_write_checkpoint()
1558 * above failed with error.
1560 f2fs_destroy_stats(sbi);
1562 /* destroy f2fs internal modules */
1563 f2fs_destroy_node_manager(sbi);
1564 f2fs_destroy_segment_manager(sbi);
1566 f2fs_destroy_post_read_wq(sbi);
1570 sb->s_fs_info = NULL;
1571 if (sbi->s_chksum_driver)
1572 crypto_free_shash(sbi->s_chksum_driver);
1573 kfree(sbi->raw_super);
1575 destroy_device_list(sbi);
1576 f2fs_destroy_page_array_cache(sbi);
1577 f2fs_destroy_xattr_caches(sbi);
1578 mempool_destroy(sbi->write_io_dummy);
1580 for (i = 0; i < MAXQUOTAS; i++)
1581 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1583 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1584 destroy_percpu_info(sbi);
1585 f2fs_destroy_iostat(sbi);
1586 for (i = 0; i < NR_PAGE_TYPE; i++)
1587 kvfree(sbi->write_io[i]);
1588 #ifdef CONFIG_UNICODE
1589 utf8_unload(sb->s_encoding);
1594 int f2fs_sync_fs(struct super_block *sb, int sync)
1596 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1599 if (unlikely(f2fs_cp_error(sbi)))
1601 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1604 trace_f2fs_sync_fs(sb, sync);
1606 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1610 err = f2fs_issue_checkpoint(sbi);
1615 static int f2fs_freeze(struct super_block *sb)
1617 if (f2fs_readonly(sb))
1620 /* IO error happened before */
1621 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1624 /* must be clean, since sync_filesystem() was already called */
1625 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1628 /* ensure no checkpoint required */
1629 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list))
1634 static int f2fs_unfreeze(struct super_block *sb)
1640 static int f2fs_statfs_project(struct super_block *sb,
1641 kprojid_t projid, struct kstatfs *buf)
1644 struct dquot *dquot;
1648 qid = make_kqid_projid(projid);
1649 dquot = dqget(sb, qid);
1651 return PTR_ERR(dquot);
1652 spin_lock(&dquot->dq_dqb_lock);
1654 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1655 dquot->dq_dqb.dqb_bhardlimit);
1657 limit >>= sb->s_blocksize_bits;
1659 if (limit && buf->f_blocks > limit) {
1660 curblock = (dquot->dq_dqb.dqb_curspace +
1661 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1662 buf->f_blocks = limit;
1663 buf->f_bfree = buf->f_bavail =
1664 (buf->f_blocks > curblock) ?
1665 (buf->f_blocks - curblock) : 0;
1668 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1669 dquot->dq_dqb.dqb_ihardlimit);
1671 if (limit && buf->f_files > limit) {
1672 buf->f_files = limit;
1674 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1675 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1678 spin_unlock(&dquot->dq_dqb_lock);
1684 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1686 struct super_block *sb = dentry->d_sb;
1687 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1688 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1689 block_t total_count, user_block_count, start_count;
1690 u64 avail_node_count;
1692 total_count = le64_to_cpu(sbi->raw_super->block_count);
1693 user_block_count = sbi->user_block_count;
1694 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1695 buf->f_type = F2FS_SUPER_MAGIC;
1696 buf->f_bsize = sbi->blocksize;
1698 buf->f_blocks = total_count - start_count;
1699 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1700 sbi->current_reserved_blocks;
1702 spin_lock(&sbi->stat_lock);
1703 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1706 buf->f_bfree -= sbi->unusable_block_count;
1707 spin_unlock(&sbi->stat_lock);
1709 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1710 buf->f_bavail = buf->f_bfree -
1711 F2FS_OPTION(sbi).root_reserved_blocks;
1715 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1717 if (avail_node_count > user_block_count) {
1718 buf->f_files = user_block_count;
1719 buf->f_ffree = buf->f_bavail;
1721 buf->f_files = avail_node_count;
1722 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1726 buf->f_namelen = F2FS_NAME_LEN;
1727 buf->f_fsid = u64_to_fsid(id);
1730 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1731 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1732 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1738 static inline void f2fs_show_quota_options(struct seq_file *seq,
1739 struct super_block *sb)
1742 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1744 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1747 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1758 seq_printf(seq, ",jqfmt=%s", fmtname);
1761 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1762 seq_show_option(seq, "usrjquota",
1763 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1765 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1766 seq_show_option(seq, "grpjquota",
1767 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1769 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1770 seq_show_option(seq, "prjjquota",
1771 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1775 #ifdef CONFIG_F2FS_FS_COMPRESSION
1776 static inline void f2fs_show_compress_options(struct seq_file *seq,
1777 struct super_block *sb)
1779 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1783 if (!f2fs_sb_has_compression(sbi))
1786 switch (F2FS_OPTION(sbi).compress_algorithm) {
1796 case COMPRESS_LZORLE:
1797 algtype = "lzo-rle";
1800 seq_printf(seq, ",compress_algorithm=%s", algtype);
1802 if (F2FS_OPTION(sbi).compress_level)
1803 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1805 seq_printf(seq, ",compress_log_size=%u",
1806 F2FS_OPTION(sbi).compress_log_size);
1808 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1809 seq_printf(seq, ",compress_extension=%s",
1810 F2FS_OPTION(sbi).extensions[i]);
1813 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1814 seq_printf(seq, ",nocompress_extension=%s",
1815 F2FS_OPTION(sbi).noextensions[i]);
1818 if (F2FS_OPTION(sbi).compress_chksum)
1819 seq_puts(seq, ",compress_chksum");
1821 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1822 seq_printf(seq, ",compress_mode=%s", "fs");
1823 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1824 seq_printf(seq, ",compress_mode=%s", "user");
1826 if (test_opt(sbi, COMPRESS_CACHE))
1827 seq_puts(seq, ",compress_cache");
1831 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1833 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1835 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1836 seq_printf(seq, ",background_gc=%s", "sync");
1837 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1838 seq_printf(seq, ",background_gc=%s", "on");
1839 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1840 seq_printf(seq, ",background_gc=%s", "off");
1842 if (test_opt(sbi, GC_MERGE))
1843 seq_puts(seq, ",gc_merge");
1845 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1846 seq_puts(seq, ",disable_roll_forward");
1847 if (test_opt(sbi, NORECOVERY))
1848 seq_puts(seq, ",norecovery");
1849 if (test_opt(sbi, DISCARD))
1850 seq_puts(seq, ",discard");
1852 seq_puts(seq, ",nodiscard");
1853 if (test_opt(sbi, NOHEAP))
1854 seq_puts(seq, ",no_heap");
1856 seq_puts(seq, ",heap");
1857 #ifdef CONFIG_F2FS_FS_XATTR
1858 if (test_opt(sbi, XATTR_USER))
1859 seq_puts(seq, ",user_xattr");
1861 seq_puts(seq, ",nouser_xattr");
1862 if (test_opt(sbi, INLINE_XATTR))
1863 seq_puts(seq, ",inline_xattr");
1865 seq_puts(seq, ",noinline_xattr");
1866 if (test_opt(sbi, INLINE_XATTR_SIZE))
1867 seq_printf(seq, ",inline_xattr_size=%u",
1868 F2FS_OPTION(sbi).inline_xattr_size);
1870 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1871 if (test_opt(sbi, POSIX_ACL))
1872 seq_puts(seq, ",acl");
1874 seq_puts(seq, ",noacl");
1876 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1877 seq_puts(seq, ",disable_ext_identify");
1878 if (test_opt(sbi, INLINE_DATA))
1879 seq_puts(seq, ",inline_data");
1881 seq_puts(seq, ",noinline_data");
1882 if (test_opt(sbi, INLINE_DENTRY))
1883 seq_puts(seq, ",inline_dentry");
1885 seq_puts(seq, ",noinline_dentry");
1886 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1887 seq_puts(seq, ",flush_merge");
1888 if (test_opt(sbi, NOBARRIER))
1889 seq_puts(seq, ",nobarrier");
1890 if (test_opt(sbi, FASTBOOT))
1891 seq_puts(seq, ",fastboot");
1892 if (test_opt(sbi, EXTENT_CACHE))
1893 seq_puts(seq, ",extent_cache");
1895 seq_puts(seq, ",noextent_cache");
1896 if (test_opt(sbi, DATA_FLUSH))
1897 seq_puts(seq, ",data_flush");
1899 seq_puts(seq, ",mode=");
1900 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1901 seq_puts(seq, "adaptive");
1902 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1903 seq_puts(seq, "lfs");
1904 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1905 seq_puts(seq, "fragment:segment");
1906 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1907 seq_puts(seq, "fragment:block");
1908 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1909 if (test_opt(sbi, RESERVE_ROOT))
1910 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1911 F2FS_OPTION(sbi).root_reserved_blocks,
1912 from_kuid_munged(&init_user_ns,
1913 F2FS_OPTION(sbi).s_resuid),
1914 from_kgid_munged(&init_user_ns,
1915 F2FS_OPTION(sbi).s_resgid));
