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>
37 #define CREATE_TRACE_POINTS
38 #include <trace/events/f2fs.h>
40 static struct kmem_cache *f2fs_inode_cachep;
42 #ifdef CONFIG_F2FS_FAULT_INJECTION
44 const char *f2fs_fault_name[FAULT_MAX] = {
45 [FAULT_KMALLOC] = "kmalloc",
46 [FAULT_KVMALLOC] = "kvmalloc",
47 [FAULT_PAGE_ALLOC] = "page alloc",
48 [FAULT_PAGE_GET] = "page get",
49 [FAULT_ALLOC_NID] = "alloc nid",
50 [FAULT_ORPHAN] = "orphan",
51 [FAULT_BLOCK] = "no more block",
52 [FAULT_DIR_DEPTH] = "too big dir depth",
53 [FAULT_EVICT_INODE] = "evict_inode fail",
54 [FAULT_TRUNCATE] = "truncate fail",
55 [FAULT_READ_IO] = "read IO error",
56 [FAULT_CHECKPOINT] = "checkpoint error",
57 [FAULT_DISCARD] = "discard error",
58 [FAULT_WRITE_IO] = "write IO error",
59 [FAULT_SLAB_ALLOC] = "slab alloc",
62 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
65 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
68 atomic_set(&ffi->inject_ops, 0);
69 ffi->inject_rate = rate;
73 ffi->inject_type = type;
76 memset(ffi, 0, sizeof(struct f2fs_fault_info));
80 /* f2fs-wide shrinker description */
81 static struct shrinker f2fs_shrinker_info = {
82 .scan_objects = f2fs_shrink_scan,
83 .count_objects = f2fs_shrink_count,
84 .seeks = DEFAULT_SEEKS,
89 Opt_disable_roll_forward,
100 Opt_disable_ext_identify,
103 Opt_inline_xattr_size,
141 Opt_test_dummy_encryption,
143 Opt_checkpoint_disable,
144 Opt_checkpoint_disable_cap,
145 Opt_checkpoint_disable_cap_perc,
146 Opt_checkpoint_enable,
147 Opt_checkpoint_merge,
148 Opt_nocheckpoint_merge,
149 Opt_compress_algorithm,
150 Opt_compress_log_size,
151 Opt_compress_extension,
152 Opt_nocompress_extension,
163 static match_table_t f2fs_tokens = {
164 {Opt_gc_background, "background_gc=%s"},
165 {Opt_disable_roll_forward, "disable_roll_forward"},
166 {Opt_norecovery, "norecovery"},
167 {Opt_discard, "discard"},
168 {Opt_nodiscard, "nodiscard"},
169 {Opt_noheap, "no_heap"},
171 {Opt_user_xattr, "user_xattr"},
172 {Opt_nouser_xattr, "nouser_xattr"},
174 {Opt_noacl, "noacl"},
175 {Opt_active_logs, "active_logs=%u"},
176 {Opt_disable_ext_identify, "disable_ext_identify"},
177 {Opt_inline_xattr, "inline_xattr"},
178 {Opt_noinline_xattr, "noinline_xattr"},
179 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
180 {Opt_inline_data, "inline_data"},
181 {Opt_inline_dentry, "inline_dentry"},
182 {Opt_noinline_dentry, "noinline_dentry"},
183 {Opt_flush_merge, "flush_merge"},
184 {Opt_noflush_merge, "noflush_merge"},
185 {Opt_nobarrier, "nobarrier"},
186 {Opt_fastboot, "fastboot"},
187 {Opt_extent_cache, "extent_cache"},
188 {Opt_noextent_cache, "noextent_cache"},
189 {Opt_noinline_data, "noinline_data"},
190 {Opt_data_flush, "data_flush"},
191 {Opt_reserve_root, "reserve_root=%u"},
192 {Opt_resgid, "resgid=%u"},
193 {Opt_resuid, "resuid=%u"},
194 {Opt_mode, "mode=%s"},
195 {Opt_io_size_bits, "io_bits=%u"},
196 {Opt_fault_injection, "fault_injection=%u"},
197 {Opt_fault_type, "fault_type=%u"},
198 {Opt_lazytime, "lazytime"},
199 {Opt_nolazytime, "nolazytime"},
200 {Opt_quota, "quota"},
201 {Opt_noquota, "noquota"},
202 {Opt_usrquota, "usrquota"},
203 {Opt_grpquota, "grpquota"},
204 {Opt_prjquota, "prjquota"},
205 {Opt_usrjquota, "usrjquota=%s"},
206 {Opt_grpjquota, "grpjquota=%s"},
207 {Opt_prjjquota, "prjjquota=%s"},
208 {Opt_offusrjquota, "usrjquota="},
209 {Opt_offgrpjquota, "grpjquota="},
210 {Opt_offprjjquota, "prjjquota="},
211 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
212 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
213 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
214 {Opt_whint, "whint_mode=%s"},
215 {Opt_alloc, "alloc_mode=%s"},
216 {Opt_fsync, "fsync_mode=%s"},
217 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
218 {Opt_test_dummy_encryption, "test_dummy_encryption"},
219 {Opt_inlinecrypt, "inlinecrypt"},
220 {Opt_checkpoint_disable, "checkpoint=disable"},
221 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
222 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
223 {Opt_checkpoint_enable, "checkpoint=enable"},
224 {Opt_checkpoint_merge, "checkpoint_merge"},
225 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
226 {Opt_compress_algorithm, "compress_algorithm=%s"},
227 {Opt_compress_log_size, "compress_log_size=%u"},
228 {Opt_compress_extension, "compress_extension=%s"},
229 {Opt_nocompress_extension, "nocompress_extension=%s"},
230 {Opt_compress_chksum, "compress_chksum"},
231 {Opt_compress_mode, "compress_mode=%s"},
232 {Opt_compress_cache, "compress_cache"},
234 {Opt_gc_merge, "gc_merge"},
235 {Opt_nogc_merge, "nogc_merge"},
236 {Opt_discard_unit, "discard_unit=%s"},
240 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
242 struct va_format vaf;
248 level = printk_get_level(fmt);
249 vaf.fmt = printk_skip_level(fmt);
251 printk("%c%cF2FS-fs (%s): %pV\n",
252 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
257 #ifdef CONFIG_UNICODE
258 static const struct f2fs_sb_encodings {
262 } f2fs_sb_encoding_map[] = {
263 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
266 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
267 const struct f2fs_sb_encodings **encoding,
270 __u16 magic = le16_to_cpu(sb->s_encoding);
273 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
274 if (magic == f2fs_sb_encoding_map[i].magic)
277 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
280 *encoding = &f2fs_sb_encoding_map[i];
281 *flags = le16_to_cpu(sb->s_encoding_flags);
286 struct kmem_cache *f2fs_cf_name_slab;
287 static int __init f2fs_create_casefold_cache(void)
289 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
291 if (!f2fs_cf_name_slab)
296 static void f2fs_destroy_casefold_cache(void)
298 kmem_cache_destroy(f2fs_cf_name_slab);
301 static int __init f2fs_create_casefold_cache(void) { return 0; }
302 static void f2fs_destroy_casefold_cache(void) { }
305 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
307 block_t limit = min((sbi->user_block_count << 1) / 1000,
308 sbi->user_block_count - sbi->reserved_blocks);
311 if (test_opt(sbi, RESERVE_ROOT) &&
312 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
313 F2FS_OPTION(sbi).root_reserved_blocks = limit;
314 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
315 F2FS_OPTION(sbi).root_reserved_blocks);
317 if (!test_opt(sbi, RESERVE_ROOT) &&
318 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
319 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
320 !gid_eq(F2FS_OPTION(sbi).s_resgid,
321 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
322 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
323 from_kuid_munged(&init_user_ns,
324 F2FS_OPTION(sbi).s_resuid),
325 from_kgid_munged(&init_user_ns,
326 F2FS_OPTION(sbi).s_resgid));
329 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
331 if (!F2FS_OPTION(sbi).unusable_cap_perc)
334 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
335 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
337 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
338 F2FS_OPTION(sbi).unusable_cap_perc;
340 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
341 F2FS_OPTION(sbi).unusable_cap,
342 F2FS_OPTION(sbi).unusable_cap_perc);
345 static void init_once(void *foo)
347 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
349 inode_init_once(&fi->vfs_inode);
353 static const char * const quotatypes[] = INITQFNAMES;
354 #define QTYPE2NAME(t) (quotatypes[t])
355 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
358 struct f2fs_sb_info *sbi = F2FS_SB(sb);
362 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
363 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
366 if (f2fs_sb_has_quota_ino(sbi)) {
367 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
371 qname = match_strdup(args);
373 f2fs_err(sbi, "Not enough memory for storing quotafile name");
376 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
377 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
380 f2fs_err(sbi, "%s quota file already specified",
384 if (strchr(qname, '/')) {
385 f2fs_err(sbi, "quotafile must be on filesystem root");
388 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
396 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
398 struct f2fs_sb_info *sbi = F2FS_SB(sb);
400 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
401 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
404 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
405 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
409 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
412 * We do the test below only for project quotas. 'usrquota' and
413 * 'grpquota' mount options are allowed even without quota feature
414 * to support legacy quotas in quota files.
416 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
417 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
420 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
421 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
422 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
423 if (test_opt(sbi, USRQUOTA) &&
424 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
425 clear_opt(sbi, USRQUOTA);
427 if (test_opt(sbi, GRPQUOTA) &&
428 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
429 clear_opt(sbi, GRPQUOTA);
431 if (test_opt(sbi, PRJQUOTA) &&
432 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
433 clear_opt(sbi, PRJQUOTA);
435 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
436 test_opt(sbi, PRJQUOTA)) {
437 f2fs_err(sbi, "old and new quota format mixing");
441 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
442 f2fs_err(sbi, "journaled quota format not specified");
447 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
448 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
449 F2FS_OPTION(sbi).s_jquota_fmt = 0;
455 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
457 const substring_t *arg,
460 struct f2fs_sb_info *sbi = F2FS_SB(sb);
461 #ifdef CONFIG_FS_ENCRYPTION
464 if (!f2fs_sb_has_encrypt(sbi)) {
465 f2fs_err(sbi, "Encrypt feature is off");
470 * This mount option is just for testing, and it's not worthwhile to
471 * implement the extra complexity (e.g. RCU protection) that would be
472 * needed to allow it to be set or changed during remount. We do allow
473 * it to be specified during remount, but only if there is no change.
475 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
476 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
479 err = fscrypt_set_test_dummy_encryption(
480 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
484 "Can't change test_dummy_encryption on remount");
485 else if (err == -EINVAL)
486 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
489 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
493 f2fs_warn(sbi, "Test dummy encryption mode enabled");
495 f2fs_warn(sbi, "Test dummy encryption mount option ignored");
500 #ifdef CONFIG_F2FS_FS_COMPRESSION
502 * 1. The same extension name cannot not appear in both compress and non-compress extension
504 * 2. If the compress extension specifies all files, the types specified by the non-compress
505 * extension will be treated as special cases and will not be compressed.
506 * 3. Don't allow the non-compress extension specifies all files.
508 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
510 unsigned char (*ext)[F2FS_EXTENSION_LEN];
511 unsigned char (*noext)[F2FS_EXTENSION_LEN];
512 int ext_cnt, noext_cnt, index = 0, no_index = 0;
514 ext = F2FS_OPTION(sbi).extensions;
515 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
516 noext = F2FS_OPTION(sbi).noextensions;
517 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
522 for (no_index = 0; no_index < noext_cnt; no_index++) {
523 if (!strcasecmp("*", noext[no_index])) {
524 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
527 for (index = 0; index < ext_cnt; index++) {
528 if (!strcasecmp(ext[index], noext[no_index])) {
529 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
538 #ifdef CONFIG_F2FS_FS_LZ4
539 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
541 #ifdef CONFIG_F2FS_FS_LZ4HC
545 if (strlen(str) == 3) {
546 F2FS_OPTION(sbi).compress_level = 0;
550 #ifdef CONFIG_F2FS_FS_LZ4HC
554 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
557 if (kstrtouint(str + 1, 10, &level))
560 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
561 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
565 F2FS_OPTION(sbi).compress_level = level;
568 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
574 #ifdef CONFIG_F2FS_FS_ZSTD
575 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
580 if (strlen(str) == len) {
581 F2FS_OPTION(sbi).compress_level = 0;
588 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
591 if (kstrtouint(str + 1, 10, &level))
594 if (!level || level > ZSTD_maxCLevel()) {
595 f2fs_info(sbi, "invalid zstd compress level: %d", level);
599 F2FS_OPTION(sbi).compress_level = level;
605 static int parse_options(struct super_block *sb, char *options, bool is_remount)
607 struct f2fs_sb_info *sbi = F2FS_SB(sb);
608 substring_t args[MAX_OPT_ARGS];
609 #ifdef CONFIG_F2FS_FS_COMPRESSION
610 unsigned char (*ext)[F2FS_EXTENSION_LEN];
611 unsigned char (*noext)[F2FS_EXTENSION_LEN];
612 int ext_cnt, noext_cnt;
623 while ((p = strsep(&options, ",")) != NULL) {
629 * Initialize args struct so we know whether arg was
630 * found; some options take optional arguments.
