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/fs_context.h>
12 #include <linux/sched/mm.h>
13 #include <linux/statfs.h>
14 #include <linux/buffer_head.h>
15 #include <linux/kthread.h>
16 #include <linux/parser.h>
17 #include <linux/mount.h>
18 #include <linux/seq_file.h>
19 #include <linux/proc_fs.h>
20 #include <linux/random.h>
21 #include <linux/exportfs.h>
22 #include <linux/blkdev.h>
23 #include <linux/quotaops.h>
24 #include <linux/f2fs_fs.h>
25 #include <linux/sysfs.h>
26 #include <linux/quota.h>
27 #include <linux/unicode.h>
28 #include <linux/part_stat.h>
29 #include <linux/zstd.h>
30 #include <linux/lz4.h>
39 #define CREATE_TRACE_POINTS
40 #include <trace/events/f2fs.h>
42 static struct kmem_cache *f2fs_inode_cachep;
44 #ifdef CONFIG_F2FS_FAULT_INJECTION
46 const char *f2fs_fault_name[FAULT_MAX] = {
47 [FAULT_KMALLOC] = "kmalloc",
48 [FAULT_KVMALLOC] = "kvmalloc",
49 [FAULT_PAGE_ALLOC] = "page alloc",
50 [FAULT_PAGE_GET] = "page get",
51 [FAULT_ALLOC_NID] = "alloc nid",
52 [FAULT_ORPHAN] = "orphan",
53 [FAULT_BLOCK] = "no more block",
54 [FAULT_DIR_DEPTH] = "too big dir depth",
55 [FAULT_EVICT_INODE] = "evict_inode fail",
56 [FAULT_TRUNCATE] = "truncate fail",
57 [FAULT_READ_IO] = "read IO error",
58 [FAULT_CHECKPOINT] = "checkpoint error",
59 [FAULT_DISCARD] = "discard error",
60 [FAULT_WRITE_IO] = "write IO error",
61 [FAULT_SLAB_ALLOC] = "slab alloc",
62 [FAULT_DQUOT_INIT] = "dquot initialize",
63 [FAULT_LOCK_OP] = "lock_op",
66 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
69 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
72 atomic_set(&ffi->inject_ops, 0);
73 ffi->inject_rate = rate;
77 ffi->inject_type = type;
80 memset(ffi, 0, sizeof(struct f2fs_fault_info));
84 /* f2fs-wide shrinker description */
85 static struct shrinker f2fs_shrinker_info = {
86 .scan_objects = f2fs_shrink_scan,
87 .count_objects = f2fs_shrink_count,
88 .seeks = DEFAULT_SEEKS,
93 Opt_disable_roll_forward,
104 Opt_disable_ext_identify,
107 Opt_inline_xattr_size,
144 Opt_test_dummy_encryption,
146 Opt_checkpoint_disable,
147 Opt_checkpoint_disable_cap,
148 Opt_checkpoint_disable_cap_perc,
149 Opt_checkpoint_enable,
150 Opt_checkpoint_merge,
151 Opt_nocheckpoint_merge,
152 Opt_compress_algorithm,
153 Opt_compress_log_size,
154 Opt_compress_extension,
155 Opt_nocompress_extension,
167 static match_table_t f2fs_tokens = {
168 {Opt_gc_background, "background_gc=%s"},
169 {Opt_disable_roll_forward, "disable_roll_forward"},
170 {Opt_norecovery, "norecovery"},
171 {Opt_discard, "discard"},
172 {Opt_nodiscard, "nodiscard"},
173 {Opt_noheap, "no_heap"},
175 {Opt_user_xattr, "user_xattr"},
176 {Opt_nouser_xattr, "nouser_xattr"},
178 {Opt_noacl, "noacl"},
179 {Opt_active_logs, "active_logs=%u"},
180 {Opt_disable_ext_identify, "disable_ext_identify"},
181 {Opt_inline_xattr, "inline_xattr"},
182 {Opt_noinline_xattr, "noinline_xattr"},
183 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
184 {Opt_inline_data, "inline_data"},
185 {Opt_inline_dentry, "inline_dentry"},
186 {Opt_noinline_dentry, "noinline_dentry"},
187 {Opt_flush_merge, "flush_merge"},
188 {Opt_noflush_merge, "noflush_merge"},
189 {Opt_nobarrier, "nobarrier"},
190 {Opt_fastboot, "fastboot"},
191 {Opt_extent_cache, "extent_cache"},
192 {Opt_noextent_cache, "noextent_cache"},
193 {Opt_noinline_data, "noinline_data"},
194 {Opt_data_flush, "data_flush"},
195 {Opt_reserve_root, "reserve_root=%u"},
196 {Opt_resgid, "resgid=%u"},
197 {Opt_resuid, "resuid=%u"},
198 {Opt_mode, "mode=%s"},
199 {Opt_io_size_bits, "io_bits=%u"},
200 {Opt_fault_injection, "fault_injection=%u"},
201 {Opt_fault_type, "fault_type=%u"},
202 {Opt_lazytime, "lazytime"},
203 {Opt_nolazytime, "nolazytime"},
204 {Opt_quota, "quota"},
205 {Opt_noquota, "noquota"},
206 {Opt_usrquota, "usrquota"},
207 {Opt_grpquota, "grpquota"},
208 {Opt_prjquota, "prjquota"},
209 {Opt_usrjquota, "usrjquota=%s"},
210 {Opt_grpjquota, "grpjquota=%s"},
211 {Opt_prjjquota, "prjjquota=%s"},
212 {Opt_offusrjquota, "usrjquota="},
213 {Opt_offgrpjquota, "grpjquota="},
214 {Opt_offprjjquota, "prjjquota="},
215 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
216 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
217 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
218 {Opt_alloc, "alloc_mode=%s"},
219 {Opt_fsync, "fsync_mode=%s"},
220 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
221 {Opt_test_dummy_encryption, "test_dummy_encryption"},
222 {Opt_inlinecrypt, "inlinecrypt"},
223 {Opt_checkpoint_disable, "checkpoint=disable"},
224 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
225 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
226 {Opt_checkpoint_enable, "checkpoint=enable"},
227 {Opt_checkpoint_merge, "checkpoint_merge"},
228 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
229 {Opt_compress_algorithm, "compress_algorithm=%s"},
230 {Opt_compress_log_size, "compress_log_size=%u"},
231 {Opt_compress_extension, "compress_extension=%s"},
232 {Opt_nocompress_extension, "nocompress_extension=%s"},
233 {Opt_compress_chksum, "compress_chksum"},
234 {Opt_compress_mode, "compress_mode=%s"},
235 {Opt_compress_cache, "compress_cache"},
237 {Opt_gc_merge, "gc_merge"},
238 {Opt_nogc_merge, "nogc_merge"},
239 {Opt_discard_unit, "discard_unit=%s"},
240 {Opt_memory_mode, "memory=%s"},
244 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
246 struct va_format vaf;
252 level = printk_get_level(fmt);
253 vaf.fmt = printk_skip_level(fmt);
255 printk("%c%cF2FS-fs (%s): %pV\n",
256 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
261 #if IS_ENABLED(CONFIG_UNICODE)
262 static const struct f2fs_sb_encodings {
265 unsigned int version;
266 } f2fs_sb_encoding_map[] = {
267 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
270 static const struct f2fs_sb_encodings *
271 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
273 __u16 magic = le16_to_cpu(sb->s_encoding);
276 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
277 if (magic == f2fs_sb_encoding_map[i].magic)
278 return &f2fs_sb_encoding_map[i];
283 struct kmem_cache *f2fs_cf_name_slab;
284 static int __init f2fs_create_casefold_cache(void)
286 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
288 if (!f2fs_cf_name_slab)
293 static void f2fs_destroy_casefold_cache(void)
295 kmem_cache_destroy(f2fs_cf_name_slab);
298 static int __init f2fs_create_casefold_cache(void) { return 0; }
299 static void f2fs_destroy_casefold_cache(void) { }
302 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
304 block_t limit = min((sbi->user_block_count >> 3),
305 sbi->user_block_count - sbi->reserved_blocks);
308 if (test_opt(sbi, RESERVE_ROOT) &&
309 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
310 F2FS_OPTION(sbi).root_reserved_blocks = limit;
311 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
312 F2FS_OPTION(sbi).root_reserved_blocks);
314 if (!test_opt(sbi, RESERVE_ROOT) &&
315 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
316 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
317 !gid_eq(F2FS_OPTION(sbi).s_resgid,
318 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
319 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
320 from_kuid_munged(&init_user_ns,
321 F2FS_OPTION(sbi).s_resuid),
322 from_kgid_munged(&init_user_ns,
323 F2FS_OPTION(sbi).s_resgid));
326 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
328 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
329 unsigned int avg_vblocks;
330 unsigned int wanted_reserved_segments;
331 block_t avail_user_block_count;
333 if (!F2FS_IO_ALIGNED(sbi))
336 /* average valid block count in section in worst case */
337 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
340 * we need enough free space when migrating one section in worst case
342 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
343 reserved_segments(sbi);
344 wanted_reserved_segments -= reserved_segments(sbi);
346 avail_user_block_count = sbi->user_block_count -
347 sbi->current_reserved_blocks -
348 F2FS_OPTION(sbi).root_reserved_blocks;
350 if (wanted_reserved_segments * sbi->blocks_per_seg >
351 avail_user_block_count) {
352 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
353 wanted_reserved_segments,
354 avail_user_block_count >> sbi->log_blocks_per_seg);
358 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
360 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
361 wanted_reserved_segments);
366 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
368 if (!F2FS_OPTION(sbi).unusable_cap_perc)
371 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
372 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
374 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
375 F2FS_OPTION(sbi).unusable_cap_perc;
377 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
378 F2FS_OPTION(sbi).unusable_cap,
379 F2FS_OPTION(sbi).unusable_cap_perc);
382 static void init_once(void *foo)
384 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
386 inode_init_once(&fi->vfs_inode);
390 static const char * const quotatypes[] = INITQFNAMES;
391 #define QTYPE2NAME(t) (quotatypes[t])
392 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
395 struct f2fs_sb_info *sbi = F2FS_SB(sb);
399 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
400 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
403 if (f2fs_sb_has_quota_ino(sbi)) {
404 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
408 qname = match_strdup(args);
410 f2fs_err(sbi, "Not enough memory for storing quotafile name");
413 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
414 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
417 f2fs_err(sbi, "%s quota file already specified",
421 if (strchr(qname, '/')) {
422 f2fs_err(sbi, "quotafile must be on filesystem root");
425 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
433 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
435 struct f2fs_sb_info *sbi = F2FS_SB(sb);
437 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
438 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
441 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
442 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
446 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
449 * We do the test below only for project quotas. 'usrquota' and
450 * 'grpquota' mount options are allowed even without quota feature
451 * to support legacy quotas in quota files.