1916 if (F2FS_IO_SIZE_BITS(sbi))
1917 seq_printf(seq, ",io_bits=%u",
1918 F2FS_OPTION(sbi).write_io_size_bits);
1919 #ifdef CONFIG_F2FS_FAULT_INJECTION
1920 if (test_opt(sbi, FAULT_INJECTION)) {
1921 seq_printf(seq, ",fault_injection=%u",
1922 F2FS_OPTION(sbi).fault_info.inject_rate);
1923 seq_printf(seq, ",fault_type=%u",
1924 F2FS_OPTION(sbi).fault_info.inject_type);
1928 if (test_opt(sbi, QUOTA))
1929 seq_puts(seq, ",quota");
1930 if (test_opt(sbi, USRQUOTA))
1931 seq_puts(seq, ",usrquota");
1932 if (test_opt(sbi, GRPQUOTA))
1933 seq_puts(seq, ",grpquota");
1934 if (test_opt(sbi, PRJQUOTA))
1935 seq_puts(seq, ",prjquota");
1937 f2fs_show_quota_options(seq, sbi->sb);
1938 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1939 seq_printf(seq, ",whint_mode=%s", "user-based");
1940 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1941 seq_printf(seq, ",whint_mode=%s", "fs-based");
1943 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1945 if (sbi->sb->s_flags & SB_INLINECRYPT)
1946 seq_puts(seq, ",inlinecrypt");
1948 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1949 seq_printf(seq, ",alloc_mode=%s", "default");
1950 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1951 seq_printf(seq, ",alloc_mode=%s", "reuse");
1953 if (test_opt(sbi, DISABLE_CHECKPOINT))
1954 seq_printf(seq, ",checkpoint=disable:%u",
1955 F2FS_OPTION(sbi).unusable_cap);
1956 if (test_opt(sbi, MERGE_CHECKPOINT))
1957 seq_puts(seq, ",checkpoint_merge");
1959 seq_puts(seq, ",nocheckpoint_merge");
1960 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1961 seq_printf(seq, ",fsync_mode=%s", "posix");
1962 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1963 seq_printf(seq, ",fsync_mode=%s", "strict");
1964 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1965 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1967 #ifdef CONFIG_F2FS_FS_COMPRESSION
1968 f2fs_show_compress_options(seq, sbi->sb);
1971 if (test_opt(sbi, ATGC))
1972 seq_puts(seq, ",atgc");
1974 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1975 seq_printf(seq, ",discard_unit=%s", "block");
1976 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1977 seq_printf(seq, ",discard_unit=%s", "segment");
1978 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1979 seq_printf(seq, ",discard_unit=%s", "section");
1984 static void default_options(struct f2fs_sb_info *sbi)
1986 /* init some FS parameters */
1987 if (f2fs_sb_has_readonly(sbi))
1988 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
1990 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
1992 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1993 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1994 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1995 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1996 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1997 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1998 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1999 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2000 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2001 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2002 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2004 sbi->sb->s_flags &= ~SB_INLINECRYPT;
2006 set_opt(sbi, INLINE_XATTR);
2007 set_opt(sbi, INLINE_DATA);
2008 set_opt(sbi, INLINE_DENTRY);
2009 set_opt(sbi, EXTENT_CACHE);
2010 set_opt(sbi, NOHEAP);
2011 clear_opt(sbi, DISABLE_CHECKPOINT);
2012 set_opt(sbi, MERGE_CHECKPOINT);
2013 F2FS_OPTION(sbi).unusable_cap = 0;
2014 sbi->sb->s_flags |= SB_LAZYTIME;
2015 set_opt(sbi, FLUSH_MERGE);
2016 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2017 set_opt(sbi, DISCARD);
2018 if (f2fs_sb_has_blkzoned(sbi)) {
2019 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2020 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2022 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2023 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2026 #ifdef CONFIG_F2FS_FS_XATTR
2027 set_opt(sbi, XATTR_USER);
2029 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2030 set_opt(sbi, POSIX_ACL);
2033 f2fs_build_fault_attr(sbi, 0, 0);
2037 static int f2fs_enable_quotas(struct super_block *sb);
2040 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2042 unsigned int s_flags = sbi->sb->s_flags;
2043 struct cp_control cpc;
2048 if (s_flags & SB_RDONLY) {
2049 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2052 sbi->sb->s_flags |= SB_ACTIVE;
2054 f2fs_update_time(sbi, DISABLE_TIME);
2056 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2057 down_write(&sbi->gc_lock);
2058 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
2059 if (err == -ENODATA) {
2063 if (err && err != -EAGAIN)
2067 ret = sync_filesystem(sbi->sb);
2069 err = ret ? ret : err;
2073 unusable = f2fs_get_unusable_blocks(sbi);
2074 if (f2fs_disable_cp_again(sbi, unusable)) {
2079 down_write(&sbi->gc_lock);
2080 cpc.reason = CP_PAUSE;
2081 set_sbi_flag(sbi, SBI_CP_DISABLED);
2082 err = f2fs_write_checkpoint(sbi, &cpc);
2086 spin_lock(&sbi->stat_lock);
2087 sbi->unusable_block_count = unusable;
2088 spin_unlock(&sbi->stat_lock);
2091 up_write(&sbi->gc_lock);
2093 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2097 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2099 int retry = DEFAULT_RETRY_IO_COUNT;
2101 /* we should flush all the data to keep data consistency */
2103 sync_inodes_sb(sbi->sb);
2105 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
2106 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2108 if (unlikely(retry < 0))
2109 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2111 down_write(&sbi->gc_lock);
2112 f2fs_dirty_to_prefree(sbi);
2114 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2115 set_sbi_flag(sbi, SBI_IS_DIRTY);
2116 up_write(&sbi->gc_lock);
2118 f2fs_sync_fs(sbi->sb, 1);
2121 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2123 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2124 struct f2fs_mount_info org_mount_opt;
2125 unsigned long old_sb_flags;
2127 bool need_restart_gc = false, need_stop_gc = false;
2128 bool need_restart_ckpt = false, need_stop_ckpt = false;
2129 bool need_restart_flush = false, need_stop_flush = false;
2130 bool need_restart_discard = false, need_stop_discard = false;
2131 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
2132 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2133 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2134 bool no_atgc = !test_opt(sbi, ATGC);
2135 bool no_discard = !test_opt(sbi, DISCARD);
2136 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2137 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2138 struct discard_cmd_control *dcc;
2144 * Save the old mount options in case we
2145 * need to restore them.
2147 org_mount_opt = sbi->mount_opt;
2148 old_sb_flags = sb->s_flags;
2151 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2152 for (i = 0; i < MAXQUOTAS; i++) {
2153 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2154 org_mount_opt.s_qf_names[i] =
2155 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2157 if (!org_mount_opt.s_qf_names[i]) {
2158 for (j = 0; j < i; j++)
2159 kfree(org_mount_opt.s_qf_names[j]);
2163 org_mount_opt.s_qf_names[i] = NULL;
2168 /* recover superblocks we couldn't write due to previous RO mount */
2169 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2170 err = f2fs_commit_super(sbi, false);
2171 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2174 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2177 default_options(sbi);
2179 /* parse mount options */
2180 err = parse_options(sb, data, true);
2185 * Previous and new state of filesystem is RO,
2186 * so skip checking GC and FLUSH_MERGE conditions.
2188 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2191 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2197 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2198 err = dquot_suspend(sb, -1);
2201 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2202 /* dquot_resume needs RW */
2203 sb->s_flags &= ~SB_RDONLY;
2204 if (sb_any_quota_suspended(sb)) {
2205 dquot_resume(sb, -1);
2206 } else if (f2fs_sb_has_quota_ino(sbi)) {
2207 err = f2fs_enable_quotas(sb);
2213 /* disallow enable atgc dynamically */
2214 if (no_atgc == !!test_opt(sbi, ATGC)) {
2216 f2fs_warn(sbi, "switch atgc option is not allowed");
2220 /* disallow enable/disable extent_cache dynamically */
2221 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
2223 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2227 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2229 f2fs_warn(sbi, "switch io_bits option is not allowed");
2233 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2235 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2239 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2241 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2245 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2247 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2252 * We stop the GC thread if FS is mounted as RO
2253 * or if background_gc = off is passed in mount
2254 * option. Also sync the filesystem.
