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/sched/mm.h>
12 #include <linux/statfs.h>
13 #include <linux/buffer_head.h>
14 #include <linux/kthread.h>
15 #include <linux/parser.h>
16 #include <linux/mount.h>
17 #include <linux/seq_file.h>
18 #include <linux/proc_fs.h>
19 #include <linux/random.h>
20 #include <linux/exportfs.h>
21 #include <linux/blkdev.h>
22 #include <linux/quotaops.h>
23 #include <linux/f2fs_fs.h>
24 #include <linux/sysfs.h>
25 #include <linux/quota.h>
26 #include <linux/unicode.h>
27 #include <linux/part_stat.h>
28 #include <linux/zstd.h>
29 #include <linux/lz4.h>
38 #define CREATE_TRACE_POINTS
39 #include <trace/events/f2fs.h>
41 static struct kmem_cache *f2fs_inode_cachep;
43 #ifdef CONFIG_F2FS_FAULT_INJECTION
45 const char *f2fs_fault_name[FAULT_MAX] = {
46 [FAULT_KMALLOC] = "kmalloc",
47 [FAULT_KVMALLOC] = "kvmalloc",
48 [FAULT_PAGE_ALLOC] = "page alloc",
49 [FAULT_PAGE_GET] = "page get",
50 [FAULT_ALLOC_NID] = "alloc nid",
51 [FAULT_ORPHAN] = "orphan",
52 [FAULT_BLOCK] = "no more block",
53 [FAULT_DIR_DEPTH] = "too big dir depth",
54 [FAULT_EVICT_INODE] = "evict_inode fail",
55 [FAULT_TRUNCATE] = "truncate fail",
56 [FAULT_READ_IO] = "read IO error",
57 [FAULT_CHECKPOINT] = "checkpoint error",
58 [FAULT_DISCARD] = "discard error",
59 [FAULT_WRITE_IO] = "write IO error",
60 [FAULT_SLAB_ALLOC] = "slab alloc",
61 [FAULT_DQUOT_INIT] = "dquot initialize",
64 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
67 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
70 atomic_set(&ffi->inject_ops, 0);
71 ffi->inject_rate = rate;
75 ffi->inject_type = type;
78 memset(ffi, 0, sizeof(struct f2fs_fault_info));
82 /* f2fs-wide shrinker description */
83 static struct shrinker f2fs_shrinker_info = {
84 .scan_objects = f2fs_shrink_scan,
85 .count_objects = f2fs_shrink_count,
86 .seeks = DEFAULT_SEEKS,
91 Opt_disable_roll_forward,
102 Opt_disable_ext_identify,
105 Opt_inline_xattr_size,
143 Opt_test_dummy_encryption,
145 Opt_checkpoint_disable,
146 Opt_checkpoint_disable_cap,
147 Opt_checkpoint_disable_cap_perc,
148 Opt_checkpoint_enable,
149 Opt_checkpoint_merge,
150 Opt_nocheckpoint_merge,
151 Opt_compress_algorithm,
152 Opt_compress_log_size,
153 Opt_compress_extension,
154 Opt_nocompress_extension,
165 static match_table_t f2fs_tokens = {
166 {Opt_gc_background, "background_gc=%s"},
167 {Opt_disable_roll_forward, "disable_roll_forward"},
168 {Opt_norecovery, "norecovery"},
169 {Opt_discard, "discard"},
170 {Opt_nodiscard, "nodiscard"},
171 {Opt_noheap, "no_heap"},
173 {Opt_user_xattr, "user_xattr"},
174 {Opt_nouser_xattr, "nouser_xattr"},
176 {Opt_noacl, "noacl"},
177 {Opt_active_logs, "active_logs=%u"},
178 {Opt_disable_ext_identify, "disable_ext_identify"},
179 {Opt_inline_xattr, "inline_xattr"},
180 {Opt_noinline_xattr, "noinline_xattr"},
181 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
182 {Opt_inline_data, "inline_data"},
183 {Opt_inline_dentry, "inline_dentry"},
184 {Opt_noinline_dentry, "noinline_dentry"},
185 {Opt_flush_merge, "flush_merge"},
186 {Opt_noflush_merge, "noflush_merge"},
187 {Opt_nobarrier, "nobarrier"},
188 {Opt_fastboot, "fastboot"},
189 {Opt_extent_cache, "extent_cache"},
190 {Opt_noextent_cache, "noextent_cache"},
191 {Opt_noinline_data, "noinline_data"},
192 {Opt_data_flush, "data_flush"},
193 {Opt_reserve_root, "reserve_root=%u"},
194 {Opt_resgid, "resgid=%u"},
195 {Opt_resuid, "resuid=%u"},
196 {Opt_mode, "mode=%s"},
197 {Opt_io_size_bits, "io_bits=%u"},
198 {Opt_fault_injection, "fault_injection=%u"},
199 {Opt_fault_type, "fault_type=%u"},
200 {Opt_lazytime, "lazytime"},
201 {Opt_nolazytime, "nolazytime"},
202 {Opt_quota, "quota"},
203 {Opt_noquota, "noquota"},
204 {Opt_usrquota, "usrquota"},
205 {Opt_grpquota, "grpquota"},
206 {Opt_prjquota, "prjquota"},
207 {Opt_usrjquota, "usrjquota=%s"},
208 {Opt_grpjquota, "grpjquota=%s"},
209 {Opt_prjjquota, "prjjquota=%s"},
210 {Opt_offusrjquota, "usrjquota="},
211 {Opt_offgrpjquota, "grpjquota="},
212 {Opt_offprjjquota, "prjjquota="},
213 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
214 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
215 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
216 {Opt_whint, "whint_mode=%s"},
217 {Opt_alloc, "alloc_mode=%s"},
218 {Opt_fsync, "fsync_mode=%s"},
219 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
220 {Opt_test_dummy_encryption, "test_dummy_encryption"},
221 {Opt_inlinecrypt, "inlinecrypt"},
222 {Opt_checkpoint_disable, "checkpoint=disable"},
223 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
224 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
225 {Opt_checkpoint_enable, "checkpoint=enable"},
226 {Opt_checkpoint_merge, "checkpoint_merge"},
227 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
228 {Opt_compress_algorithm, "compress_algorithm=%s"},
229 {Opt_compress_log_size, "compress_log_size=%u"},
230 {Opt_compress_extension, "compress_extension=%s"},
231 {Opt_nocompress_extension, "nocompress_extension=%s"},
232 {Opt_compress_chksum, "compress_chksum"},
233 {Opt_compress_mode, "compress_mode=%s"},
234 {Opt_compress_cache, "compress_cache"},
236 {Opt_gc_merge, "gc_merge"},
237 {Opt_nogc_merge, "nogc_merge"},
238 {Opt_discard_unit, "discard_unit=%s"},
242 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
244 struct va_format vaf;
250 level = printk_get_level(fmt);
251 vaf.fmt = printk_skip_level(fmt);
253 printk("%c%cF2FS-fs (%s): %pV\n",
254 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
259 #if IS_ENABLED(CONFIG_UNICODE)
260 static const struct f2fs_sb_encodings {
263 unsigned int version;
264 } f2fs_sb_encoding_map[] = {
265 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
268 static const struct f2fs_sb_encodings *
269 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
271 __u16 magic = le16_to_cpu(sb->s_encoding);
274 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
275 if (magic == f2fs_sb_encoding_map[i].magic)
276 return &f2fs_sb_encoding_map[i];
281 struct kmem_cache *f2fs_cf_name_slab;
282 static int __init f2fs_create_casefold_cache(void)
284 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
286 if (!f2fs_cf_name_slab)
291 static void f2fs_destroy_casefold_cache(void)
293 kmem_cache_destroy(f2fs_cf_name_slab);
296 static int __init f2fs_create_casefold_cache(void) { return 0; }
297 static void f2fs_destroy_casefold_cache(void) { }
300 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
302 block_t limit = min((sbi->user_block_count << 1) / 1000,
303 sbi->user_block_count - sbi->reserved_blocks);
306 if (test_opt(sbi, RESERVE_ROOT) &&
307 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
308 F2FS_OPTION(sbi).root_reserved_blocks = limit;
309 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
310 F2FS_OPTION(sbi).root_reserved_blocks);
312 if (!test_opt(sbi, RESERVE_ROOT) &&
313 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
314 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
315 !gid_eq(F2FS_OPTION(sbi).s_resgid,
316 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
317 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
318 from_kuid_munged(&init_user_ns,
319 F2FS_OPTION(sbi).s_resuid),
320 from_kgid_munged(&init_user_ns,
321 F2FS_OPTION(sbi).s_resgid));
324 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
326 if (!F2FS_OPTION(sbi).unusable_cap_perc)
329 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
330 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
332 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
333 F2FS_OPTION(sbi).unusable_cap_perc;
335 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
336 F2FS_OPTION(sbi).unusable_cap,
337 F2FS_OPTION(sbi).unusable_cap_perc);
340 static void init_once(void *foo)
342 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
344 inode_init_once(&fi->vfs_inode);
348 static const char * const quotatypes[] = INITQFNAMES;
349 #define QTYPE2NAME(t) (quotatypes[t])
350 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
353 struct f2fs_sb_info *sbi = F2FS_SB(sb);
357 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
358 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
361 if (f2fs_sb_has_quota_ino(sbi)) {
362 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
366 qname = match_strdup(args);
368 f2fs_err(sbi, "Not enough memory for storing quotafile name");
371 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
372 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
375 f2fs_err(sbi, "%s quota file already specified",
379 if (strchr(qname, '/')) {
380 f2fs_err(sbi, "quotafile must be on filesystem root");
383 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
391 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
393 struct f2fs_sb_info *sbi = F2FS_SB(sb);
395 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
396 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
399 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
400 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
404 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
407 * We do the test below only for project quotas. 'usrquota' and
408 * 'grpquota' mount options are allowed even without quota feature
409 * to support legacy quotas in quota files.
411 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
412 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
415 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
416 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
417 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
418 if (test_opt(sbi, USRQUOTA) &&
419 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
420 clear_opt(sbi, USRQUOTA);
422 if (test_opt(sbi, GRPQUOTA) &&
423 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
424 clear_opt(sbi, GRPQUOTA);
426 if (test_opt(sbi, PRJQUOTA) &&
427 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
428 clear_opt(sbi, PRJQUOTA);
430 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
431 test_opt(sbi, PRJQUOTA)) {
432 f2fs_err(sbi, "old and new quota format mixing");
436 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
437 f2fs_err(sbi, "journaled quota format not specified");
442 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
443 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
444 F2FS_OPTION(sbi).s_jquota_fmt = 0;
450 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
452 const substring_t *arg,
455 struct f2fs_sb_info *sbi = F2FS_SB(sb);
456 #ifdef CONFIG_FS_ENCRYPTION
459 if (!f2fs_sb_has_encrypt(sbi)) {
460 f2fs_err(sbi, "Encrypt feature is off");
465 * This mount option is just for testing, and it's not worthwhile to
466 * implement the extra complexity (e.g. RCU protection) that would be
467 * needed to allow it to be set or changed during remount. We do allow
468 * it to be specified during remount, but only if there is no change.
470 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
471 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
474 err = fscrypt_set_test_dummy_encryption(
475 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
479 "Can't change test_dummy_encryption on remount");
480 else if (err == -EINVAL)
481 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
484 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
488 f2fs_warn(sbi, "Test dummy encryption mode enabled");
490 f2fs_warn(sbi, "Test dummy encryption mount option ignored");
495 #ifdef CONFIG_F2FS_FS_COMPRESSION
497 * 1. The same extension name cannot not appear in both compress and non-compress extension
499 * 2. If the compress extension specifies all files, the types specified by the non-compress
500 * extension will be treated as special cases and will not be compressed.
501 * 3. Don't allow the non-compress extension specifies all files.
503 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
505 unsigned char (*ext)[F2FS_EXTENSION_LEN];
506 unsigned char (*noext)[F2FS_EXTENSION_LEN];
507 int ext_cnt, noext_cnt, index = 0, no_index = 0;
509 ext = F2FS_OPTION(sbi).extensions;
510 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
511 noext = F2FS_OPTION(sbi).noextensions;
512 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
517 for (no_index = 0; no_index < noext_cnt; no_index++) {
518 if (!strcasecmp("*", noext[no_index])) {
519 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
522 for (index = 0; index < ext_cnt; index++) {
523 if (!strcasecmp(ext[index], noext[no_index])) {
524 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
533 #ifdef CONFIG_F2FS_FS_LZ4
534 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
536 #ifdef CONFIG_F2FS_FS_LZ4HC
540 if (strlen(str) == 3) {
541 F2FS_OPTION(sbi).compress_level = 0;
545 #ifdef CONFIG_F2FS_FS_LZ4HC
549 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
552 if (kstrtouint(str + 1, 10, &level))
555 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
556 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
560 F2FS_OPTION(sbi).compress_level = level;
563 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
569 #ifdef CONFIG_F2FS_FS_ZSTD
570 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
575 if (strlen(str) == len) {
576 F2FS_OPTION(sbi).compress_level = 0;
583 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
586 if (kstrtouint(str + 1, 10, &level))
589 if (!level || level > zstd_max_clevel()) {
590 f2fs_info(sbi, "invalid zstd compress level: %d", level);
594 F2FS_OPTION(sbi).compress_level = level;
600 static int parse_options(struct super_block *sb, char *options, bool is_remount)
602 struct f2fs_sb_info *sbi = F2FS_SB(sb);
603 substring_t args[MAX_OPT_ARGS];
604 #ifdef CONFIG_F2FS_FS_COMPRESSION
605 unsigned char (*ext)[F2FS_EXTENSION_LEN];
606 unsigned char (*noext)[F2FS_EXTENSION_LEN];
607 int ext_cnt, noext_cnt;
618 while ((p = strsep(&options, ",")) != NULL) {
624 * Initialize args struct so we know whether arg was
625 * found; some options take optional arguments.