453 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
454 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
457 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
458 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
459 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
460 if (test_opt(sbi, USRQUOTA) &&
461 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
462 clear_opt(sbi, USRQUOTA);
464 if (test_opt(sbi, GRPQUOTA) &&
465 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
466 clear_opt(sbi, GRPQUOTA);
468 if (test_opt(sbi, PRJQUOTA) &&
469 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
470 clear_opt(sbi, PRJQUOTA);
472 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
473 test_opt(sbi, PRJQUOTA)) {
474 f2fs_err(sbi, "old and new quota format mixing");
478 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
479 f2fs_err(sbi, "journaled quota format not specified");
484 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
485 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
486 F2FS_OPTION(sbi).s_jquota_fmt = 0;
492 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
494 const substring_t *arg,
497 struct f2fs_sb_info *sbi = F2FS_SB(sb);
498 struct fs_parameter param = {
499 .type = fs_value_is_string,
500 .string = arg->from ? arg->from : "",
502 struct fscrypt_dummy_policy *policy =
503 &F2FS_OPTION(sbi).dummy_enc_policy;
506 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
507 f2fs_warn(sbi, "test_dummy_encryption option not supported");
511 if (!f2fs_sb_has_encrypt(sbi)) {
512 f2fs_err(sbi, "Encrypt feature is off");
517 * This mount option is just for testing, and it's not worthwhile to
518 * implement the extra complexity (e.g. RCU protection) that would be
519 * needed to allow it to be set or changed during remount. We do allow
520 * it to be specified during remount, but only if there is no change.
522 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
523 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
527 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
531 "Can't change test_dummy_encryption on remount");
532 else if (err == -EINVAL)
533 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
536 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
540 err = fscrypt_add_test_dummy_key(sb, policy);
542 f2fs_warn(sbi, "Error adding test dummy encryption key [%d]",
546 f2fs_warn(sbi, "Test dummy encryption mode enabled");
550 #ifdef CONFIG_F2FS_FS_COMPRESSION
552 * 1. The same extension name cannot not appear in both compress and non-compress extension
554 * 2. If the compress extension specifies all files, the types specified by the non-compress
555 * extension will be treated as special cases and will not be compressed.
556 * 3. Don't allow the non-compress extension specifies all files.
558 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
560 unsigned char (*ext)[F2FS_EXTENSION_LEN];
561 unsigned char (*noext)[F2FS_EXTENSION_LEN];
562 int ext_cnt, noext_cnt, index = 0, no_index = 0;
564 ext = F2FS_OPTION(sbi).extensions;
565 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
566 noext = F2FS_OPTION(sbi).noextensions;
567 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
572 for (no_index = 0; no_index < noext_cnt; no_index++) {
573 if (!strcasecmp("*", noext[no_index])) {
574 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
577 for (index = 0; index < ext_cnt; index++) {
578 if (!strcasecmp(ext[index], noext[no_index])) {
579 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
588 #ifdef CONFIG_F2FS_FS_LZ4
589 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
591 #ifdef CONFIG_F2FS_FS_LZ4HC
595 if (strlen(str) == 3) {
596 F2FS_OPTION(sbi).compress_level = 0;
600 #ifdef CONFIG_F2FS_FS_LZ4HC
604 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
607 if (kstrtouint(str + 1, 10, &level))
610 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
611 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
615 F2FS_OPTION(sbi).compress_level = level;
618 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
624 #ifdef CONFIG_F2FS_FS_ZSTD
625 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
630 if (strlen(str) == len) {
631 F2FS_OPTION(sbi).compress_level = 0;
638 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
641 if (kstrtouint(str + 1, 10, &level))
644 if (!level || level > zstd_max_clevel()) {
645 f2fs_info(sbi, "invalid zstd compress level: %d", level);
649 F2FS_OPTION(sbi).compress_level = level;
655 static int parse_options(struct super_block *sb, char *options, bool is_remount)
657 struct f2fs_sb_info *sbi = F2FS_SB(sb);
658 substring_t args[MAX_OPT_ARGS];
659 #ifdef CONFIG_F2FS_FS_COMPRESSION
660 unsigned char (*ext)[F2FS_EXTENSION_LEN];
661 unsigned char (*noext)[F2FS_EXTENSION_LEN];
662 int ext_cnt, noext_cnt;
673 while ((p = strsep(&options, ",")) != NULL) {
679 * Initialize args struct so we know whether arg was
680 * found; some options take optional arguments.
682 args[0].to = args[0].from = NULL;
683 token = match_token(p, f2fs_tokens, args);
686 case Opt_gc_background:
687 name = match_strdup(&args[0]);
691 if (!strcmp(name, "on")) {
692 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
693 } else if (!strcmp(name, "off")) {
694 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
695 } else if (!strcmp(name, "sync")) {
696 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
703 case Opt_disable_roll_forward:
704 set_opt(sbi, DISABLE_ROLL_FORWARD);
707 /* this option mounts f2fs with ro */
708 set_opt(sbi, NORECOVERY);
709 if (!f2fs_readonly(sb))
713 if (!f2fs_hw_support_discard(sbi)) {
714 f2fs_warn(sbi, "device does not support discard");
717 set_opt(sbi, DISCARD);
720 if (f2fs_hw_should_discard(sbi)) {
721 f2fs_warn(sbi, "discard is required for zoned block devices");
724 clear_opt(sbi, DISCARD);
727 set_opt(sbi, NOHEAP);
730 clear_opt(sbi, NOHEAP);
732 #ifdef CONFIG_F2FS_FS_XATTR
734 set_opt(sbi, XATTR_USER);
736 case Opt_nouser_xattr:
737 clear_opt(sbi, XATTR_USER);
739 case Opt_inline_xattr:
740 set_opt(sbi, INLINE_XATTR);
742 case Opt_noinline_xattr:
743 clear_opt(sbi, INLINE_XATTR);
745 case Opt_inline_xattr_size:
746 if (args->from && match_int(args, &arg))
748 set_opt(sbi, INLINE_XATTR_SIZE);
749 F2FS_OPTION(sbi).inline_xattr_size = arg;
753 f2fs_info(sbi, "user_xattr options not supported");
755 case Opt_nouser_xattr:
756 f2fs_info(sbi, "nouser_xattr options not supported");
758 case Opt_inline_xattr:
759 f2fs_info(sbi, "inline_xattr options not supported");
761 case Opt_noinline_xattr:
762 f2fs_info(sbi, "noinline_xattr options not supported");
765 #ifdef CONFIG_F2FS_FS_POSIX_ACL
767 set_opt(sbi, POSIX_ACL);
770 clear_opt(sbi, POSIX_ACL);
774 f2fs_info(sbi, "acl options not supported");
777 f2fs_info(sbi, "noacl options not supported");
780 case Opt_active_logs:
781 if (args->from && match_int(args, &arg))
783 if (arg != 2 && arg != 4 &&
784 arg != NR_CURSEG_PERSIST_TYPE)
786 F2FS_OPTION(sbi).active_logs = arg;
788 case Opt_disable_ext_identify:
789 set_opt(sbi, DISABLE_EXT_IDENTIFY);
791 case Opt_inline_data:
792 set_opt(sbi, INLINE_DATA);
794 case Opt_inline_dentry:
795 set_opt(sbi, INLINE_DENTRY);
797 case Opt_noinline_dentry:
798 clear_opt(sbi, INLINE_DENTRY);
800 case Opt_flush_merge:
801 set_opt(sbi, FLUSH_MERGE);
803 case Opt_noflush_merge:
804 clear_opt(sbi, FLUSH_MERGE);
807 set_opt(sbi, NOBARRIER);
810 set_opt(sbi, FASTBOOT);
812 case Opt_extent_cache:
813 set_opt(sbi, READ_EXTENT_CACHE);
815 case Opt_noextent_cache:
816 clear_opt(sbi, READ_EXTENT_CACHE);
818 case Opt_noinline_data:
819 clear_opt(sbi, INLINE_DATA);
822 set_opt(sbi, DATA_FLUSH);
824 case Opt_reserve_root:
825 if (args->from && match_int(args, &arg))
827 if (test_opt(sbi, RESERVE_ROOT)) {
828 f2fs_info(sbi, "Preserve previous reserve_root=%u",
829 F2FS_OPTION(sbi).root_reserved_blocks);
831 F2FS_OPTION(sbi).root_reserved_blocks = arg;
832 set_opt(sbi, RESERVE_ROOT);
836 if (args->from && match_int(args, &arg))
838 uid = make_kuid(current_user_ns(), arg);
839 if (!uid_valid(uid)) {
840 f2fs_err(sbi, "Invalid uid value %d", arg);
843 F2FS_OPTION(sbi).s_resuid = uid;
846 if (args->from && match_int(args, &arg))
848 gid = make_kgid(current_user_ns(), arg);
849 if (!gid_valid(gid)) {
850 f2fs_err(sbi, "Invalid gid value %d", arg);
853 F2FS_OPTION(sbi).s_resgid = gid;