632 args[0].to = args[0].from = NULL;
633 token = match_token(p, f2fs_tokens, args);
636 case Opt_gc_background:
637 name = match_strdup(&args[0]);
641 if (!strcmp(name, "on")) {
642 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
643 } else if (!strcmp(name, "off")) {
644 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
645 } else if (!strcmp(name, "sync")) {
646 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
653 case Opt_disable_roll_forward:
654 set_opt(sbi, DISABLE_ROLL_FORWARD);
657 /* this option mounts f2fs with ro */
658 set_opt(sbi, NORECOVERY);
659 if (!f2fs_readonly(sb))
663 set_opt(sbi, DISCARD);
666 if (f2fs_sb_has_blkzoned(sbi)) {
667 f2fs_warn(sbi, "discard is required for zoned block devices");
670 clear_opt(sbi, DISCARD);
673 set_opt(sbi, NOHEAP);
676 clear_opt(sbi, NOHEAP);
678 #ifdef CONFIG_F2FS_FS_XATTR
680 set_opt(sbi, XATTR_USER);
682 case Opt_nouser_xattr:
683 clear_opt(sbi, XATTR_USER);
685 case Opt_inline_xattr:
686 set_opt(sbi, INLINE_XATTR);
688 case Opt_noinline_xattr:
689 clear_opt(sbi, INLINE_XATTR);
691 case Opt_inline_xattr_size:
692 if (args->from && match_int(args, &arg))
694 set_opt(sbi, INLINE_XATTR_SIZE);
695 F2FS_OPTION(sbi).inline_xattr_size = arg;
699 f2fs_info(sbi, "user_xattr options not supported");
701 case Opt_nouser_xattr:
702 f2fs_info(sbi, "nouser_xattr options not supported");
704 case Opt_inline_xattr:
705 f2fs_info(sbi, "inline_xattr options not supported");
707 case Opt_noinline_xattr:
708 f2fs_info(sbi, "noinline_xattr options not supported");
711 #ifdef CONFIG_F2FS_FS_POSIX_ACL
713 set_opt(sbi, POSIX_ACL);
716 clear_opt(sbi, POSIX_ACL);
720 f2fs_info(sbi, "acl options not supported");
723 f2fs_info(sbi, "noacl options not supported");
726 case Opt_active_logs:
727 if (args->from && match_int(args, &arg))
729 if (arg != 2 && arg != 4 &&
730 arg != NR_CURSEG_PERSIST_TYPE)
732 F2FS_OPTION(sbi).active_logs = arg;
734 case Opt_disable_ext_identify:
735 set_opt(sbi, DISABLE_EXT_IDENTIFY);
737 case Opt_inline_data:
738 set_opt(sbi, INLINE_DATA);
740 case Opt_inline_dentry:
741 set_opt(sbi, INLINE_DENTRY);
743 case Opt_noinline_dentry:
744 clear_opt(sbi, INLINE_DENTRY);
746 case Opt_flush_merge:
747 set_opt(sbi, FLUSH_MERGE);
749 case Opt_noflush_merge:
750 clear_opt(sbi, FLUSH_MERGE);
753 set_opt(sbi, NOBARRIER);
756 set_opt(sbi, FASTBOOT);
758 case Opt_extent_cache:
759 set_opt(sbi, EXTENT_CACHE);
761 case Opt_noextent_cache:
762 clear_opt(sbi, EXTENT_CACHE);
764 case Opt_noinline_data:
765 clear_opt(sbi, INLINE_DATA);
768 set_opt(sbi, DATA_FLUSH);
770 case Opt_reserve_root:
771 if (args->from && match_int(args, &arg))
773 if (test_opt(sbi, RESERVE_ROOT)) {
774 f2fs_info(sbi, "Preserve previous reserve_root=%u",
775 F2FS_OPTION(sbi).root_reserved_blocks);
777 F2FS_OPTION(sbi).root_reserved_blocks = arg;
778 set_opt(sbi, RESERVE_ROOT);
782 if (args->from && match_int(args, &arg))
784 uid = make_kuid(current_user_ns(), arg);
785 if (!uid_valid(uid)) {
786 f2fs_err(sbi, "Invalid uid value %d", arg);
789 F2FS_OPTION(sbi).s_resuid = uid;
792 if (args->from && match_int(args, &arg))
794 gid = make_kgid(current_user_ns(), arg);
795 if (!gid_valid(gid)) {
796 f2fs_err(sbi, "Invalid gid value %d", arg);
799 F2FS_OPTION(sbi).s_resgid = gid;
802 name = match_strdup(&args[0]);
806 if (!strcmp(name, "adaptive")) {
807 if (f2fs_sb_has_blkzoned(sbi)) {
808 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
812 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
813 } else if (!strcmp(name, "lfs")) {
814 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
821 case Opt_io_size_bits:
822 if (args->from && match_int(args, &arg))
824 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
825 f2fs_warn(sbi, "Not support %d, larger than %d",
826 1 << arg, BIO_MAX_VECS);
829 F2FS_OPTION(sbi).write_io_size_bits = arg;
831 #ifdef CONFIG_F2FS_FAULT_INJECTION
832 case Opt_fault_injection:
833 if (args->from && match_int(args, &arg))
835 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
836 set_opt(sbi, FAULT_INJECTION);
840 if (args->from && match_int(args, &arg))
842 f2fs_build_fault_attr(sbi, 0, arg);
843 set_opt(sbi, FAULT_INJECTION);
846 case Opt_fault_injection:
847 f2fs_info(sbi, "fault_injection options not supported");
851 f2fs_info(sbi, "fault_type options not supported");
855 sb->s_flags |= SB_LAZYTIME;
858 sb->s_flags &= ~SB_LAZYTIME;
863 set_opt(sbi, USRQUOTA);
866 set_opt(sbi, GRPQUOTA);
869 set_opt(sbi, PRJQUOTA);
872 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
877 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
882 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
886 case Opt_offusrjquota:
887 ret = f2fs_clear_qf_name(sb, USRQUOTA);
891 case Opt_offgrpjquota:
892 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
896 case Opt_offprjjquota:
897 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
901 case Opt_jqfmt_vfsold:
902 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
904 case Opt_jqfmt_vfsv0:
905 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
907 case Opt_jqfmt_vfsv1:
908 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
911 clear_opt(sbi, QUOTA);
912 clear_opt(sbi, USRQUOTA);
913 clear_opt(sbi, GRPQUOTA);
914 clear_opt(sbi, PRJQUOTA);
924 case Opt_offusrjquota:
925 case Opt_offgrpjquota:
926 case Opt_offprjjquota:
927 case Opt_jqfmt_vfsold:
928 case Opt_jqfmt_vfsv0:
929 case Opt_jqfmt_vfsv1:
931 f2fs_info(sbi, "quota operations not supported");
935 name = match_strdup(&args[0]);
938 if (!strcmp(name, "user-based")) {
939 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
940 } else if (!strcmp(name, "off")) {
941 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
942 } else if (!strcmp(name, "fs-based")) {
943 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
951 name = match_strdup(&args[0]);
955 if (!strcmp(name, "default")) {
956 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
957 } else if (!strcmp(name, "reuse")) {
958 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
966 name = match_strdup(&args[0]);
969 if (!strcmp(name, "posix")) {
970 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
971 } else if (!strcmp(name, "strict")) {
972 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
973 } else if (!strcmp(name, "nobarrier")) {
974 F2FS_OPTION(sbi).fsync_mode =
975 FSYNC_MODE_NOBARRIER;
982 case Opt_test_dummy_encryption:
983 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
988 case Opt_inlinecrypt:
989 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
990 sb->s_flags |= SB_INLINECRYPT;
992 f2fs_info(sbi, "inline encryption not supported");
995 case Opt_checkpoint_disable_cap_perc:
996 if (args->from && match_int(args, &arg))
998 if (arg < 0 || arg > 100)
1000 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1001 set_opt(sbi, DISABLE_CHECKPOINT);
1003 case Opt_checkpoint_disable_cap:
1004 if (args->from && match_int(args, &arg))
1006 F2FS_OPTION(sbi).unusable_cap = arg;
1007 set_opt(sbi, DISABLE_CHECKPOINT);
1009 case Opt_checkpoint_disable:
1010 set_opt(sbi, DISABLE_CHECKPOINT);
1012 case Opt_checkpoint_enable:
1013 clear_opt(sbi, DISABLE_CHECKPOINT);
1015 case Opt_checkpoint_merge:
1016 set_opt(sbi, MERGE_CHECKPOINT);
1018 case Opt_nocheckpoint_merge:
1019 clear_opt(sbi, MERGE_CHECKPOINT);
1021 #ifdef CONFIG_F2FS_FS_COMPRESSION
1022 case Opt_compress_algorithm:
1023 if (!f2fs_sb_has_compression(sbi)) {
1024 f2fs_info(sbi, "Image doesn't support compression");
1027 name = match_strdup(&args[0]);
1030 if (!strcmp(name, "lzo")) {
1031 #ifdef CONFIG_F2FS_FS_LZO
1032 F2FS_OPTION(sbi).compress_level = 0;
1033 F2FS_OPTION(sbi).compress_algorithm =
1036 f2fs_info(sbi, "kernel doesn't support lzo compression");
1038 } else if (!strncmp(name, "lz4", 3)) {
1039 #ifdef CONFIG_F2FS_FS_LZ4
1040 ret = f2fs_set_lz4hc_level(sbi, name);
1045 F2FS_OPTION(sbi).compress_algorithm =
1048 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1050 } else if (!strncmp(name, "zstd", 4)) {
1051 #ifdef CONFIG_F2FS_FS_ZSTD
1052 ret = f2fs_set_zstd_level(sbi, name);
1057 F2FS_OPTION(sbi).compress_algorithm =
1060 f2fs_info(sbi, "kernel doesn't support zstd compression");
1062 } else if (!strcmp(name, "lzo-rle")) {
1063 #ifdef CONFIG_F2FS_FS_LZORLE
1064 F2FS_OPTION(sbi).compress_level = 0;
1065 F2FS_OPTION(sbi).compress_algorithm =
1068 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1076 case Opt_compress_log_size:
1077 if (!f2fs_sb_has_compression(sbi)) {
1078 f2fs_info(sbi, "Image doesn't support compression");
1081 if (args->from && match_int(args, &arg))
1083 if (arg < MIN_COMPRESS_LOG_SIZE ||
1084 arg > MAX_COMPRESS_LOG_SIZE) {
1086 "Compress cluster log size is out of range");
1089 F2FS_OPTION(sbi).compress_log_size = arg;
1091 case Opt_compress_extension:
1092 if (!f2fs_sb_has_compression(sbi)) {
1093 f2fs_info(sbi, "Image doesn't support compression");
1096 name = match_strdup(&args[0]);
1100 ext = F2FS_OPTION(sbi).extensions;
1101 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1103 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1104 ext_cnt >= COMPRESS_EXT_NUM) {
1106 "invalid extension length/number");
1111 strcpy(ext[ext_cnt], name);
1112 F2FS_OPTION(sbi).compress_ext_cnt++;
1115 case Opt_nocompress_extension:
1116 if (!f2fs_sb_has_compression(sbi)) {
1117 f2fs_info(sbi, "Image doesn't support compression");
1120 name = match_strdup(&args[0]);
1124 noext = F2FS_OPTION(sbi).noextensions;
1125 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1127 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1128 noext_cnt >= COMPRESS_EXT_NUM) {
1130 "invalid extension length/number");
1135 strcpy(noext[noext_cnt], name);
1136 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1139 case Opt_compress_chksum:
1140 F2FS_OPTION(sbi).compress_chksum = true;
1142 case Opt_compress_mode:
1143 name = match_strdup(&args[0]);
1146 if (!strcmp(name, "fs")) {
1147 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1148 } else if (!strcmp(name, "user")) {
1149 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1156 case Opt_compress_cache:
1157 set_opt(sbi, COMPRESS_CACHE);
1160 case Opt_compress_algorithm:
1161 case Opt_compress_log_size:
1162 case Opt_compress_extension:
1163 case Opt_nocompress_extension:
1164 case Opt_compress_chksum:
1165 case Opt_compress_mode:
1166 case Opt_compress_cache:
1167 f2fs_info(sbi, "compression options not supported");
1174 set_opt(sbi, GC_MERGE);
1176 case Opt_nogc_merge:
1177 clear_opt(sbi, GC_MERGE);
1179 case Opt_discard_unit:
1180 name = match_strdup(&args[0]);
1183 if (!strcmp(name, "block")) {
1184 F2FS_OPTION(sbi).discard_unit =
1186 } else if (!strcmp(name, "segment")) {
1187 F2FS_OPTION(sbi).discard_unit =
1188 DISCARD_UNIT_SEGMENT;
1189 } else if (!strcmp(name, "section")) {
1190 F2FS_OPTION(sbi).discard_unit =
1191 DISCARD_UNIT_SECTION;
1199 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1206 if (f2fs_check_quota_options(sbi))
1209 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1210 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1213 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1214 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1218 #ifndef CONFIG_UNICODE
1219 if (f2fs_sb_has_casefold(sbi)) {
1221 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1226 * The BLKZONED feature indicates that the drive was formatted with
1227 * zone alignment optimization. This is optional for host-aware
1228 * devices, but mandatory for host-managed zoned block devices.