627 args[0].to = args[0].from = NULL;
628 token = match_token(p, f2fs_tokens, args);
631 case Opt_gc_background:
632 name = match_strdup(&args[0]);
636 if (!strcmp(name, "on")) {
637 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
638 } else if (!strcmp(name, "off")) {
639 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
640 } else if (!strcmp(name, "sync")) {
641 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
648 case Opt_disable_roll_forward:
649 set_opt(sbi, DISABLE_ROLL_FORWARD);
652 /* this option mounts f2fs with ro */
653 set_opt(sbi, NORECOVERY);
654 if (!f2fs_readonly(sb))
658 if (!f2fs_hw_support_discard(sbi)) {
659 f2fs_warn(sbi, "device does not support discard");
662 set_opt(sbi, DISCARD);
665 if (f2fs_hw_should_discard(sbi)) {
666 f2fs_warn(sbi, "discard is required for zoned block devices");
669 clear_opt(sbi, DISCARD);
672 set_opt(sbi, NOHEAP);
675 clear_opt(sbi, NOHEAP);
677 #ifdef CONFIG_F2FS_FS_XATTR
679 set_opt(sbi, XATTR_USER);
681 case Opt_nouser_xattr:
682 clear_opt(sbi, XATTR_USER);
684 case Opt_inline_xattr:
685 set_opt(sbi, INLINE_XATTR);
687 case Opt_noinline_xattr:
688 clear_opt(sbi, INLINE_XATTR);
690 case Opt_inline_xattr_size:
691 if (args->from && match_int(args, &arg))
693 set_opt(sbi, INLINE_XATTR_SIZE);
694 F2FS_OPTION(sbi).inline_xattr_size = arg;
698 f2fs_info(sbi, "user_xattr options not supported");
700 case Opt_nouser_xattr:
701 f2fs_info(sbi, "nouser_xattr options not supported");
703 case Opt_inline_xattr:
704 f2fs_info(sbi, "inline_xattr options not supported");
706 case Opt_noinline_xattr:
707 f2fs_info(sbi, "noinline_xattr options not supported");
710 #ifdef CONFIG_F2FS_FS_POSIX_ACL
712 set_opt(sbi, POSIX_ACL);
715 clear_opt(sbi, POSIX_ACL);
719 f2fs_info(sbi, "acl options not supported");
722 f2fs_info(sbi, "noacl options not supported");
725 case Opt_active_logs:
726 if (args->from && match_int(args, &arg))
728 if (arg != 2 && arg != 4 &&
729 arg != NR_CURSEG_PERSIST_TYPE)
731 F2FS_OPTION(sbi).active_logs = arg;
733 case Opt_disable_ext_identify:
734 set_opt(sbi, DISABLE_EXT_IDENTIFY);
736 case Opt_inline_data:
737 set_opt(sbi, INLINE_DATA);
739 case Opt_inline_dentry:
740 set_opt(sbi, INLINE_DENTRY);
742 case Opt_noinline_dentry:
743 clear_opt(sbi, INLINE_DENTRY);
745 case Opt_flush_merge:
746 set_opt(sbi, FLUSH_MERGE);
748 case Opt_noflush_merge:
749 clear_opt(sbi, FLUSH_MERGE);
752 set_opt(sbi, NOBARRIER);
755 set_opt(sbi, FASTBOOT);
757 case Opt_extent_cache:
758 set_opt(sbi, EXTENT_CACHE);
760 case Opt_noextent_cache:
761 clear_opt(sbi, EXTENT_CACHE);
763 case Opt_noinline_data:
764 clear_opt(sbi, INLINE_DATA);
767 set_opt(sbi, DATA_FLUSH);
769 case Opt_reserve_root:
770 if (args->from && match_int(args, &arg))
772 if (test_opt(sbi, RESERVE_ROOT)) {
773 f2fs_info(sbi, "Preserve previous reserve_root=%u",
774 F2FS_OPTION(sbi).root_reserved_blocks);
776 F2FS_OPTION(sbi).root_reserved_blocks = arg;
777 set_opt(sbi, RESERVE_ROOT);
781 if (args->from && match_int(args, &arg))
783 uid = make_kuid(current_user_ns(), arg);
784 if (!uid_valid(uid)) {
785 f2fs_err(sbi, "Invalid uid value %d", arg);
788 F2FS_OPTION(sbi).s_resuid = uid;
791 if (args->from && match_int(args, &arg))
793 gid = make_kgid(current_user_ns(), arg);
794 if (!gid_valid(gid)) {
795 f2fs_err(sbi, "Invalid gid value %d", arg);
798 F2FS_OPTION(sbi).s_resgid = gid;
801 name = match_strdup(&args[0]);
805 if (!strcmp(name, "adaptive")) {
806 if (f2fs_sb_has_blkzoned(sbi)) {
807 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
811 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
812 } else if (!strcmp(name, "lfs")) {
813 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
814 } else if (!strcmp(name, "fragment:segment")) {
815 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
816 } else if (!strcmp(name, "fragment:block")) {
817 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
824 case Opt_io_size_bits:
825 if (args->from && match_int(args, &arg))
827 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
828 f2fs_warn(sbi, "Not support %d, larger than %d",
829 1 << arg, BIO_MAX_VECS);
832 F2FS_OPTION(sbi).write_io_size_bits = arg;
834 #ifdef CONFIG_F2FS_FAULT_INJECTION
835 case Opt_fault_injection:
836 if (args->from && match_int(args, &arg))
838 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
839 set_opt(sbi, FAULT_INJECTION);
843 if (args->from && match_int(args, &arg))
845 f2fs_build_fault_attr(sbi, 0, arg);
846 set_opt(sbi, FAULT_INJECTION);
849 case Opt_fault_injection:
850 f2fs_info(sbi, "fault_injection options not supported");
854 f2fs_info(sbi, "fault_type options not supported");
858 sb->s_flags |= SB_LAZYTIME;
861 sb->s_flags &= ~SB_LAZYTIME;
866 set_opt(sbi, USRQUOTA);
869 set_opt(sbi, GRPQUOTA);
872 set_opt(sbi, PRJQUOTA);
875 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
880 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
885 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
889 case Opt_offusrjquota:
890 ret = f2fs_clear_qf_name(sb, USRQUOTA);
894 case Opt_offgrpjquota:
895 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
899 case Opt_offprjjquota:
900 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
904 case Opt_jqfmt_vfsold:
905 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
907 case Opt_jqfmt_vfsv0:
908 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
910 case Opt_jqfmt_vfsv1:
911 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
914 clear_opt(sbi, QUOTA);
915 clear_opt(sbi, USRQUOTA);
916 clear_opt(sbi, GRPQUOTA);
917 clear_opt(sbi, PRJQUOTA);
927 case Opt_offusrjquota:
928 case Opt_offgrpjquota:
929 case Opt_offprjjquota:
930 case Opt_jqfmt_vfsold:
931 case Opt_jqfmt_vfsv0:
932 case Opt_jqfmt_vfsv1:
934 f2fs_info(sbi, "quota operations not supported");
938 name = match_strdup(&args[0]);
941 if (!strcmp(name, "user-based")) {
942 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
943 } else if (!strcmp(name, "off")) {
944 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
945 } else if (!strcmp(name, "fs-based")) {
946 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
954 name = match_strdup(&args[0]);
958 if (!strcmp(name, "default")) {
959 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
960 } else if (!strcmp(name, "reuse")) {
961 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
969 name = match_strdup(&args[0]);
972 if (!strcmp(name, "posix")) {
973 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
974 } else if (!strcmp(name, "strict")) {
975 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
976 } else if (!strcmp(name, "nobarrier")) {
977 F2FS_OPTION(sbi).fsync_mode =
978 FSYNC_MODE_NOBARRIER;
985 case Opt_test_dummy_encryption:
986 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
991 case Opt_inlinecrypt:
992 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
993 sb->s_flags |= SB_INLINECRYPT;
995 f2fs_info(sbi, "inline encryption not supported");
998 case Opt_checkpoint_disable_cap_perc:
999 if (args->from && match_int(args, &arg))
1001 if (arg < 0 || arg > 100)
1003 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1004 set_opt(sbi, DISABLE_CHECKPOINT);
1006 case Opt_checkpoint_disable_cap:
1007 if (args->from && match_int(args, &arg))
1009 F2FS_OPTION(sbi).unusable_cap = arg;
1010 set_opt(sbi, DISABLE_CHECKPOINT);
1012 case Opt_checkpoint_disable:
1013 set_opt(sbi, DISABLE_CHECKPOINT);
1015 case Opt_checkpoint_enable:
1016 clear_opt(sbi, DISABLE_CHECKPOINT);
1018 case Opt_checkpoint_merge:
1019 set_opt(sbi, MERGE_CHECKPOINT);
1021 case Opt_nocheckpoint_merge:
1022 clear_opt(sbi, MERGE_CHECKPOINT);
1024 #ifdef CONFIG_F2FS_FS_COMPRESSION
1025 case Opt_compress_algorithm:
1026 if (!f2fs_sb_has_compression(sbi)) {
1027 f2fs_info(sbi, "Image doesn't support compression");
1030 name = match_strdup(&args[0]);
1033 if (!strcmp(name, "lzo")) {
1034 #ifdef CONFIG_F2FS_FS_LZO
1035 F2FS_OPTION(sbi).compress_level = 0;
1036 F2FS_OPTION(sbi).compress_algorithm =
1039 f2fs_info(sbi, "kernel doesn't support lzo compression");
1041 } else if (!strncmp(name, "lz4", 3)) {
1042 #ifdef CONFIG_F2FS_FS_LZ4
1043 ret = f2fs_set_lz4hc_level(sbi, name);
1048 F2FS_OPTION(sbi).compress_algorithm =
1051 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1053 } else if (!strncmp(name, "zstd", 4)) {
1054 #ifdef CONFIG_F2FS_FS_ZSTD
1055 ret = f2fs_set_zstd_level(sbi, name);
1060 F2FS_OPTION(sbi).compress_algorithm =
1063 f2fs_info(sbi, "kernel doesn't support zstd compression");
1065 } else if (!strcmp(name, "lzo-rle")) {
1066 #ifdef CONFIG_F2FS_FS_LZORLE
1067 F2FS_OPTION(sbi).compress_level = 0;
1068 F2FS_OPTION(sbi).compress_algorithm =
1071 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1079 case Opt_compress_log_size:
1080 if (!f2fs_sb_has_compression(sbi)) {
1081 f2fs_info(sbi, "Image doesn't support compression");
1084 if (args->from && match_int(args, &arg))
1086 if (arg < MIN_COMPRESS_LOG_SIZE ||
1087 arg > MAX_COMPRESS_LOG_SIZE) {
1089 "Compress cluster log size is out of range");
1092 F2FS_OPTION(sbi).compress_log_size = arg;
1094 case Opt_compress_extension:
1095 if (!f2fs_sb_has_compression(sbi)) {
1096 f2fs_info(sbi, "Image doesn't support compression");
1099 name = match_strdup(&args[0]);
1103 ext = F2FS_OPTION(sbi).extensions;
1104 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1106 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1107 ext_cnt >= COMPRESS_EXT_NUM) {
1109 "invalid extension length/number");
1114 strcpy(ext[ext_cnt], name);
1115 F2FS_OPTION(sbi).compress_ext_cnt++;
1118 case Opt_nocompress_extension:
1119 if (!f2fs_sb_has_compression(sbi)) {
1120 f2fs_info(sbi, "Image doesn't support compression");
1123 name = match_strdup(&args[0]);
1127 noext = F2FS_OPTION(sbi).noextensions;
1128 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1130 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1131 noext_cnt >= COMPRESS_EXT_NUM) {
1133 "invalid extension length/number");
1138 strcpy(noext[noext_cnt], name);
1139 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1142 case Opt_compress_chksum:
1143 F2FS_OPTION(sbi).compress_chksum = true;
1145 case Opt_compress_mode:
1146 name = match_strdup(&args[0]);
1149 if (!strcmp(name, "fs")) {
1150 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1151 } else if (!strcmp(name, "user")) {
1152 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1159 case Opt_compress_cache:
1160 set_opt(sbi, COMPRESS_CACHE);
1163 case Opt_compress_algorithm:
1164 case Opt_compress_log_size:
1165 case Opt_compress_extension:
1166 case Opt_nocompress_extension:
1167 case Opt_compress_chksum:
1168 case Opt_compress_mode:
1169 case Opt_compress_cache:
1170 f2fs_info(sbi, "compression options not supported");
1177 set_opt(sbi, GC_MERGE);
1179 case Opt_nogc_merge:
1180 clear_opt(sbi, GC_MERGE);
1182 case Opt_discard_unit:
1183 name = match_strdup(&args[0]);
1186 if (!strcmp(name, "block")) {
1187 F2FS_OPTION(sbi).discard_unit =
1189 } else if (!strcmp(name, "segment")) {
1190 F2FS_OPTION(sbi).discard_unit =
1191 DISCARD_UNIT_SEGMENT;
1192 } else if (!strcmp(name, "section")) {
1193 F2FS_OPTION(sbi).discard_unit =
1194 DISCARD_UNIT_SECTION;
1202 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1209 if (f2fs_check_quota_options(sbi))
1212 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1213 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1216 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1217 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1221 #if !IS_ENABLED(CONFIG_UNICODE)
1222 if (f2fs_sb_has_casefold(sbi)) {
1224 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1229 * The BLKZONED feature indicates that the drive was formatted with
1230 * zone alignment optimization. This is optional for host-aware
1231 * devices, but mandatory for host-managed zoned block devices.