856 name = match_strdup(&args[0]);
860 if (!strcmp(name, "adaptive")) {
861 if (f2fs_sb_has_blkzoned(sbi)) {
862 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
866 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
867 } else if (!strcmp(name, "lfs")) {
868 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
869 } else if (!strcmp(name, "fragment:segment")) {
870 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
871 } else if (!strcmp(name, "fragment:block")) {
872 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
879 case Opt_io_size_bits:
880 if (args->from && match_int(args, &arg))
882 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
883 f2fs_warn(sbi, "Not support %d, larger than %d",
884 1 << arg, BIO_MAX_VECS);
887 F2FS_OPTION(sbi).write_io_size_bits = arg;
889 #ifdef CONFIG_F2FS_FAULT_INJECTION
890 case Opt_fault_injection:
891 if (args->from && match_int(args, &arg))
893 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
894 set_opt(sbi, FAULT_INJECTION);
898 if (args->from && match_int(args, &arg))
900 f2fs_build_fault_attr(sbi, 0, arg);
901 set_opt(sbi, FAULT_INJECTION);
904 case Opt_fault_injection:
905 f2fs_info(sbi, "fault_injection options not supported");
909 f2fs_info(sbi, "fault_type options not supported");
913 sb->s_flags |= SB_LAZYTIME;
916 sb->s_flags &= ~SB_LAZYTIME;
921 set_opt(sbi, USRQUOTA);
924 set_opt(sbi, GRPQUOTA);
927 set_opt(sbi, PRJQUOTA);
930 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
935 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
940 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
944 case Opt_offusrjquota:
945 ret = f2fs_clear_qf_name(sb, USRQUOTA);
949 case Opt_offgrpjquota:
950 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
954 case Opt_offprjjquota:
955 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
959 case Opt_jqfmt_vfsold:
960 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
962 case Opt_jqfmt_vfsv0:
963 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
965 case Opt_jqfmt_vfsv1:
966 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
969 clear_opt(sbi, QUOTA);
970 clear_opt(sbi, USRQUOTA);
971 clear_opt(sbi, GRPQUOTA);
972 clear_opt(sbi, PRJQUOTA);
982 case Opt_offusrjquota:
983 case Opt_offgrpjquota:
984 case Opt_offprjjquota:
985 case Opt_jqfmt_vfsold:
986 case Opt_jqfmt_vfsv0:
987 case Opt_jqfmt_vfsv1:
989 f2fs_info(sbi, "quota operations not supported");
993 name = match_strdup(&args[0]);
997 if (!strcmp(name, "default")) {
998 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
999 } else if (!strcmp(name, "reuse")) {
1000 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1008 name = match_strdup(&args[0]);
1011 if (!strcmp(name, "posix")) {
1012 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1013 } else if (!strcmp(name, "strict")) {
1014 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1015 } else if (!strcmp(name, "nobarrier")) {
1016 F2FS_OPTION(sbi).fsync_mode =
1017 FSYNC_MODE_NOBARRIER;
1024 case Opt_test_dummy_encryption:
1025 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1030 case Opt_inlinecrypt:
1031 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1032 sb->s_flags |= SB_INLINECRYPT;
1034 f2fs_info(sbi, "inline encryption not supported");
1037 case Opt_checkpoint_disable_cap_perc:
1038 if (args->from && match_int(args, &arg))
1040 if (arg < 0 || arg > 100)
1042 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1043 set_opt(sbi, DISABLE_CHECKPOINT);
1045 case Opt_checkpoint_disable_cap:
1046 if (args->from && match_int(args, &arg))
1048 F2FS_OPTION(sbi).unusable_cap = arg;
1049 set_opt(sbi, DISABLE_CHECKPOINT);
1051 case Opt_checkpoint_disable:
1052 set_opt(sbi, DISABLE_CHECKPOINT);
1054 case Opt_checkpoint_enable:
1055 clear_opt(sbi, DISABLE_CHECKPOINT);
1057 case Opt_checkpoint_merge:
1058 set_opt(sbi, MERGE_CHECKPOINT);
1060 case Opt_nocheckpoint_merge:
1061 clear_opt(sbi, MERGE_CHECKPOINT);
1063 #ifdef CONFIG_F2FS_FS_COMPRESSION
1064 case Opt_compress_algorithm:
1065 if (!f2fs_sb_has_compression(sbi)) {
1066 f2fs_info(sbi, "Image doesn't support compression");
1069 name = match_strdup(&args[0]);
1072 if (!strcmp(name, "lzo")) {
1073 #ifdef CONFIG_F2FS_FS_LZO
1074 F2FS_OPTION(sbi).compress_level = 0;
1075 F2FS_OPTION(sbi).compress_algorithm =
1078 f2fs_info(sbi, "kernel doesn't support lzo compression");
1080 } else if (!strncmp(name, "lz4", 3)) {
1081 #ifdef CONFIG_F2FS_FS_LZ4
1082 ret = f2fs_set_lz4hc_level(sbi, name);
1087 F2FS_OPTION(sbi).compress_algorithm =
1090 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1092 } else if (!strncmp(name, "zstd", 4)) {
1093 #ifdef CONFIG_F2FS_FS_ZSTD
1094 ret = f2fs_set_zstd_level(sbi, name);
1099 F2FS_OPTION(sbi).compress_algorithm =
1102 f2fs_info(sbi, "kernel doesn't support zstd compression");
1104 } else if (!strcmp(name, "lzo-rle")) {
1105 #ifdef CONFIG_F2FS_FS_LZORLE
1106 F2FS_OPTION(sbi).compress_level = 0;
1107 F2FS_OPTION(sbi).compress_algorithm =
1110 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1118 case Opt_compress_log_size:
1119 if (!f2fs_sb_has_compression(sbi)) {
1120 f2fs_info(sbi, "Image doesn't support compression");
1123 if (args->from && match_int(args, &arg))
1125 if (arg < MIN_COMPRESS_LOG_SIZE ||
1126 arg > MAX_COMPRESS_LOG_SIZE) {
1128 "Compress cluster log size is out of range");
1131 F2FS_OPTION(sbi).compress_log_size = arg;
1133 case Opt_compress_extension:
1134 if (!f2fs_sb_has_compression(sbi)) {
1135 f2fs_info(sbi, "Image doesn't support compression");
1138 name = match_strdup(&args[0]);
1142 ext = F2FS_OPTION(sbi).extensions;
1143 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1145 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1146 ext_cnt >= COMPRESS_EXT_NUM) {
1148 "invalid extension length/number");
1153 strcpy(ext[ext_cnt], name);
1154 F2FS_OPTION(sbi).compress_ext_cnt++;
1157 case Opt_nocompress_extension:
1158 if (!f2fs_sb_has_compression(sbi)) {
1159 f2fs_info(sbi, "Image doesn't support compression");
1162 name = match_strdup(&args[0]);
1166 noext = F2FS_OPTION(sbi).noextensions;
1167 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1169 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1170 noext_cnt >= COMPRESS_EXT_NUM) {
1172 "invalid extension length/number");
1177 strcpy(noext[noext_cnt], name);
1178 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1181 case Opt_compress_chksum:
1182 F2FS_OPTION(sbi).compress_chksum = true;
1184 case Opt_compress_mode:
1185 name = match_strdup(&args[0]);
1188 if (!strcmp(name, "fs")) {
1189 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1190 } else if (!strcmp(name, "user")) {
1191 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1198 case Opt_compress_cache:
1199 set_opt(sbi, COMPRESS_CACHE);
1202 case Opt_compress_algorithm:
1203 case Opt_compress_log_size:
1204 case Opt_compress_extension:
1205 case Opt_nocompress_extension:
1206 case Opt_compress_chksum:
1207 case Opt_compress_mode:
1208 case Opt_compress_cache:
1209 f2fs_info(sbi, "compression options not supported");
1216 set_opt(sbi, GC_MERGE);
1218 case Opt_nogc_merge:
1219 clear_opt(sbi, GC_MERGE);
1221 case Opt_discard_unit:
1222 name = match_strdup(&args[0]);
1225 if (!strcmp(name, "block")) {
1226 F2FS_OPTION(sbi).discard_unit =
1228 } else if (!strcmp(name, "segment")) {
1229 F2FS_OPTION(sbi).discard_unit =
1230 DISCARD_UNIT_SEGMENT;
1231 } else if (!strcmp(name, "section")) {
1232 F2FS_OPTION(sbi).discard_unit =
1233 DISCARD_UNIT_SECTION;
1240 case Opt_memory_mode:
1241 name = match_strdup(&args[0]);
1244 if (!strcmp(name, "normal")) {
1245 F2FS_OPTION(sbi).memory_mode =
1247 } else if (!strcmp(name, "low")) {
1248 F2FS_OPTION(sbi).memory_mode =
1257 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1264 if (f2fs_check_quota_options(sbi))
1267 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1268 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1271 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1272 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1276 #if !IS_ENABLED(CONFIG_UNICODE)
1277 if (f2fs_sb_has_casefold(sbi)) {
1279 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1284 * The BLKZONED feature indicates that the drive was formatted with
1285 * zone alignment optimization. This is optional for host-aware
1286 * devices, but mandatory for host-managed zoned block devices.