1230 #ifndef CONFIG_BLK_DEV_ZONED
1231 if (f2fs_sb_has_blkzoned(sbi)) {
1232 f2fs_err(sbi, "Zoned block device support is not enabled");
1236 if (f2fs_sb_has_blkzoned(sbi)) {
1237 if (F2FS_OPTION(sbi).discard_unit !=
1238 DISCARD_UNIT_SECTION) {
1239 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1240 F2FS_OPTION(sbi).discard_unit =
1241 DISCARD_UNIT_SECTION;
1245 #ifdef CONFIG_F2FS_FS_COMPRESSION
1246 if (f2fs_test_compress_extension(sbi)) {
1247 f2fs_err(sbi, "invalid compress or nocompress extension");
1252 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1253 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1254 F2FS_IO_SIZE_KB(sbi));
1258 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1259 int min_size, max_size;
1261 if (!f2fs_sb_has_extra_attr(sbi) ||
1262 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1263 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1266 if (!test_opt(sbi, INLINE_XATTR)) {
1267 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1271 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1272 max_size = MAX_INLINE_XATTR_SIZE;
1274 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1275 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1276 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1277 min_size, max_size);
1282 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1283 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1287 /* Not pass down write hints if the number of active logs is lesser
1288 * than NR_CURSEG_PERSIST_TYPE.
1290 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
1291 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1293 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1294 f2fs_err(sbi, "Allow to mount readonly mode only");
1300 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1302 struct f2fs_inode_info *fi;
1304 fi = f2fs_kmem_cache_alloc(f2fs_inode_cachep,
1305 GFP_F2FS_ZERO, false, F2FS_SB(sb));
1309 init_once((void *) fi);
1311 /* Initialize f2fs-specific inode info */
1312 atomic_set(&fi->dirty_pages, 0);
1313 atomic_set(&fi->i_compr_blocks, 0);
1314 init_rwsem(&fi->i_sem);
1315 spin_lock_init(&fi->i_size_lock);
1316 INIT_LIST_HEAD(&fi->dirty_list);
1317 INIT_LIST_HEAD(&fi->gdirty_list);
1318 INIT_LIST_HEAD(&fi->inmem_ilist);
1319 INIT_LIST_HEAD(&fi->inmem_pages);
1320 mutex_init(&fi->inmem_lock);
1321 init_rwsem(&fi->i_gc_rwsem[READ]);
1322 init_rwsem(&fi->i_gc_rwsem[WRITE]);
1323 init_rwsem(&fi->i_mmap_sem);
1324 init_rwsem(&fi->i_xattr_sem);
1326 /* Will be used by directory only */
1327 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1329 return &fi->vfs_inode;
1332 static int f2fs_drop_inode(struct inode *inode)
1334 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1338 * during filesystem shutdown, if checkpoint is disabled,
1339 * drop useless meta/node dirty pages.
1341 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1342 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1343 inode->i_ino == F2FS_META_INO(sbi)) {
1344 trace_f2fs_drop_inode(inode, 1);
1350 * This is to avoid a deadlock condition like below.
1351 * writeback_single_inode(inode)
1352 * - f2fs_write_data_page
1353 * - f2fs_gc -> iput -> evict
1354 * - inode_wait_for_writeback(inode)
1356 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1357 if (!inode->i_nlink && !is_bad_inode(inode)) {
1358 /* to avoid evict_inode call simultaneously */
1359 atomic_inc(&inode->i_count);
1360 spin_unlock(&inode->i_lock);
1362 /* some remained atomic pages should discarded */
1363 if (f2fs_is_atomic_file(inode))
1364 f2fs_drop_inmem_pages(inode);
1366 /* should remain fi->extent_tree for writepage */
1367 f2fs_destroy_extent_node(inode);
1369 sb_start_intwrite(inode->i_sb);
1370 f2fs_i_size_write(inode, 0);
1372 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1373 inode, NULL, 0, DATA);
1374 truncate_inode_pages_final(inode->i_mapping);
1376 if (F2FS_HAS_BLOCKS(inode))
1377 f2fs_truncate(inode);
1379 sb_end_intwrite(inode->i_sb);
1381 spin_lock(&inode->i_lock);
1382 atomic_dec(&inode->i_count);
1384 trace_f2fs_drop_inode(inode, 0);
1387 ret = generic_drop_inode(inode);
1389 ret = fscrypt_drop_inode(inode);
1390 trace_f2fs_drop_inode(inode, ret);
1394 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1396 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1399 spin_lock(&sbi->inode_lock[DIRTY_META]);
1400 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1403 set_inode_flag(inode, FI_DIRTY_INODE);
1404 stat_inc_dirty_inode(sbi, DIRTY_META);
1406 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1407 list_add_tail(&F2FS_I(inode)->gdirty_list,
1408 &sbi->inode_list[DIRTY_META]);
1409 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1411 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1415 void f2fs_inode_synced(struct inode *inode)
1417 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1419 spin_lock(&sbi->inode_lock[DIRTY_META]);
1420 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1421 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1424 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1425 list_del_init(&F2FS_I(inode)->gdirty_list);
1426 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1428 clear_inode_flag(inode, FI_DIRTY_INODE);
1429 clear_inode_flag(inode, FI_AUTO_RECOVER);
1430 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1431 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1435 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1437 * We should call set_dirty_inode to write the dirty inode through write_inode.
1439 static void f2fs_dirty_inode(struct inode *inode, int flags)
1441 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1443 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1444 inode->i_ino == F2FS_META_INO(sbi))
1447 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1448 clear_inode_flag(inode, FI_AUTO_RECOVER);
1450 f2fs_inode_dirtied(inode, false);
1453 static void f2fs_free_inode(struct inode *inode)
1455 fscrypt_free_inode(inode);
1456 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1459 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1461 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1462 percpu_counter_destroy(&sbi->total_valid_inode_count);
1465 static void destroy_device_list(struct f2fs_sb_info *sbi)
1469 for (i = 0; i < sbi->s_ndevs; i++) {
1470 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1471 #ifdef CONFIG_BLK_DEV_ZONED
1472 kvfree(FDEV(i).blkz_seq);
1473 kfree(FDEV(i).zone_capacity_blocks);
1479 static void f2fs_put_super(struct super_block *sb)
1481 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1485 /* unregister procfs/sysfs entries in advance to avoid race case */
1486 f2fs_unregister_sysfs(sbi);
1488 f2fs_quota_off_umount(sb);
1490 /* prevent remaining shrinker jobs */
1491 mutex_lock(&sbi->umount_mutex);
1494 * flush all issued checkpoints and stop checkpoint issue thread.
1495 * after then, all checkpoints should be done by each process context.
1497 f2fs_stop_ckpt_thread(sbi);
1500 * We don't need to do checkpoint when superblock is clean.
1501 * But, the previous checkpoint was not done by umount, it needs to do
1502 * clean checkpoint again.
1504 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1505 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1506 struct cp_control cpc = {
1507 .reason = CP_UMOUNT,
1509 f2fs_write_checkpoint(sbi, &cpc);
1512 /* be sure to wait for any on-going discard commands */
1513 dropped = f2fs_issue_discard_timeout(sbi);
1515 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1516 !sbi->discard_blks && !dropped) {
1517 struct cp_control cpc = {
1518 .reason = CP_UMOUNT | CP_TRIMMED,
1520 f2fs_write_checkpoint(sbi, &cpc);
1524 * normally superblock is clean, so we need to release this.
1525 * In addition, EIO will skip do checkpoint, we need this as well.
1527 f2fs_release_ino_entry(sbi, true);
1529 f2fs_leave_shrinker(sbi);
1530 mutex_unlock(&sbi->umount_mutex);
1532 /* our cp_error case, we can wait for any writeback page */
1533 f2fs_flush_merged_writes(sbi);
1535 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1537 f2fs_bug_on(sbi, sbi->fsync_node_num);
1539 f2fs_destroy_compress_inode(sbi);
1541 iput(sbi->node_inode);
1542 sbi->node_inode = NULL;
1544 iput(sbi->meta_inode);
1545 sbi->meta_inode = NULL;
1548 * iput() can update stat information, if f2fs_write_checkpoint()
1549 * above failed with error.
1551 f2fs_destroy_stats(sbi);
1553 /* destroy f2fs internal modules */
1554 f2fs_destroy_node_manager(sbi);
1555 f2fs_destroy_segment_manager(sbi);
1557 f2fs_destroy_post_read_wq(sbi);
1561 sb->s_fs_info = NULL;
1562 if (sbi->s_chksum_driver)
1563 crypto_free_shash(sbi->s_chksum_driver);
1564 kfree(sbi->raw_super);
1566 destroy_device_list(sbi);
1567 f2fs_destroy_page_array_cache(sbi);
1568 f2fs_destroy_xattr_caches(sbi);
1569 mempool_destroy(sbi->write_io_dummy);
1571 for (i = 0; i < MAXQUOTAS; i++)
1572 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1574 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1575 destroy_percpu_info(sbi);
1576 for (i = 0; i < NR_PAGE_TYPE; i++)
1577 kvfree(sbi->write_io[i]);
1578 #ifdef CONFIG_UNICODE
1579 utf8_unload(sb->s_encoding);
1584 int f2fs_sync_fs(struct super_block *sb, int sync)
1586 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1589 if (unlikely(f2fs_cp_error(sbi)))
1591 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1594 trace_f2fs_sync_fs(sb, sync);
1596 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1600 err = f2fs_issue_checkpoint(sbi);
1605 static int f2fs_freeze(struct super_block *sb)
1607 if (f2fs_readonly(sb))
1610 /* IO error happened before */
1611 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1614 /* must be clean, since sync_filesystem() was already called */
1615 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1618 /* ensure no checkpoint required */
1619 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list))
1624 static int f2fs_unfreeze(struct super_block *sb)
1630 static int f2fs_statfs_project(struct super_block *sb,
1631 kprojid_t projid, struct kstatfs *buf)
1634 struct dquot *dquot;
1638 qid = make_kqid_projid(projid);
1639 dquot = dqget(sb, qid);
1641 return PTR_ERR(dquot);
1642 spin_lock(&dquot->dq_dqb_lock);
1644 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1645 dquot->dq_dqb.dqb_bhardlimit);
1647 limit >>= sb->s_blocksize_bits;
1649 if (limit && buf->f_blocks > limit) {
1650 curblock = (dquot->dq_dqb.dqb_curspace +
1651 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1652 buf->f_blocks = limit;
1653 buf->f_bfree = buf->f_bavail =
1654 (buf->f_blocks > curblock) ?