2256 if ((*flags & SB_RDONLY) ||
2257 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2258 !test_opt(sbi, GC_MERGE))) {
2259 if (sbi->gc_thread) {
2260 f2fs_stop_gc_thread(sbi);
2261 need_restart_gc = true;
2263 } else if (!sbi->gc_thread) {
2264 err = f2fs_start_gc_thread(sbi);
2267 need_stop_gc = true;
2270 if (*flags & SB_RDONLY ||
2271 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
2274 set_sbi_flag(sbi, SBI_IS_DIRTY);
2275 set_sbi_flag(sbi, SBI_IS_CLOSE);
2276 f2fs_sync_fs(sb, 1);
2277 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2280 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2281 !test_opt(sbi, MERGE_CHECKPOINT)) {
2282 f2fs_stop_ckpt_thread(sbi);
2283 need_restart_ckpt = true;
2285 err = f2fs_start_ckpt_thread(sbi);
2288 "Failed to start F2FS issue_checkpoint_thread (%d)",
2292 need_stop_ckpt = true;
2296 * We stop issue flush thread if FS is mounted as RO
2297 * or if flush_merge is not passed in mount option.
2299 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2300 clear_opt(sbi, FLUSH_MERGE);
2301 f2fs_destroy_flush_cmd_control(sbi, false);
2302 need_restart_flush = true;
2304 err = f2fs_create_flush_cmd_control(sbi);
2307 need_stop_flush = true;
2310 if (no_discard == !!test_opt(sbi, DISCARD)) {
2311 if (test_opt(sbi, DISCARD)) {
2312 err = f2fs_start_discard_thread(sbi);
2315 need_stop_discard = true;
2317 dcc = SM_I(sbi)->dcc_info;
2318 f2fs_stop_discard_thread(sbi);
2319 if (atomic_read(&dcc->discard_cmd_cnt))
2320 f2fs_issue_discard_timeout(sbi);
2321 need_restart_discard = true;
2325 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2326 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2327 err = f2fs_disable_checkpoint(sbi);
2329 goto restore_discard;
2331 f2fs_enable_checkpoint(sbi);
2337 /* Release old quota file names */
2338 for (i = 0; i < MAXQUOTAS; i++)
2339 kfree(org_mount_opt.s_qf_names[i]);
2341 /* Update the POSIXACL Flag */
2342 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2343 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2345 limit_reserve_root(sbi);
2346 adjust_unusable_cap_perc(sbi);
2347 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2350 if (need_restart_discard) {
2351 if (f2fs_start_discard_thread(sbi))
2352 f2fs_warn(sbi, "discard has been stopped");
2353 } else if (need_stop_discard) {
2354 f2fs_stop_discard_thread(sbi);
2357 if (need_restart_flush) {
2358 if (f2fs_create_flush_cmd_control(sbi))
2359 f2fs_warn(sbi, "background flush thread has stopped");
2360 } else if (need_stop_flush) {
2361 clear_opt(sbi, FLUSH_MERGE);
2362 f2fs_destroy_flush_cmd_control(sbi, false);
2365 if (need_restart_ckpt) {
2366 if (f2fs_start_ckpt_thread(sbi))
2367 f2fs_warn(sbi, "background ckpt thread has stopped");
2368 } else if (need_stop_ckpt) {
2369 f2fs_stop_ckpt_thread(sbi);
2372 if (need_restart_gc) {
2373 if (f2fs_start_gc_thread(sbi))
2374 f2fs_warn(sbi, "background gc thread has stopped");
2375 } else if (need_stop_gc) {
2376 f2fs_stop_gc_thread(sbi);
2380 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2381 for (i = 0; i < MAXQUOTAS; i++) {
2382 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2383 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2386 sbi->mount_opt = org_mount_opt;
2387 sb->s_flags = old_sb_flags;
2392 /* Read data from quotafile */
2393 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2394 size_t len, loff_t off)
2396 struct inode *inode = sb_dqopt(sb)->files[type];
2397 struct address_space *mapping = inode->i_mapping;
2398 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2399 int offset = off & (sb->s_blocksize - 1);
2402 loff_t i_size = i_size_read(inode);
2409 if (off + len > i_size)
2412 while (toread > 0) {
2413 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2415 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2417 if (PTR_ERR(page) == -ENOMEM) {
2418 congestion_wait(BLK_RW_ASYNC,
2419 DEFAULT_IO_TIMEOUT);
2422 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2423 return PTR_ERR(page);
2428 if (unlikely(page->mapping != mapping)) {
2429 f2fs_put_page(page, 1);
2432 if (unlikely(!PageUptodate(page))) {
2433 f2fs_put_page(page, 1);
2434 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2438 kaddr = kmap_atomic(page);
2439 memcpy(data, kaddr + offset, tocopy);
2440 kunmap_atomic(kaddr);
2441 f2fs_put_page(page, 1);
2451 /* Write to quotafile */
2452 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2453 const char *data, size_t len, loff_t off)
2455 struct inode *inode = sb_dqopt(sb)->files[type];
2456 struct address_space *mapping = inode->i_mapping;
2457 const struct address_space_operations *a_ops = mapping->a_ops;
2458 int offset = off & (sb->s_blocksize - 1);
2459 size_t towrite = len;
2461 void *fsdata = NULL;
2466 while (towrite > 0) {
2467 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2470 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2472 if (unlikely(err)) {
2473 if (err == -ENOMEM) {
2474 congestion_wait(BLK_RW_ASYNC,
2475 DEFAULT_IO_TIMEOUT);
2478 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2482 kaddr = kmap_atomic(page);
2483 memcpy(kaddr + offset, data, tocopy);
2484 kunmap_atomic(kaddr);
2485 flush_dcache_page(page);
2487 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2498 inode->i_mtime = inode->i_ctime = current_time(inode);
2499 f2fs_mark_inode_dirty_sync(inode, false);
2500 return len - towrite;
2503 int f2fs_dquot_initialize(struct inode *inode)
2505 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) {
2506 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT);
2510 return dquot_initialize(inode);
2513 static struct dquot **f2fs_get_dquots(struct inode *inode)
2515 return F2FS_I(inode)->i_dquot;
2518 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2520 return &F2FS_I(inode)->i_reserved_quota;
2523 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2525 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2526 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2530 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2531 F2FS_OPTION(sbi).s_jquota_fmt, type);
2534 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2539 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2540 err = f2fs_enable_quotas(sbi->sb);
2542 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2548 for (i = 0; i < MAXQUOTAS; i++) {
2549 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2550 err = f2fs_quota_on_mount(sbi, i);
2555 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2562 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2565 struct inode *qf_inode;
2566 unsigned long qf_inum;
2569 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2571 qf_inum = f2fs_qf_ino(sb, type);
2575 qf_inode = f2fs_iget(sb, qf_inum);
2576 if (IS_ERR(qf_inode)) {
2577 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2578 return PTR_ERR(qf_inode);
2581 /* Don't account quota for quota files to avoid recursion */
2582 qf_inode->i_flags |= S_NOQUOTA;
2583 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2588 static int f2fs_enable_quotas(struct super_block *sb)
2590 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2592 unsigned long qf_inum;
2593 bool quota_mopt[MAXQUOTAS] = {
2594 test_opt(sbi, USRQUOTA),
2595 test_opt(sbi, GRPQUOTA),
2596 test_opt(sbi, PRJQUOTA),
2599 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2600 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2604 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2606 for (type = 0; type < MAXQUOTAS; type++) {
2607 qf_inum = f2fs_qf_ino(sb, type);
2609 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2610 DQUOT_USAGE_ENABLED |
2611 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2613 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2615 for (type--; type >= 0; type--)
2616 dquot_quota_off(sb, type);
2617 set_sbi_flag(F2FS_SB(sb),
2618 SBI_QUOTA_NEED_REPAIR);
2626 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2628 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2629 struct address_space *mapping = dqopt->files[type]->i_mapping;
2632 ret = dquot_writeback_dquots(sbi->sb, type);
2636 ret = filemap_fdatawrite(mapping);
2640 /* if we are using journalled quota */
2641 if (is_journalled_quota(sbi))
2644 ret = filemap_fdatawait(mapping);
2646 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2649 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2653 int f2fs_quota_sync(struct super_block *sb, int type)
2655 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2656 struct quota_info *dqopt = sb_dqopt(sb);
2661 * Now when everything is written we can discard the pagecache so
2662 * that userspace sees the changes.
2664 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2666 if (type != -1 && cnt != type)
2669 if (!sb_has_quota_active(sb, type))
2672 inode_lock(dqopt->files[cnt]);
2677 * down_read(quota_sem)
2678 * dquot_writeback_dquots()
2681 * down_read(quota_sem)
2684 down_read(&sbi->quota_sem);
2686 ret = f2fs_quota_sync_file(sbi, cnt);
2688 up_read(&sbi->quota_sem);
2689 f2fs_unlock_op(sbi);
2691 inode_unlock(dqopt->files[cnt]);
2699 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2700 const struct path *path)
2702 struct inode *inode;
2705 /* if quota sysfile exists, deny enabling quota with specific file */
2706 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2707 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2711 err = f2fs_quota_sync(sb, type);
2715 err = dquot_quota_on(sb, type, format_id, path);
2719 inode = d_inode(path->dentry);
2722 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2723 f2fs_set_inode_flags(inode);
2724 inode_unlock(inode);
2725 f2fs_mark_inode_dirty_sync(inode, false);
2730 static int __f2fs_quota_off(struct super_block *sb, int type)
2732 struct inode *inode = sb_dqopt(sb)->files[type];
2735 if (!inode || !igrab(inode))
2736 return dquot_quota_off(sb, type);
2738 err = f2fs_quota_sync(sb, type);
2742 err = dquot_quota_off(sb, type);
2743 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2747 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2748 f2fs_set_inode_flags(inode);
2749 inode_unlock(inode);
2750 f2fs_mark_inode_dirty_sync(inode, false);
2756 static int f2fs_quota_off(struct super_block *sb, int type)
2758 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2761 err = __f2fs_quota_off(sb, type);
2764 * quotactl can shutdown journalled quota, result in inconsistence
2765 * between quota record and fs data by following updates, tag the
2766 * flag to let fsck be aware of it.