1288 #ifndef CONFIG_BLK_DEV_ZONED
1289 if (f2fs_sb_has_blkzoned(sbi)) {
1290 f2fs_err(sbi, "Zoned block device support is not enabled");
1294 if (f2fs_sb_has_blkzoned(sbi)) {
1295 if (F2FS_OPTION(sbi).discard_unit !=
1296 DISCARD_UNIT_SECTION) {
1297 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1298 F2FS_OPTION(sbi).discard_unit =
1299 DISCARD_UNIT_SECTION;
1303 #ifdef CONFIG_F2FS_FS_COMPRESSION
1304 if (f2fs_test_compress_extension(sbi)) {
1305 f2fs_err(sbi, "invalid compress or nocompress extension");
1310 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1311 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1312 F2FS_IO_SIZE_KB(sbi));
1316 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1317 int min_size, max_size;
1319 if (!f2fs_sb_has_extra_attr(sbi) ||
1320 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1321 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1324 if (!test_opt(sbi, INLINE_XATTR)) {
1325 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1329 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1330 max_size = MAX_INLINE_XATTR_SIZE;
1332 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1333 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1334 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1335 min_size, max_size);
1340 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1341 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1345 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1346 f2fs_err(sbi, "LFS not compatible with ATGC");
1350 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1351 f2fs_err(sbi, "Allow to mount readonly mode only");
1357 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1359 struct f2fs_inode_info *fi;
1361 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) {
1362 f2fs_show_injection_info(F2FS_SB(sb), FAULT_SLAB_ALLOC);
1366 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1370 init_once((void *) fi);
1372 /* Initialize f2fs-specific inode info */
1373 atomic_set(&fi->dirty_pages, 0);
1374 atomic_set(&fi->i_compr_blocks, 0);
1375 init_f2fs_rwsem(&fi->i_sem);
1376 spin_lock_init(&fi->i_size_lock);
1377 INIT_LIST_HEAD(&fi->dirty_list);
1378 INIT_LIST_HEAD(&fi->gdirty_list);
1379 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1380 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1381 init_f2fs_rwsem(&fi->i_xattr_sem);
1383 /* Will be used by directory only */
1384 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1386 return &fi->vfs_inode;
1389 static int f2fs_drop_inode(struct inode *inode)
1391 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1395 * during filesystem shutdown, if checkpoint is disabled,
1396 * drop useless meta/node dirty pages.
1398 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1399 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1400 inode->i_ino == F2FS_META_INO(sbi)) {
1401 trace_f2fs_drop_inode(inode, 1);
1407 * This is to avoid a deadlock condition like below.
1408 * writeback_single_inode(inode)
1409 * - f2fs_write_data_page
1410 * - f2fs_gc -> iput -> evict
1411 * - inode_wait_for_writeback(inode)
1413 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1414 if (!inode->i_nlink && !is_bad_inode(inode)) {
1415 /* to avoid evict_inode call simultaneously */
1416 atomic_inc(&inode->i_count);
1417 spin_unlock(&inode->i_lock);
1419 /* should remain fi->extent_tree for writepage */
1420 f2fs_destroy_extent_node(inode);
1422 sb_start_intwrite(inode->i_sb);
1423 f2fs_i_size_write(inode, 0);
1425 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1426 inode, NULL, 0, DATA);
1427 truncate_inode_pages_final(inode->i_mapping);
1429 if (F2FS_HAS_BLOCKS(inode))
1430 f2fs_truncate(inode);
1432 sb_end_intwrite(inode->i_sb);
1434 spin_lock(&inode->i_lock);
1435 atomic_dec(&inode->i_count);
1437 trace_f2fs_drop_inode(inode, 0);
1440 ret = generic_drop_inode(inode);
1442 ret = fscrypt_drop_inode(inode);
1443 trace_f2fs_drop_inode(inode, ret);
1447 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1449 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1452 spin_lock(&sbi->inode_lock[DIRTY_META]);
1453 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1456 set_inode_flag(inode, FI_DIRTY_INODE);
1457 stat_inc_dirty_inode(sbi, DIRTY_META);
1459 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1460 list_add_tail(&F2FS_I(inode)->gdirty_list,
1461 &sbi->inode_list[DIRTY_META]);
1462 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1464 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1468 void f2fs_inode_synced(struct inode *inode)
1470 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1472 spin_lock(&sbi->inode_lock[DIRTY_META]);
1473 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1474 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1477 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1478 list_del_init(&F2FS_I(inode)->gdirty_list);
1479 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1481 clear_inode_flag(inode, FI_DIRTY_INODE);
1482 clear_inode_flag(inode, FI_AUTO_RECOVER);
1483 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1484 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1488 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1490 * We should call set_dirty_inode to write the dirty inode through write_inode.
1492 static void f2fs_dirty_inode(struct inode *inode, int flags)
1494 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1496 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1497 inode->i_ino == F2FS_META_INO(sbi))
1500 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1501 clear_inode_flag(inode, FI_AUTO_RECOVER);
1503 f2fs_inode_dirtied(inode, false);
1506 static void f2fs_free_inode(struct inode *inode)
1508 fscrypt_free_inode(inode);
1509 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1512 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1514 percpu_counter_destroy(&sbi->total_valid_inode_count);
1515 percpu_counter_destroy(&sbi->rf_node_block_count);
1516 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1519 static void destroy_device_list(struct f2fs_sb_info *sbi)
1523 for (i = 0; i < sbi->s_ndevs; i++) {
1524 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1525 #ifdef CONFIG_BLK_DEV_ZONED
1526 kvfree(FDEV(i).blkz_seq);
1532 static void f2fs_put_super(struct super_block *sb)
1534 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1538 /* unregister procfs/sysfs entries in advance to avoid race case */
1539 f2fs_unregister_sysfs(sbi);
1541 f2fs_quota_off_umount(sb);
1543 /* prevent remaining shrinker jobs */
1544 mutex_lock(&sbi->umount_mutex);
1547 * flush all issued checkpoints and stop checkpoint issue thread.
1548 * after then, all checkpoints should be done by each process context.
1550 f2fs_stop_ckpt_thread(sbi);
1553 * We don't need to do checkpoint when superblock is clean.
1554 * But, the previous checkpoint was not done by umount, it needs to do
1555 * clean checkpoint again.
1557 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1558 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1559 struct cp_control cpc = {
1560 .reason = CP_UMOUNT,
1562 f2fs_write_checkpoint(sbi, &cpc);
1565 /* be sure to wait for any on-going discard commands */
1566 dropped = f2fs_issue_discard_timeout(sbi);
1568 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1569 !sbi->discard_blks && !dropped) {
1570 struct cp_control cpc = {
1571 .reason = CP_UMOUNT | CP_TRIMMED,
1573 f2fs_write_checkpoint(sbi, &cpc);
1577 * normally superblock is clean, so we need to release this.
1578 * In addition, EIO will skip do checkpoint, we need this as well.
1580 f2fs_release_ino_entry(sbi, true);
1582 f2fs_leave_shrinker(sbi);
1583 mutex_unlock(&sbi->umount_mutex);
1585 /* our cp_error case, we can wait for any writeback page */
1586 f2fs_flush_merged_writes(sbi);
1588 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1590 f2fs_bug_on(sbi, sbi->fsync_node_num);
1592 f2fs_destroy_compress_inode(sbi);
1594 iput(sbi->node_inode);
1595 sbi->node_inode = NULL;
1597 iput(sbi->meta_inode);
1598 sbi->meta_inode = NULL;
1601 * iput() can update stat information, if f2fs_write_checkpoint()
1602 * above failed with error.
1604 f2fs_destroy_stats(sbi);
1606 /* destroy f2fs internal modules */
1607 f2fs_destroy_node_manager(sbi);
1608 f2fs_destroy_segment_manager(sbi);
1610 f2fs_destroy_post_read_wq(sbi);
1614 sb->s_fs_info = NULL;
1615 if (sbi->s_chksum_driver)
1616 crypto_free_shash(sbi->s_chksum_driver);
1617 kfree(sbi->raw_super);
1619 destroy_device_list(sbi);
1620 f2fs_destroy_page_array_cache(sbi);
1621 f2fs_destroy_xattr_caches(sbi);
1622 mempool_destroy(sbi->write_io_dummy);
1624 for (i = 0; i < MAXQUOTAS; i++)
1625 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1627 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1628 destroy_percpu_info(sbi);
1629 f2fs_destroy_iostat(sbi);
1630 for (i = 0; i < NR_PAGE_TYPE; i++)
1631 kvfree(sbi->write_io[i]);
1632 #if IS_ENABLED(CONFIG_UNICODE)
1633 utf8_unload(sb->s_encoding);
1638 int f2fs_sync_fs(struct super_block *sb, int sync)
1640 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1643 if (unlikely(f2fs_cp_error(sbi)))
1645 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1648 trace_f2fs_sync_fs(sb, sync);
1650 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1654 err = f2fs_issue_checkpoint(sbi);
1659 static int f2fs_freeze(struct super_block *sb)
1661 if (f2fs_readonly(sb))
1664 /* IO error happened before */
1665 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1668 /* must be clean, since sync_filesystem() was already called */
1669 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1672 /* Let's flush checkpoints and stop the thread. */
1673 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1675 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1676 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1680 static int f2fs_unfreeze(struct super_block *sb)
1682 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1687 static int f2fs_statfs_project(struct super_block *sb,
1688 kprojid_t projid, struct kstatfs *buf)
1691 struct dquot *dquot;
1695 qid = make_kqid_projid(projid);
1696 dquot = dqget(sb, qid);
1698 return PTR_ERR(dquot);
1699 spin_lock(&dquot->dq_dqb_lock);
1701 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1702 dquot->dq_dqb.dqb_bhardlimit);
1704 limit >>= sb->s_blocksize_bits;
1706 if (limit && buf->f_blocks > limit) {
1707 curblock = (dquot->dq_dqb.dqb_curspace +
1708 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1709 buf->f_blocks = limit;
1710 buf->f_bfree = buf->f_bavail =
1711 (buf->f_blocks > curblock) ?