1655 (buf->f_blocks - curblock) : 0;
1658 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1659 dquot->dq_dqb.dqb_ihardlimit);
1661 if (limit && buf->f_files > limit) {
1662 buf->f_files = limit;
1664 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1665 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1668 spin_unlock(&dquot->dq_dqb_lock);
1674 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1676 struct super_block *sb = dentry->d_sb;
1677 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1678 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1679 block_t total_count, user_block_count, start_count;
1680 u64 avail_node_count;
1682 total_count = le64_to_cpu(sbi->raw_super->block_count);
1683 user_block_count = sbi->user_block_count;
1684 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1685 buf->f_type = F2FS_SUPER_MAGIC;
1686 buf->f_bsize = sbi->blocksize;
1688 buf->f_blocks = total_count - start_count;
1689 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1690 sbi->current_reserved_blocks;
1692 spin_lock(&sbi->stat_lock);
1693 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1696 buf->f_bfree -= sbi->unusable_block_count;
1697 spin_unlock(&sbi->stat_lock);
1699 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1700 buf->f_bavail = buf->f_bfree -
1701 F2FS_OPTION(sbi).root_reserved_blocks;
1705 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1707 if (avail_node_count > user_block_count) {
1708 buf->f_files = user_block_count;
1709 buf->f_ffree = buf->f_bavail;
1711 buf->f_files = avail_node_count;
1712 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1716 buf->f_namelen = F2FS_NAME_LEN;
1717 buf->f_fsid = u64_to_fsid(id);
1720 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1721 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1722 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1728 static inline void f2fs_show_quota_options(struct seq_file *seq,
1729 struct super_block *sb)
1732 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1734 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1737 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1748 seq_printf(seq, ",jqfmt=%s", fmtname);
1751 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1752 seq_show_option(seq, "usrjquota",
1753 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1755 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1756 seq_show_option(seq, "grpjquota",
1757 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1759 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1760 seq_show_option(seq, "prjjquota",
1761 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1765 #ifdef CONFIG_F2FS_FS_COMPRESSION
1766 static inline void f2fs_show_compress_options(struct seq_file *seq,
1767 struct super_block *sb)
1769 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1773 if (!f2fs_sb_has_compression(sbi))
1776 switch (F2FS_OPTION(sbi).compress_algorithm) {
1786 case COMPRESS_LZORLE:
1787 algtype = "lzo-rle";
1790 seq_printf(seq, ",compress_algorithm=%s", algtype);
1792 if (F2FS_OPTION(sbi).compress_level)
1793 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1795 seq_printf(seq, ",compress_log_size=%u",
1796 F2FS_OPTION(sbi).compress_log_size);
1798 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1799 seq_printf(seq, ",compress_extension=%s",
1800 F2FS_OPTION(sbi).extensions[i]);
1803 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1804 seq_printf(seq, ",nocompress_extension=%s",
1805 F2FS_OPTION(sbi).noextensions[i]);
1808 if (F2FS_OPTION(sbi).compress_chksum)
1809 seq_puts(seq, ",compress_chksum");
1811 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1812 seq_printf(seq, ",compress_mode=%s", "fs");
1813 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1814 seq_printf(seq, ",compress_mode=%s", "user");
1816 if (test_opt(sbi, COMPRESS_CACHE))
1817 seq_puts(seq, ",compress_cache");
1821 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1823 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1825 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1826 seq_printf(seq, ",background_gc=%s", "sync");
1827 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1828 seq_printf(seq, ",background_gc=%s", "on");
1829 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1830 seq_printf(seq, ",background_gc=%s", "off");
1832 if (test_opt(sbi, GC_MERGE))
1833 seq_puts(seq, ",gc_merge");
1835 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1836 seq_puts(seq, ",disable_roll_forward");
1837 if (test_opt(sbi, NORECOVERY))
1838 seq_puts(seq, ",norecovery");
1839 if (test_opt(sbi, DISCARD))
1840 seq_puts(seq, ",discard");
1842 seq_puts(seq, ",nodiscard");
1843 if (test_opt(sbi, NOHEAP))
1844 seq_puts(seq, ",no_heap");
1846 seq_puts(seq, ",heap");
1847 #ifdef CONFIG_F2FS_FS_XATTR
1848 if (test_opt(sbi, XATTR_USER))
1849 seq_puts(seq, ",user_xattr");
1851 seq_puts(seq, ",nouser_xattr");
1852 if (test_opt(sbi, INLINE_XATTR))
1853 seq_puts(seq, ",inline_xattr");
1855 seq_puts(seq, ",noinline_xattr");
1856 if (test_opt(sbi, INLINE_XATTR_SIZE))
1857 seq_printf(seq, ",inline_xattr_size=%u",
1858 F2FS_OPTION(sbi).inline_xattr_size);
1860 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1861 if (test_opt(sbi, POSIX_ACL))
1862 seq_puts(seq, ",acl");
1864 seq_puts(seq, ",noacl");
1866 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1867 seq_puts(seq, ",disable_ext_identify");
1868 if (test_opt(sbi, INLINE_DATA))
1869 seq_puts(seq, ",inline_data");
1871 seq_puts(seq, ",noinline_data");
1872 if (test_opt(sbi, INLINE_DENTRY))
1873 seq_puts(seq, ",inline_dentry");
1875 seq_puts(seq, ",noinline_dentry");
1876 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1877 seq_puts(seq, ",flush_merge");
1878 if (test_opt(sbi, NOBARRIER))
1879 seq_puts(seq, ",nobarrier");
1880 if (test_opt(sbi, FASTBOOT))
1881 seq_puts(seq, ",fastboot");
1882 if (test_opt(sbi, EXTENT_CACHE))
1883 seq_puts(seq, ",extent_cache");
1885 seq_puts(seq, ",noextent_cache");
1886 if (test_opt(sbi, DATA_FLUSH))
1887 seq_puts(seq, ",data_flush");
1889 seq_puts(seq, ",mode=");
1890 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1891 seq_puts(seq, "adaptive");
1892 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1893 seq_puts(seq, "lfs");
1894 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1895 if (test_opt(sbi, RESERVE_ROOT))
1896 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1897 F2FS_OPTION(sbi).root_reserved_blocks,
1898 from_kuid_munged(&init_user_ns,
1899 F2FS_OPTION(sbi).s_resuid),
1900 from_kgid_munged(&init_user_ns,
1901 F2FS_OPTION(sbi).s_resgid));
1902 if (F2FS_IO_SIZE_BITS(sbi))
1903 seq_printf(seq, ",io_bits=%u",
1904 F2FS_OPTION(sbi).write_io_size_bits);
1905 #ifdef CONFIG_F2FS_FAULT_INJECTION
1906 if (test_opt(sbi, FAULT_INJECTION)) {
1907 seq_printf(seq, ",fault_injection=%u",
1908 F2FS_OPTION(sbi).fault_info.inject_rate);
1909 seq_printf(seq, ",fault_type=%u",
1910 F2FS_OPTION(sbi).fault_info.inject_type);
1914 if (test_opt(sbi, QUOTA))
1915 seq_puts(seq, ",quota");
1916 if (test_opt(sbi, USRQUOTA))
1917 seq_puts(seq, ",usrquota");
1918 if (test_opt(sbi, GRPQUOTA))
1919 seq_puts(seq, ",grpquota");
1920 if (test_opt(sbi, PRJQUOTA))
1921 seq_puts(seq, ",prjquota");
1923 f2fs_show_quota_options(seq, sbi->sb);
1924 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1925 seq_printf(seq, ",whint_mode=%s", "user-based");
1926 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1927 seq_printf(seq, ",whint_mode=%s", "fs-based");
1929 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1931 if (sbi->sb->s_flags & SB_INLINECRYPT)
1932 seq_puts(seq, ",inlinecrypt");
1934 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1935 seq_printf(seq, ",alloc_mode=%s", "default");
1936 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1937 seq_printf(seq, ",alloc_mode=%s", "reuse");
1939 if (test_opt(sbi, DISABLE_CHECKPOINT))
1940 seq_printf(seq, ",checkpoint=disable:%u",
1941 F2FS_OPTION(sbi).unusable_cap);
1942 if (test_opt(sbi, MERGE_CHECKPOINT))
1943 seq_puts(seq, ",checkpoint_merge");
1945 seq_puts(seq, ",nocheckpoint_merge");
1946 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1947 seq_printf(seq, ",fsync_mode=%s", "posix");
1948 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1949 seq_printf(seq, ",fsync_mode=%s", "strict");
1950 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1951 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1953 #ifdef CONFIG_F2FS_FS_COMPRESSION
1954 f2fs_show_compress_options(seq, sbi->sb);
1957 if (test_opt(sbi, ATGC))
1958 seq_puts(seq, ",atgc");
1960 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1961 seq_printf(seq, ",discard_unit=%s", "block");
1962 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1963 seq_printf(seq, ",discard_unit=%s", "segment");
1964 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1965 seq_printf(seq, ",discard_unit=%s", "section");
1970 static void default_options(struct f2fs_sb_info *sbi)
1972 /* init some FS parameters */
1973 if (f2fs_sb_has_readonly(sbi))
1974 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
1976 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
1978 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1979 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1980 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1981 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1982 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1983 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1984 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1985 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1986 F2FS_OPTION(sbi).compress_ext_cnt = 0;
1987 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1988 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1990 sbi->sb->s_flags &= ~SB_INLINECRYPT;
1992 set_opt(sbi, INLINE_XATTR);
1993 set_opt(sbi, INLINE_DATA);
1994 set_opt(sbi, INLINE_DENTRY);
1995 set_opt(sbi, EXTENT_CACHE);
1996 set_opt(sbi, NOHEAP);
1997 clear_opt(sbi, DISABLE_CHECKPOINT);
1998 set_opt(sbi, MERGE_CHECKPOINT);
1999 F2FS_OPTION(sbi).unusable_cap = 0;
2000 sbi->sb->s_flags |= SB_LAZYTIME;
2001 set_opt(sbi, FLUSH_MERGE);
2002 set_opt(sbi, DISCARD);
2003 if (f2fs_sb_has_blkzoned(sbi)) {
2004 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2005 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2007 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2008 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2011 #ifdef CONFIG_F2FS_FS_XATTR
2012 set_opt(sbi, XATTR_USER);
2014 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2015 set_opt(sbi, POSIX_ACL);
2018 f2fs_build_fault_attr(sbi, 0, 0);
2022 static int f2fs_enable_quotas(struct super_block *sb);
2025 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2027 unsigned int s_flags = sbi->sb->s_flags;
2028 struct cp_control cpc;
2033 if (s_flags & SB_RDONLY) {
2034 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2037 sbi->sb->s_flags |= SB_ACTIVE;
2039 f2fs_update_time(sbi, DISABLE_TIME);
2041 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2042 down_write(&sbi->gc_lock);
2043 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
2044 if (err == -ENODATA) {
2048 if (err && err != -EAGAIN)
2052 ret = sync_filesystem(sbi->sb);
2054 err = ret ? ret : err;
2058 unusable = f2fs_get_unusable_blocks(sbi);
2059 if (f2fs_disable_cp_again(sbi, unusable)) {
2064 down_write(&sbi->gc_lock);
2065 cpc.reason = CP_PAUSE;
2066 set_sbi_flag(sbi, SBI_CP_DISABLED);
2067 err = f2fs_write_checkpoint(sbi, &cpc);
2071 spin_lock(&sbi->stat_lock);
2072 sbi->unusable_block_count = unusable;
2073 spin_unlock(&sbi->stat_lock);
2076 up_write(&sbi->gc_lock);
2078 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2082 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2084 /* we should flush all the data to keep data consistency */
2085 sync_inodes_sb(sbi->sb);
2087 down_write(&sbi->gc_lock);
2088 f2fs_dirty_to_prefree(sbi);
2090 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2091 set_sbi_flag(sbi, SBI_IS_DIRTY);
2092 up_write(&sbi->gc_lock);
2094 f2fs_sync_fs(sbi->sb, 1);
2097 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2099 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2100 struct f2fs_mount_info org_mount_opt;
2101 unsigned long old_sb_flags;
2103 bool need_restart_gc = false, need_stop_gc = false;
2104 bool need_restart_ckpt = false, need_stop_ckpt = false;
2105 bool need_restart_flush = false, need_stop_flush = false;
2106 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
2107 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2108 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2109 bool no_atgc = !test_opt(sbi, ATGC);
2110 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2111 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2117 * Save the old mount options in case we
2118 * need to restore them.
2120 org_mount_opt = sbi->mount_opt;
2121 old_sb_flags = sb->s_flags;
2124 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2125 for (i = 0; i < MAXQUOTAS; i++) {
2126 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2127 org_mount_opt.s_qf_names[i] =
2128 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2130 if (!org_mount_opt.s_qf_names[i]) {
2131 for (j = 0; j < i; j++)
2132 kfree(org_mount_opt.s_qf_names[j]);
2136 org_mount_opt.s_qf_names[i] = NULL;
2141 /* recover superblocks we couldn't write due to previous RO mount */
2142 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2143 err = f2fs_commit_super(sbi, false);
2144 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2147 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2150 default_options(sbi);
2152 /* parse mount options */
2153 err = parse_options(sb, data, true);
2158 * Previous and new state of filesystem is RO,
2159 * so skip checking GC and FLUSH_MERGE conditions.
2161 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2164 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2170 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2171 err = dquot_suspend(sb, -1);
2174 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2175 /* dquot_resume needs RW */
2176 sb->s_flags &= ~SB_RDONLY;
2177 if (sb_any_quota_suspended(sb)) {
2178 dquot_resume(sb, -1);
2179 } else if (f2fs_sb_has_quota_ino(sbi)) {
2180 err = f2fs_enable_quotas(sb);
2186 /* disallow enable atgc dynamically */
2187 if (no_atgc == !!test_opt(sbi, ATGC)) {
2189 f2fs_warn(sbi, "switch atgc option is not allowed");
2193 /* disallow enable/disable extent_cache dynamically */
2194 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
2196 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2200 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2202 f2fs_warn(sbi, "switch io_bits option is not allowed");
2206 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2208 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2212 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2214 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2218 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2220 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2225 * We stop the GC thread if FS is mounted as RO
2226 * or if background_gc = off is passed in mount
2227 * option. Also sync the filesystem.