1233 #ifndef CONFIG_BLK_DEV_ZONED
1234 if (f2fs_sb_has_blkzoned(sbi)) {
1235 f2fs_err(sbi, "Zoned block device support is not enabled");
1239 if (f2fs_sb_has_blkzoned(sbi)) {
1240 if (F2FS_OPTION(sbi).discard_unit !=
1241 DISCARD_UNIT_SECTION) {
1242 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1243 F2FS_OPTION(sbi).discard_unit =
1244 DISCARD_UNIT_SECTION;
1248 #ifdef CONFIG_F2FS_FS_COMPRESSION
1249 if (f2fs_test_compress_extension(sbi)) {
1250 f2fs_err(sbi, "invalid compress or nocompress extension");
1255 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1256 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1257 F2FS_IO_SIZE_KB(sbi));
1261 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1262 int min_size, max_size;
1264 if (!f2fs_sb_has_extra_attr(sbi) ||
1265 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1266 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1269 if (!test_opt(sbi, INLINE_XATTR)) {
1270 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1274 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1275 max_size = MAX_INLINE_XATTR_SIZE;
1277 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1278 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1279 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1280 min_size, max_size);
1285 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1286 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1290 /* Not pass down write hints if the number of active logs is lesser
1291 * than NR_CURSEG_PERSIST_TYPE.
1293 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE)
1294 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1296 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1297 f2fs_err(sbi, "Allow to mount readonly mode only");
1303 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1305 struct f2fs_inode_info *fi;
1307 fi = f2fs_kmem_cache_alloc(f2fs_inode_cachep,
1308 GFP_F2FS_ZERO, false, F2FS_SB(sb));
1312 init_once((void *) fi);
1314 /* Initialize f2fs-specific inode info */
1315 atomic_set(&fi->dirty_pages, 0);
1316 atomic_set(&fi->i_compr_blocks, 0);
1317 init_rwsem(&fi->i_sem);
1318 spin_lock_init(&fi->i_size_lock);
1319 INIT_LIST_HEAD(&fi->dirty_list);
1320 INIT_LIST_HEAD(&fi->gdirty_list);
1321 INIT_LIST_HEAD(&fi->inmem_ilist);
1322 INIT_LIST_HEAD(&fi->inmem_pages);
1323 mutex_init(&fi->inmem_lock);
1324 init_rwsem(&fi->i_gc_rwsem[READ]);
1325 init_rwsem(&fi->i_gc_rwsem[WRITE]);
1326 init_rwsem(&fi->i_xattr_sem);
1328 /* Will be used by directory only */
1329 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1331 return &fi->vfs_inode;
1334 static int f2fs_drop_inode(struct inode *inode)
1336 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1340 * during filesystem shutdown, if checkpoint is disabled,
1341 * drop useless meta/node dirty pages.
1343 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1344 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1345 inode->i_ino == F2FS_META_INO(sbi)) {
1346 trace_f2fs_drop_inode(inode, 1);
1352 * This is to avoid a deadlock condition like below.
1353 * writeback_single_inode(inode)
1354 * - f2fs_write_data_page
1355 * - f2fs_gc -> iput -> evict
1356 * - inode_wait_for_writeback(inode)
1358 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1359 if (!inode->i_nlink && !is_bad_inode(inode)) {
1360 /* to avoid evict_inode call simultaneously */
1361 atomic_inc(&inode->i_count);
1362 spin_unlock(&inode->i_lock);
1364 /* some remained atomic pages should discarded */
1365 if (f2fs_is_atomic_file(inode))
1366 f2fs_drop_inmem_pages(inode);
1368 /* should remain fi->extent_tree for writepage */
1369 f2fs_destroy_extent_node(inode);
1371 sb_start_intwrite(inode->i_sb);
1372 f2fs_i_size_write(inode, 0);
1374 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1375 inode, NULL, 0, DATA);
1376 truncate_inode_pages_final(inode->i_mapping);
1378 if (F2FS_HAS_BLOCKS(inode))
1379 f2fs_truncate(inode);
1381 sb_end_intwrite(inode->i_sb);
1383 spin_lock(&inode->i_lock);
1384 atomic_dec(&inode->i_count);
1386 trace_f2fs_drop_inode(inode, 0);
1389 ret = generic_drop_inode(inode);
1391 ret = fscrypt_drop_inode(inode);
1392 trace_f2fs_drop_inode(inode, ret);
1396 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1398 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1401 spin_lock(&sbi->inode_lock[DIRTY_META]);
1402 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1405 set_inode_flag(inode, FI_DIRTY_INODE);
1406 stat_inc_dirty_inode(sbi, DIRTY_META);
1408 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1409 list_add_tail(&F2FS_I(inode)->gdirty_list,
1410 &sbi->inode_list[DIRTY_META]);
1411 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1413 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1417 void f2fs_inode_synced(struct inode *inode)
1419 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1421 spin_lock(&sbi->inode_lock[DIRTY_META]);
1422 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1423 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1426 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1427 list_del_init(&F2FS_I(inode)->gdirty_list);
1428 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1430 clear_inode_flag(inode, FI_DIRTY_INODE);
1431 clear_inode_flag(inode, FI_AUTO_RECOVER);
1432 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1433 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1437 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1439 * We should call set_dirty_inode to write the dirty inode through write_inode.
1441 static void f2fs_dirty_inode(struct inode *inode, int flags)
1443 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1445 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1446 inode->i_ino == F2FS_META_INO(sbi))
1449 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1450 clear_inode_flag(inode, FI_AUTO_RECOVER);
1452 f2fs_inode_dirtied(inode, false);
1455 static void f2fs_free_inode(struct inode *inode)
1457 fscrypt_free_inode(inode);
1458 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1461 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1463 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1464 percpu_counter_destroy(&sbi->total_valid_inode_count);
1467 static void destroy_device_list(struct f2fs_sb_info *sbi)
1471 for (i = 0; i < sbi->s_ndevs; i++) {
1472 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1473 #ifdef CONFIG_BLK_DEV_ZONED
1474 kvfree(FDEV(i).blkz_seq);
1475 kfree(FDEV(i).zone_capacity_blocks);
1481 static void f2fs_put_super(struct super_block *sb)
1483 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1487 /* unregister procfs/sysfs entries in advance to avoid race case */
1488 f2fs_unregister_sysfs(sbi);
1490 f2fs_quota_off_umount(sb);
1492 /* prevent remaining shrinker jobs */
1493 mutex_lock(&sbi->umount_mutex);
1496 * flush all issued checkpoints and stop checkpoint issue thread.
1497 * after then, all checkpoints should be done by each process context.
1499 f2fs_stop_ckpt_thread(sbi);
1502 * We don't need to do checkpoint when superblock is clean.
1503 * But, the previous checkpoint was not done by umount, it needs to do
1504 * clean checkpoint again.
1506 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1507 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1508 struct cp_control cpc = {
1509 .reason = CP_UMOUNT,
1511 f2fs_write_checkpoint(sbi, &cpc);
1514 /* be sure to wait for any on-going discard commands */
1515 dropped = f2fs_issue_discard_timeout(sbi);
1517 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1518 !sbi->discard_blks && !dropped) {
1519 struct cp_control cpc = {
1520 .reason = CP_UMOUNT | CP_TRIMMED,
1522 f2fs_write_checkpoint(sbi, &cpc);
1526 * normally superblock is clean, so we need to release this.
1527 * In addition, EIO will skip do checkpoint, we need this as well.
1529 f2fs_release_ino_entry(sbi, true);
1531 f2fs_leave_shrinker(sbi);
1532 mutex_unlock(&sbi->umount_mutex);
1534 /* our cp_error case, we can wait for any writeback page */
1535 f2fs_flush_merged_writes(sbi);
1537 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1539 f2fs_bug_on(sbi, sbi->fsync_node_num);
1541 f2fs_destroy_compress_inode(sbi);
1543 iput(sbi->node_inode);
1544 sbi->node_inode = NULL;
1546 iput(sbi->meta_inode);
1547 sbi->meta_inode = NULL;
1550 * iput() can update stat information, if f2fs_write_checkpoint()
1551 * above failed with error.
1553 f2fs_destroy_stats(sbi);
1555 /* destroy f2fs internal modules */
1556 f2fs_destroy_node_manager(sbi);
1557 f2fs_destroy_segment_manager(sbi);
1559 f2fs_destroy_post_read_wq(sbi);
1563 sb->s_fs_info = NULL;
1564 if (sbi->s_chksum_driver)
1565 crypto_free_shash(sbi->s_chksum_driver);
1566 kfree(sbi->raw_super);
1568 destroy_device_list(sbi);
1569 f2fs_destroy_page_array_cache(sbi);
1570 f2fs_destroy_xattr_caches(sbi);
1571 mempool_destroy(sbi->write_io_dummy);
1573 for (i = 0; i < MAXQUOTAS; i++)
1574 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1576 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1577 destroy_percpu_info(sbi);
1578 f2fs_destroy_iostat(sbi);
1579 for (i = 0; i < NR_PAGE_TYPE; i++)
1580 kvfree(sbi->write_io[i]);
1581 #if IS_ENABLED(CONFIG_UNICODE)
1582 utf8_unload(sb->s_encoding);
1587 int f2fs_sync_fs(struct super_block *sb, int sync)
1589 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1592 if (unlikely(f2fs_cp_error(sbi)))
1594 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1597 trace_f2fs_sync_fs(sb, sync);
1599 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1603 err = f2fs_issue_checkpoint(sbi);
1608 static int f2fs_freeze(struct super_block *sb)
1610 if (f2fs_readonly(sb))
1613 /* IO error happened before */
1614 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1617 /* must be clean, since sync_filesystem() was already called */
1618 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1621 /* ensure no checkpoint required */
1622 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list))
1627 static int f2fs_unfreeze(struct super_block *sb)
1633 static int f2fs_statfs_project(struct super_block *sb,
1634 kprojid_t projid, struct kstatfs *buf)
1637 struct dquot *dquot;
1641 qid = make_kqid_projid(projid);
1642 dquot = dqget(sb, qid);
1644 return PTR_ERR(dquot);
1645 spin_lock(&dquot->dq_dqb_lock);
1647 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1648 dquot->dq_dqb.dqb_bhardlimit);
1650 limit >>= sb->s_blocksize_bits;
1652 if (limit && buf->f_blocks > limit) {
1653 curblock = (dquot->dq_dqb.dqb_curspace +
1654 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1655 buf->f_blocks = limit;
1656 buf->f_bfree = buf->f_bavail =
1657 (buf->f_blocks > curblock) ?