2768 if (is_journalled_quota(sbi))
2769 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2773 void f2fs_quota_off_umount(struct super_block *sb)
2778 for (type = 0; type < MAXQUOTAS; type++) {
2779 err = __f2fs_quota_off(sb, type);
2781 int ret = dquot_quota_off(sb, type);
2783 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2785 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2789 * In case of checkpoint=disable, we must flush quota blocks.
2790 * This can cause NULL exception for node_inode in end_io, since
2791 * put_super already dropped it.
2793 sync_filesystem(sb);
2796 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2798 struct quota_info *dqopt = sb_dqopt(sb);
2801 for (type = 0; type < MAXQUOTAS; type++) {
2802 if (!dqopt->files[type])
2804 f2fs_inode_synced(dqopt->files[type]);
2808 static int f2fs_dquot_commit(struct dquot *dquot)
2810 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2813 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2814 ret = dquot_commit(dquot);
2816 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2817 up_read(&sbi->quota_sem);
2821 static int f2fs_dquot_acquire(struct dquot *dquot)
2823 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2826 down_read(&sbi->quota_sem);
2827 ret = dquot_acquire(dquot);
2829 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2830 up_read(&sbi->quota_sem);
2834 static int f2fs_dquot_release(struct dquot *dquot)
2836 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2837 int ret = dquot_release(dquot);
2840 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2844 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2846 struct super_block *sb = dquot->dq_sb;
2847 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2848 int ret = dquot_mark_dquot_dirty(dquot);
2850 /* if we are using journalled quota */
2851 if (is_journalled_quota(sbi))
2852 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2857 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2859 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2860 int ret = dquot_commit_info(sb, type);
2863 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2867 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2869 *projid = F2FS_I(inode)->i_projid;
2873 static const struct dquot_operations f2fs_quota_operations = {
2874 .get_reserved_space = f2fs_get_reserved_space,
2875 .write_dquot = f2fs_dquot_commit,
2876 .acquire_dquot = f2fs_dquot_acquire,
2877 .release_dquot = f2fs_dquot_release,
2878 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2879 .write_info = f2fs_dquot_commit_info,
2880 .alloc_dquot = dquot_alloc,
2881 .destroy_dquot = dquot_destroy,
2882 .get_projid = f2fs_get_projid,
2883 .get_next_id = dquot_get_next_id,
2886 static const struct quotactl_ops f2fs_quotactl_ops = {
2887 .quota_on = f2fs_quota_on,
2888 .quota_off = f2fs_quota_off,
2889 .quota_sync = f2fs_quota_sync,
2890 .get_state = dquot_get_state,
2891 .set_info = dquot_set_dqinfo,
2892 .get_dqblk = dquot_get_dqblk,
2893 .set_dqblk = dquot_set_dqblk,
2894 .get_nextdqblk = dquot_get_next_dqblk,
2897 int f2fs_dquot_initialize(struct inode *inode)
2902 int f2fs_quota_sync(struct super_block *sb, int type)
2907 void f2fs_quota_off_umount(struct super_block *sb)
2912 static const struct super_operations f2fs_sops = {
2913 .alloc_inode = f2fs_alloc_inode,
2914 .free_inode = f2fs_free_inode,
2915 .drop_inode = f2fs_drop_inode,
2916 .write_inode = f2fs_write_inode,
2917 .dirty_inode = f2fs_dirty_inode,
2918 .show_options = f2fs_show_options,
2920 .quota_read = f2fs_quota_read,
2921 .quota_write = f2fs_quota_write,
2922 .get_dquots = f2fs_get_dquots,
2924 .evict_inode = f2fs_evict_inode,
2925 .put_super = f2fs_put_super,
2926 .sync_fs = f2fs_sync_fs,
2927 .freeze_fs = f2fs_freeze,
2928 .unfreeze_fs = f2fs_unfreeze,
2929 .statfs = f2fs_statfs,
2930 .remount_fs = f2fs_remount,
2933 #ifdef CONFIG_FS_ENCRYPTION
2934 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2936 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2937 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2941 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2944 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2947 * Encrypting the root directory is not allowed because fsck
2948 * expects lost+found directory to exist and remain unencrypted
2949 * if LOST_FOUND feature is enabled.
2952 if (f2fs_sb_has_lost_found(sbi) &&
2953 inode->i_ino == F2FS_ROOT_INO(sbi))
2956 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2957 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2958 ctx, len, fs_data, XATTR_CREATE);
2961 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
2963 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
2966 static bool f2fs_has_stable_inodes(struct super_block *sb)
2971 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2972 int *ino_bits_ret, int *lblk_bits_ret)
2974 *ino_bits_ret = 8 * sizeof(nid_t);
2975 *lblk_bits_ret = 8 * sizeof(block_t);
2978 static int f2fs_get_num_devices(struct super_block *sb)
2980 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2982 if (f2fs_is_multi_device(sbi))
2983 return sbi->s_ndevs;
2987 static void f2fs_get_devices(struct super_block *sb,
2988 struct request_queue **devs)
2990 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2993 for (i = 0; i < sbi->s_ndevs; i++)
2994 devs[i] = bdev_get_queue(FDEV(i).bdev);
2997 static const struct fscrypt_operations f2fs_cryptops = {
2998 .key_prefix = "f2fs:",
2999 .get_context = f2fs_get_context,
3000 .set_context = f2fs_set_context,
3001 .get_dummy_policy = f2fs_get_dummy_policy,
3002 .empty_dir = f2fs_empty_dir,
3003 .has_stable_inodes = f2fs_has_stable_inodes,
3004 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
3005 .get_num_devices = f2fs_get_num_devices,
3006 .get_devices = f2fs_get_devices,
3010 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3011 u64 ino, u32 generation)
3013 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3014 struct inode *inode;
3016 if (f2fs_check_nid_range(sbi, ino))
3017 return ERR_PTR(-ESTALE);
3020 * f2fs_iget isn't quite right if the inode is currently unallocated!
3021 * However f2fs_iget currently does appropriate checks to handle stale
3022 * inodes so everything is OK.