2229 if ((*flags & SB_RDONLY) ||
2230 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2231 !test_opt(sbi, GC_MERGE))) {
2232 if (sbi->gc_thread) {
2233 f2fs_stop_gc_thread(sbi);
2234 need_restart_gc = true;
2236 } else if (!sbi->gc_thread) {
2237 err = f2fs_start_gc_thread(sbi);
2240 need_stop_gc = true;
2243 if (*flags & SB_RDONLY ||
2244 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
2247 set_sbi_flag(sbi, SBI_IS_DIRTY);
2248 set_sbi_flag(sbi, SBI_IS_CLOSE);
2249 f2fs_sync_fs(sb, 1);
2250 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2253 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2254 !test_opt(sbi, MERGE_CHECKPOINT)) {
2255 f2fs_stop_ckpt_thread(sbi);
2256 need_restart_ckpt = true;
2258 err = f2fs_start_ckpt_thread(sbi);
2261 "Failed to start F2FS issue_checkpoint_thread (%d)",
2265 need_stop_ckpt = true;
2269 * We stop issue flush thread if FS is mounted as RO
2270 * or if flush_merge is not passed in mount option.
2272 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2273 clear_opt(sbi, FLUSH_MERGE);
2274 f2fs_destroy_flush_cmd_control(sbi, false);
2275 need_restart_flush = true;
2277 err = f2fs_create_flush_cmd_control(sbi);
2280 need_stop_flush = true;
2283 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2284 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2285 err = f2fs_disable_checkpoint(sbi);
2289 f2fs_enable_checkpoint(sbi);
2295 /* Release old quota file names */
2296 for (i = 0; i < MAXQUOTAS; i++)
2297 kfree(org_mount_opt.s_qf_names[i]);
2299 /* Update the POSIXACL Flag */
2300 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2301 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2303 limit_reserve_root(sbi);
2304 adjust_unusable_cap_perc(sbi);
2305 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2308 if (need_restart_flush) {
2309 if (f2fs_create_flush_cmd_control(sbi))
2310 f2fs_warn(sbi, "background flush thread has stopped");
2311 } else if (need_stop_flush) {
2312 clear_opt(sbi, FLUSH_MERGE);
2313 f2fs_destroy_flush_cmd_control(sbi, false);
2316 if (need_restart_ckpt) {
2317 if (f2fs_start_ckpt_thread(sbi))
2318 f2fs_warn(sbi, "background ckpt thread has stopped");
2319 } else if (need_stop_ckpt) {
2320 f2fs_stop_ckpt_thread(sbi);
2323 if (need_restart_gc) {
2324 if (f2fs_start_gc_thread(sbi))
2325 f2fs_warn(sbi, "background gc thread has stopped");
2326 } else if (need_stop_gc) {
2327 f2fs_stop_gc_thread(sbi);
2331 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2332 for (i = 0; i < MAXQUOTAS; i++) {
2333 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2334 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2337 sbi->mount_opt = org_mount_opt;
2338 sb->s_flags = old_sb_flags;
2343 /* Read data from quotafile */
2344 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2345 size_t len, loff_t off)
2347 struct inode *inode = sb_dqopt(sb)->files[type];
2348 struct address_space *mapping = inode->i_mapping;
2349 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2350 int offset = off & (sb->s_blocksize - 1);
2353 loff_t i_size = i_size_read(inode);
2360 if (off + len > i_size)
2363 while (toread > 0) {
2364 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2366 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2368 if (PTR_ERR(page) == -ENOMEM) {
2369 congestion_wait(BLK_RW_ASYNC,
2370 DEFAULT_IO_TIMEOUT);
2373 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2374 return PTR_ERR(page);
2379 if (unlikely(page->mapping != mapping)) {
2380 f2fs_put_page(page, 1);
2383 if (unlikely(!PageUptodate(page))) {
2384 f2fs_put_page(page, 1);
2385 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2389 kaddr = kmap_atomic(page);
2390 memcpy(data, kaddr + offset, tocopy);
2391 kunmap_atomic(kaddr);
2392 f2fs_put_page(page, 1);
2402 /* Write to quotafile */
2403 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2404 const char *data, size_t len, loff_t off)
2406 struct inode *inode = sb_dqopt(sb)->files[type];
2407 struct address_space *mapping = inode->i_mapping;
2408 const struct address_space_operations *a_ops = mapping->a_ops;
2409 int offset = off & (sb->s_blocksize - 1);
2410 size_t towrite = len;
2412 void *fsdata = NULL;
2417 while (towrite > 0) {
2418 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2421 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2423 if (unlikely(err)) {
2424 if (err == -ENOMEM) {
2425 congestion_wait(BLK_RW_ASYNC,
2426 DEFAULT_IO_TIMEOUT);
2429 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2433 kaddr = kmap_atomic(page);
2434 memcpy(kaddr + offset, data, tocopy);
2435 kunmap_atomic(kaddr);
2436 flush_dcache_page(page);
2438 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2449 inode->i_mtime = inode->i_ctime = current_time(inode);
2450 f2fs_mark_inode_dirty_sync(inode, false);
2451 return len - towrite;
2454 static struct dquot **f2fs_get_dquots(struct inode *inode)
2456 return F2FS_I(inode)->i_dquot;
2459 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2461 return &F2FS_I(inode)->i_reserved_quota;
2464 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2466 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2467 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2471 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2472 F2FS_OPTION(sbi).s_jquota_fmt, type);
2475 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2480 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2481 err = f2fs_enable_quotas(sbi->sb);
2483 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2489 for (i = 0; i < MAXQUOTAS; i++) {
2490 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2491 err = f2fs_quota_on_mount(sbi, i);
2496 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2503 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2506 struct inode *qf_inode;
2507 unsigned long qf_inum;
2510 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2512 qf_inum = f2fs_qf_ino(sb, type);
2516 qf_inode = f2fs_iget(sb, qf_inum);
2517 if (IS_ERR(qf_inode)) {
2518 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2519 return PTR_ERR(qf_inode);
2522 /* Don't account quota for quota files to avoid recursion */
2523 qf_inode->i_flags |= S_NOQUOTA;
2524 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2529 static int f2fs_enable_quotas(struct super_block *sb)
2531 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2533 unsigned long qf_inum;
2534 bool quota_mopt[MAXQUOTAS] = {
2535 test_opt(sbi, USRQUOTA),
2536 test_opt(sbi, GRPQUOTA),
2537 test_opt(sbi, PRJQUOTA),
2540 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2541 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2545 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2547 for (type = 0; type < MAXQUOTAS; type++) {
2548 qf_inum = f2fs_qf_ino(sb, type);
2550 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2551 DQUOT_USAGE_ENABLED |
2552 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2554 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2556 for (type--; type >= 0; type--)
2557 dquot_quota_off(sb, type);
2558 set_sbi_flag(F2FS_SB(sb),
2559 SBI_QUOTA_NEED_REPAIR);
2567 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2569 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2570 struct address_space *mapping = dqopt->files[type]->i_mapping;
2573 ret = dquot_writeback_dquots(sbi->sb, type);
2577 ret = filemap_fdatawrite(mapping);
2581 /* if we are using journalled quota */
2582 if (is_journalled_quota(sbi))
2585 ret = filemap_fdatawait(mapping);
2587 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2590 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2594 int f2fs_quota_sync(struct super_block *sb, int type)
2596 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2597 struct quota_info *dqopt = sb_dqopt(sb);
2602 * Now when everything is written we can discard the pagecache so
2603 * that userspace sees the changes.
2605 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2607 if (type != -1 && cnt != type)
2610 if (!sb_has_quota_active(sb, type))
2613 inode_lock(dqopt->files[cnt]);
2618 * down_read(quota_sem)
2619 * dquot_writeback_dquots()
2622 * down_read(quota_sem)
2625 down_read(&sbi->quota_sem);
2627 ret = f2fs_quota_sync_file(sbi, cnt);
2629 up_read(&sbi->quota_sem);
2630 f2fs_unlock_op(sbi);
2632 inode_unlock(dqopt->files[cnt]);
2640 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2641 const struct path *path)
2643 struct inode *inode;
2646 /* if quota sysfile exists, deny enabling quota with specific file */
2647 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2648 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2652 err = f2fs_quota_sync(sb, type);
2656 err = dquot_quota_on(sb, type, format_id, path);
2660 inode = d_inode(path->dentry);
2663 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2664 f2fs_set_inode_flags(inode);
2665 inode_unlock(inode);
2666 f2fs_mark_inode_dirty_sync(inode, false);
2671 static int __f2fs_quota_off(struct super_block *sb, int type)
2673 struct inode *inode = sb_dqopt(sb)->files[type];
2676 if (!inode || !igrab(inode))
2677 return dquot_quota_off(sb, type);
2679 err = f2fs_quota_sync(sb, type);
2683 err = dquot_quota_off(sb, type);
2684 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2688 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2689 f2fs_set_inode_flags(inode);
2690 inode_unlock(inode);
2691 f2fs_mark_inode_dirty_sync(inode, false);
2697 static int f2fs_quota_off(struct super_block *sb, int type)
2699 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2702 err = __f2fs_quota_off(sb, type);
2705 * quotactl can shutdown journalled quota, result in inconsistence
2706 * between quota record and fs data by following updates, tag the
2707 * flag to let fsck be aware of it.
2709 if (is_journalled_quota(sbi))
2710 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2714 void f2fs_quota_off_umount(struct super_block *sb)
2719 for (type = 0; type < MAXQUOTAS; type++) {
2720 err = __f2fs_quota_off(sb, type);
2722 int ret = dquot_quota_off(sb, type);
2724 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2726 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2730 * In case of checkpoint=disable, we must flush quota blocks.
2731 * This can cause NULL exception for node_inode in end_io, since
2732 * put_super already dropped it.
2734 sync_filesystem(sb);
2737 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2739 struct quota_info *dqopt = sb_dqopt(sb);
2742 for (type = 0; type < MAXQUOTAS; type++) {
2743 if (!dqopt->files[type])
2745 f2fs_inode_synced(dqopt->files[type]);
2749 static int f2fs_dquot_commit(struct dquot *dquot)
2751 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2754 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2755 ret = dquot_commit(dquot);
2757 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2758 up_read(&sbi->quota_sem);
2762 static int f2fs_dquot_acquire(struct dquot *dquot)
2764 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2767 down_read(&sbi->quota_sem);
2768 ret = dquot_acquire(dquot);
2770 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2771 up_read(&sbi->quota_sem);
2775 static int f2fs_dquot_release(struct dquot *dquot)
2777 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2778 int ret = dquot_release(dquot);
2781 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2785 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2787 struct super_block *sb = dquot->dq_sb;
2788 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2789 int ret = dquot_mark_dquot_dirty(dquot);
2791 /* if we are using journalled quota */
2792 if (is_journalled_quota(sbi))
2793 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2798 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2800 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2801 int ret = dquot_commit_info(sb, type);
2804 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2808 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2810 *projid = F2FS_I(inode)->i_projid;
2814 static const struct dquot_operations f2fs_quota_operations = {
2815 .get_reserved_space = f2fs_get_reserved_space,
2816 .write_dquot = f2fs_dquot_commit,
2817 .acquire_dquot = f2fs_dquot_acquire,
2818 .release_dquot = f2fs_dquot_release,
2819 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2820 .write_info = f2fs_dquot_commit_info,
2821 .alloc_dquot = dquot_alloc,
2822 .destroy_dquot = dquot_destroy,
2823 .get_projid = f2fs_get_projid,
2824 .get_next_id = dquot_get_next_id,
2827 static const struct quotactl_ops f2fs_quotactl_ops = {
2828 .quota_on = f2fs_quota_on,
2829 .quota_off = f2fs_quota_off,
2830 .quota_sync = f2fs_quota_sync,
2831 .get_state = dquot_get_state,
2832 .set_info = dquot_set_dqinfo,
2833 .get_dqblk = dquot_get_dqblk,
2834 .set_dqblk = dquot_set_dqblk,
2835 .get_nextdqblk = dquot_get_next_dqblk,
2838 int f2fs_quota_sync(struct super_block *sb, int type)
2843 void f2fs_quota_off_umount(struct super_block *sb)
2848 static const struct super_operations f2fs_sops = {
2849 .alloc_inode = f2fs_alloc_inode,
2850 .free_inode = f2fs_free_inode,
2851 .drop_inode = f2fs_drop_inode,
2852 .write_inode = f2fs_write_inode,
2853 .dirty_inode = f2fs_dirty_inode,
2854 .show_options = f2fs_show_options,
2856 .quota_read = f2fs_quota_read,
2857 .quota_write = f2fs_quota_write,
2858 .get_dquots = f2fs_get_dquots,
2860 .evict_inode = f2fs_evict_inode,
2861 .put_super = f2fs_put_super,
2862 .sync_fs = f2fs_sync_fs,
2863 .freeze_fs = f2fs_freeze,
2864 .unfreeze_fs = f2fs_unfreeze,
2865 .statfs = f2fs_statfs,
2866 .remount_fs = f2fs_remount,
2869 #ifdef CONFIG_FS_ENCRYPTION
2870 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2872 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2873 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2877 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2880 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2883 * Encrypting the root directory is not allowed because fsck
2884 * expects lost+found directory to exist and remain unencrypted
2885 * if LOST_FOUND feature is enabled.