1658 (buf->f_blocks - curblock) : 0;
1661 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1662 dquot->dq_dqb.dqb_ihardlimit);
1664 if (limit && buf->f_files > limit) {
1665 buf->f_files = limit;
1667 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1668 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1671 spin_unlock(&dquot->dq_dqb_lock);
1677 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1679 struct super_block *sb = dentry->d_sb;
1680 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1681 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1682 block_t total_count, user_block_count, start_count;
1683 u64 avail_node_count;
1685 total_count = le64_to_cpu(sbi->raw_super->block_count);
1686 user_block_count = sbi->user_block_count;
1687 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1688 buf->f_type = F2FS_SUPER_MAGIC;
1689 buf->f_bsize = sbi->blocksize;
1691 buf->f_blocks = total_count - start_count;
1692 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1693 sbi->current_reserved_blocks;
1695 spin_lock(&sbi->stat_lock);
1696 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1699 buf->f_bfree -= sbi->unusable_block_count;
1700 spin_unlock(&sbi->stat_lock);
1702 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1703 buf->f_bavail = buf->f_bfree -
1704 F2FS_OPTION(sbi).root_reserved_blocks;
1708 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1710 if (avail_node_count > user_block_count) {
1711 buf->f_files = user_block_count;
1712 buf->f_ffree = buf->f_bavail;
1714 buf->f_files = avail_node_count;
1715 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1719 buf->f_namelen = F2FS_NAME_LEN;
1720 buf->f_fsid = u64_to_fsid(id);
1723 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1724 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1725 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1731 static inline void f2fs_show_quota_options(struct seq_file *seq,
1732 struct super_block *sb)
1735 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1737 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1740 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1751 seq_printf(seq, ",jqfmt=%s", fmtname);
1754 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1755 seq_show_option(seq, "usrjquota",
1756 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1758 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1759 seq_show_option(seq, "grpjquota",
1760 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1762 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1763 seq_show_option(seq, "prjjquota",
1764 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1768 #ifdef CONFIG_F2FS_FS_COMPRESSION
1769 static inline void f2fs_show_compress_options(struct seq_file *seq,
1770 struct super_block *sb)
1772 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1776 if (!f2fs_sb_has_compression(sbi))
1779 switch (F2FS_OPTION(sbi).compress_algorithm) {
1789 case COMPRESS_LZORLE:
1790 algtype = "lzo-rle";
1793 seq_printf(seq, ",compress_algorithm=%s", algtype);
1795 if (F2FS_OPTION(sbi).compress_level)
1796 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1798 seq_printf(seq, ",compress_log_size=%u",
1799 F2FS_OPTION(sbi).compress_log_size);
1801 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1802 seq_printf(seq, ",compress_extension=%s",
1803 F2FS_OPTION(sbi).extensions[i]);
1806 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1807 seq_printf(seq, ",nocompress_extension=%s",
1808 F2FS_OPTION(sbi).noextensions[i]);
1811 if (F2FS_OPTION(sbi).compress_chksum)
1812 seq_puts(seq, ",compress_chksum");
1814 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1815 seq_printf(seq, ",compress_mode=%s", "fs");
1816 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1817 seq_printf(seq, ",compress_mode=%s", "user");
1819 if (test_opt(sbi, COMPRESS_CACHE))
1820 seq_puts(seq, ",compress_cache");
1824 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1826 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1828 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1829 seq_printf(seq, ",background_gc=%s", "sync");
1830 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1831 seq_printf(seq, ",background_gc=%s", "on");
1832 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1833 seq_printf(seq, ",background_gc=%s", "off");
1835 if (test_opt(sbi, GC_MERGE))
1836 seq_puts(seq, ",gc_merge");
1838 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1839 seq_puts(seq, ",disable_roll_forward");
1840 if (test_opt(sbi, NORECOVERY))
1841 seq_puts(seq, ",norecovery");
1842 if (test_opt(sbi, DISCARD))
1843 seq_puts(seq, ",discard");
1845 seq_puts(seq, ",nodiscard");
1846 if (test_opt(sbi, NOHEAP))
1847 seq_puts(seq, ",no_heap");
1849 seq_puts(seq, ",heap");
1850 #ifdef CONFIG_F2FS_FS_XATTR
1851 if (test_opt(sbi, XATTR_USER))
1852 seq_puts(seq, ",user_xattr");
1854 seq_puts(seq, ",nouser_xattr");
1855 if (test_opt(sbi, INLINE_XATTR))
1856 seq_puts(seq, ",inline_xattr");
1858 seq_puts(seq, ",noinline_xattr");
1859 if (test_opt(sbi, INLINE_XATTR_SIZE))
1860 seq_printf(seq, ",inline_xattr_size=%u",
1861 F2FS_OPTION(sbi).inline_xattr_size);
1863 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1864 if (test_opt(sbi, POSIX_ACL))
1865 seq_puts(seq, ",acl");
1867 seq_puts(seq, ",noacl");
1869 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1870 seq_puts(seq, ",disable_ext_identify");
1871 if (test_opt(sbi, INLINE_DATA))
1872 seq_puts(seq, ",inline_data");
1874 seq_puts(seq, ",noinline_data");
1875 if (test_opt(sbi, INLINE_DENTRY))
1876 seq_puts(seq, ",inline_dentry");
1878 seq_puts(seq, ",noinline_dentry");
1879 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1880 seq_puts(seq, ",flush_merge");
1881 if (test_opt(sbi, NOBARRIER))
1882 seq_puts(seq, ",nobarrier");
1883 if (test_opt(sbi, FASTBOOT))
1884 seq_puts(seq, ",fastboot");
1885 if (test_opt(sbi, EXTENT_CACHE))
1886 seq_puts(seq, ",extent_cache");
1888 seq_puts(seq, ",noextent_cache");
1889 if (test_opt(sbi, DATA_FLUSH))
1890 seq_puts(seq, ",data_flush");
1892 seq_puts(seq, ",mode=");
1893 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1894 seq_puts(seq, "adaptive");
1895 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1896 seq_puts(seq, "lfs");
1897 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1898 seq_puts(seq, "fragment:segment");
1899 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1900 seq_puts(seq, "fragment:block");
1901 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1902 if (test_opt(sbi, RESERVE_ROOT))
1903 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1904 F2FS_OPTION(sbi).root_reserved_blocks,
1905 from_kuid_munged(&init_user_ns,
1906 F2FS_OPTION(sbi).s_resuid),
1907 from_kgid_munged(&init_user_ns,
1908 F2FS_OPTION(sbi).s_resgid));
1909 if (F2FS_IO_SIZE_BITS(sbi))
1910 seq_printf(seq, ",io_bits=%u",
1911 F2FS_OPTION(sbi).write_io_size_bits);
1912 #ifdef CONFIG_F2FS_FAULT_INJECTION
1913 if (test_opt(sbi, FAULT_INJECTION)) {
1914 seq_printf(seq, ",fault_injection=%u",
1915 F2FS_OPTION(sbi).fault_info.inject_rate);
1916 seq_printf(seq, ",fault_type=%u",
1917 F2FS_OPTION(sbi).fault_info.inject_type);
1921 if (test_opt(sbi, QUOTA))
1922 seq_puts(seq, ",quota");
1923 if (test_opt(sbi, USRQUOTA))
1924 seq_puts(seq, ",usrquota");
1925 if (test_opt(sbi, GRPQUOTA))
1926 seq_puts(seq, ",grpquota");
1927 if (test_opt(sbi, PRJQUOTA))
1928 seq_puts(seq, ",prjquota");
1930 f2fs_show_quota_options(seq, sbi->sb);
1931 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1932 seq_printf(seq, ",whint_mode=%s", "user-based");
1933 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1934 seq_printf(seq, ",whint_mode=%s", "fs-based");
1936 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1938 if (sbi->sb->s_flags & SB_INLINECRYPT)
1939 seq_puts(seq, ",inlinecrypt");
1941 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1942 seq_printf(seq, ",alloc_mode=%s", "default");
1943 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1944 seq_printf(seq, ",alloc_mode=%s", "reuse");
1946 if (test_opt(sbi, DISABLE_CHECKPOINT))
1947 seq_printf(seq, ",checkpoint=disable:%u",
1948 F2FS_OPTION(sbi).unusable_cap);
1949 if (test_opt(sbi, MERGE_CHECKPOINT))
1950 seq_puts(seq, ",checkpoint_merge");
1952 seq_puts(seq, ",nocheckpoint_merge");
1953 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1954 seq_printf(seq, ",fsync_mode=%s", "posix");
1955 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1956 seq_printf(seq, ",fsync_mode=%s", "strict");
1957 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1958 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1960 #ifdef CONFIG_F2FS_FS_COMPRESSION
1961 f2fs_show_compress_options(seq, sbi->sb);
1964 if (test_opt(sbi, ATGC))
1965 seq_puts(seq, ",atgc");
1967 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1968 seq_printf(seq, ",discard_unit=%s", "block");
1969 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1970 seq_printf(seq, ",discard_unit=%s", "segment");
1971 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1972 seq_printf(seq, ",discard_unit=%s", "section");
1977 static void default_options(struct f2fs_sb_info *sbi)
1979 /* init some FS parameters */
1980 if (f2fs_sb_has_readonly(sbi))
1981 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
1983 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
1985 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1986 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1987 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1988 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1989 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1990 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1991 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1992 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1993 F2FS_OPTION(sbi).compress_ext_cnt = 0;
1994 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1995 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1997 sbi->sb->s_flags &= ~SB_INLINECRYPT;
1999 set_opt(sbi, INLINE_XATTR);
2000 set_opt(sbi, INLINE_DATA);
2001 set_opt(sbi, INLINE_DENTRY);
2002 set_opt(sbi, EXTENT_CACHE);
2003 set_opt(sbi, NOHEAP);
2004 clear_opt(sbi, DISABLE_CHECKPOINT);
2005 set_opt(sbi, MERGE_CHECKPOINT);
2006 F2FS_OPTION(sbi).unusable_cap = 0;
2007 sbi->sb->s_flags |= SB_LAZYTIME;
2008 set_opt(sbi, FLUSH_MERGE);
2009 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2010 set_opt(sbi, DISCARD);
2011 if (f2fs_sb_has_blkzoned(sbi)) {
2012 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2013 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2015 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2016 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2019 #ifdef CONFIG_F2FS_FS_XATTR
2020 set_opt(sbi, XATTR_USER);
2022 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2023 set_opt(sbi, POSIX_ACL);
2026 f2fs_build_fault_attr(sbi, 0, 0);
2030 static int f2fs_enable_quotas(struct super_block *sb);
2033 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2035 unsigned int s_flags = sbi->sb->s_flags;
2036 struct cp_control cpc;
2041 if (s_flags & SB_RDONLY) {
2042 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2045 sbi->sb->s_flags |= SB_ACTIVE;
2047 f2fs_update_time(sbi, DISABLE_TIME);
2049 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2050 down_write(&sbi->gc_lock);
2051 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
2052 if (err == -ENODATA) {
2056 if (err && err != -EAGAIN)
2060 ret = sync_filesystem(sbi->sb);
2062 err = ret ? ret : err;
2066 unusable = f2fs_get_unusable_blocks(sbi);
2067 if (f2fs_disable_cp_again(sbi, unusable)) {
2072 down_write(&sbi->gc_lock);
2073 cpc.reason = CP_PAUSE;
2074 set_sbi_flag(sbi, SBI_CP_DISABLED);
2075 err = f2fs_write_checkpoint(sbi, &cpc);
2079 spin_lock(&sbi->stat_lock);
2080 sbi->unusable_block_count = unusable;
2081 spin_unlock(&sbi->stat_lock);
2084 up_write(&sbi->gc_lock);
2086 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2090 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2092 int retry = DEFAULT_RETRY_IO_COUNT;
2094 /* we should flush all the data to keep data consistency */
2096 sync_inodes_sb(sbi->sb);
2098 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
2099 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2101 if (unlikely(retry < 0))
2102 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2104 down_write(&sbi->gc_lock);
2105 f2fs_dirty_to_prefree(sbi);
2107 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2108 set_sbi_flag(sbi, SBI_IS_DIRTY);
2109 up_write(&sbi->gc_lock);
2111 f2fs_sync_fs(sbi->sb, 1);
2114 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2116 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2117 struct f2fs_mount_info org_mount_opt;
2118 unsigned long old_sb_flags;
2120 bool need_restart_gc = false, need_stop_gc = false;
2121 bool need_restart_ckpt = false, need_stop_ckpt = false;
2122 bool need_restart_flush = false, need_stop_flush = false;
2123 bool need_restart_discard = false, need_stop_discard = false;
2124 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
2125 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2126 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2127 bool no_atgc = !test_opt(sbi, ATGC);
2128 bool no_discard = !test_opt(sbi, DISCARD);
2129 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2130 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2131 struct discard_cmd_control *dcc;
2137 * Save the old mount options in case we
2138 * need to restore them.
2140 org_mount_opt = sbi->mount_opt;
2141 old_sb_flags = sb->s_flags;
2144 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2145 for (i = 0; i < MAXQUOTAS; i++) {
2146 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2147 org_mount_opt.s_qf_names[i] =
2148 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2150 if (!org_mount_opt.s_qf_names[i]) {
2151 for (j = 0; j < i; j++)
2152 kfree(org_mount_opt.s_qf_names[j]);
2156 org_mount_opt.s_qf_names[i] = NULL;
2161 /* recover superblocks we couldn't write due to previous RO mount */
2162 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2163 err = f2fs_commit_super(sbi, false);
2164 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2167 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2170 default_options(sbi);
2172 /* parse mount options */
2173 err = parse_options(sb, data, true);
2178 * Previous and new state of filesystem is RO,
2179 * so skip checking GC and FLUSH_MERGE conditions.
2181 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2184 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2190 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2191 err = dquot_suspend(sb, -1);
2194 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2195 /* dquot_resume needs RW */
2196 sb->s_flags &= ~SB_RDONLY;
2197 if (sb_any_quota_suspended(sb)) {
2198 dquot_resume(sb, -1);
2199 } else if (f2fs_sb_has_quota_ino(sbi)) {
2200 err = f2fs_enable_quotas(sb);
2206 /* disallow enable atgc dynamically */
2207 if (no_atgc == !!test_opt(sbi, ATGC)) {
2209 f2fs_warn(sbi, "switch atgc option is not allowed");
2213 /* disallow enable/disable extent_cache dynamically */
2214 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
2216 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2220 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2222 f2fs_warn(sbi, "switch io_bits option is not allowed");
2226 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2228 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2232 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2234 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2238 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2240 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2245 * We stop the GC thread if FS is mounted as RO
2246 * or if background_gc = off is passed in mount
2247 * option. Also sync the filesystem.