3024 inode = f2fs_iget(sb, ino);
3026 return ERR_CAST(inode);
3027 if (unlikely(generation && inode->i_generation != generation)) {
3028 /* we didn't find the right inode.. */
3030 return ERR_PTR(-ESTALE);
3035 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3036 int fh_len, int fh_type)
3038 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3039 f2fs_nfs_get_inode);
3042 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3043 int fh_len, int fh_type)
3045 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3046 f2fs_nfs_get_inode);
3049 static const struct export_operations f2fs_export_ops = {
3050 .fh_to_dentry = f2fs_fh_to_dentry,
3051 .fh_to_parent = f2fs_fh_to_parent,
3052 .get_parent = f2fs_get_parent,
3055 loff_t max_file_blocks(struct inode *inode)
3061 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3062 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3063 * space in inode.i_addr, it will be more safe to reassign
3067 if (inode && f2fs_compressed_file(inode))
3068 leaf_count = ADDRS_PER_BLOCK(inode);
3070 leaf_count = DEF_ADDRS_PER_BLOCK;
3072 /* two direct node blocks */
3073 result += (leaf_count * 2);
3075 /* two indirect node blocks */
3076 leaf_count *= NIDS_PER_BLOCK;
3077 result += (leaf_count * 2);
3079 /* one double indirect node block */
3080 leaf_count *= NIDS_PER_BLOCK;
3081 result += leaf_count;
3086 static int __f2fs_commit_super(struct buffer_head *bh,
3087 struct f2fs_super_block *super)
3091 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3092 set_buffer_dirty(bh);
3095 /* it's rare case, we can do fua all the time */
3096 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3099 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3100 struct buffer_head *bh)
3102 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3103 (bh->b_data + F2FS_SUPER_OFFSET);
3104 struct super_block *sb = sbi->sb;
3105 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3106 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3107 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3108 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3109 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3110 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3111 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3112 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3113 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3114 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3115 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3116 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3117 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3118 u64 main_end_blkaddr = main_blkaddr +
3119 (segment_count_main << log_blocks_per_seg);
3120 u64 seg_end_blkaddr = segment0_blkaddr +
3121 (segment_count << log_blocks_per_seg);
3123 if (segment0_blkaddr != cp_blkaddr) {
3124 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3125 segment0_blkaddr, cp_blkaddr);
3129 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3131 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3132 cp_blkaddr, sit_blkaddr,
3133 segment_count_ckpt << log_blocks_per_seg);
3137 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3139 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3140 sit_blkaddr, nat_blkaddr,
3141 segment_count_sit << log_blocks_per_seg);
3145 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3147 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3148 nat_blkaddr, ssa_blkaddr,
3149 segment_count_nat << log_blocks_per_seg);
3153 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3155 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3156 ssa_blkaddr, main_blkaddr,
3157 segment_count_ssa << log_blocks_per_seg);
3161 if (main_end_blkaddr > seg_end_blkaddr) {
3162 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3163 main_blkaddr, seg_end_blkaddr,
3164 segment_count_main << log_blocks_per_seg);
3166 } else if (main_end_blkaddr < seg_end_blkaddr) {
3170 /* fix in-memory information all the time */
3171 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3172 segment0_blkaddr) >> log_blocks_per_seg);
3174 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3175 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3178 err = __f2fs_commit_super(bh, NULL);
3179 res = err ? "failed" : "done";
3181 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3182 res, main_blkaddr, seg_end_blkaddr,
3183 segment_count_main << log_blocks_per_seg);
3190 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3191 struct buffer_head *bh)
3193 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3194 block_t total_sections, blocks_per_seg;
3195 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3196 (bh->b_data + F2FS_SUPER_OFFSET);
3197 size_t crc_offset = 0;
3200 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3201 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3202 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3206 /* Check checksum_offset and crc in superblock */
3207 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3208 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3210 offsetof(struct f2fs_super_block, crc)) {
3211 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3213 return -EFSCORRUPTED;
3215 crc = le32_to_cpu(raw_super->crc);
3216 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3217 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3218 return -EFSCORRUPTED;
3222 /* Currently, support only 4KB block size */
3223 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3224 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3225 le32_to_cpu(raw_super->log_blocksize),
3227 return -EFSCORRUPTED;
3230 /* check log blocks per segment */
3231 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3232 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3233 le32_to_cpu(raw_super->log_blocks_per_seg));
3234 return -EFSCORRUPTED;
3237 /* Currently, support 512/1024/2048/4096 bytes sector size */
3238 if (le32_to_cpu(raw_super->log_sectorsize) >
3239 F2FS_MAX_LOG_SECTOR_SIZE ||
3240 le32_to_cpu(raw_super->log_sectorsize) <
3241 F2FS_MIN_LOG_SECTOR_SIZE) {
3242 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3243 le32_to_cpu(raw_super->log_sectorsize));
3244 return -EFSCORRUPTED;
3246 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3247 le32_to_cpu(raw_super->log_sectorsize) !=
3248 F2FS_MAX_LOG_SECTOR_SIZE) {
3249 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3250 le32_to_cpu(raw_super->log_sectors_per_block),
3251 le32_to_cpu(raw_super->log_sectorsize));
3252 return -EFSCORRUPTED;
3255 segment_count = le32_to_cpu(raw_super->segment_count);
3256 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3257 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3258 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3259 total_sections = le32_to_cpu(raw_super->section_count);
3261 /* blocks_per_seg should be 512, given the above check */
3262 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3264 if (segment_count > F2FS_MAX_SEGMENT ||
3265 segment_count < F2FS_MIN_SEGMENTS) {
3266 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3267 return -EFSCORRUPTED;
3270 if (total_sections > segment_count_main || total_sections < 1 ||
3271 segs_per_sec > segment_count || !segs_per_sec) {
3272 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3273 segment_count, total_sections, segs_per_sec);
3274 return -EFSCORRUPTED;
3277 if (segment_count_main != total_sections * segs_per_sec) {
3278 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3279 segment_count_main, total_sections, segs_per_sec);
3280 return -EFSCORRUPTED;
3283 if ((segment_count / segs_per_sec) < total_sections) {
3284 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3285 segment_count, segs_per_sec, total_sections);
3286 return -EFSCORRUPTED;
3289 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3290 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3291 segment_count, le64_to_cpu(raw_super->block_count));
3292 return -EFSCORRUPTED;
3295 if (RDEV(0).path[0]) {
3296 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3299 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3300 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3303 if (segment_count != dev_seg_count) {
3304 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3305 segment_count, dev_seg_count);
3306 return -EFSCORRUPTED;
3309 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3310 !bdev_is_zoned(sbi->sb->s_bdev)) {
3311 f2fs_info(sbi, "Zoned block device path is missing");
3312 return -EFSCORRUPTED;
3316 if (secs_per_zone > total_sections || !secs_per_zone) {
3317 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3318 secs_per_zone, total_sections);
3319 return -EFSCORRUPTED;
3321 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3322 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3323 (le32_to_cpu(raw_super->extension_count) +
3324 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3325 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3326 le32_to_cpu(raw_super->extension_count),
3327 raw_super->hot_ext_count,
3328 F2FS_MAX_EXTENSION);
3329 return -EFSCORRUPTED;
3332 if (le32_to_cpu(raw_super->cp_payload) >=
3333 (blocks_per_seg - F2FS_CP_PACKS -
3334 NR_CURSEG_PERSIST_TYPE)) {
3335 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3336 le32_to_cpu(raw_super->cp_payload),
3337 blocks_per_seg - F2FS_CP_PACKS -
3338 NR_CURSEG_PERSIST_TYPE);
3339 return -EFSCORRUPTED;
3342 /* check reserved ino info */
3343 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3344 le32_to_cpu(raw_super->meta_ino) != 2 ||
3345 le32_to_cpu(raw_super->root_ino) != 3) {
3346 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3347 le32_to_cpu(raw_super->node_ino),
3348 le32_to_cpu(raw_super->meta_ino),
3349 le32_to_cpu(raw_super->root_ino));
3350 return -EFSCORRUPTED;
3353 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3354 if (sanity_check_area_boundary(sbi, bh))
3355 return -EFSCORRUPTED;
3360 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3362 unsigned int total, fsmeta;
3363 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3364 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3365 unsigned int ovp_segments, reserved_segments;
3366 unsigned int main_segs, blocks_per_seg;
3367 unsigned int sit_segs, nat_segs;
3368 unsigned int sit_bitmap_size, nat_bitmap_size;
3369 unsigned int log_blocks_per_seg;
3370 unsigned int segment_count_main;
3371 unsigned int cp_pack_start_sum, cp_payload;
3372 block_t user_block_count, valid_user_blocks;
3373 block_t avail_node_count, valid_node_count;
3374 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3377 total = le32_to_cpu(raw_super->segment_count);
3378 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3379 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3381 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3383 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3384 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3386 if (unlikely(fsmeta >= total))
3389 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3390 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3392 if (!f2fs_sb_has_readonly(sbi) &&
3393 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3394 ovp_segments == 0 || reserved_segments == 0)) {