1712 (buf->f_blocks - curblock) : 0;
1715 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1716 dquot->dq_dqb.dqb_ihardlimit);
1718 if (limit && buf->f_files > limit) {
1719 buf->f_files = limit;
1721 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1722 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1725 spin_unlock(&dquot->dq_dqb_lock);
1731 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1733 struct super_block *sb = dentry->d_sb;
1734 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1735 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1736 block_t total_count, user_block_count, start_count;
1737 u64 avail_node_count;
1738 unsigned int total_valid_node_count;
1740 total_count = le64_to_cpu(sbi->raw_super->block_count);
1741 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1742 buf->f_type = F2FS_SUPER_MAGIC;
1743 buf->f_bsize = sbi->blocksize;
1745 buf->f_blocks = total_count - start_count;
1747 spin_lock(&sbi->stat_lock);
1749 user_block_count = sbi->user_block_count;
1750 total_valid_node_count = valid_node_count(sbi);
1751 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1752 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1753 sbi->current_reserved_blocks;
1755 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1758 buf->f_bfree -= sbi->unusable_block_count;
1759 spin_unlock(&sbi->stat_lock);
1761 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1762 buf->f_bavail = buf->f_bfree -
1763 F2FS_OPTION(sbi).root_reserved_blocks;
1767 if (avail_node_count > user_block_count) {
1768 buf->f_files = user_block_count;
1769 buf->f_ffree = buf->f_bavail;
1771 buf->f_files = avail_node_count;
1772 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1776 buf->f_namelen = F2FS_NAME_LEN;
1777 buf->f_fsid = u64_to_fsid(id);
1780 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1781 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1782 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1788 static inline void f2fs_show_quota_options(struct seq_file *seq,
1789 struct super_block *sb)
1792 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1794 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1797 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1808 seq_printf(seq, ",jqfmt=%s", fmtname);
1811 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1812 seq_show_option(seq, "usrjquota",
1813 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1815 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1816 seq_show_option(seq, "grpjquota",
1817 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1819 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1820 seq_show_option(seq, "prjjquota",
1821 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1825 #ifdef CONFIG_F2FS_FS_COMPRESSION
1826 static inline void f2fs_show_compress_options(struct seq_file *seq,
1827 struct super_block *sb)
1829 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1833 if (!f2fs_sb_has_compression(sbi))
1836 switch (F2FS_OPTION(sbi).compress_algorithm) {
1846 case COMPRESS_LZORLE:
1847 algtype = "lzo-rle";
1850 seq_printf(seq, ",compress_algorithm=%s", algtype);
1852 if (F2FS_OPTION(sbi).compress_level)
1853 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1855 seq_printf(seq, ",compress_log_size=%u",
1856 F2FS_OPTION(sbi).compress_log_size);
1858 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1859 seq_printf(seq, ",compress_extension=%s",
1860 F2FS_OPTION(sbi).extensions[i]);
1863 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1864 seq_printf(seq, ",nocompress_extension=%s",
1865 F2FS_OPTION(sbi).noextensions[i]);
1868 if (F2FS_OPTION(sbi).compress_chksum)
1869 seq_puts(seq, ",compress_chksum");
1871 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1872 seq_printf(seq, ",compress_mode=%s", "fs");
1873 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1874 seq_printf(seq, ",compress_mode=%s", "user");
1876 if (test_opt(sbi, COMPRESS_CACHE))
1877 seq_puts(seq, ",compress_cache");
1881 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1883 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1885 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1886 seq_printf(seq, ",background_gc=%s", "sync");
1887 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1888 seq_printf(seq, ",background_gc=%s", "on");
1889 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1890 seq_printf(seq, ",background_gc=%s", "off");
1892 if (test_opt(sbi, GC_MERGE))
1893 seq_puts(seq, ",gc_merge");
1895 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1896 seq_puts(seq, ",disable_roll_forward");
1897 if (test_opt(sbi, NORECOVERY))
1898 seq_puts(seq, ",norecovery");
1899 if (test_opt(sbi, DISCARD))
1900 seq_puts(seq, ",discard");
1902 seq_puts(seq, ",nodiscard");
1903 if (test_opt(sbi, NOHEAP))
1904 seq_puts(seq, ",no_heap");
1906 seq_puts(seq, ",heap");
1907 #ifdef CONFIG_F2FS_FS_XATTR
1908 if (test_opt(sbi, XATTR_USER))
1909 seq_puts(seq, ",user_xattr");
1911 seq_puts(seq, ",nouser_xattr");
1912 if (test_opt(sbi, INLINE_XATTR))
1913 seq_puts(seq, ",inline_xattr");
1915 seq_puts(seq, ",noinline_xattr");
1916 if (test_opt(sbi, INLINE_XATTR_SIZE))
1917 seq_printf(seq, ",inline_xattr_size=%u",
1918 F2FS_OPTION(sbi).inline_xattr_size);
1920 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1921 if (test_opt(sbi, POSIX_ACL))
1922 seq_puts(seq, ",acl");
1924 seq_puts(seq, ",noacl");
1926 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1927 seq_puts(seq, ",disable_ext_identify");
1928 if (test_opt(sbi, INLINE_DATA))
1929 seq_puts(seq, ",inline_data");
1931 seq_puts(seq, ",noinline_data");
1932 if (test_opt(sbi, INLINE_DENTRY))
1933 seq_puts(seq, ",inline_dentry");
1935 seq_puts(seq, ",noinline_dentry");
1936 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1937 seq_puts(seq, ",flush_merge");
1938 if (test_opt(sbi, NOBARRIER))
1939 seq_puts(seq, ",nobarrier");
1940 if (test_opt(sbi, FASTBOOT))
1941 seq_puts(seq, ",fastboot");
1942 if (test_opt(sbi, READ_EXTENT_CACHE))
1943 seq_puts(seq, ",extent_cache");
1945 seq_puts(seq, ",noextent_cache");
1946 if (test_opt(sbi, DATA_FLUSH))
1947 seq_puts(seq, ",data_flush");
1949 seq_puts(seq, ",mode=");
1950 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1951 seq_puts(seq, "adaptive");
1952 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1953 seq_puts(seq, "lfs");
1954 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1955 seq_puts(seq, "fragment:segment");
1956 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1957 seq_puts(seq, "fragment:block");
1958 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1959 if (test_opt(sbi, RESERVE_ROOT))
1960 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1961 F2FS_OPTION(sbi).root_reserved_blocks,
1962 from_kuid_munged(&init_user_ns,
1963 F2FS_OPTION(sbi).s_resuid),
1964 from_kgid_munged(&init_user_ns,
1965 F2FS_OPTION(sbi).s_resgid));
1966 if (F2FS_IO_SIZE_BITS(sbi))
1967 seq_printf(seq, ",io_bits=%u",
1968 F2FS_OPTION(sbi).write_io_size_bits);
1969 #ifdef CONFIG_F2FS_FAULT_INJECTION
1970 if (test_opt(sbi, FAULT_INJECTION)) {
1971 seq_printf(seq, ",fault_injection=%u",
1972 F2FS_OPTION(sbi).fault_info.inject_rate);
1973 seq_printf(seq, ",fault_type=%u",
1974 F2FS_OPTION(sbi).fault_info.inject_type);
1978 if (test_opt(sbi, QUOTA))
1979 seq_puts(seq, ",quota");
1980 if (test_opt(sbi, USRQUOTA))
1981 seq_puts(seq, ",usrquota");
1982 if (test_opt(sbi, GRPQUOTA))
1983 seq_puts(seq, ",grpquota");
1984 if (test_opt(sbi, PRJQUOTA))
1985 seq_puts(seq, ",prjquota");
1987 f2fs_show_quota_options(seq, sbi->sb);
1989 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1991 if (sbi->sb->s_flags & SB_INLINECRYPT)
1992 seq_puts(seq, ",inlinecrypt");
1994 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1995 seq_printf(seq, ",alloc_mode=%s", "default");
1996 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1997 seq_printf(seq, ",alloc_mode=%s", "reuse");
1999 if (test_opt(sbi, DISABLE_CHECKPOINT))
2000 seq_printf(seq, ",checkpoint=disable:%u",
2001 F2FS_OPTION(sbi).unusable_cap);
2002 if (test_opt(sbi, MERGE_CHECKPOINT))
2003 seq_puts(seq, ",checkpoint_merge");
2005 seq_puts(seq, ",nocheckpoint_merge");
2006 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2007 seq_printf(seq, ",fsync_mode=%s", "posix");
2008 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2009 seq_printf(seq, ",fsync_mode=%s", "strict");
2010 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2011 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2013 #ifdef CONFIG_F2FS_FS_COMPRESSION
2014 f2fs_show_compress_options(seq, sbi->sb);
2017 if (test_opt(sbi, ATGC))
2018 seq_puts(seq, ",atgc");
2020 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2021 seq_printf(seq, ",discard_unit=%s", "block");