2249 if ((*flags & SB_RDONLY) ||
2250 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2251 !test_opt(sbi, GC_MERGE))) {
2252 if (sbi->gc_thread) {
2253 f2fs_stop_gc_thread(sbi);
2254 need_restart_gc = true;
2256 } else if (!sbi->gc_thread) {
2257 err = f2fs_start_gc_thread(sbi);
2260 need_stop_gc = true;
2263 if (*flags & SB_RDONLY ||
2264 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
2267 set_sbi_flag(sbi, SBI_IS_DIRTY);
2268 set_sbi_flag(sbi, SBI_IS_CLOSE);
2269 f2fs_sync_fs(sb, 1);
2270 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2273 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2274 !test_opt(sbi, MERGE_CHECKPOINT)) {
2275 f2fs_stop_ckpt_thread(sbi);
2276 need_restart_ckpt = true;
2278 err = f2fs_start_ckpt_thread(sbi);
2281 "Failed to start F2FS issue_checkpoint_thread (%d)",
2285 need_stop_ckpt = true;
2289 * We stop issue flush thread if FS is mounted as RO
2290 * or if flush_merge is not passed in mount option.
2292 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2293 clear_opt(sbi, FLUSH_MERGE);
2294 f2fs_destroy_flush_cmd_control(sbi, false);
2295 need_restart_flush = true;
2297 err = f2fs_create_flush_cmd_control(sbi);
2300 need_stop_flush = true;
2303 if (no_discard == !!test_opt(sbi, DISCARD)) {
2304 if (test_opt(sbi, DISCARD)) {
2305 err = f2fs_start_discard_thread(sbi);
2308 need_stop_discard = true;
2310 dcc = SM_I(sbi)->dcc_info;
2311 f2fs_stop_discard_thread(sbi);
2312 if (atomic_read(&dcc->discard_cmd_cnt))
2313 f2fs_issue_discard_timeout(sbi);
2314 need_restart_discard = true;
2318 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2319 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2320 err = f2fs_disable_checkpoint(sbi);
2322 goto restore_discard;
2324 f2fs_enable_checkpoint(sbi);
2330 /* Release old quota file names */
2331 for (i = 0; i < MAXQUOTAS; i++)
2332 kfree(org_mount_opt.s_qf_names[i]);
2334 /* Update the POSIXACL Flag */
2335 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2336 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2338 limit_reserve_root(sbi);
2339 adjust_unusable_cap_perc(sbi);
2340 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2343 if (need_restart_discard) {
2344 if (f2fs_start_discard_thread(sbi))
2345 f2fs_warn(sbi, "discard has been stopped");
2346 } else if (need_stop_discard) {
2347 f2fs_stop_discard_thread(sbi);
2350 if (need_restart_flush) {
2351 if (f2fs_create_flush_cmd_control(sbi))
2352 f2fs_warn(sbi, "background flush thread has stopped");
2353 } else if (need_stop_flush) {
2354 clear_opt(sbi, FLUSH_MERGE);
2355 f2fs_destroy_flush_cmd_control(sbi, false);
2358 if (need_restart_ckpt) {
2359 if (f2fs_start_ckpt_thread(sbi))
2360 f2fs_warn(sbi, "background ckpt thread has stopped");
2361 } else if (need_stop_ckpt) {
2362 f2fs_stop_ckpt_thread(sbi);
2365 if (need_restart_gc) {
2366 if (f2fs_start_gc_thread(sbi))
2367 f2fs_warn(sbi, "background gc thread has stopped");
2368 } else if (need_stop_gc) {
2369 f2fs_stop_gc_thread(sbi);
2373 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2374 for (i = 0; i < MAXQUOTAS; i++) {
2375 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2376 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2379 sbi->mount_opt = org_mount_opt;
2380 sb->s_flags = old_sb_flags;
2385 /* Read data from quotafile */
2386 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2387 size_t len, loff_t off)
2389 struct inode *inode = sb_dqopt(sb)->files[type];
2390 struct address_space *mapping = inode->i_mapping;
2391 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2392 int offset = off & (sb->s_blocksize - 1);
2395 loff_t i_size = i_size_read(inode);
2402 if (off + len > i_size)
2405 while (toread > 0) {
2406 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2408 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2410 if (PTR_ERR(page) == -ENOMEM) {
2411 memalloc_retry_wait(GFP_NOFS);
2414 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2415 return PTR_ERR(page);
2420 if (unlikely(page->mapping != mapping)) {
2421 f2fs_put_page(page, 1);
2424 if (unlikely(!PageUptodate(page))) {
2425 f2fs_put_page(page, 1);
2426 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2430 kaddr = kmap_atomic(page);
2431 memcpy(data, kaddr + offset, tocopy);
2432 kunmap_atomic(kaddr);
2433 f2fs_put_page(page, 1);
2443 /* Write to quotafile */
2444 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2445 const char *data, size_t len, loff_t off)
2447 struct inode *inode = sb_dqopt(sb)->files[type];
2448 struct address_space *mapping = inode->i_mapping;
2449 const struct address_space_operations *a_ops = mapping->a_ops;
2450 int offset = off & (sb->s_blocksize - 1);
2451 size_t towrite = len;
2453 void *fsdata = NULL;
2458 while (towrite > 0) {
2459 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2462 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2464 if (unlikely(err)) {
2465 if (err == -ENOMEM) {
2466 congestion_wait(BLK_RW_ASYNC,
2467 DEFAULT_IO_TIMEOUT);
2470 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2474 kaddr = kmap_atomic(page);
2475 memcpy(kaddr + offset, data, tocopy);
2476 kunmap_atomic(kaddr);
2477 flush_dcache_page(page);
2479 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2490 inode->i_mtime = inode->i_ctime = current_time(inode);
2491 f2fs_mark_inode_dirty_sync(inode, false);
2492 return len - towrite;
2495 int f2fs_dquot_initialize(struct inode *inode)
2497 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) {
2498 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT);
2502 return dquot_initialize(inode);
2505 static struct dquot **f2fs_get_dquots(struct inode *inode)
2507 return F2FS_I(inode)->i_dquot;
2510 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2512 return &F2FS_I(inode)->i_reserved_quota;
2515 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2517 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2518 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2522 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2523 F2FS_OPTION(sbi).s_jquota_fmt, type);
2526 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2531 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2532 err = f2fs_enable_quotas(sbi->sb);
2534 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2540 for (i = 0; i < MAXQUOTAS; i++) {
2541 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2542 err = f2fs_quota_on_mount(sbi, i);
2547 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2554 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2557 struct inode *qf_inode;
2558 unsigned long qf_inum;
2561 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2563 qf_inum = f2fs_qf_ino(sb, type);
2567 qf_inode = f2fs_iget(sb, qf_inum);
2568 if (IS_ERR(qf_inode)) {
2569 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2570 return PTR_ERR(qf_inode);
2573 /* Don't account quota for quota files to avoid recursion */
2574 qf_inode->i_flags |= S_NOQUOTA;
2575 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2580 static int f2fs_enable_quotas(struct super_block *sb)
2582 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2584 unsigned long qf_inum;
2585 bool quota_mopt[MAXQUOTAS] = {
2586 test_opt(sbi, USRQUOTA),
2587 test_opt(sbi, GRPQUOTA),
2588 test_opt(sbi, PRJQUOTA),
2591 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2592 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2596 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2598 for (type = 0; type < MAXQUOTAS; type++) {
2599 qf_inum = f2fs_qf_ino(sb, type);
2601 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2602 DQUOT_USAGE_ENABLED |
2603 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2605 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2607 for (type--; type >= 0; type--)
2608 dquot_quota_off(sb, type);
2609 set_sbi_flag(F2FS_SB(sb),
2610 SBI_QUOTA_NEED_REPAIR);
2618 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2620 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2621 struct address_space *mapping = dqopt->files[type]->i_mapping;
2624 ret = dquot_writeback_dquots(sbi->sb, type);
2628 ret = filemap_fdatawrite(mapping);
2632 /* if we are using journalled quota */
2633 if (is_journalled_quota(sbi))
2636 ret = filemap_fdatawait(mapping);
2638 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2641 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2645 int f2fs_quota_sync(struct super_block *sb, int type)
2647 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2648 struct quota_info *dqopt = sb_dqopt(sb);
2653 * Now when everything is written we can discard the pagecache so
2654 * that userspace sees the changes.
2656 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2658 if (type != -1 && cnt != type)
2661 if (!sb_has_quota_active(sb, type))
2664 inode_lock(dqopt->files[cnt]);
2669 * down_read(quota_sem)
2670 * dquot_writeback_dquots()
2673 * down_read(quota_sem)
2676 down_read(&sbi->quota_sem);
2678 ret = f2fs_quota_sync_file(sbi, cnt);
2680 up_read(&sbi->quota_sem);
2681 f2fs_unlock_op(sbi);
2683 inode_unlock(dqopt->files[cnt]);
2691 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2692 const struct path *path)
2694 struct inode *inode;
2697 /* if quota sysfile exists, deny enabling quota with specific file */
2698 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2699 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2703 err = f2fs_quota_sync(sb, type);
2707 err = dquot_quota_on(sb, type, format_id, path);
2711 inode = d_inode(path->dentry);
2714 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2715 f2fs_set_inode_flags(inode);
2716 inode_unlock(inode);
2717 f2fs_mark_inode_dirty_sync(inode, false);
2722 static int __f2fs_quota_off(struct super_block *sb, int type)
2724 struct inode *inode = sb_dqopt(sb)->files[type];
2727 if (!inode || !igrab(inode))
2728 return dquot_quota_off(sb, type);
2730 err = f2fs_quota_sync(sb, type);
2734 err = dquot_quota_off(sb, type);
2735 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2739 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2740 f2fs_set_inode_flags(inode);
2741 inode_unlock(inode);
2742 f2fs_mark_inode_dirty_sync(inode, false);
2748 static int f2fs_quota_off(struct super_block *sb, int type)
2750 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2753 err = __f2fs_quota_off(sb, type);
2756 * quotactl can shutdown journalled quota, result in inconsistence
2757 * between quota record and fs data by following updates, tag the
2758 * flag to let fsck be aware of it.
2760 if (is_journalled_quota(sbi))
2761 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2765 void f2fs_quota_off_umount(struct super_block *sb)
2770 for (type = 0; type < MAXQUOTAS; type++) {
2771 err = __f2fs_quota_off(sb, type);
2773 int ret = dquot_quota_off(sb, type);
2775 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2777 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2781 * In case of checkpoint=disable, we must flush quota blocks.
2782 * This can cause NULL exception for node_inode in end_io, since
2783 * put_super already dropped it.
2785 sync_filesystem(sb);
2788 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2790 struct quota_info *dqopt = sb_dqopt(sb);
2793 for (type = 0; type < MAXQUOTAS; type++) {
2794 if (!dqopt->files[type])
2796 f2fs_inode_synced(dqopt->files[type]);
2800 static int f2fs_dquot_commit(struct dquot *dquot)
2802 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2805 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2806 ret = dquot_commit(dquot);
2808 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2809 up_read(&sbi->quota_sem);
2813 static int f2fs_dquot_acquire(struct dquot *dquot)
2815 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2818 down_read(&sbi->quota_sem);
2819 ret = dquot_acquire(dquot);
2821 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2822 up_read(&sbi->quota_sem);
2826 static int f2fs_dquot_release(struct dquot *dquot)
2828 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2829 int ret = dquot_release(dquot);
2832 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2836 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2838 struct super_block *sb = dquot->dq_sb;
2839 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2840 int ret = dquot_mark_dquot_dirty(dquot);
2842 /* if we are using journalled quota */
2843 if (is_journalled_quota(sbi))
2844 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2849 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2851 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2852 int ret = dquot_commit_info(sb, type);
2855 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2859 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2861 *projid = F2FS_I(inode)->i_projid;
2865 static const struct dquot_operations f2fs_quota_operations = {
2866 .get_reserved_space = f2fs_get_reserved_space,
2867 .write_dquot = f2fs_dquot_commit,
2868 .acquire_dquot = f2fs_dquot_acquire,
2869 .release_dquot = f2fs_dquot_release,
2870 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2871 .write_info = f2fs_dquot_commit_info,
2872 .alloc_dquot = dquot_alloc,
2873 .destroy_dquot = dquot_destroy,
2874 .get_projid = f2fs_get_projid,
2875 .get_next_id = dquot_get_next_id,
2878 static const struct quotactl_ops f2fs_quotactl_ops = {
2879 .quota_on = f2fs_quota_on,
2880 .quota_off = f2fs_quota_off,
2881 .quota_sync = f2fs_quota_sync,
2882 .get_state = dquot_get_state,
2883 .set_info = dquot_set_dqinfo,
2884 .get_dqblk = dquot_get_dqblk,
2885 .set_dqblk = dquot_set_dqblk,
2886 .get_nextdqblk = dquot_get_next_dqblk,
2889 int f2fs_dquot_initialize(struct inode *inode)
2894 int f2fs_quota_sync(struct super_block *sb, int type)
2899 void f2fs_quota_off_umount(struct super_block *sb)
2904 static const struct super_operations f2fs_sops = {
2905 .alloc_inode = f2fs_alloc_inode,
2906 .free_inode = f2fs_free_inode,
2907 .drop_inode = f2fs_drop_inode,
2908 .write_inode = f2fs_write_inode,
2909 .dirty_inode = f2fs_dirty_inode,
2910 .show_options = f2fs_show_options,
2912 .quota_read = f2fs_quota_read,
2913 .quota_write = f2fs_quota_write,
2914 .get_dquots = f2fs_get_dquots,
2916 .evict_inode = f2fs_evict_inode,
2917 .put_super = f2fs_put_super,
2918 .sync_fs = f2fs_sync_fs,
2919 .freeze_fs = f2fs_freeze,
2920 .unfreeze_fs = f2fs_unfreeze,
2921 .statfs = f2fs_statfs,
2922 .remount_fs = f2fs_remount,
2925 #ifdef CONFIG_FS_ENCRYPTION
2926 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2928 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2929 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2933 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2936 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2939 * Encrypting the root directory is not allowed because fsck
2940 * expects lost+found directory to exist and remain unencrypted
2941 * if LOST_FOUND feature is enabled.