3395 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3398 user_block_count = le64_to_cpu(ckpt->user_block_count);
3399 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3400 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3401 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3402 if (!user_block_count || user_block_count >=
3403 segment_count_main << log_blocks_per_seg) {
3404 f2fs_err(sbi, "Wrong user_block_count: %u",
3409 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3410 if (valid_user_blocks > user_block_count) {
3411 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3412 valid_user_blocks, user_block_count);
3416 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3417 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3418 if (valid_node_count > avail_node_count) {
3419 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3420 valid_node_count, avail_node_count);
3424 main_segs = le32_to_cpu(raw_super->segment_count_main);
3425 blocks_per_seg = sbi->blocks_per_seg;
3427 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3428 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3429 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3432 if (f2fs_sb_has_readonly(sbi))
3435 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3436 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3437 le32_to_cpu(ckpt->cur_node_segno[j])) {
3438 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3440 le32_to_cpu(ckpt->cur_node_segno[i]));
3446 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3447 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3448 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3451 if (f2fs_sb_has_readonly(sbi))
3454 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3455 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3456 le32_to_cpu(ckpt->cur_data_segno[j])) {
3457 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3459 le32_to_cpu(ckpt->cur_data_segno[i]));
3464 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3465 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3466 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3467 le32_to_cpu(ckpt->cur_data_segno[j])) {
3468 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3470 le32_to_cpu(ckpt->cur_node_segno[i]));
3476 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3477 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3479 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3480 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3481 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3482 sit_bitmap_size, nat_bitmap_size);
3486 cp_pack_start_sum = __start_sum_addr(sbi);
3487 cp_payload = __cp_payload(sbi);
3488 if (cp_pack_start_sum < cp_payload + 1 ||
3489 cp_pack_start_sum > blocks_per_seg - 1 -
3490 NR_CURSEG_PERSIST_TYPE) {
3491 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3496 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3497 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3498 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3499 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3500 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3501 le32_to_cpu(ckpt->checksum_offset));
3505 nat_blocks = nat_segs << log_blocks_per_seg;
3506 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3507 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3508 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3509 (cp_payload + F2FS_CP_PACKS +
3510 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3511 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3512 cp_payload, nat_bits_blocks);
3516 if (unlikely(f2fs_cp_error(sbi))) {
3517 f2fs_err(sbi, "A bug case: need to run fsck");
3523 static void init_sb_info(struct f2fs_sb_info *sbi)
3525 struct f2fs_super_block *raw_super = sbi->raw_super;
3528 sbi->log_sectors_per_block =
3529 le32_to_cpu(raw_super->log_sectors_per_block);
3530 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3531 sbi->blocksize = 1 << sbi->log_blocksize;
3532 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3533 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3534 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3535 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3536 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3537 sbi->total_node_count =
3538 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3539 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3540 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3541 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3542 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3543 sbi->cur_victim_sec = NULL_SECNO;
3544 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3545 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3546 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3547 sbi->migration_granularity = sbi->segs_per_sec;
3548 sbi->seq_file_ra_mul = MIN_RA_MUL;
3549 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3550 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3552 sbi->dir_level = DEF_DIR_LEVEL;
3553 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3554 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3555 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3556 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3557 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3558 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3559 DEF_UMOUNT_DISCARD_TIMEOUT;
3560 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3562 for (i = 0; i < NR_COUNT_TYPE; i++)
3563 atomic_set(&sbi->nr_pages[i], 0);
3565 for (i = 0; i < META; i++)
3566 atomic_set(&sbi->wb_sync_req[i], 0);
3568 INIT_LIST_HEAD(&sbi->s_list);
3569 mutex_init(&sbi->umount_mutex);
3570 init_rwsem(&sbi->io_order_lock);
3571 spin_lock_init(&sbi->cp_lock);
3573 sbi->dirty_device = 0;
3574 spin_lock_init(&sbi->dev_lock);
3576 init_rwsem(&sbi->sb_lock);
3577 init_rwsem(&sbi->pin_sem);
3580 static int init_percpu_info(struct f2fs_sb_info *sbi)
3584 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3588 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3591 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3596 #ifdef CONFIG_BLK_DEV_ZONED
3598 struct f2fs_report_zones_args {
3599 struct f2fs_dev_info *dev;
3600 bool zone_cap_mismatch;
3603 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3606 struct f2fs_report_zones_args *rz_args = data;
3608 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3611 set_bit(idx, rz_args->dev->blkz_seq);
3612 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >>
3613 F2FS_LOG_SECTORS_PER_BLOCK;
3614 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch)
3615 rz_args->zone_cap_mismatch = true;
3620 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3622 struct block_device *bdev = FDEV(devi).bdev;
3623 sector_t nr_sectors = bdev_nr_sectors(bdev);
3624 struct f2fs_report_zones_args rep_zone_arg;
3627 if (!f2fs_sb_has_blkzoned(sbi))
3630 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3631 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3633 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3634 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3635 __ilog2_u32(sbi->blocks_per_blkz))
3637 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3638 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3639 sbi->log_blocks_per_blkz;
3640 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3641 FDEV(devi).nr_blkz++;
3643 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3644 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3645 * sizeof(unsigned long),
3647 if (!FDEV(devi).blkz_seq)
3650 /* Get block zones type and zone-capacity */
3651 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi,
3652 FDEV(devi).nr_blkz * sizeof(block_t),
3654 if (!FDEV(devi).zone_capacity_blocks)
3657 rep_zone_arg.dev = &FDEV(devi);
3658 rep_zone_arg.zone_cap_mismatch = false;
3660 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3665 if (!rep_zone_arg.zone_cap_mismatch) {
3666 kfree(FDEV(devi).zone_capacity_blocks);
3667 FDEV(devi).zone_capacity_blocks = NULL;
3675 * Read f2fs raw super block.
3676 * Because we have two copies of super block, so read both of them
3677 * to get the first valid one. If any one of them is broken, we pass
3678 * them recovery flag back to the caller.
3680 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3681 struct f2fs_super_block **raw_super,
3682 int *valid_super_block, int *recovery)
3684 struct super_block *sb = sbi->sb;
3686 struct buffer_head *bh;
3687 struct f2fs_super_block *super;
3690 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3694 for (block = 0; block < 2; block++) {
3695 bh = sb_bread(sb, block);
3697 f2fs_err(sbi, "Unable to read %dth superblock",
3704 /* sanity checking of raw super */
3705 err = sanity_check_raw_super(sbi, bh);
3707 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3715 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3717 *valid_super_block = block;
3723 /* No valid superblock */
3732 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3734 struct buffer_head *bh;
3738 if ((recover && f2fs_readonly(sbi->sb)) ||
3739 bdev_read_only(sbi->sb->s_bdev)) {
3740 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3744 /* we should update superblock crc here */
3745 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3746 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3747 offsetof(struct f2fs_super_block, crc));
3748 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3751 /* write back-up superblock first */
3752 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3755 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3758 /* if we are in recovery path, skip writing valid superblock */
3762 /* write current valid superblock */
3763 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3766 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3771 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3773 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3774 unsigned int max_devices = MAX_DEVICES;
3775 unsigned int logical_blksize;
3778 /* Initialize single device information */
3779 if (!RDEV(0).path[0]) {
3780 if (!bdev_is_zoned(sbi->sb->s_bdev))
3786 * Initialize multiple devices information, or single
3787 * zoned block device information.
3789 sbi->devs = f2fs_kzalloc(sbi,
3790 array_size(max_devices,
3791 sizeof(struct f2fs_dev_info)),
3796 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3797 sbi->aligned_blksize = true;
3799 for (i = 0; i < max_devices; i++) {
3801 if (i > 0 && !RDEV(i).path[0])
3804 if (max_devices == 1) {
3805 /* Single zoned block device mount */
3807 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3808 sbi->sb->s_mode, sbi->sb->s_type);
3810 /* Multi-device mount */
3811 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3812 FDEV(i).total_segments =
3813 le32_to_cpu(RDEV(i).total_segments);
3815 FDEV(i).start_blk = 0;
3816 FDEV(i).end_blk = FDEV(i).start_blk +
3817 (FDEV(i).total_segments <<
3818 sbi->log_blocks_per_seg) - 1 +
3819 le32_to_cpu(raw_super->segment0_blkaddr);
3821 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3822 FDEV(i).end_blk = FDEV(i).start_blk +