2022 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2023 seq_printf(seq, ",discard_unit=%s", "segment");
2024 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2025 seq_printf(seq, ",discard_unit=%s", "section");
2027 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2028 seq_printf(seq, ",memory=%s", "normal");
2029 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2030 seq_printf(seq, ",memory=%s", "low");
2035 static void default_options(struct f2fs_sb_info *sbi)
2037 /* init some FS parameters */
2038 if (f2fs_sb_has_readonly(sbi))
2039 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2041 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2043 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2044 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2045 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2046 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2047 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2048 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2049 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2050 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2051 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2052 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2053 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2055 sbi->sb->s_flags &= ~SB_INLINECRYPT;
2057 set_opt(sbi, INLINE_XATTR);
2058 set_opt(sbi, INLINE_DATA);
2059 set_opt(sbi, INLINE_DENTRY);
2060 set_opt(sbi, READ_EXTENT_CACHE);
2061 set_opt(sbi, NOHEAP);
2062 clear_opt(sbi, DISABLE_CHECKPOINT);
2063 set_opt(sbi, MERGE_CHECKPOINT);
2064 F2FS_OPTION(sbi).unusable_cap = 0;
2065 sbi->sb->s_flags |= SB_LAZYTIME;
2066 set_opt(sbi, FLUSH_MERGE);
2067 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2068 set_opt(sbi, DISCARD);
2069 if (f2fs_sb_has_blkzoned(sbi)) {
2070 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2071 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2073 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2074 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2077 #ifdef CONFIG_F2FS_FS_XATTR
2078 set_opt(sbi, XATTR_USER);
2080 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2081 set_opt(sbi, POSIX_ACL);
2084 f2fs_build_fault_attr(sbi, 0, 0);
2088 static int f2fs_enable_quotas(struct super_block *sb);
2091 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2093 unsigned int s_flags = sbi->sb->s_flags;
2094 struct cp_control cpc;
2095 unsigned int gc_mode = sbi->gc_mode;
2100 if (s_flags & SB_RDONLY) {
2101 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2104 sbi->sb->s_flags |= SB_ACTIVE;
2106 /* check if we need more GC first */
2107 unusable = f2fs_get_unusable_blocks(sbi);
2108 if (!f2fs_disable_cp_again(sbi, unusable))
2111 f2fs_update_time(sbi, DISABLE_TIME);
2113 sbi->gc_mode = GC_URGENT_HIGH;
2115 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2116 struct f2fs_gc_control gc_control = {
2117 .victim_segno = NULL_SEGNO,
2118 .init_gc_type = FG_GC,
2119 .should_migrate_blocks = false,
2120 .err_gc_skipped = true,
2121 .nr_free_secs = 1 };
2123 f2fs_down_write(&sbi->gc_lock);
2124 err = f2fs_gc(sbi, &gc_control);
2125 if (err == -ENODATA) {
2129 if (err && err != -EAGAIN)
2133 ret = sync_filesystem(sbi->sb);
2135 err = ret ? ret : err;
2139 unusable = f2fs_get_unusable_blocks(sbi);
2140 if (f2fs_disable_cp_again(sbi, unusable)) {
2146 f2fs_down_write(&sbi->gc_lock);
2147 cpc.reason = CP_PAUSE;
2148 set_sbi_flag(sbi, SBI_CP_DISABLED);
2149 err = f2fs_write_checkpoint(sbi, &cpc);
2153 spin_lock(&sbi->stat_lock);
2154 sbi->unusable_block_count = unusable;
2155 spin_unlock(&sbi->stat_lock);
2158 f2fs_up_write(&sbi->gc_lock);
2160 sbi->gc_mode = gc_mode;
2161 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2165 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2167 int retry = DEFAULT_RETRY_IO_COUNT;
2169 /* we should flush all the data to keep data consistency */
2171 sync_inodes_sb(sbi->sb);
2172 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2173 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2175 if (unlikely(retry < 0))
2176 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2178 f2fs_down_write(&sbi->gc_lock);
2179 f2fs_dirty_to_prefree(sbi);
2181 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2182 set_sbi_flag(sbi, SBI_IS_DIRTY);
2183 f2fs_up_write(&sbi->gc_lock);
2185 f2fs_sync_fs(sbi->sb, 1);
2187 /* Let's ensure there's no pending checkpoint anymore */
2188 f2fs_flush_ckpt_thread(sbi);
2191 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2193 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2194 struct f2fs_mount_info org_mount_opt;
2195 unsigned long old_sb_flags;
2197 bool need_restart_gc = false, need_stop_gc = false;
2198 bool need_restart_ckpt = false, need_stop_ckpt = false;
2199 bool need_restart_flush = false, need_stop_flush = false;
2200 bool need_restart_discard = false, need_stop_discard = false;
2201 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2202 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2203 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2204 bool no_atgc = !test_opt(sbi, ATGC);
2205 bool no_discard = !test_opt(sbi, DISCARD);
2206 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2207 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2208 struct discard_cmd_control *dcc;
2214 * Save the old mount options in case we
2215 * need to restore them.
2217 org_mount_opt = sbi->mount_opt;
2218 old_sb_flags = sb->s_flags;
2221 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2222 for (i = 0; i < MAXQUOTAS; i++) {
2223 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2224 org_mount_opt.s_qf_names[i] =
2225 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2227 if (!org_mount_opt.s_qf_names[i]) {
2228 for (j = 0; j < i; j++)
2229 kfree(org_mount_opt.s_qf_names[j]);
2233 org_mount_opt.s_qf_names[i] = NULL;
2238 /* recover superblocks we couldn't write due to previous RO mount */
2239 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2240 err = f2fs_commit_super(sbi, false);
2241 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2244 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2247 default_options(sbi);
2249 /* parse mount options */
2250 err = parse_options(sb, data, true);
2255 * Previous and new state of filesystem is RO,
2256 * so skip checking GC and FLUSH_MERGE conditions.
2258 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2261 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2267 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2268 err = dquot_suspend(sb, -1);
2271 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2272 /* dquot_resume needs RW */
2273 sb->s_flags &= ~SB_RDONLY;
2274 if (sb_any_quota_suspended(sb)) {
2275 dquot_resume(sb, -1);
2276 } else if (f2fs_sb_has_quota_ino(sbi)) {
2277 err = f2fs_enable_quotas(sb);
2283 /* disallow enable atgc dynamically */
2284 if (no_atgc == !!test_opt(sbi, ATGC)) {
2286 f2fs_warn(sbi, "switch atgc option is not allowed");
2290 /* disallow enable/disable extent_cache dynamically */
2291 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2293 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2297 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2299 f2fs_warn(sbi, "switch io_bits option is not allowed");
2303 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2305 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2309 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2311 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2315 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2317 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2322 * We stop the GC thread if FS is mounted as RO
2323 * or if background_gc = off is passed in mount
2324 * option. Also sync the filesystem.
2326 if ((*flags & SB_RDONLY) ||
2327 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2328 !test_opt(sbi, GC_MERGE))) {
2329 if (sbi->gc_thread) {
2330 f2fs_stop_gc_thread(sbi);
2331 need_restart_gc = true;
2333 } else if (!sbi->gc_thread) {
2334 err = f2fs_start_gc_thread(sbi);
2337 need_stop_gc = true;
2340 if (*flags & SB_RDONLY) {
2343 set_sbi_flag(sbi, SBI_IS_DIRTY);
2344 set_sbi_flag(sbi, SBI_IS_CLOSE);
2345 f2fs_sync_fs(sb, 1);
2346 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2349 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2350 !test_opt(sbi, MERGE_CHECKPOINT)) {
2351 f2fs_stop_ckpt_thread(sbi);
2352 need_restart_ckpt = true;
2354 /* Flush if the prevous checkpoint, if exists. */
2355 f2fs_flush_ckpt_thread(sbi);
2357 err = f2fs_start_ckpt_thread(sbi);
2360 "Failed to start F2FS issue_checkpoint_thread (%d)",
2364 need_stop_ckpt = true;
2368 * We stop issue flush thread if FS is mounted as RO
2369 * or if flush_merge is not passed in mount option.