2888 if (f2fs_sb_has_lost_found(sbi) &&
2889 inode->i_ino == F2FS_ROOT_INO(sbi))
2892 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2893 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2894 ctx, len, fs_data, XATTR_CREATE);
2897 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
2899 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
2902 static bool f2fs_has_stable_inodes(struct super_block *sb)
2907 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2908 int *ino_bits_ret, int *lblk_bits_ret)
2910 *ino_bits_ret = 8 * sizeof(nid_t);
2911 *lblk_bits_ret = 8 * sizeof(block_t);
2914 static int f2fs_get_num_devices(struct super_block *sb)
2916 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2918 if (f2fs_is_multi_device(sbi))
2919 return sbi->s_ndevs;
2923 static void f2fs_get_devices(struct super_block *sb,
2924 struct request_queue **devs)
2926 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2929 for (i = 0; i < sbi->s_ndevs; i++)
2930 devs[i] = bdev_get_queue(FDEV(i).bdev);
2933 static const struct fscrypt_operations f2fs_cryptops = {
2934 .key_prefix = "f2fs:",
2935 .get_context = f2fs_get_context,
2936 .set_context = f2fs_set_context,
2937 .get_dummy_policy = f2fs_get_dummy_policy,
2938 .empty_dir = f2fs_empty_dir,
2939 .max_namelen = F2FS_NAME_LEN,
2940 .has_stable_inodes = f2fs_has_stable_inodes,
2941 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2942 .get_num_devices = f2fs_get_num_devices,
2943 .get_devices = f2fs_get_devices,
2947 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2948 u64 ino, u32 generation)
2950 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2951 struct inode *inode;
2953 if (f2fs_check_nid_range(sbi, ino))
2954 return ERR_PTR(-ESTALE);
2957 * f2fs_iget isn't quite right if the inode is currently unallocated!
2958 * However f2fs_iget currently does appropriate checks to handle stale
2959 * inodes so everything is OK.
2961 inode = f2fs_iget(sb, ino);
2963 return ERR_CAST(inode);
2964 if (unlikely(generation && inode->i_generation != generation)) {
2965 /* we didn't find the right inode.. */
2967 return ERR_PTR(-ESTALE);
2972 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2973 int fh_len, int fh_type)
2975 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2976 f2fs_nfs_get_inode);
2979 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2980 int fh_len, int fh_type)
2982 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2983 f2fs_nfs_get_inode);
2986 static const struct export_operations f2fs_export_ops = {
2987 .fh_to_dentry = f2fs_fh_to_dentry,
2988 .fh_to_parent = f2fs_fh_to_parent,
2989 .get_parent = f2fs_get_parent,
2992 loff_t max_file_blocks(struct inode *inode)
2998 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2999 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3000 * space in inode.i_addr, it will be more safe to reassign
3004 if (inode && f2fs_compressed_file(inode))
3005 leaf_count = ADDRS_PER_BLOCK(inode);
3007 leaf_count = DEF_ADDRS_PER_BLOCK;
3009 /* two direct node blocks */
3010 result += (leaf_count * 2);
3012 /* two indirect node blocks */
3013 leaf_count *= NIDS_PER_BLOCK;
3014 result += (leaf_count * 2);
3016 /* one double indirect node block */
3017 leaf_count *= NIDS_PER_BLOCK;
3018 result += leaf_count;
3023 static int __f2fs_commit_super(struct buffer_head *bh,
3024 struct f2fs_super_block *super)
3028 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3029 set_buffer_dirty(bh);
3032 /* it's rare case, we can do fua all the time */
3033 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3036 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3037 struct buffer_head *bh)
3039 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3040 (bh->b_data + F2FS_SUPER_OFFSET);
3041 struct super_block *sb = sbi->sb;
3042 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3043 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3044 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3045 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3046 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3047 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3048 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3049 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3050 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3051 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3052 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3053 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3054 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3055 u64 main_end_blkaddr = main_blkaddr +
3056 (segment_count_main << log_blocks_per_seg);
3057 u64 seg_end_blkaddr = segment0_blkaddr +
3058 (segment_count << log_blocks_per_seg);
3060 if (segment0_blkaddr != cp_blkaddr) {
3061 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3062 segment0_blkaddr, cp_blkaddr);
3066 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3068 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3069 cp_blkaddr, sit_blkaddr,
3070 segment_count_ckpt << log_blocks_per_seg);
3074 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3076 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3077 sit_blkaddr, nat_blkaddr,
3078 segment_count_sit << log_blocks_per_seg);
3082 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3084 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3085 nat_blkaddr, ssa_blkaddr,
3086 segment_count_nat << log_blocks_per_seg);
3090 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3092 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3093 ssa_blkaddr, main_blkaddr,
3094 segment_count_ssa << log_blocks_per_seg);
3098 if (main_end_blkaddr > seg_end_blkaddr) {
3099 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3100 main_blkaddr, seg_end_blkaddr,
3101 segment_count_main << log_blocks_per_seg);
3103 } else if (main_end_blkaddr < seg_end_blkaddr) {
3107 /* fix in-memory information all the time */
3108 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3109 segment0_blkaddr) >> log_blocks_per_seg);
3111 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3112 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3115 err = __f2fs_commit_super(bh, NULL);
3116 res = err ? "failed" : "done";
3118 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3119 res, main_blkaddr, seg_end_blkaddr,
3120 segment_count_main << log_blocks_per_seg);
3127 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3128 struct buffer_head *bh)
3130 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3131 block_t total_sections, blocks_per_seg;
3132 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3133 (bh->b_data + F2FS_SUPER_OFFSET);
3134 size_t crc_offset = 0;
3137 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3138 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3139 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3143 /* Check checksum_offset and crc in superblock */
3144 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3145 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3147 offsetof(struct f2fs_super_block, crc)) {
3148 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3150 return -EFSCORRUPTED;
3152 crc = le32_to_cpu(raw_super->crc);
3153 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3154 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3155 return -EFSCORRUPTED;
3159 /* Currently, support only 4KB block size */
3160 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3161 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3162 le32_to_cpu(raw_super->log_blocksize),
3164 return -EFSCORRUPTED;
3167 /* check log blocks per segment */
3168 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3169 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3170 le32_to_cpu(raw_super->log_blocks_per_seg));
3171 return -EFSCORRUPTED;
3174 /* Currently, support 512/1024/2048/4096 bytes sector size */
3175 if (le32_to_cpu(raw_super->log_sectorsize) >
3176 F2FS_MAX_LOG_SECTOR_SIZE ||
3177 le32_to_cpu(raw_super->log_sectorsize) <
3178 F2FS_MIN_LOG_SECTOR_SIZE) {
3179 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3180 le32_to_cpu(raw_super->log_sectorsize));
3181 return -EFSCORRUPTED;
3183 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3184 le32_to_cpu(raw_super->log_sectorsize) !=
3185 F2FS_MAX_LOG_SECTOR_SIZE) {
3186 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3187 le32_to_cpu(raw_super->log_sectors_per_block),
3188 le32_to_cpu(raw_super->log_sectorsize));
3189 return -EFSCORRUPTED;
3192 segment_count = le32_to_cpu(raw_super->segment_count);
3193 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3194 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3195 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3196 total_sections = le32_to_cpu(raw_super->section_count);
3198 /* blocks_per_seg should be 512, given the above check */
3199 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3201 if (segment_count > F2FS_MAX_SEGMENT ||
3202 segment_count < F2FS_MIN_SEGMENTS) {
3203 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3204 return -EFSCORRUPTED;
3207 if (total_sections > segment_count_main || total_sections < 1 ||
3208 segs_per_sec > segment_count || !segs_per_sec) {
3209 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3210 segment_count, total_sections, segs_per_sec);
3211 return -EFSCORRUPTED;
3214 if (segment_count_main != total_sections * segs_per_sec) {
3215 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3216 segment_count_main, total_sections, segs_per_sec);
3217 return -EFSCORRUPTED;
3220 if ((segment_count / segs_per_sec) < total_sections) {
3221 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3222 segment_count, segs_per_sec, total_sections);
3223 return -EFSCORRUPTED;
3226 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3227 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3228 segment_count, le64_to_cpu(raw_super->block_count));
3229 return -EFSCORRUPTED;
3232 if (RDEV(0).path[0]) {
3233 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3236 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3237 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3240 if (segment_count != dev_seg_count) {
3241 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3242 segment_count, dev_seg_count);
3243 return -EFSCORRUPTED;
3246 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3247 !bdev_is_zoned(sbi->sb->s_bdev)) {
3248 f2fs_info(sbi, "Zoned block device path is missing");
3249 return -EFSCORRUPTED;
3253 if (secs_per_zone > total_sections || !secs_per_zone) {
3254 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3255 secs_per_zone, total_sections);
3256 return -EFSCORRUPTED;
3258 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3259 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3260 (le32_to_cpu(raw_super->extension_count) +
3261 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3262 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3263 le32_to_cpu(raw_super->extension_count),
3264 raw_super->hot_ext_count,
3265 F2FS_MAX_EXTENSION);
3266 return -EFSCORRUPTED;
3269 if (le32_to_cpu(raw_super->cp_payload) >=
3270 (blocks_per_seg - F2FS_CP_PACKS -
3271 NR_CURSEG_PERSIST_TYPE)) {
3272 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3273 le32_to_cpu(raw_super->cp_payload),
3274 blocks_per_seg - F2FS_CP_PACKS -
3275 NR_CURSEG_PERSIST_TYPE);
3276 return -EFSCORRUPTED;
3279 /* check reserved ino info */
3280 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3281 le32_to_cpu(raw_super->meta_ino) != 2 ||
3282 le32_to_cpu(raw_super->root_ino) != 3) {
3283 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3284 le32_to_cpu(raw_super->node_ino),
3285 le32_to_cpu(raw_super->meta_ino),
3286 le32_to_cpu(raw_super->root_ino));
3287 return -EFSCORRUPTED;
3290 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3291 if (sanity_check_area_boundary(sbi, bh))
3292 return -EFSCORRUPTED;
3297 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3299 unsigned int total, fsmeta;
3300 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3301 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3302 unsigned int ovp_segments, reserved_segments;
3303 unsigned int main_segs, blocks_per_seg;
3304 unsigned int sit_segs, nat_segs;
3305 unsigned int sit_bitmap_size, nat_bitmap_size;
3306 unsigned int log_blocks_per_seg;
3307 unsigned int segment_count_main;
3308 unsigned int cp_pack_start_sum, cp_payload;
3309 block_t user_block_count, valid_user_blocks;
3310 block_t avail_node_count, valid_node_count;
3311 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3314 total = le32_to_cpu(raw_super->segment_count);
3315 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3316 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3318 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3320 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3321 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3323 if (unlikely(fsmeta >= total))
3326 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3327 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3329 if (!f2fs_sb_has_readonly(sbi) &&
3330 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3331 ovp_segments == 0 || reserved_segments == 0)) {
3332 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3335 user_block_count = le64_to_cpu(ckpt->user_block_count);
3336 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3337 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3338 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3339 if (!user_block_count || user_block_count >=
3340 segment_count_main << log_blocks_per_seg) {
3341 f2fs_err(sbi, "Wrong user_block_count: %u",
3346 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3347 if (valid_user_blocks > user_block_count) {
3348 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3349 valid_user_blocks, user_block_count);
3353 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3354 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3355 if (valid_node_count > avail_node_count) {
3356 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3357 valid_node_count, avail_node_count);
3361 main_segs = le32_to_cpu(raw_super->segment_count_main);
3362 blocks_per_seg = sbi->blocks_per_seg;
3364 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3365 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3366 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3369 if (f2fs_sb_has_readonly(sbi))
3372 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3373 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3374 le32_to_cpu(ckpt->cur_node_segno[j])) {
3375 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3377 le32_to_cpu(ckpt->cur_node_segno[i]));
3383 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3384 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3385 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3388 if (f2fs_sb_has_readonly(sbi))
3391 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3392 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3393 le32_to_cpu(ckpt->cur_data_segno[j])) {
3394 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3396 le32_to_cpu(ckpt->cur_data_segno[i]));
3401 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3402 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3403 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3404 le32_to_cpu(ckpt->cur_data_segno[j])) {
3405 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3407 le32_to_cpu(ckpt->cur_node_segno[i]));
3413 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3414 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3416 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3417 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3418 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3419 sit_bitmap_size, nat_bitmap_size);
3423 cp_pack_start_sum = __start_sum_addr(sbi);
3424 cp_payload = __cp_payload(sbi);
3425 if (cp_pack_start_sum < cp_payload + 1 ||
3426 cp_pack_start_sum > blocks_per_seg - 1 -
3427 NR_CURSEG_PERSIST_TYPE) {
3428 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3433 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3434 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3435 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3436 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3437 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3438 le32_to_cpu(ckpt->checksum_offset));
3442 nat_blocks = nat_segs << log_blocks_per_seg;
3443 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3444 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3445 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3446 (cp_payload + F2FS_CP_PACKS +
3447 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3448 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3449 cp_payload, nat_bits_blocks);
3450 return -EFSCORRUPTED;
3453 if (unlikely(f2fs_cp_error(sbi))) {
3454 f2fs_err(sbi, "A bug case: need to run fsck");
3460 static void init_sb_info(struct f2fs_sb_info *sbi)
3462 struct f2fs_super_block *raw_super = sbi->raw_super;
3465 sbi->log_sectors_per_block =
3466 le32_to_cpu(raw_super->log_sectors_per_block);
3467 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3468 sbi->blocksize = 1 << sbi->log_blocksize;
3469 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3470 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3471 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3472 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3473 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3474 sbi->total_node_count =
3475 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3476 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3477 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3478 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3479 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3480 sbi->cur_victim_sec = NULL_SECNO;
3481 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3482 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3483 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3484 sbi->migration_granularity = sbi->segs_per_sec;
3485 sbi->seq_file_ra_mul = MIN_RA_MUL;
3487 sbi->dir_level = DEF_DIR_LEVEL;
3488 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3489 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3490 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3491 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3492 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3493 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3494 DEF_UMOUNT_DISCARD_TIMEOUT;
3495 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3497 for (i = 0; i < NR_COUNT_TYPE; i++)
3498 atomic_set(&sbi->nr_pages[i], 0);
3500 for (i = 0; i < META; i++)
3501 atomic_set(&sbi->wb_sync_req[i], 0);
3503 INIT_LIST_HEAD(&sbi->s_list);
3504 mutex_init(&sbi->umount_mutex);
3505 init_rwsem(&sbi->io_order_lock);
3506 spin_lock_init(&sbi->cp_lock);
3508 sbi->dirty_device = 0;
3509 spin_lock_init(&sbi->dev_lock);
3511 init_rwsem(&sbi->sb_lock);
3512 init_rwsem(&sbi->pin_sem);
3515 static int init_percpu_info(struct f2fs_sb_info *sbi)
3519 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3523 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3526 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3531 #ifdef CONFIG_BLK_DEV_ZONED
3533 struct f2fs_report_zones_args {
3534 struct f2fs_dev_info *dev;
3535 bool zone_cap_mismatch;
3538 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3541 struct f2fs_report_zones_args *rz_args = data;
3543 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3546 set_bit(idx, rz_args->dev->blkz_seq);
3547 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >>
3548 F2FS_LOG_SECTORS_PER_BLOCK;
3549 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch)
3550 rz_args->zone_cap_mismatch = true;
3555 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3557 struct block_device *bdev = FDEV(devi).bdev;
3558 sector_t nr_sectors = bdev_nr_sectors(bdev);
3559 struct f2fs_report_zones_args rep_zone_arg;
3562 if (!f2fs_sb_has_blkzoned(sbi))
3565 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3566 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3568 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3569 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3570 __ilog2_u32(sbi->blocks_per_blkz))
3572 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3573 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3574 sbi->log_blocks_per_blkz;
3575 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3576 FDEV(devi).nr_blkz++;
3578 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3579 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3580 * sizeof(unsigned long),
3582 if (!FDEV(devi).blkz_seq)
3585 /* Get block zones type and zone-capacity */
3586 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi,
3587 FDEV(devi).nr_blkz * sizeof(block_t),
3589 if (!FDEV(devi).zone_capacity_blocks)
3592 rep_zone_arg.dev = &FDEV(devi);
3593 rep_zone_arg.zone_cap_mismatch = false;
3595 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3600 if (!rep_zone_arg.zone_cap_mismatch) {
3601 kfree(FDEV(devi).zone_capacity_blocks);
3602 FDEV(devi).zone_capacity_blocks = NULL;
3610 * Read f2fs raw super block.