3823 (FDEV(i).total_segments <<
3824 sbi->log_blocks_per_seg) - 1;
3826 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3827 sbi->sb->s_mode, sbi->sb->s_type);
3829 if (IS_ERR(FDEV(i).bdev))
3830 return PTR_ERR(FDEV(i).bdev);
3832 /* to release errored devices */
3833 sbi->s_ndevs = i + 1;
3835 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
3836 sbi->aligned_blksize = false;
3838 #ifdef CONFIG_BLK_DEV_ZONED
3839 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3840 !f2fs_sb_has_blkzoned(sbi)) {
3841 f2fs_err(sbi, "Zoned block device feature not enabled");
3844 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3845 if (init_blkz_info(sbi, i)) {
3846 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3849 if (max_devices == 1)
3851 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3853 FDEV(i).total_segments,
3854 FDEV(i).start_blk, FDEV(i).end_blk,
3855 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3856 "Host-aware" : "Host-managed");
3860 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3862 FDEV(i).total_segments,
3863 FDEV(i).start_blk, FDEV(i).end_blk);
3866 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3870 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3872 #ifdef CONFIG_UNICODE
3873 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3874 const struct f2fs_sb_encodings *encoding_info;
3875 struct unicode_map *encoding;
3876 __u16 encoding_flags;
3878 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3881 "Encoding requested by superblock is unknown");
3885 encoding = utf8_load(encoding_info->version);
3886 if (IS_ERR(encoding)) {
3888 "can't mount with superblock charset: %s-%s "
3889 "not supported by the kernel. flags: 0x%x.",
3890 encoding_info->name, encoding_info->version,
3892 return PTR_ERR(encoding);
3894 f2fs_info(sbi, "Using encoding defined by superblock: "
3895 "%s-%s with flags 0x%hx", encoding_info->name,
3896 encoding_info->version?:"\b", encoding_flags);
3898 sbi->sb->s_encoding = encoding;
3899 sbi->sb->s_encoding_flags = encoding_flags;
3902 if (f2fs_sb_has_casefold(sbi)) {
3903 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3910 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3912 struct f2fs_sm_info *sm_i = SM_I(sbi);
3914 /* adjust parameters according to the volume size */
3915 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3916 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3917 if (f2fs_block_unit_discard(sbi))
3918 sm_i->dcc_info->discard_granularity = 1;
3919 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3922 sbi->readdir_ra = 1;
3925 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3927 struct f2fs_sb_info *sbi;
3928 struct f2fs_super_block *raw_super;
3931 bool skip_recovery = false, need_fsck = false;
3932 char *options = NULL;
3933 int recovery, i, valid_super_block;
3934 struct curseg_info *seg_i;
3940 valid_super_block = -1;
3943 /* allocate memory for f2fs-specific super block info */
3944 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3950 /* Load the checksum driver */
3951 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3952 if (IS_ERR(sbi->s_chksum_driver)) {
3953 f2fs_err(sbi, "Cannot load crc32 driver.");
3954 err = PTR_ERR(sbi->s_chksum_driver);
3955 sbi->s_chksum_driver = NULL;
3959 /* set a block size */
3960 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3961 f2fs_err(sbi, "unable to set blocksize");
3965 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3970 sb->s_fs_info = sbi;
3971 sbi->raw_super = raw_super;
3973 /* precompute checksum seed for metadata */
3974 if (f2fs_sb_has_inode_chksum(sbi))
3975 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3976 sizeof(raw_super->uuid));
3978 default_options(sbi);
3979 /* parse mount options */
3980 options = kstrdup((const char *)data, GFP_KERNEL);
3981 if (data && !options) {
3986 err = parse_options(sb, options, false);
3990 sb->s_maxbytes = max_file_blocks(NULL) <<
3991 le32_to_cpu(raw_super->log_blocksize);
3992 sb->s_max_links = F2FS_LINK_MAX;
3994 err = f2fs_setup_casefold(sbi);
3999 sb->dq_op = &f2fs_quota_operations;
4000 sb->s_qcop = &f2fs_quotactl_ops;
4001 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4003 if (f2fs_sb_has_quota_ino(sbi)) {
4004 for (i = 0; i < MAXQUOTAS; i++) {
4005 if (f2fs_qf_ino(sbi->sb, i))
4006 sbi->nquota_files++;
4011 sb->s_op = &f2fs_sops;
4012 #ifdef CONFIG_FS_ENCRYPTION
4013 sb->s_cop = &f2fs_cryptops;
4015 #ifdef CONFIG_FS_VERITY
4016 sb->s_vop = &f2fs_verityops;
4018 sb->s_xattr = f2fs_xattr_handlers;
4019 sb->s_export_op = &f2fs_export_ops;
4020 sb->s_magic = F2FS_SUPER_MAGIC;
4021 sb->s_time_gran = 1;
4022 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4023 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4024 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4025 sb->s_iflags |= SB_I_CGROUPWB;
4027 /* init f2fs-specific super block info */
4028 sbi->valid_super_block = valid_super_block;
4029 init_rwsem(&sbi->gc_lock);
4030 mutex_init(&sbi->writepages);
4031 init_rwsem(&sbi->cp_global_sem);
4032 init_rwsem(&sbi->node_write);
4033 init_rwsem(&sbi->node_change);
4035 /* disallow all the data/node/meta page writes */
4036 set_sbi_flag(sbi, SBI_POR_DOING);
4037 spin_lock_init(&sbi->stat_lock);
4039 for (i = 0; i < NR_PAGE_TYPE; i++) {
4040 int n = (i == META) ? 1 : NR_TEMP_TYPE;
4046 sizeof(struct f2fs_bio_info)),
4048 if (!sbi->write_io[i]) {
4053 for (j = HOT; j < n; j++) {
4054 init_rwsem(&sbi->write_io[i][j].io_rwsem);
4055 sbi->write_io[i][j].sbi = sbi;
4056 sbi->write_io[i][j].bio = NULL;
4057 spin_lock_init(&sbi->write_io[i][j].io_lock);
4058 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
4059 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
4060 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
4064 init_rwsem(&sbi->cp_rwsem);
4065 init_rwsem(&sbi->quota_sem);
4066 init_waitqueue_head(&sbi->cp_wait);
4069 err = f2fs_init_iostat(sbi);
4073 err = init_percpu_info(sbi);
4077 if (F2FS_IO_ALIGNED(sbi)) {
4078 sbi->write_io_dummy =
4079 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4080 if (!sbi->write_io_dummy) {
4086 /* init per sbi slab cache */
4087 err = f2fs_init_xattr_caches(sbi);
4090 err = f2fs_init_page_array_cache(sbi);
4092 goto free_xattr_cache;
4094 /* get an inode for meta space */
4095 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4096 if (IS_ERR(sbi->meta_inode)) {
4097 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4098 err = PTR_ERR(sbi->meta_inode);
4099 goto free_page_array_cache;
4102 err = f2fs_get_valid_checkpoint(sbi);
4104 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4105 goto free_meta_inode;
4108 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4109 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4110 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4111 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4112 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4115 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4116 set_sbi_flag(sbi, SBI_NEED_FSCK);
4118 /* Initialize device list */
4119 err = f2fs_scan_devices(sbi);
4121 f2fs_err(sbi, "Failed to find devices");
4125 err = f2fs_init_post_read_wq(sbi);
4127 f2fs_err(sbi, "Failed to initialize post read workqueue");
4131 sbi->total_valid_node_count =
4132 le32_to_cpu(sbi->ckpt->valid_node_count);
4133 percpu_counter_set(&sbi->total_valid_inode_count,
4134 le32_to_cpu(sbi->ckpt->valid_inode_count));
4135 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4136 sbi->total_valid_block_count =
4137 le64_to_cpu(sbi->ckpt->valid_block_count);
4138 sbi->last_valid_block_count = sbi->total_valid_block_count;
4139 sbi->reserved_blocks = 0;
4140 sbi->current_reserved_blocks = 0;
4141 limit_reserve_root(sbi);
4142 adjust_unusable_cap_perc(sbi);
4144 for (i = 0; i < NR_INODE_TYPE; i++) {
4145 INIT_LIST_HEAD(&sbi->inode_list[i]);
4146 spin_lock_init(&sbi->inode_lock[i]);
4148 mutex_init(&sbi->flush_lock);
4150 f2fs_init_extent_cache_info(sbi);
4152 f2fs_init_ino_entry_info(sbi);
4154 f2fs_init_fsync_node_info(sbi);
4156 /* setup checkpoint request control and start checkpoint issue thread */
4157 f2fs_init_ckpt_req_control(sbi);
4158 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4159 test_opt(sbi, MERGE_CHECKPOINT)) {
4160 err = f2fs_start_ckpt_thread(sbi);
4163 "Failed to start F2FS issue_checkpoint_thread (%d)",
4165 goto stop_ckpt_thread;
4169 /* setup f2fs internal modules */
4170 err = f2fs_build_segment_manager(sbi);
4172 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4176 err = f2fs_build_node_manager(sbi);
4178 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4183 /* For write statistics */
4184 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4186 /* Read accumulated write IO statistics if exists */
4187 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4188 if (__exist_node_summaries(sbi))
4189 sbi->kbytes_written =
4190 le64_to_cpu(seg_i->journal->info.kbytes_written);
4192 f2fs_build_gc_manager(sbi);
4194 err = f2fs_build_stats(sbi);
4198 /* get an inode for node space */
4199 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4200 if (IS_ERR(sbi->node_inode)) {
4201 f2fs_err(sbi, "Failed to read node inode");
4202 err = PTR_ERR(sbi->node_inode);
4206 /* read root inode and dentry */
4207 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4209 f2fs_err(sbi, "Failed to read root inode");
4210 err = PTR_ERR(root);
4211 goto free_node_inode;
4213 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4214 !root->i_size || !root->i_nlink) {
4217 goto free_node_inode;
4220 sb->s_root = d_make_root(root); /* allocate root dentry */
4223 goto free_node_inode;
4226 err = f2fs_init_compress_inode(sbi);
4228 goto free_root_inode;
4230 err = f2fs_register_sysfs(sbi);
4232 goto free_compress_inode;
4235 /* Enable quota usage during mount */
4236 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4237 err = f2fs_enable_quotas(sb);
4239 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4242 /* if there are any orphan inodes, free them */
4243 err = f2fs_recover_orphan_inodes(sbi);
4247 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4248 goto reset_checkpoint;
4250 /* recover fsynced data */
4251 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4252 !test_opt(sbi, NORECOVERY)) {
4254 * mount should be failed, when device has readonly mode, and
4255 * previous checkpoint was not done by clean system shutdown.
4257 if (f2fs_hw_is_readonly(sbi)) {
4258 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4259 err = f2fs_recover_fsync_data(sbi, true);
4262 f2fs_err(sbi, "Need to recover fsync data, but "
4263 "write access unavailable, please try "
4264 "mount w/ disable_roll_forward or norecovery");
4269 f2fs_info(sbi, "write access unavailable, skipping recovery");
4270 goto reset_checkpoint;
4274 set_sbi_flag(sbi, SBI_NEED_FSCK);
4277 goto reset_checkpoint;
4279 err = f2fs_recover_fsync_data(sbi, false);
4282 skip_recovery = true;
4284 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4289 err = f2fs_recover_fsync_data(sbi, true);
4291 if (!f2fs_readonly(sb) && err > 0) {
4293 f2fs_err(sbi, "Need to recover fsync data");
4299 * If the f2fs is not readonly and fsync data recovery succeeds,
4300 * check zoned block devices' write pointer consistency.