2371 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2372 clear_opt(sbi, FLUSH_MERGE);
2373 f2fs_destroy_flush_cmd_control(sbi, false);
2374 need_restart_flush = true;
2376 err = f2fs_create_flush_cmd_control(sbi);
2379 need_stop_flush = true;
2382 if (no_discard == !!test_opt(sbi, DISCARD)) {
2383 if (test_opt(sbi, DISCARD)) {
2384 err = f2fs_start_discard_thread(sbi);
2387 need_stop_discard = true;
2389 dcc = SM_I(sbi)->dcc_info;
2390 f2fs_stop_discard_thread(sbi);
2391 if (atomic_read(&dcc->discard_cmd_cnt))
2392 f2fs_issue_discard_timeout(sbi);
2393 need_restart_discard = true;
2397 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2398 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2399 err = f2fs_disable_checkpoint(sbi);
2401 goto restore_discard;
2403 f2fs_enable_checkpoint(sbi);
2409 /* Release old quota file names */
2410 for (i = 0; i < MAXQUOTAS; i++)
2411 kfree(org_mount_opt.s_qf_names[i]);
2413 /* Update the POSIXACL Flag */
2414 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2415 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2417 limit_reserve_root(sbi);
2418 adjust_unusable_cap_perc(sbi);
2419 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2422 if (need_restart_discard) {
2423 if (f2fs_start_discard_thread(sbi))
2424 f2fs_warn(sbi, "discard has been stopped");
2425 } else if (need_stop_discard) {
2426 f2fs_stop_discard_thread(sbi);
2429 if (need_restart_flush) {
2430 if (f2fs_create_flush_cmd_control(sbi))
2431 f2fs_warn(sbi, "background flush thread has stopped");
2432 } else if (need_stop_flush) {
2433 clear_opt(sbi, FLUSH_MERGE);
2434 f2fs_destroy_flush_cmd_control(sbi, false);
2437 if (need_restart_ckpt) {
2438 if (f2fs_start_ckpt_thread(sbi))
2439 f2fs_warn(sbi, "background ckpt thread has stopped");
2440 } else if (need_stop_ckpt) {
2441 f2fs_stop_ckpt_thread(sbi);
2444 if (need_restart_gc) {
2445 if (f2fs_start_gc_thread(sbi))
2446 f2fs_warn(sbi, "background gc thread has stopped");
2447 } else if (need_stop_gc) {
2448 f2fs_stop_gc_thread(sbi);
2452 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2453 for (i = 0; i < MAXQUOTAS; i++) {
2454 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2455 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2458 sbi->mount_opt = org_mount_opt;
2459 sb->s_flags = old_sb_flags;
2464 /* Read data from quotafile */
2465 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2466 size_t len, loff_t off)
2468 struct inode *inode = sb_dqopt(sb)->files[type];
2469 struct address_space *mapping = inode->i_mapping;
2470 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2471 int offset = off & (sb->s_blocksize - 1);
2474 loff_t i_size = i_size_read(inode);
2480 if (off + len > i_size)
2483 while (toread > 0) {
2484 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2486 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2488 if (PTR_ERR(page) == -ENOMEM) {
2489 memalloc_retry_wait(GFP_NOFS);
2492 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2493 return PTR_ERR(page);
2498 if (unlikely(page->mapping != mapping)) {
2499 f2fs_put_page(page, 1);
2502 if (unlikely(!PageUptodate(page))) {
2503 f2fs_put_page(page, 1);
2504 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2508 memcpy_from_page(data, page, offset, tocopy);
2509 f2fs_put_page(page, 1);
2519 /* Write to quotafile */
2520 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2521 const char *data, size_t len, loff_t off)
2523 struct inode *inode = sb_dqopt(sb)->files[type];
2524 struct address_space *mapping = inode->i_mapping;
2525 const struct address_space_operations *a_ops = mapping->a_ops;
2526 int offset = off & (sb->s_blocksize - 1);
2527 size_t towrite = len;
2529 void *fsdata = NULL;
2533 while (towrite > 0) {
2534 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2537 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2539 if (unlikely(err)) {
2540 if (err == -ENOMEM) {
2541 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2544 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2548 memcpy_to_page(page, offset, data, tocopy);
2550 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2561 inode->i_mtime = inode->i_ctime = current_time(inode);
2562 f2fs_mark_inode_dirty_sync(inode, false);
2563 return len - towrite;
2566 int f2fs_dquot_initialize(struct inode *inode)
2568 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) {
2569 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT);
2573 return dquot_initialize(inode);
2576 static struct dquot **f2fs_get_dquots(struct inode *inode)
2578 return F2FS_I(inode)->i_dquot;
2581 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2583 return &F2FS_I(inode)->i_reserved_quota;
2586 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2588 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2589 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2593 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2594 F2FS_OPTION(sbi).s_jquota_fmt, type);
2597 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2602 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2603 err = f2fs_enable_quotas(sbi->sb);
2605 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2611 for (i = 0; i < MAXQUOTAS; i++) {
2612 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2613 err = f2fs_quota_on_mount(sbi, i);
2618 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2625 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2628 struct inode *qf_inode;
2629 unsigned long qf_inum;
2632 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2634 qf_inum = f2fs_qf_ino(sb, type);
2638 qf_inode = f2fs_iget(sb, qf_inum);
2639 if (IS_ERR(qf_inode)) {
2640 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2641 return PTR_ERR(qf_inode);
2644 /* Don't account quota for quota files to avoid recursion */
2645 qf_inode->i_flags |= S_NOQUOTA;
2646 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2651 static int f2fs_enable_quotas(struct super_block *sb)
2653 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2655 unsigned long qf_inum;
2656 bool quota_mopt[MAXQUOTAS] = {
2657 test_opt(sbi, USRQUOTA),
2658 test_opt(sbi, GRPQUOTA),
2659 test_opt(sbi, PRJQUOTA),
2662 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2663 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2667 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2669 for (type = 0; type < MAXQUOTAS; type++) {
2670 qf_inum = f2fs_qf_ino(sb, type);
2672 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2673 DQUOT_USAGE_ENABLED |
2674 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2676 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2678 for (type--; type >= 0; type--)
2679 dquot_quota_off(sb, type);
2680 set_sbi_flag(F2FS_SB(sb),
2681 SBI_QUOTA_NEED_REPAIR);
2689 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2691 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2692 struct address_space *mapping = dqopt->files[type]->i_mapping;
2695 ret = dquot_writeback_dquots(sbi->sb, type);
2699 ret = filemap_fdatawrite(mapping);
2703 /* if we are using journalled quota */
2704 if (is_journalled_quota(sbi))
2707 ret = filemap_fdatawait(mapping);
2709 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2712 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2716 int f2fs_quota_sync(struct super_block *sb, int type)
2718 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2719 struct quota_info *dqopt = sb_dqopt(sb);
2724 * Now when everything is written we can discard the pagecache so
2725 * that userspace sees the changes.
2727 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2729 if (type != -1 && cnt != type)
2732 if (!sb_has_quota_active(sb, cnt))
2735 if (!f2fs_sb_has_quota_ino(sbi))
2736 inode_lock(dqopt->files[cnt]);
2741 * f2fs_down_read(quota_sem)
2742 * dquot_writeback_dquots()
2745 * f2fs_down_read(quota_sem)
2748 f2fs_down_read(&sbi->quota_sem);
2750 ret = f2fs_quota_sync_file(sbi, cnt);
2752 f2fs_up_read(&sbi->quota_sem);
2753 f2fs_unlock_op(sbi);
2755 if (!f2fs_sb_has_quota_ino(sbi))
2756 inode_unlock(dqopt->files[cnt]);
2764 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2765 const struct path *path)
2767 struct inode *inode;
2770 /* if quota sysfile exists, deny enabling quota with specific file */
2771 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2772 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2776 err = f2fs_quota_sync(sb, type);
2780 err = dquot_quota_on(sb, type, format_id, path);
2784 inode = d_inode(path->dentry);
2787 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2788 f2fs_set_inode_flags(inode);
2789 inode_unlock(inode);
2790 f2fs_mark_inode_dirty_sync(inode, false);
2795 static int __f2fs_quota_off(struct super_block *sb, int type)
2797 struct inode *inode = sb_dqopt(sb)->files[type];
2800 if (!inode || !igrab(inode))
2801 return dquot_quota_off(sb, type);
2803 err = f2fs_quota_sync(sb, type);
2807 err = dquot_quota_off(sb, type);
2808 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2812 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2813 f2fs_set_inode_flags(inode);
2814 inode_unlock(inode);
2815 f2fs_mark_inode_dirty_sync(inode, false);
2821 static int f2fs_quota_off(struct super_block *sb, int type)
2823 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2826 err = __f2fs_quota_off(sb, type);
2829 * quotactl can shutdown journalled quota, result in inconsistence
2830 * between quota record and fs data by following updates, tag the
2831 * flag to let fsck be aware of it.
2833 if (is_journalled_quota(sbi))
2834 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2838 void f2fs_quota_off_umount(struct super_block *sb)
2843 for (type = 0; type < MAXQUOTAS; type++) {
2844 err = __f2fs_quota_off(sb, type);
2846 int ret = dquot_quota_off(sb, type);
2848 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2850 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2854 * In case of checkpoint=disable, we must flush quota blocks.
2855 * This can cause NULL exception for node_inode in end_io, since
2856 * put_super already dropped it.
2858 sync_filesystem(sb);
2861 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2863 struct quota_info *dqopt = sb_dqopt(sb);
2866 for (type = 0; type < MAXQUOTAS; type++) {
2867 if (!dqopt->files[type])
2869 f2fs_inode_synced(dqopt->files[type]);
2873 static int f2fs_dquot_commit(struct dquot *dquot)
2875 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2878 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2879 ret = dquot_commit(dquot);
2881 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2882 f2fs_up_read(&sbi->quota_sem);
2886 static int f2fs_dquot_acquire(struct dquot *dquot)
2888 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2891 f2fs_down_read(&sbi->quota_sem);
2892 ret = dquot_acquire(dquot);
2894 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2895 f2fs_up_read(&sbi->quota_sem);
2899 static int f2fs_dquot_release(struct dquot *dquot)
2901 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2902 int ret = dquot_release(dquot);
2905 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2909 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2911 struct super_block *sb = dquot->dq_sb;
2912 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2913 int ret = dquot_mark_dquot_dirty(dquot);
2915 /* if we are using journalled quota */
2916 if (is_journalled_quota(sbi))
2917 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2922 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2924 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2925 int ret = dquot_commit_info(sb, type);
2928 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2932 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2934 *projid = F2FS_I(inode)->i_projid;
2938 static const struct dquot_operations f2fs_quota_operations = {
2939 .get_reserved_space = f2fs_get_reserved_space,
2940 .write_dquot = f2fs_dquot_commit,
2941 .acquire_dquot = f2fs_dquot_acquire,
2942 .release_dquot = f2fs_dquot_release,
2943 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2944 .write_info = f2fs_dquot_commit_info,
2945 .alloc_dquot = dquot_alloc,
2946 .destroy_dquot = dquot_destroy,
2947 .get_projid = f2fs_get_projid,
2948 .get_next_id = dquot_get_next_id,
2951 static const struct quotactl_ops f2fs_quotactl_ops = {
2952 .quota_on = f2fs_quota_on,
2953 .quota_off = f2fs_quota_off,
2954 .quota_sync = f2fs_quota_sync,
2955 .get_state = dquot_get_state,
2956 .set_info = dquot_set_dqinfo,
2957 .get_dqblk = dquot_get_dqblk,
2958 .set_dqblk = dquot_set_dqblk,
2959 .get_nextdqblk = dquot_get_next_dqblk,
2962 int f2fs_dquot_initialize(struct inode *inode)
2967 int f2fs_quota_sync(struct super_block *sb, int type)
2972 void f2fs_quota_off_umount(struct super_block *sb)
2977 static const struct super_operations f2fs_sops = {
2978 .alloc_inode = f2fs_alloc_inode,
2979 .free_inode = f2fs_free_inode,
2980 .drop_inode = f2fs_drop_inode,
2981 .write_inode = f2fs_write_inode,
2982 .dirty_inode = f2fs_dirty_inode,
2983 .show_options = f2fs_show_options,
2985 .quota_read = f2fs_quota_read,
2986 .quota_write = f2fs_quota_write,
2987 .get_dquots = f2fs_get_dquots,
2989 .evict_inode = f2fs_evict_inode,
2990 .put_super = f2fs_put_super,
2991 .sync_fs = f2fs_sync_fs,
2992 .freeze_fs = f2fs_freeze,
2993 .unfreeze_fs = f2fs_unfreeze,
2994 .statfs = f2fs_statfs,
2995 .remount_fs = f2fs_remount,
2998 #ifdef CONFIG_FS_ENCRYPTION
2999 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3001 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3002 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3006 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3009 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3012 * Encrypting the root directory is not allowed because fsck
3013 * expects lost+found directory to exist and remain unencrypted
3014 * if LOST_FOUND feature is enabled.