3611 * Because we have two copies of super block, so read both of them
3612 * to get the first valid one. If any one of them is broken, we pass
3613 * them recovery flag back to the caller.
3615 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3616 struct f2fs_super_block **raw_super,
3617 int *valid_super_block, int *recovery)
3619 struct super_block *sb = sbi->sb;
3621 struct buffer_head *bh;
3622 struct f2fs_super_block *super;
3625 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3629 for (block = 0; block < 2; block++) {
3630 bh = sb_bread(sb, block);
3632 f2fs_err(sbi, "Unable to read %dth superblock",
3639 /* sanity checking of raw super */
3640 err = sanity_check_raw_super(sbi, bh);
3642 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3650 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3652 *valid_super_block = block;
3658 /* No valid superblock */
3667 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3669 struct buffer_head *bh;
3673 if ((recover && f2fs_readonly(sbi->sb)) ||
3674 bdev_read_only(sbi->sb->s_bdev)) {
3675 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3679 /* we should update superblock crc here */
3680 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3681 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3682 offsetof(struct f2fs_super_block, crc));
3683 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3686 /* write back-up superblock first */
3687 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3690 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3693 /* if we are in recovery path, skip writing valid superblock */
3697 /* write current valid superblock */
3698 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3701 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3706 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3708 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3709 unsigned int max_devices = MAX_DEVICES;
3712 /* Initialize single device information */
3713 if (!RDEV(0).path[0]) {
3714 if (!bdev_is_zoned(sbi->sb->s_bdev))
3720 * Initialize multiple devices information, or single
3721 * zoned block device information.
3723 sbi->devs = f2fs_kzalloc(sbi,
3724 array_size(max_devices,
3725 sizeof(struct f2fs_dev_info)),
3730 for (i = 0; i < max_devices; i++) {
3732 if (i > 0 && !RDEV(i).path[0])
3735 if (max_devices == 1) {
3736 /* Single zoned block device mount */
3738 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3739 sbi->sb->s_mode, sbi->sb->s_type);
3741 /* Multi-device mount */
3742 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3743 FDEV(i).total_segments =
3744 le32_to_cpu(RDEV(i).total_segments);
3746 FDEV(i).start_blk = 0;
3747 FDEV(i).end_blk = FDEV(i).start_blk +
3748 (FDEV(i).total_segments <<
3749 sbi->log_blocks_per_seg) - 1 +
3750 le32_to_cpu(raw_super->segment0_blkaddr);
3752 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3753 FDEV(i).end_blk = FDEV(i).start_blk +
3754 (FDEV(i).total_segments <<
3755 sbi->log_blocks_per_seg) - 1;
3757 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3758 sbi->sb->s_mode, sbi->sb->s_type);
3760 if (IS_ERR(FDEV(i).bdev))
3761 return PTR_ERR(FDEV(i).bdev);
3763 /* to release errored devices */
3764 sbi->s_ndevs = i + 1;
3766 #ifdef CONFIG_BLK_DEV_ZONED
3767 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3768 !f2fs_sb_has_blkzoned(sbi)) {
3769 f2fs_err(sbi, "Zoned block device feature not enabled");
3772 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3773 if (init_blkz_info(sbi, i)) {
3774 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3777 if (max_devices == 1)
3779 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3781 FDEV(i).total_segments,
3782 FDEV(i).start_blk, FDEV(i).end_blk,
3783 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3784 "Host-aware" : "Host-managed");
3788 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3790 FDEV(i).total_segments,
3791 FDEV(i).start_blk, FDEV(i).end_blk);
3794 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3798 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3800 #ifdef CONFIG_UNICODE
3801 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3802 const struct f2fs_sb_encodings *encoding_info;
3803 struct unicode_map *encoding;
3804 __u16 encoding_flags;
3806 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3809 "Encoding requested by superblock is unknown");
3813 encoding = utf8_load(encoding_info->version);
3814 if (IS_ERR(encoding)) {
3816 "can't mount with superblock charset: %s-%s "
3817 "not supported by the kernel. flags: 0x%x.",
3818 encoding_info->name, encoding_info->version,
3820 return PTR_ERR(encoding);
3822 f2fs_info(sbi, "Using encoding defined by superblock: "
3823 "%s-%s with flags 0x%hx", encoding_info->name,
3824 encoding_info->version?:"\b", encoding_flags);
3826 sbi->sb->s_encoding = encoding;
3827 sbi->sb->s_encoding_flags = encoding_flags;
3830 if (f2fs_sb_has_casefold(sbi)) {
3831 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3838 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3840 struct f2fs_sm_info *sm_i = SM_I(sbi);
3842 /* adjust parameters according to the volume size */
3843 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3844 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3845 if (f2fs_block_unit_discard(sbi))
3846 sm_i->dcc_info->discard_granularity = 1;
3847 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3850 sbi->readdir_ra = 1;
3853 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3855 struct f2fs_sb_info *sbi;
3856 struct f2fs_super_block *raw_super;
3859 bool skip_recovery = false, need_fsck = false;
3860 char *options = NULL;
3861 int recovery, i, valid_super_block;
3862 struct curseg_info *seg_i;
3868 valid_super_block = -1;
3871 /* allocate memory for f2fs-specific super block info */
3872 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3878 /* Load the checksum driver */
3879 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3880 if (IS_ERR(sbi->s_chksum_driver)) {
3881 f2fs_err(sbi, "Cannot load crc32 driver.");
3882 err = PTR_ERR(sbi->s_chksum_driver);
3883 sbi->s_chksum_driver = NULL;
3887 /* set a block size */
3888 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3889 f2fs_err(sbi, "unable to set blocksize");
3893 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3898 sb->s_fs_info = sbi;
3899 sbi->raw_super = raw_super;
3901 /* precompute checksum seed for metadata */
3902 if (f2fs_sb_has_inode_chksum(sbi))
3903 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3904 sizeof(raw_super->uuid));
3906 default_options(sbi);
3907 /* parse mount options */
3908 options = kstrdup((const char *)data, GFP_KERNEL);
3909 if (data && !options) {
3914 err = parse_options(sb, options, false);
3918 sb->s_maxbytes = max_file_blocks(NULL) <<
3919 le32_to_cpu(raw_super->log_blocksize);
3920 sb->s_max_links = F2FS_LINK_MAX;
3922 err = f2fs_setup_casefold(sbi);
3927 sb->dq_op = &f2fs_quota_operations;
3928 sb->s_qcop = &f2fs_quotactl_ops;
3929 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3931 if (f2fs_sb_has_quota_ino(sbi)) {
3932 for (i = 0; i < MAXQUOTAS; i++) {
3933 if (f2fs_qf_ino(sbi->sb, i))
3934 sbi->nquota_files++;
3939 sb->s_op = &f2fs_sops;
3940 #ifdef CONFIG_FS_ENCRYPTION
3941 sb->s_cop = &f2fs_cryptops;
3943 #ifdef CONFIG_FS_VERITY
3944 sb->s_vop = &f2fs_verityops;
3946 sb->s_xattr = f2fs_xattr_handlers;
3947 sb->s_export_op = &f2fs_export_ops;
3948 sb->s_magic = F2FS_SUPER_MAGIC;
3949 sb->s_time_gran = 1;
3950 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3951 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3952 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3953 sb->s_iflags |= SB_I_CGROUPWB;
3955 /* init f2fs-specific super block info */
3956 sbi->valid_super_block = valid_super_block;
3957 init_rwsem(&sbi->gc_lock);
3958 mutex_init(&sbi->writepages);
3959 init_rwsem(&sbi->cp_global_sem);
3960 init_rwsem(&sbi->node_write);
3961 init_rwsem(&sbi->node_change);
3963 /* disallow all the data/node/meta page writes */
3964 set_sbi_flag(sbi, SBI_POR_DOING);
3965 spin_lock_init(&sbi->stat_lock);
3967 /* init iostat info */
3968 spin_lock_init(&sbi->iostat_lock);
3969 sbi->iostat_enable = false;
3970 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3972 for (i = 0; i < NR_PAGE_TYPE; i++) {
3973 int n = (i == META) ? 1 : NR_TEMP_TYPE;
3979 sizeof(struct f2fs_bio_info)),
3981 if (!sbi->write_io[i]) {
3986 for (j = HOT; j < n; j++) {
3987 init_rwsem(&sbi->write_io[i][j].io_rwsem);
3988 sbi->write_io[i][j].sbi = sbi;
3989 sbi->write_io[i][j].bio = NULL;
3990 spin_lock_init(&sbi->write_io[i][j].io_lock);
3991 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3992 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3993 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3997 init_rwsem(&sbi->cp_rwsem);
3998 init_rwsem(&sbi->quota_sem);
3999 init_waitqueue_head(&sbi->cp_wait);
4002 err = init_percpu_info(sbi);
4006 if (F2FS_IO_ALIGNED(sbi)) {
4007 sbi->write_io_dummy =
4008 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4009 if (!sbi->write_io_dummy) {
4015 /* init per sbi slab cache */
4016 err = f2fs_init_xattr_caches(sbi);
4019 err = f2fs_init_page_array_cache(sbi);
4021 goto free_xattr_cache;
4023 /* get an inode for meta space */
4024 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4025 if (IS_ERR(sbi->meta_inode)) {
4026 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4027 err = PTR_ERR(sbi->meta_inode);
4028 goto free_page_array_cache;
4031 err = f2fs_get_valid_checkpoint(sbi);
4033 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4034 goto free_meta_inode;
4037 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4038 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4039 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4040 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4041 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4044 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4045 set_sbi_flag(sbi, SBI_NEED_FSCK);
4047 /* Initialize device list */
4048 err = f2fs_scan_devices(sbi);
4050 f2fs_err(sbi, "Failed to find devices");
4054 err = f2fs_init_post_read_wq(sbi);
4056 f2fs_err(sbi, "Failed to initialize post read workqueue");
4060 sbi->total_valid_node_count =
4061 le32_to_cpu(sbi->ckpt->valid_node_count);
4062 percpu_counter_set(&sbi->total_valid_inode_count,
4063 le32_to_cpu(sbi->ckpt->valid_inode_count));
4064 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4065 sbi->total_valid_block_count =
4066 le64_to_cpu(sbi->ckpt->valid_block_count);
4067 sbi->last_valid_block_count = sbi->total_valid_block_count;
4068 sbi->reserved_blocks = 0;
4069 sbi->current_reserved_blocks = 0;
4070 limit_reserve_root(sbi);
4071 adjust_unusable_cap_perc(sbi);
4073 for (i = 0; i < NR_INODE_TYPE; i++) {
4074 INIT_LIST_HEAD(&sbi->inode_list[i]);
4075 spin_lock_init(&sbi->inode_lock[i]);
4077 mutex_init(&sbi->flush_lock);
4079 f2fs_init_extent_cache_info(sbi);
4081 f2fs_init_ino_entry_info(sbi);
4083 f2fs_init_fsync_node_info(sbi);
4085 /* setup checkpoint request control and start checkpoint issue thread */
4086 f2fs_init_ckpt_req_control(sbi);
4087 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4088 test_opt(sbi, MERGE_CHECKPOINT)) {
4089 err = f2fs_start_ckpt_thread(sbi);
4092 "Failed to start F2FS issue_checkpoint_thread (%d)",
4094 goto stop_ckpt_thread;
4098 /* setup f2fs internal modules */
4099 err = f2fs_build_segment_manager(sbi);
4101 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4105 err = f2fs_build_node_manager(sbi);
4107 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4112 /* For write statistics */
4113 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4115 /* Read accumulated write IO statistics if exists */
4116 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4117 if (__exist_node_summaries(sbi))
4118 sbi->kbytes_written =
4119 le64_to_cpu(seg_i->journal->info.kbytes_written);
4121 f2fs_build_gc_manager(sbi);
4123 err = f2fs_build_stats(sbi);
4127 /* get an inode for node space */
4128 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4129 if (IS_ERR(sbi->node_inode)) {
4130 f2fs_err(sbi, "Failed to read node inode");
4131 err = PTR_ERR(sbi->node_inode);
4135 /* read root inode and dentry */
4136 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4138 f2fs_err(sbi, "Failed to read root inode");
4139 err = PTR_ERR(root);
4140 goto free_node_inode;
4142 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4143 !root->i_size || !root->i_nlink) {
4146 goto free_node_inode;
4149 sb->s_root = d_make_root(root); /* allocate root dentry */
4152 goto free_node_inode;
4155 err = f2fs_init_compress_inode(sbi);
4157 goto free_root_inode;
4159 err = f2fs_register_sysfs(sbi);
4161 goto free_compress_inode;
4164 /* Enable quota usage during mount */
4165 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4166 err = f2fs_enable_quotas(sb);
4168 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4171 /* if there are any orphan inodes, free them */
4172 err = f2fs_recover_orphan_inodes(sbi);
4176 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4177 goto reset_checkpoint;
4179 /* recover fsynced data */
4180 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4181 !test_opt(sbi, NORECOVERY)) {
4183 * mount should be failed, when device has readonly mode, and
4184 * previous checkpoint was not done by clean system shutdown.