2944 if (f2fs_sb_has_lost_found(sbi) &&
2945 inode->i_ino == F2FS_ROOT_INO(sbi))
2948 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2949 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2950 ctx, len, fs_data, XATTR_CREATE);
2953 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
2955 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
2958 static bool f2fs_has_stable_inodes(struct super_block *sb)
2963 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2964 int *ino_bits_ret, int *lblk_bits_ret)
2966 *ino_bits_ret = 8 * sizeof(nid_t);
2967 *lblk_bits_ret = 8 * sizeof(block_t);
2970 static int f2fs_get_num_devices(struct super_block *sb)
2972 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2974 if (f2fs_is_multi_device(sbi))
2975 return sbi->s_ndevs;
2979 static void f2fs_get_devices(struct super_block *sb,
2980 struct request_queue **devs)
2982 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2985 for (i = 0; i < sbi->s_ndevs; i++)
2986 devs[i] = bdev_get_queue(FDEV(i).bdev);
2989 static const struct fscrypt_operations f2fs_cryptops = {
2990 .key_prefix = "f2fs:",
2991 .get_context = f2fs_get_context,
2992 .set_context = f2fs_set_context,
2993 .get_dummy_policy = f2fs_get_dummy_policy,
2994 .empty_dir = f2fs_empty_dir,
2995 .has_stable_inodes = f2fs_has_stable_inodes,
2996 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2997 .get_num_devices = f2fs_get_num_devices,
2998 .get_devices = f2fs_get_devices,
3002 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3003 u64 ino, u32 generation)
3005 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3006 struct inode *inode;
3008 if (f2fs_check_nid_range(sbi, ino))
3009 return ERR_PTR(-ESTALE);
3012 * f2fs_iget isn't quite right if the inode is currently unallocated!
3013 * However f2fs_iget currently does appropriate checks to handle stale
3014 * inodes so everything is OK.
3016 inode = f2fs_iget(sb, ino);
3018 return ERR_CAST(inode);
3019 if (unlikely(generation && inode->i_generation != generation)) {
3020 /* we didn't find the right inode.. */
3022 return ERR_PTR(-ESTALE);
3027 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3028 int fh_len, int fh_type)
3030 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3031 f2fs_nfs_get_inode);
3034 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3035 int fh_len, int fh_type)
3037 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3038 f2fs_nfs_get_inode);
3041 static const struct export_operations f2fs_export_ops = {
3042 .fh_to_dentry = f2fs_fh_to_dentry,
3043 .fh_to_parent = f2fs_fh_to_parent,
3044 .get_parent = f2fs_get_parent,
3047 loff_t max_file_blocks(struct inode *inode)
3053 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3054 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3055 * space in inode.i_addr, it will be more safe to reassign
3059 if (inode && f2fs_compressed_file(inode))
3060 leaf_count = ADDRS_PER_BLOCK(inode);
3062 leaf_count = DEF_ADDRS_PER_BLOCK;
3064 /* two direct node blocks */
3065 result += (leaf_count * 2);
3067 /* two indirect node blocks */
3068 leaf_count *= NIDS_PER_BLOCK;
3069 result += (leaf_count * 2);
3071 /* one double indirect node block */
3072 leaf_count *= NIDS_PER_BLOCK;
3073 result += leaf_count;
3078 static int __f2fs_commit_super(struct buffer_head *bh,
3079 struct f2fs_super_block *super)
3083 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3084 set_buffer_dirty(bh);
3087 /* it's rare case, we can do fua all the time */
3088 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3091 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3092 struct buffer_head *bh)
3094 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3095 (bh->b_data + F2FS_SUPER_OFFSET);
3096 struct super_block *sb = sbi->sb;
3097 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3098 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3099 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3100 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3101 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3102 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3103 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3104 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3105 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3106 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3107 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3108 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3109 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3110 u64 main_end_blkaddr = main_blkaddr +
3111 (segment_count_main << log_blocks_per_seg);
3112 u64 seg_end_blkaddr = segment0_blkaddr +
3113 (segment_count << log_blocks_per_seg);
3115 if (segment0_blkaddr != cp_blkaddr) {
3116 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3117 segment0_blkaddr, cp_blkaddr);
3121 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3123 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3124 cp_blkaddr, sit_blkaddr,
3125 segment_count_ckpt << log_blocks_per_seg);
3129 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3131 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3132 sit_blkaddr, nat_blkaddr,
3133 segment_count_sit << log_blocks_per_seg);
3137 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3139 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3140 nat_blkaddr, ssa_blkaddr,
3141 segment_count_nat << log_blocks_per_seg);
3145 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3147 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3148 ssa_blkaddr, main_blkaddr,
3149 segment_count_ssa << log_blocks_per_seg);
3153 if (main_end_blkaddr > seg_end_blkaddr) {
3154 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3155 main_blkaddr, seg_end_blkaddr,
3156 segment_count_main << log_blocks_per_seg);
3158 } else if (main_end_blkaddr < seg_end_blkaddr) {
3162 /* fix in-memory information all the time */
3163 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3164 segment0_blkaddr) >> log_blocks_per_seg);
3166 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3167 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3170 err = __f2fs_commit_super(bh, NULL);
3171 res = err ? "failed" : "done";
3173 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3174 res, main_blkaddr, seg_end_blkaddr,
3175 segment_count_main << log_blocks_per_seg);
3182 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3183 struct buffer_head *bh)
3185 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3186 block_t total_sections, blocks_per_seg;
3187 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3188 (bh->b_data + F2FS_SUPER_OFFSET);
3189 size_t crc_offset = 0;
3192 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3193 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3194 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3198 /* Check checksum_offset and crc in superblock */
3199 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3200 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3202 offsetof(struct f2fs_super_block, crc)) {
3203 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3205 return -EFSCORRUPTED;
3207 crc = le32_to_cpu(raw_super->crc);
3208 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3209 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3210 return -EFSCORRUPTED;
3214 /* Currently, support only 4KB block size */
3215 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3216 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3217 le32_to_cpu(raw_super->log_blocksize),
3219 return -EFSCORRUPTED;
3222 /* check log blocks per segment */
3223 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3224 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3225 le32_to_cpu(raw_super->log_blocks_per_seg));
3226 return -EFSCORRUPTED;
3229 /* Currently, support 512/1024/2048/4096 bytes sector size */
3230 if (le32_to_cpu(raw_super->log_sectorsize) >
3231 F2FS_MAX_LOG_SECTOR_SIZE ||
3232 le32_to_cpu(raw_super->log_sectorsize) <
3233 F2FS_MIN_LOG_SECTOR_SIZE) {
3234 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3235 le32_to_cpu(raw_super->log_sectorsize));
3236 return -EFSCORRUPTED;
3238 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3239 le32_to_cpu(raw_super->log_sectorsize) !=
3240 F2FS_MAX_LOG_SECTOR_SIZE) {
3241 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3242 le32_to_cpu(raw_super->log_sectors_per_block),
3243 le32_to_cpu(raw_super->log_sectorsize));
3244 return -EFSCORRUPTED;
3247 segment_count = le32_to_cpu(raw_super->segment_count);
3248 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3249 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3250 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3251 total_sections = le32_to_cpu(raw_super->section_count);
3253 /* blocks_per_seg should be 512, given the above check */
3254 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3256 if (segment_count > F2FS_MAX_SEGMENT ||
3257 segment_count < F2FS_MIN_SEGMENTS) {
3258 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3259 return -EFSCORRUPTED;
3262 if (total_sections > segment_count_main || total_sections < 1 ||
3263 segs_per_sec > segment_count || !segs_per_sec) {
3264 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3265 segment_count, total_sections, segs_per_sec);
3266 return -EFSCORRUPTED;
3269 if (segment_count_main != total_sections * segs_per_sec) {
3270 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3271 segment_count_main, total_sections, segs_per_sec);
3272 return -EFSCORRUPTED;
3275 if ((segment_count / segs_per_sec) < total_sections) {
3276 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3277 segment_count, segs_per_sec, total_sections);
3278 return -EFSCORRUPTED;
3281 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3282 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3283 segment_count, le64_to_cpu(raw_super->block_count));
3284 return -EFSCORRUPTED;
3287 if (RDEV(0).path[0]) {
3288 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3291 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3292 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3295 if (segment_count != dev_seg_count) {
3296 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3297 segment_count, dev_seg_count);
3298 return -EFSCORRUPTED;
3301 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3302 !bdev_is_zoned(sbi->sb->s_bdev)) {
3303 f2fs_info(sbi, "Zoned block device path is missing");
3304 return -EFSCORRUPTED;
3308 if (secs_per_zone > total_sections || !secs_per_zone) {
3309 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3310 secs_per_zone, total_sections);
3311 return -EFSCORRUPTED;
3313 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3314 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3315 (le32_to_cpu(raw_super->extension_count) +
3316 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3317 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3318 le32_to_cpu(raw_super->extension_count),
3319 raw_super->hot_ext_count,
3320 F2FS_MAX_EXTENSION);
3321 return -EFSCORRUPTED;
3324 if (le32_to_cpu(raw_super->cp_payload) >=
3325 (blocks_per_seg - F2FS_CP_PACKS -
3326 NR_CURSEG_PERSIST_TYPE)) {
3327 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3328 le32_to_cpu(raw_super->cp_payload),
3329 blocks_per_seg - F2FS_CP_PACKS -
3330 NR_CURSEG_PERSIST_TYPE);
3331 return -EFSCORRUPTED;
3334 /* check reserved ino info */
3335 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3336 le32_to_cpu(raw_super->meta_ino) != 2 ||
3337 le32_to_cpu(raw_super->root_ino) != 3) {
3338 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3339 le32_to_cpu(raw_super->node_ino),
3340 le32_to_cpu(raw_super->meta_ino),
3341 le32_to_cpu(raw_super->root_ino));
3342 return -EFSCORRUPTED;
3345 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3346 if (sanity_check_area_boundary(sbi, bh))
3347 return -EFSCORRUPTED;
3352 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3354 unsigned int total, fsmeta;
3355 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3356 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3357 unsigned int ovp_segments, reserved_segments;
3358 unsigned int main_segs, blocks_per_seg;
3359 unsigned int sit_segs, nat_segs;
3360 unsigned int sit_bitmap_size, nat_bitmap_size;
3361 unsigned int log_blocks_per_seg;
3362 unsigned int segment_count_main;
3363 unsigned int cp_pack_start_sum, cp_payload;
3364 block_t user_block_count, valid_user_blocks;
3365 block_t avail_node_count, valid_node_count;
3366 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3369 total = le32_to_cpu(raw_super->segment_count);
3370 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3371 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3373 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3375 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3376 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3378 if (unlikely(fsmeta >= total))
3381 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3382 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3384 if (!f2fs_sb_has_readonly(sbi) &&
3385 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3386 ovp_segments == 0 || reserved_segments == 0)) {
3387 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3390 user_block_count = le64_to_cpu(ckpt->user_block_count);
3391 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3392 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3393 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3394 if (!user_block_count || user_block_count >=
3395 segment_count_main << log_blocks_per_seg) {
3396 f2fs_err(sbi, "Wrong user_block_count: %u",
3401 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3402 if (valid_user_blocks > user_block_count) {
3403 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3404 valid_user_blocks, user_block_count);
3408 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3409 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3410 if (valid_node_count > avail_node_count) {
3411 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3412 valid_node_count, avail_node_count);
3416 main_segs = le32_to_cpu(raw_super->segment_count_main);
3417 blocks_per_seg = sbi->blocks_per_seg;
3419 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3420 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3421 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3424 if (f2fs_sb_has_readonly(sbi))
3427 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3428 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3429 le32_to_cpu(ckpt->cur_node_segno[j])) {
3430 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3432 le32_to_cpu(ckpt->cur_node_segno[i]));
3438 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3439 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3440 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3443 if (f2fs_sb_has_readonly(sbi))
3446 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3447 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3448 le32_to_cpu(ckpt->cur_data_segno[j])) {
3449 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3451 le32_to_cpu(ckpt->cur_data_segno[i]));
3456 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3457 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3458 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3459 le32_to_cpu(ckpt->cur_data_segno[j])) {
3460 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3462 le32_to_cpu(ckpt->cur_node_segno[i]));
3468 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3469 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3471 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3472 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3473 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3474 sit_bitmap_size, nat_bitmap_size);
3478 cp_pack_start_sum = __start_sum_addr(sbi);
3479 cp_payload = __cp_payload(sbi);
3480 if (cp_pack_start_sum < cp_payload + 1 ||
3481 cp_pack_start_sum > blocks_per_seg - 1 -
3482 NR_CURSEG_PERSIST_TYPE) {
3483 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3488 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3489 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3490 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3491 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3492 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3493 le32_to_cpu(ckpt->checksum_offset));
3497 nat_blocks = nat_segs << log_blocks_per_seg;
3498 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3499 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3500 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3501 (cp_payload + F2FS_CP_PACKS +
3502 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3503 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3504 cp_payload, nat_bits_blocks);
3508 if (unlikely(f2fs_cp_error(sbi))) {
3509 f2fs_err(sbi, "A bug case: need to run fsck");
3515 static void init_sb_info(struct f2fs_sb_info *sbi)
3517 struct f2fs_super_block *raw_super = sbi->raw_super;
3520 sbi->log_sectors_per_block =
3521 le32_to_cpu(raw_super->log_sectors_per_block);
3522 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3523 sbi->blocksize = 1 << sbi->log_blocksize;
3524 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3525 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3526 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3527 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3528 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3529 sbi->total_node_count =
3530 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3531 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3532 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3533 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3534 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3535 sbi->cur_victim_sec = NULL_SECNO;
3536 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3537 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3538 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3539 sbi->migration_granularity = sbi->segs_per_sec;
3540 sbi->seq_file_ra_mul = MIN_RA_MUL;
3541 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3542 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3544 sbi->dir_level = DEF_DIR_LEVEL;
3545 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3546 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3547 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3548 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3549 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3550 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3551 DEF_UMOUNT_DISCARD_TIMEOUT;
3552 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3554 for (i = 0; i < NR_COUNT_TYPE; i++)
3555 atomic_set(&sbi->nr_pages[i], 0);
3557 for (i = 0; i < META; i++)
3558 atomic_set(&sbi->wb_sync_req[i], 0);
3560 INIT_LIST_HEAD(&sbi->s_list);
3561 mutex_init(&sbi->umount_mutex);
3562 init_rwsem(&sbi->io_order_lock);
3563 spin_lock_init(&sbi->cp_lock);
3565 sbi->dirty_device = 0;
3566 spin_lock_init(&sbi->dev_lock);
3568 init_rwsem(&sbi->sb_lock);
3569 init_rwsem(&sbi->pin_sem);
3572 static int init_percpu_info(struct f2fs_sb_info *sbi)
3576 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3580 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3583 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3588 #ifdef CONFIG_BLK_DEV_ZONED
3590 struct f2fs_report_zones_args {
3591 struct f2fs_dev_info *dev;
3592 bool zone_cap_mismatch;
3595 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3598 struct f2fs_report_zones_args *rz_args = data;
3600 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3603 set_bit(idx, rz_args->dev->blkz_seq);
3604 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >>
3605 F2FS_LOG_SECTORS_PER_BLOCK;
3606 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch)
3607 rz_args->zone_cap_mismatch = true;
3612 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3614 struct block_device *bdev = FDEV(devi).bdev;
3615 sector_t nr_sectors = bdev_nr_sectors(bdev);
3616 struct f2fs_report_zones_args rep_zone_arg;
3619 if (!f2fs_sb_has_blkzoned(sbi))
3622 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3623 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3625 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3626 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3627 __ilog2_u32(sbi->blocks_per_blkz))
3629 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3630 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3631 sbi->log_blocks_per_blkz;
3632 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3633 FDEV(devi).nr_blkz++;
3635 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3636 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3637 * sizeof(unsigned long),
3639 if (!FDEV(devi).blkz_seq)
3642 /* Get block zones type and zone-capacity */
3643 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi,
3644 FDEV(devi).nr_blkz * sizeof(block_t),
3646 if (!FDEV(devi).zone_capacity_blocks)
3649 rep_zone_arg.dev = &FDEV(devi);
3650 rep_zone_arg.zone_cap_mismatch = false;
3652 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3657 if (!rep_zone_arg.zone_cap_mismatch) {
3658 kfree(FDEV(devi).zone_capacity_blocks);
3659 FDEV(devi).zone_capacity_blocks = NULL;
3667 * Read f2fs raw super block.