4302 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4303 err = f2fs_check_write_pointer(sbi);
4309 f2fs_init_inmem_curseg(sbi);
4311 /* f2fs_recover_fsync_data() cleared this already */
4312 clear_sbi_flag(sbi, SBI_POR_DOING);
4314 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4315 err = f2fs_disable_checkpoint(sbi);
4317 goto sync_free_meta;
4318 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4319 f2fs_enable_checkpoint(sbi);
4323 * If filesystem is not mounted as read-only then
4324 * do start the gc_thread.
4326 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4327 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4328 /* After POR, we can run background GC thread.*/
4329 err = f2fs_start_gc_thread(sbi);
4331 goto sync_free_meta;
4335 /* recover broken superblock */
4337 err = f2fs_commit_super(sbi, true);
4338 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4339 sbi->valid_super_block ? 1 : 2, err);
4342 f2fs_join_shrinker(sbi);
4344 f2fs_tuning_parameters(sbi);
4346 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4347 cur_cp_version(F2FS_CKPT(sbi)));
4348 f2fs_update_time(sbi, CP_TIME);
4349 f2fs_update_time(sbi, REQ_TIME);
4350 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4354 /* safe to flush all the data */
4355 sync_filesystem(sbi->sb);
4360 f2fs_truncate_quota_inode_pages(sb);
4361 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4362 f2fs_quota_off_umount(sbi->sb);
4365 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4366 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4367 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4368 * falls into an infinite loop in f2fs_sync_meta_pages().
4370 truncate_inode_pages_final(META_MAPPING(sbi));
4371 /* evict some inodes being cached by GC */
4373 f2fs_unregister_sysfs(sbi);
4374 free_compress_inode:
4375 f2fs_destroy_compress_inode(sbi);
4380 f2fs_release_ino_entry(sbi, true);
4381 truncate_inode_pages_final(NODE_MAPPING(sbi));
4382 iput(sbi->node_inode);
4383 sbi->node_inode = NULL;
4385 f2fs_destroy_stats(sbi);
4387 /* stop discard thread before destroying node manager */
4388 f2fs_stop_discard_thread(sbi);
4389 f2fs_destroy_node_manager(sbi);
4391 f2fs_destroy_segment_manager(sbi);
4392 f2fs_destroy_post_read_wq(sbi);
4394 f2fs_stop_ckpt_thread(sbi);
4396 destroy_device_list(sbi);
4399 make_bad_inode(sbi->meta_inode);
4400 iput(sbi->meta_inode);
4401 sbi->meta_inode = NULL;
4402 free_page_array_cache:
4403 f2fs_destroy_page_array_cache(sbi);
4405 f2fs_destroy_xattr_caches(sbi);
4407 mempool_destroy(sbi->write_io_dummy);
4409 destroy_percpu_info(sbi);
4411 f2fs_destroy_iostat(sbi);
4413 for (i = 0; i < NR_PAGE_TYPE; i++)
4414 kvfree(sbi->write_io[i]);
4416 #ifdef CONFIG_UNICODE
4417 utf8_unload(sb->s_encoding);
4418 sb->s_encoding = NULL;
4422 for (i = 0; i < MAXQUOTAS; i++)
4423 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4425 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4430 if (sbi->s_chksum_driver)
4431 crypto_free_shash(sbi->s_chksum_driver);
4434 /* give only one another chance */
4435 if (retry_cnt > 0 && skip_recovery) {
4437 shrink_dcache_sb(sb);
4443 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4444 const char *dev_name, void *data)
4446 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4449 static void kill_f2fs_super(struct super_block *sb)
4452 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4454 set_sbi_flag(sbi, SBI_IS_CLOSE);
4455 f2fs_stop_gc_thread(sbi);
4456 f2fs_stop_discard_thread(sbi);
4458 #ifdef CONFIG_F2FS_FS_COMPRESSION
4460 * latter evict_inode() can bypass checking and invalidating
4461 * compress inode cache.
4463 if (test_opt(sbi, COMPRESS_CACHE))
4464 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4467 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4468 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4469 struct cp_control cpc = {
4470 .reason = CP_UMOUNT,
4472 f2fs_write_checkpoint(sbi, &cpc);
4475 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4476 sb->s_flags &= ~SB_RDONLY;
4478 kill_block_super(sb);
4481 static struct file_system_type f2fs_fs_type = {
4482 .owner = THIS_MODULE,
4484 .mount = f2fs_mount,
4485 .kill_sb = kill_f2fs_super,
4486 .fs_flags = FS_REQUIRES_DEV,
4488 MODULE_ALIAS_FS("f2fs");
4490 static int __init init_inodecache(void)
4492 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4493 sizeof(struct f2fs_inode_info), 0,
4494 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4495 if (!f2fs_inode_cachep)
4500 static void destroy_inodecache(void)
4503 * Make sure all delayed rcu free inodes are flushed before we
4507 kmem_cache_destroy(f2fs_inode_cachep);
4510 static int __init init_f2fs_fs(void)
4514 if (PAGE_SIZE != F2FS_BLKSIZE) {
4515 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4516 PAGE_SIZE, F2FS_BLKSIZE);
4520 err = init_inodecache();
4523 err = f2fs_create_node_manager_caches();
4525 goto free_inodecache;
4526 err = f2fs_create_segment_manager_caches();
4528 goto free_node_manager_caches;
4529 err = f2fs_create_checkpoint_caches();
4531 goto free_segment_manager_caches;
4532 err = f2fs_create_recovery_cache();
4534 goto free_checkpoint_caches;
4535 err = f2fs_create_extent_cache();
4537 goto free_recovery_cache;
4538 err = f2fs_create_garbage_collection_cache();
4540 goto free_extent_cache;
4541 err = f2fs_init_sysfs();
4543 goto free_garbage_collection_cache;
4544 err = register_shrinker(&f2fs_shrinker_info);
4547 err = register_filesystem(&f2fs_fs_type);
4550 f2fs_create_root_stats();
4551 err = f2fs_init_post_read_processing();
4553 goto free_root_stats;
4554 err = f2fs_init_iostat_processing();
4556 goto free_post_read;
4557 err = f2fs_init_bio_entry_cache();
4560 err = f2fs_init_bioset();
4562 goto free_bio_enrty_cache;
4563 err = f2fs_init_compress_mempool();
4566 err = f2fs_init_compress_cache();
4568 goto free_compress_mempool;
4569 err = f2fs_create_casefold_cache();
4571 goto free_compress_cache;
4573 free_compress_cache:
4574 f2fs_destroy_compress_cache();
4575 free_compress_mempool:
4576 f2fs_destroy_compress_mempool();
4578 f2fs_destroy_bioset();
4579 free_bio_enrty_cache:
4580 f2fs_destroy_bio_entry_cache();
4582 f2fs_destroy_iostat_processing();
4584 f2fs_destroy_post_read_processing();
4586 f2fs_destroy_root_stats();
4587 unregister_filesystem(&f2fs_fs_type);
4589 unregister_shrinker(&f2fs_shrinker_info);
4592 free_garbage_collection_cache:
4593 f2fs_destroy_garbage_collection_cache();
4595 f2fs_destroy_extent_cache();
4596 free_recovery_cache:
4597 f2fs_destroy_recovery_cache();
4598 free_checkpoint_caches:
4599 f2fs_destroy_checkpoint_caches();
4600 free_segment_manager_caches:
4601 f2fs_destroy_segment_manager_caches();
4602 free_node_manager_caches:
4603 f2fs_destroy_node_manager_caches();
4605 destroy_inodecache();
4610 static void __exit exit_f2fs_fs(void)
4612 f2fs_destroy_casefold_cache();
4613 f2fs_destroy_compress_cache();
4614 f2fs_destroy_compress_mempool();
4615 f2fs_destroy_bioset();
4616 f2fs_destroy_bio_entry_cache();
4617 f2fs_destroy_iostat_processing();
4618 f2fs_destroy_post_read_processing();
4619 f2fs_destroy_root_stats();
4620 unregister_filesystem(&f2fs_fs_type);
4621 unregister_shrinker(&f2fs_shrinker_info);
4623 f2fs_destroy_garbage_collection_cache();
4624 f2fs_destroy_extent_cache();
4625 f2fs_destroy_recovery_cache();
4626 f2fs_destroy_checkpoint_caches();
4627 f2fs_destroy_segment_manager_caches();
4628 f2fs_destroy_node_manager_caches();
4629 destroy_inodecache();
4632 module_init(init_f2fs_fs)
4633 module_exit(exit_f2fs_fs)
4635 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4636 MODULE_DESCRIPTION("Flash Friendly File System");
4637 MODULE_LICENSE("GPL");
4638 MODULE_SOFTDEP("pre: crc32");