4186 if (f2fs_hw_is_readonly(sbi)) {
4187 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4188 err = f2fs_recover_fsync_data(sbi, true);
4191 f2fs_err(sbi, "Need to recover fsync data, but "
4192 "write access unavailable, please try "
4193 "mount w/ disable_roll_forward or norecovery");
4198 f2fs_info(sbi, "write access unavailable, skipping recovery");
4199 goto reset_checkpoint;
4203 set_sbi_flag(sbi, SBI_NEED_FSCK);
4206 goto reset_checkpoint;
4208 err = f2fs_recover_fsync_data(sbi, false);
4211 skip_recovery = true;
4213 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4218 err = f2fs_recover_fsync_data(sbi, true);
4220 if (!f2fs_readonly(sb) && err > 0) {
4222 f2fs_err(sbi, "Need to recover fsync data");
4228 * If the f2fs is not readonly and fsync data recovery succeeds,
4229 * check zoned block devices' write pointer consistency.
4231 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4232 err = f2fs_check_write_pointer(sbi);
4238 f2fs_init_inmem_curseg(sbi);
4240 /* f2fs_recover_fsync_data() cleared this already */
4241 clear_sbi_flag(sbi, SBI_POR_DOING);
4243 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4244 err = f2fs_disable_checkpoint(sbi);
4246 goto sync_free_meta;
4247 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4248 f2fs_enable_checkpoint(sbi);
4252 * If filesystem is not mounted as read-only then
4253 * do start the gc_thread.
4255 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4256 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4257 /* After POR, we can run background GC thread.*/
4258 err = f2fs_start_gc_thread(sbi);
4260 goto sync_free_meta;
4264 /* recover broken superblock */
4266 err = f2fs_commit_super(sbi, true);
4267 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4268 sbi->valid_super_block ? 1 : 2, err);
4271 f2fs_join_shrinker(sbi);
4273 f2fs_tuning_parameters(sbi);
4275 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4276 cur_cp_version(F2FS_CKPT(sbi)));
4277 f2fs_update_time(sbi, CP_TIME);
4278 f2fs_update_time(sbi, REQ_TIME);
4279 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4283 /* safe to flush all the data */
4284 sync_filesystem(sbi->sb);
4289 f2fs_truncate_quota_inode_pages(sb);
4290 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4291 f2fs_quota_off_umount(sbi->sb);
4294 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4295 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4296 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4297 * falls into an infinite loop in f2fs_sync_meta_pages().
4299 truncate_inode_pages_final(META_MAPPING(sbi));
4300 /* evict some inodes being cached by GC */
4302 f2fs_unregister_sysfs(sbi);
4303 free_compress_inode:
4304 f2fs_destroy_compress_inode(sbi);
4309 f2fs_release_ino_entry(sbi, true);
4310 truncate_inode_pages_final(NODE_MAPPING(sbi));
4311 iput(sbi->node_inode);
4312 sbi->node_inode = NULL;
4314 f2fs_destroy_stats(sbi);
4316 f2fs_destroy_node_manager(sbi);
4318 f2fs_destroy_segment_manager(sbi);
4319 f2fs_destroy_post_read_wq(sbi);
4321 f2fs_stop_ckpt_thread(sbi);
4323 destroy_device_list(sbi);
4326 make_bad_inode(sbi->meta_inode);
4327 iput(sbi->meta_inode);
4328 sbi->meta_inode = NULL;
4329 free_page_array_cache:
4330 f2fs_destroy_page_array_cache(sbi);
4332 f2fs_destroy_xattr_caches(sbi);
4334 mempool_destroy(sbi->write_io_dummy);
4336 destroy_percpu_info(sbi);
4338 for (i = 0; i < NR_PAGE_TYPE; i++)
4339 kvfree(sbi->write_io[i]);
4341 #ifdef CONFIG_UNICODE
4342 utf8_unload(sb->s_encoding);
4343 sb->s_encoding = NULL;
4347 for (i = 0; i < MAXQUOTAS; i++)
4348 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4350 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4355 if (sbi->s_chksum_driver)
4356 crypto_free_shash(sbi->s_chksum_driver);
4359 /* give only one another chance */
4360 if (retry_cnt > 0 && skip_recovery) {
4362 shrink_dcache_sb(sb);
4368 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4369 const char *dev_name, void *data)
4371 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4374 static void kill_f2fs_super(struct super_block *sb)
4377 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4379 set_sbi_flag(sbi, SBI_IS_CLOSE);
4380 f2fs_stop_gc_thread(sbi);
4381 f2fs_stop_discard_thread(sbi);
4383 #ifdef CONFIG_F2FS_FS_COMPRESSION
4385 * latter evict_inode() can bypass checking and invalidating
4386 * compress inode cache.
4388 if (test_opt(sbi, COMPRESS_CACHE))
4389 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4392 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4393 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4394 struct cp_control cpc = {
4395 .reason = CP_UMOUNT,
4397 f2fs_write_checkpoint(sbi, &cpc);
4400 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4401 sb->s_flags &= ~SB_RDONLY;
4403 kill_block_super(sb);
4406 static struct file_system_type f2fs_fs_type = {
4407 .owner = THIS_MODULE,
4409 .mount = f2fs_mount,
4410 .kill_sb = kill_f2fs_super,
4411 .fs_flags = FS_REQUIRES_DEV,
4413 MODULE_ALIAS_FS("f2fs");
4415 static int __init init_inodecache(void)
4417 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4418 sizeof(struct f2fs_inode_info), 0,
4419 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4420 if (!f2fs_inode_cachep)
4425 static void destroy_inodecache(void)
4428 * Make sure all delayed rcu free inodes are flushed before we
4432 kmem_cache_destroy(f2fs_inode_cachep);
4435 static int __init init_f2fs_fs(void)
4439 if (PAGE_SIZE != F2FS_BLKSIZE) {
4440 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4441 PAGE_SIZE, F2FS_BLKSIZE);
4445 err = init_inodecache();
4448 err = f2fs_create_node_manager_caches();
4450 goto free_inodecache;
4451 err = f2fs_create_segment_manager_caches();
4453 goto free_node_manager_caches;
4454 err = f2fs_create_checkpoint_caches();
4456 goto free_segment_manager_caches;
4457 err = f2fs_create_recovery_cache();
4459 goto free_checkpoint_caches;
4460 err = f2fs_create_extent_cache();
4462 goto free_recovery_cache;
4463 err = f2fs_create_garbage_collection_cache();
4465 goto free_extent_cache;
4466 err = f2fs_init_sysfs();
4468 goto free_garbage_collection_cache;
4469 err = register_shrinker(&f2fs_shrinker_info);
4472 err = register_filesystem(&f2fs_fs_type);
4475 f2fs_create_root_stats();
4476 err = f2fs_init_post_read_processing();
4478 goto free_root_stats;
4479 err = f2fs_init_bio_entry_cache();
4481 goto free_post_read;
4482 err = f2fs_init_bioset();
4484 goto free_bio_enrty_cache;
4485 err = f2fs_init_compress_mempool();
4488 err = f2fs_init_compress_cache();
4490 goto free_compress_mempool;
4491 err = f2fs_create_casefold_cache();
4493 goto free_compress_cache;
4495 free_compress_cache:
4496 f2fs_destroy_compress_cache();
4497 free_compress_mempool:
4498 f2fs_destroy_compress_mempool();
4500 f2fs_destroy_bioset();
4501 free_bio_enrty_cache:
4502 f2fs_destroy_bio_entry_cache();
4504 f2fs_destroy_post_read_processing();
4506 f2fs_destroy_root_stats();
4507 unregister_filesystem(&f2fs_fs_type);
4509 unregister_shrinker(&f2fs_shrinker_info);
4512 free_garbage_collection_cache:
4513 f2fs_destroy_garbage_collection_cache();
4515 f2fs_destroy_extent_cache();
4516 free_recovery_cache:
4517 f2fs_destroy_recovery_cache();
4518 free_checkpoint_caches:
4519 f2fs_destroy_checkpoint_caches();
4520 free_segment_manager_caches:
4521 f2fs_destroy_segment_manager_caches();
4522 free_node_manager_caches:
4523 f2fs_destroy_node_manager_caches();
4525 destroy_inodecache();
4530 static void __exit exit_f2fs_fs(void)
4532 f2fs_destroy_casefold_cache();
4533 f2fs_destroy_compress_cache();
4534 f2fs_destroy_compress_mempool();
4535 f2fs_destroy_bioset();
4536 f2fs_destroy_bio_entry_cache();
4537 f2fs_destroy_post_read_processing();
4538 f2fs_destroy_root_stats();
4539 unregister_filesystem(&f2fs_fs_type);
4540 unregister_shrinker(&f2fs_shrinker_info);
4542 f2fs_destroy_garbage_collection_cache();
4543 f2fs_destroy_extent_cache();
4544 f2fs_destroy_recovery_cache();
4545 f2fs_destroy_checkpoint_caches();
4546 f2fs_destroy_segment_manager_caches();
4547 f2fs_destroy_node_manager_caches();
4548 destroy_inodecache();
4551 module_init(init_f2fs_fs)
4552 module_exit(exit_f2fs_fs)
4554 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4555 MODULE_DESCRIPTION("Flash Friendly File System");
4556 MODULE_LICENSE("GPL");
4557 MODULE_SOFTDEP("pre: crc32");