3668 * Because we have two copies of super block, so read both of them
3669 * to get the first valid one. If any one of them is broken, we pass
3670 * them recovery flag back to the caller.
3672 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3673 struct f2fs_super_block **raw_super,
3674 int *valid_super_block, int *recovery)
3676 struct super_block *sb = sbi->sb;
3678 struct buffer_head *bh;
3679 struct f2fs_super_block *super;
3682 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3686 for (block = 0; block < 2; block++) {
3687 bh = sb_bread(sb, block);
3689 f2fs_err(sbi, "Unable to read %dth superblock",
3696 /* sanity checking of raw super */
3697 err = sanity_check_raw_super(sbi, bh);
3699 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3707 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3709 *valid_super_block = block;
3715 /* No valid superblock */
3724 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3726 struct buffer_head *bh;
3730 if ((recover && f2fs_readonly(sbi->sb)) ||
3731 bdev_read_only(sbi->sb->s_bdev)) {
3732 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3736 /* we should update superblock crc here */
3737 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3738 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3739 offsetof(struct f2fs_super_block, crc));
3740 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3743 /* write back-up superblock first */
3744 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3747 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3750 /* if we are in recovery path, skip writing valid superblock */
3754 /* write current valid superblock */
3755 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3758 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3763 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3765 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3766 unsigned int max_devices = MAX_DEVICES;
3767 unsigned int logical_blksize;
3770 /* Initialize single device information */
3771 if (!RDEV(0).path[0]) {
3772 if (!bdev_is_zoned(sbi->sb->s_bdev))
3778 * Initialize multiple devices information, or single
3779 * zoned block device information.
3781 sbi->devs = f2fs_kzalloc(sbi,
3782 array_size(max_devices,
3783 sizeof(struct f2fs_dev_info)),
3788 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3789 sbi->aligned_blksize = true;
3791 for (i = 0; i < max_devices; i++) {
3793 if (i > 0 && !RDEV(i).path[0])
3796 if (max_devices == 1) {
3797 /* Single zoned block device mount */
3799 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3800 sbi->sb->s_mode, sbi->sb->s_type);
3802 /* Multi-device mount */
3803 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3804 FDEV(i).total_segments =
3805 le32_to_cpu(RDEV(i).total_segments);
3807 FDEV(i).start_blk = 0;
3808 FDEV(i).end_blk = FDEV(i).start_blk +
3809 (FDEV(i).total_segments <<
3810 sbi->log_blocks_per_seg) - 1 +
3811 le32_to_cpu(raw_super->segment0_blkaddr);
3813 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3814 FDEV(i).end_blk = FDEV(i).start_blk +
3815 (FDEV(i).total_segments <<
3816 sbi->log_blocks_per_seg) - 1;
3818 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3819 sbi->sb->s_mode, sbi->sb->s_type);
3821 if (IS_ERR(FDEV(i).bdev))
3822 return PTR_ERR(FDEV(i).bdev);
3824 /* to release errored devices */
3825 sbi->s_ndevs = i + 1;
3827 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
3828 sbi->aligned_blksize = false;
3830 #ifdef CONFIG_BLK_DEV_ZONED
3831 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3832 !f2fs_sb_has_blkzoned(sbi)) {
3833 f2fs_err(sbi, "Zoned block device feature not enabled");
3836 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3837 if (init_blkz_info(sbi, i)) {
3838 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3841 if (max_devices == 1)
3843 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3845 FDEV(i).total_segments,
3846 FDEV(i).start_blk, FDEV(i).end_blk,
3847 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3848 "Host-aware" : "Host-managed");
3852 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3854 FDEV(i).total_segments,
3855 FDEV(i).start_blk, FDEV(i).end_blk);
3858 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3862 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3864 #if IS_ENABLED(CONFIG_UNICODE)
3865 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3866 const struct f2fs_sb_encodings *encoding_info;
3867 struct unicode_map *encoding;
3868 __u16 encoding_flags;
3870 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
3871 if (!encoding_info) {
3873 "Encoding requested by superblock is unknown");
3877 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
3878 encoding = utf8_load(encoding_info->version);
3879 if (IS_ERR(encoding)) {
3881 "can't mount with superblock charset: %s-%u.%u.%u "
3882 "not supported by the kernel. flags: 0x%x.",
3883 encoding_info->name,
3884 unicode_major(encoding_info->version),
3885 unicode_minor(encoding_info->version),
3886 unicode_rev(encoding_info->version),
3888 return PTR_ERR(encoding);
3890 f2fs_info(sbi, "Using encoding defined by superblock: "
3891 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
3892 unicode_major(encoding_info->version),
3893 unicode_minor(encoding_info->version),
3894 unicode_rev(encoding_info->version),
3897 sbi->sb->s_encoding = encoding;
3898 sbi->sb->s_encoding_flags = encoding_flags;
3901 if (f2fs_sb_has_casefold(sbi)) {
3902 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3909 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3911 struct f2fs_sm_info *sm_i = SM_I(sbi);
3913 /* adjust parameters according to the volume size */
3914 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3915 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3916 if (f2fs_block_unit_discard(sbi))
3917 sm_i->dcc_info->discard_granularity = 1;
3918 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3921 sbi->readdir_ra = 1;
3924 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3926 struct f2fs_sb_info *sbi;
3927 struct f2fs_super_block *raw_super;
3930 bool skip_recovery = false, need_fsck = false;
3931 char *options = NULL;
3932 int recovery, i, valid_super_block;
3933 struct curseg_info *seg_i;
3939 valid_super_block = -1;
3942 /* allocate memory for f2fs-specific super block info */
3943 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3949 /* Load the checksum driver */
3950 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3951 if (IS_ERR(sbi->s_chksum_driver)) {
3952 f2fs_err(sbi, "Cannot load crc32 driver.");
3953 err = PTR_ERR(sbi->s_chksum_driver);
3954 sbi->s_chksum_driver = NULL;
3958 /* set a block size */
3959 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3960 f2fs_err(sbi, "unable to set blocksize");
3964 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3969 sb->s_fs_info = sbi;
3970 sbi->raw_super = raw_super;
3972 /* precompute checksum seed for metadata */
3973 if (f2fs_sb_has_inode_chksum(sbi))
3974 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3975 sizeof(raw_super->uuid));
3977 default_options(sbi);
3978 /* parse mount options */
3979 options = kstrdup((const char *)data, GFP_KERNEL);
3980 if (data && !options) {
3985 err = parse_options(sb, options, false);
3989 sb->s_maxbytes = max_file_blocks(NULL) <<
3990 le32_to_cpu(raw_super->log_blocksize);
3991 sb->s_max_links = F2FS_LINK_MAX;
3993 err = f2fs_setup_casefold(sbi);
3998 sb->dq_op = &f2fs_quota_operations;
3999 sb->s_qcop = &f2fs_quotactl_ops;
4000 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4002 if (f2fs_sb_has_quota_ino(sbi)) {
4003 for (i = 0; i < MAXQUOTAS; i++) {
4004 if (f2fs_qf_ino(sbi->sb, i))
4005 sbi->nquota_files++;
4010 sb->s_op = &f2fs_sops;
4011 #ifdef CONFIG_FS_ENCRYPTION
4012 sb->s_cop = &f2fs_cryptops;
4014 #ifdef CONFIG_FS_VERITY
4015 sb->s_vop = &f2fs_verityops;
4017 sb->s_xattr = f2fs_xattr_handlers;
4018 sb->s_export_op = &f2fs_export_ops;
4019 sb->s_magic = F2FS_SUPER_MAGIC;
4020 sb->s_time_gran = 1;
4021 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4022 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4023 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4024 sb->s_iflags |= SB_I_CGROUPWB;
4026 /* init f2fs-specific super block info */
4027 sbi->valid_super_block = valid_super_block;
4028 init_rwsem(&sbi->gc_lock);
4029 mutex_init(&sbi->writepages);
4030 init_rwsem(&sbi->cp_global_sem);
4031 init_rwsem(&sbi->node_write);
4032 init_rwsem(&sbi->node_change);
4034 /* disallow all the data/node/meta page writes */
4035 set_sbi_flag(sbi, SBI_POR_DOING);
4036 spin_lock_init(&sbi->stat_lock);
4038 for (i = 0; i < NR_PAGE_TYPE; i++) {
4039 int n = (i == META) ? 1 : NR_TEMP_TYPE;
4045 sizeof(struct f2fs_bio_info)),
4047 if (!sbi->write_io[i]) {
4052 for (j = HOT; j < n; j++) {
4053 init_rwsem(&sbi->write_io[i][j].io_rwsem);
4054 sbi->write_io[i][j].sbi = sbi;
4055 sbi->write_io[i][j].bio = NULL;
4056 spin_lock_init(&sbi->write_io[i][j].io_lock);
4057 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
4058 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
4059 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
4063 init_rwsem(&sbi->cp_rwsem);
4064 init_rwsem(&sbi->quota_sem);
4065 init_waitqueue_head(&sbi->cp_wait);
4068 err = f2fs_init_iostat(sbi);
4072 err = init_percpu_info(sbi);
4076 if (F2FS_IO_ALIGNED(sbi)) {
4077 sbi->write_io_dummy =
4078 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4079 if (!sbi->write_io_dummy) {
4085 /* init per sbi slab cache */
4086 err = f2fs_init_xattr_caches(sbi);
4089 err = f2fs_init_page_array_cache(sbi);
4091 goto free_xattr_cache;
4093 /* get an inode for meta space */
4094 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4095 if (IS_ERR(sbi->meta_inode)) {
4096 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4097 err = PTR_ERR(sbi->meta_inode);
4098 goto free_page_array_cache;
4101 err = f2fs_get_valid_checkpoint(sbi);
4103 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4104 goto free_meta_inode;
4107 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4108 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4109 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4110 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4111 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4114 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4115 set_sbi_flag(sbi, SBI_NEED_FSCK);
4117 /* Initialize device list */
4118 err = f2fs_scan_devices(sbi);
4120 f2fs_err(sbi, "Failed to find devices");
4124 err = f2fs_init_post_read_wq(sbi);
4126 f2fs_err(sbi, "Failed to initialize post read workqueue");
4130 sbi->total_valid_node_count =
4131 le32_to_cpu(sbi->ckpt->valid_node_count);
4132 percpu_counter_set(&sbi->total_valid_inode_count,
4133 le32_to_cpu(sbi->ckpt->valid_inode_count));
4134 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4135 sbi->total_valid_block_count =
4136 le64_to_cpu(sbi->ckpt->valid_block_count);
4137 sbi->last_valid_block_count = sbi->total_valid_block_count;
4138 sbi->reserved_blocks = 0;
4139 sbi->current_reserved_blocks = 0;
4140 limit_reserve_root(sbi);
4141 adjust_unusable_cap_perc(sbi);
4143 for (i = 0; i < NR_INODE_TYPE; i++) {
4144 INIT_LIST_HEAD(&sbi->inode_list[i]);
4145 spin_lock_init(&sbi->inode_lock[i]);
4147 mutex_init(&sbi->flush_lock);
4149 f2fs_init_extent_cache_info(sbi);
4151 f2fs_init_ino_entry_info(sbi);
4153 f2fs_init_fsync_node_info(sbi);
4155 /* setup checkpoint request control and start checkpoint issue thread */
4156 f2fs_init_ckpt_req_control(sbi);
4157 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4158 test_opt(sbi, MERGE_CHECKPOINT)) {
4159 err = f2fs_start_ckpt_thread(sbi);
4162 "Failed to start F2FS issue_checkpoint_thread (%d)",
4164 goto stop_ckpt_thread;
4168 /* setup f2fs internal modules */
4169 err = f2fs_build_segment_manager(sbi);
4171 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4175 err = f2fs_build_node_manager(sbi);
4177 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4182 /* For write statistics */
4183 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4185 /* Read accumulated write IO statistics if exists */
4186 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4187 if (__exist_node_summaries(sbi))
4188 sbi->kbytes_written =
4189 le64_to_cpu(seg_i->journal->info.kbytes_written);
4191 f2fs_build_gc_manager(sbi);
4193 err = f2fs_build_stats(sbi);
4197 /* get an inode for node space */
4198 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4199 if (IS_ERR(sbi->node_inode)) {
4200 f2fs_err(sbi, "Failed to read node inode");
4201 err = PTR_ERR(sbi->node_inode);
4205 /* read root inode and dentry */
4206 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4208 f2fs_err(sbi, "Failed to read root inode");
4209 err = PTR_ERR(root);
4210 goto free_node_inode;
4212 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4213 !root->i_size || !root->i_nlink) {
4216 goto free_node_inode;
4219 sb->s_root = d_make_root(root); /* allocate root dentry */
4222 goto free_node_inode;
4225 err = f2fs_init_compress_inode(sbi);
4227 goto free_root_inode;
4229 err = f2fs_register_sysfs(sbi);
4231 goto free_compress_inode;
4234 /* Enable quota usage during mount */
4235 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4236 err = f2fs_enable_quotas(sb);
4238 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4241 /* if there are any orphan inodes, free them */
4242 err = f2fs_recover_orphan_inodes(sbi);
4246 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4247 goto reset_checkpoint;
4249 /* recover fsynced data */
4250 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4251 !test_opt(sbi, NORECOVERY)) {
4253 * mount should be failed, when device has readonly mode, and
4254 * previous checkpoint was not done by clean system shutdown.
4256 if (f2fs_hw_is_readonly(sbi)) {
4257 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4258 err = f2fs_recover_fsync_data(sbi, true);
4261 f2fs_err(sbi, "Need to recover fsync data, but "
4262 "write access unavailable, please try "
4263 "mount w/ disable_roll_forward or norecovery");
4268 f2fs_info(sbi, "write access unavailable, skipping recovery");
4269 goto reset_checkpoint;
4273 set_sbi_flag(sbi, SBI_NEED_FSCK);
4276 goto reset_checkpoint;
4278 err = f2fs_recover_fsync_data(sbi, false);
4281 skip_recovery = true;
4283 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4288 err = f2fs_recover_fsync_data(sbi, true);
4290 if (!f2fs_readonly(sb) && err > 0) {
4292 f2fs_err(sbi, "Need to recover fsync data");
4298 * If the f2fs is not readonly and fsync data recovery succeeds,
4299 * check zoned block devices' write pointer consistency.
4301 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4302 err = f2fs_check_write_pointer(sbi);
4308 f2fs_init_inmem_curseg(sbi);
4310 /* f2fs_recover_fsync_data() cleared this already */
4311 clear_sbi_flag(sbi, SBI_POR_DOING);
4313 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4314 err = f2fs_disable_checkpoint(sbi);
4316 goto sync_free_meta;
4317 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4318 f2fs_enable_checkpoint(sbi);
4322 * If filesystem is not mounted as read-only then
4323 * do start the gc_thread.
4325 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4326 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4327 /* After POR, we can run background GC thread.*/
4328 err = f2fs_start_gc_thread(sbi);
4330 goto sync_free_meta;
4334 /* recover broken superblock */
4336 err = f2fs_commit_super(sbi, true);
4337 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4338 sbi->valid_super_block ? 1 : 2, err);
4341 f2fs_join_shrinker(sbi);
4343 f2fs_tuning_parameters(sbi);
4345 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4346 cur_cp_version(F2FS_CKPT(sbi)));
4347 f2fs_update_time(sbi, CP_TIME);
4348 f2fs_update_time(sbi, REQ_TIME);
4349 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4353 /* safe to flush all the data */
4354 sync_filesystem(sbi->sb);
4359 f2fs_truncate_quota_inode_pages(sb);
4360 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4361 f2fs_quota_off_umount(sbi->sb);
4364 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4365 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4366 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4367 * falls into an infinite loop in f2fs_sync_meta_pages().
4369 truncate_inode_pages_final(META_MAPPING(sbi));
4370 /* evict some inodes being cached by GC */
4372 f2fs_unregister_sysfs(sbi);
4373 free_compress_inode:
4374 f2fs_destroy_compress_inode(sbi);
4379 f2fs_release_ino_entry(sbi, true);
4380 truncate_inode_pages_final(NODE_MAPPING(sbi));
4381 iput(sbi->node_inode);
4382 sbi->node_inode = NULL;
4384 f2fs_destroy_stats(sbi);
4386 /* stop discard thread before destroying node manager */
4387 f2fs_stop_discard_thread(sbi);
4388 f2fs_destroy_node_manager(sbi);
4390 f2fs_destroy_segment_manager(sbi);
4391 f2fs_destroy_post_read_wq(sbi);
4393 f2fs_stop_ckpt_thread(sbi);
4395 destroy_device_list(sbi);
4398 make_bad_inode(sbi->meta_inode);
4399 iput(sbi->meta_inode);
4400 sbi->meta_inode = NULL;
4401 free_page_array_cache:
4402 f2fs_destroy_page_array_cache(sbi);
4404 f2fs_destroy_xattr_caches(sbi);
4406 mempool_destroy(sbi->write_io_dummy);
4408 destroy_percpu_info(sbi);
4410 f2fs_destroy_iostat(sbi);
4412 for (i = 0; i < NR_PAGE_TYPE; i++)
4413 kvfree(sbi->write_io[i]);
4415 #if IS_ENABLED(CONFIG_UNICODE)
4416 utf8_unload(sb->s_encoding);
4417 sb->s_encoding = NULL;
4421 for (i = 0; i < MAXQUOTAS; i++)
4422 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4424 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4429 if (sbi->s_chksum_driver)
4430 crypto_free_shash(sbi->s_chksum_driver);
4433 /* give only one another chance */
4434 if (retry_cnt > 0 && skip_recovery) {
4436 shrink_dcache_sb(sb);
4442 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4443 const char *dev_name, void *data)
4445 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4448 static void kill_f2fs_super(struct super_block *sb)
4451 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4453 set_sbi_flag(sbi, SBI_IS_CLOSE);
4454 f2fs_stop_gc_thread(sbi);
4455 f2fs_stop_discard_thread(sbi);
4457 #ifdef CONFIG_F2FS_FS_COMPRESSION
4459 * latter evict_inode() can bypass checking and invalidating
4460 * compress inode cache.
4462 if (test_opt(sbi, COMPRESS_CACHE))
4463 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4466 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4467 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4468 struct cp_control cpc = {
4469 .reason = CP_UMOUNT,
4471 f2fs_write_checkpoint(sbi, &cpc);
4474 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4475 sb->s_flags &= ~SB_RDONLY;
4477 kill_block_super(sb);
4480 static struct file_system_type f2fs_fs_type = {
4481 .owner = THIS_MODULE,
4483 .mount = f2fs_mount,
4484 .kill_sb = kill_f2fs_super,
4485 .fs_flags = FS_REQUIRES_DEV,
4487 MODULE_ALIAS_FS("f2fs");
4489 static int __init init_inodecache(void)
4491 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4492 sizeof(struct f2fs_inode_info), 0,
4493 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4494 if (!f2fs_inode_cachep)
4499 static void destroy_inodecache(void)
4502 * Make sure all delayed rcu free inodes are flushed before we
4506 kmem_cache_destroy(f2fs_inode_cachep);
4509 static int __init init_f2fs_fs(void)
4513 if (PAGE_SIZE != F2FS_BLKSIZE) {
4514 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4515 PAGE_SIZE, F2FS_BLKSIZE);
4519 err = init_inodecache();
4522 err = f2fs_create_node_manager_caches();
4524 goto free_inodecache;
4525 err = f2fs_create_segment_manager_caches();
4527 goto free_node_manager_caches;
4528 err = f2fs_create_checkpoint_caches();
4530 goto free_segment_manager_caches;
4531 err = f2fs_create_recovery_cache();
4533 goto free_checkpoint_caches;
4534 err = f2fs_create_extent_cache();
4536 goto free_recovery_cache;
4537 err = f2fs_create_garbage_collection_cache();
4539 goto free_extent_cache;
4540 err = f2fs_init_sysfs();
4542 goto free_garbage_collection_cache;
4543 err = register_shrinker(&f2fs_shrinker_info);
4546 err = register_filesystem(&f2fs_fs_type);
4549 f2fs_create_root_stats();
4550 err = f2fs_init_post_read_processing();
4552 goto free_root_stats;
4553 err = f2fs_init_iostat_processing();
4555 goto free_post_read;
4556 err = f2fs_init_bio_entry_cache();
4559 err = f2fs_init_bioset();
4561 goto free_bio_enrty_cache;
4562 err = f2fs_init_compress_mempool();
4565 err = f2fs_init_compress_cache();
4567 goto free_compress_mempool;
4568 err = f2fs_create_casefold_cache();
4570 goto free_compress_cache;
4572 free_compress_cache:
4573 f2fs_destroy_compress_cache();
4574 free_compress_mempool:
4575 f2fs_destroy_compress_mempool();
4577 f2fs_destroy_bioset();
4578 free_bio_enrty_cache:
4579 f2fs_destroy_bio_entry_cache();
4581 f2fs_destroy_iostat_processing();
4583 f2fs_destroy_post_read_processing();
4585 f2fs_destroy_root_stats();
4586 unregister_filesystem(&f2fs_fs_type);
4588 unregister_shrinker(&f2fs_shrinker_info);
4591 free_garbage_collection_cache:
4592 f2fs_destroy_garbage_collection_cache();
4594 f2fs_destroy_extent_cache();
4595 free_recovery_cache:
4596 f2fs_destroy_recovery_cache();
4597 free_checkpoint_caches:
4598 f2fs_destroy_checkpoint_caches();
4599 free_segment_manager_caches:
4600 f2fs_destroy_segment_manager_caches();
4601 free_node_manager_caches:
4602 f2fs_destroy_node_manager_caches();
4604 destroy_inodecache();
4609 static void __exit exit_f2fs_fs(void)
4611 f2fs_destroy_casefold_cache();
4612 f2fs_destroy_compress_cache();
4613 f2fs_destroy_compress_mempool();
4614 f2fs_destroy_bioset();
4615 f2fs_destroy_bio_entry_cache();
4616 f2fs_destroy_iostat_processing();
4617 f2fs_destroy_post_read_processing();
4618 f2fs_destroy_root_stats();
4619 unregister_filesystem(&f2fs_fs_type);
4620 unregister_shrinker(&f2fs_shrinker_info);
4622 f2fs_destroy_garbage_collection_cache();
4623 f2fs_destroy_extent_cache();
4624 f2fs_destroy_recovery_cache();
4625 f2fs_destroy_checkpoint_caches();
4626 f2fs_destroy_segment_manager_caches();
4627 f2fs_destroy_node_manager_caches();
4628 destroy_inodecache();
4631 module_init(init_f2fs_fs)
4632 module_exit(exit_f2fs_fs)
4634 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4635 MODULE_DESCRIPTION("Flash Friendly File System");
4636 MODULE_LICENSE("GPL");
4637 MODULE_SOFTDEP("pre: crc32");