3017 if (f2fs_sb_has_lost_found(sbi) &&
3018 inode->i_ino == F2FS_ROOT_INO(sbi))
3021 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3022 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3023 ctx, len, fs_data, XATTR_CREATE);
3026 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3028 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3031 static bool f2fs_has_stable_inodes(struct super_block *sb)
3036 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3037 int *ino_bits_ret, int *lblk_bits_ret)
3039 *ino_bits_ret = 8 * sizeof(nid_t);
3040 *lblk_bits_ret = 8 * sizeof(block_t);
3043 static struct block_device **f2fs_get_devices(struct super_block *sb,
3044 unsigned int *num_devs)
3046 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3047 struct block_device **devs;
3050 if (!f2fs_is_multi_device(sbi))
3053 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3055 return ERR_PTR(-ENOMEM);
3057 for (i = 0; i < sbi->s_ndevs; i++)
3058 devs[i] = FDEV(i).bdev;
3059 *num_devs = sbi->s_ndevs;
3063 static const struct fscrypt_operations f2fs_cryptops = {
3064 .key_prefix = "f2fs:",
3065 .get_context = f2fs_get_context,
3066 .set_context = f2fs_set_context,
3067 .get_dummy_policy = f2fs_get_dummy_policy,
3068 .empty_dir = f2fs_empty_dir,
3069 .has_stable_inodes = f2fs_has_stable_inodes,
3070 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
3071 .get_devices = f2fs_get_devices,
3075 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3076 u64 ino, u32 generation)
3078 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3079 struct inode *inode;
3081 if (f2fs_check_nid_range(sbi, ino))
3082 return ERR_PTR(-ESTALE);
3085 * f2fs_iget isn't quite right if the inode is currently unallocated!
3086 * However f2fs_iget currently does appropriate checks to handle stale
3087 * inodes so everything is OK.
3089 inode = f2fs_iget(sb, ino);
3091 return ERR_CAST(inode);
3092 if (unlikely(generation && inode->i_generation != generation)) {
3093 /* we didn't find the right inode.. */
3095 return ERR_PTR(-ESTALE);
3100 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3101 int fh_len, int fh_type)
3103 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3104 f2fs_nfs_get_inode);
3107 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3108 int fh_len, int fh_type)
3110 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3111 f2fs_nfs_get_inode);
3114 static const struct export_operations f2fs_export_ops = {
3115 .fh_to_dentry = f2fs_fh_to_dentry,
3116 .fh_to_parent = f2fs_fh_to_parent,
3117 .get_parent = f2fs_get_parent,
3120 loff_t max_file_blocks(struct inode *inode)
3126 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3127 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3128 * space in inode.i_addr, it will be more safe to reassign
3132 if (inode && f2fs_compressed_file(inode))
3133 leaf_count = ADDRS_PER_BLOCK(inode);
3135 leaf_count = DEF_ADDRS_PER_BLOCK;
3137 /* two direct node blocks */
3138 result += (leaf_count * 2);
3140 /* two indirect node blocks */
3141 leaf_count *= NIDS_PER_BLOCK;
3142 result += (leaf_count * 2);
3144 /* one double indirect node block */
3145 leaf_count *= NIDS_PER_BLOCK;
3146 result += leaf_count;
3151 static int __f2fs_commit_super(struct buffer_head *bh,
3152 struct f2fs_super_block *super)
3156 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3157 set_buffer_dirty(bh);
3160 /* it's rare case, we can do fua all the time */
3161 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3164 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3165 struct buffer_head *bh)
3167 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3168 (bh->b_data + F2FS_SUPER_OFFSET);
3169 struct super_block *sb = sbi->sb;
3170 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3171 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3172 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3173 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3174 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3175 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3176 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3177 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3178 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3179 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3180 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3181 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3182 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3183 u64 main_end_blkaddr = main_blkaddr +
3184 (segment_count_main << log_blocks_per_seg);
3185 u64 seg_end_blkaddr = segment0_blkaddr +
3186 (segment_count << log_blocks_per_seg);
3188 if (segment0_blkaddr != cp_blkaddr) {
3189 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3190 segment0_blkaddr, cp_blkaddr);
3194 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3196 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3197 cp_blkaddr, sit_blkaddr,
3198 segment_count_ckpt << log_blocks_per_seg);
3202 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3204 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3205 sit_blkaddr, nat_blkaddr,
3206 segment_count_sit << log_blocks_per_seg);
3210 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3212 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3213 nat_blkaddr, ssa_blkaddr,
3214 segment_count_nat << log_blocks_per_seg);
3218 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3220 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3221 ssa_blkaddr, main_blkaddr,
3222 segment_count_ssa << log_blocks_per_seg);
3226 if (main_end_blkaddr > seg_end_blkaddr) {
3227 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3228 main_blkaddr, seg_end_blkaddr,
3229 segment_count_main << log_blocks_per_seg);
3231 } else if (main_end_blkaddr < seg_end_blkaddr) {
3235 /* fix in-memory information all the time */
3236 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3237 segment0_blkaddr) >> log_blocks_per_seg);
3239 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3240 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3243 err = __f2fs_commit_super(bh, NULL);
3244 res = err ? "failed" : "done";
3246 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3247 res, main_blkaddr, seg_end_blkaddr,
3248 segment_count_main << log_blocks_per_seg);
3255 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3256 struct buffer_head *bh)
3258 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3259 block_t total_sections, blocks_per_seg;
3260 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3261 (bh->b_data + F2FS_SUPER_OFFSET);
3262 size_t crc_offset = 0;
3265 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3266 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3267 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3271 /* Check checksum_offset and crc in superblock */
3272 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3273 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3275 offsetof(struct f2fs_super_block, crc)) {
3276 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3278 return -EFSCORRUPTED;
3280 crc = le32_to_cpu(raw_super->crc);
3281 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3282 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3283 return -EFSCORRUPTED;
3287 /* Currently, support only 4KB block size */
3288 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3289 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3290 le32_to_cpu(raw_super->log_blocksize),
3292 return -EFSCORRUPTED;
3295 /* check log blocks per segment */
3296 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3297 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3298 le32_to_cpu(raw_super->log_blocks_per_seg));
3299 return -EFSCORRUPTED;
3302 /* Currently, support 512/1024/2048/4096 bytes sector size */
3303 if (le32_to_cpu(raw_super->log_sectorsize) >
3304 F2FS_MAX_LOG_SECTOR_SIZE ||
3305 le32_to_cpu(raw_super->log_sectorsize) <
3306 F2FS_MIN_LOG_SECTOR_SIZE) {
3307 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3308 le32_to_cpu(raw_super->log_sectorsize));
3309 return -EFSCORRUPTED;
3311 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3312 le32_to_cpu(raw_super->log_sectorsize) !=
3313 F2FS_MAX_LOG_SECTOR_SIZE) {
3314 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3315 le32_to_cpu(raw_super->log_sectors_per_block),
3316 le32_to_cpu(raw_super->log_sectorsize));
3317 return -EFSCORRUPTED;
3320 segment_count = le32_to_cpu(raw_super->segment_count);
3321 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3322 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3323 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3324 total_sections = le32_to_cpu(raw_super->section_count);
3326 /* blocks_per_seg should be 512, given the above check */
3327 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3329 if (segment_count > F2FS_MAX_SEGMENT ||
3330 segment_count < F2FS_MIN_SEGMENTS) {
3331 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3332 return -EFSCORRUPTED;
3335 if (total_sections > segment_count_main || total_sections < 1 ||
3336 segs_per_sec > segment_count || !segs_per_sec) {
3337 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3338 segment_count, total_sections, segs_per_sec);
3339 return -EFSCORRUPTED;
3342 if (segment_count_main != total_sections * segs_per_sec) {
3343 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3344 segment_count_main, total_sections, segs_per_sec);
3345 return -EFSCORRUPTED;
3348 if ((segment_count / segs_per_sec) < total_sections) {
3349 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3350 segment_count, segs_per_sec, total_sections);
3351 return -EFSCORRUPTED;
3354 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3355 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3356 segment_count, le64_to_cpu(raw_super->block_count));
3357 return -EFSCORRUPTED;
3360 if (RDEV(0).path[0]) {
3361 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3364 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3365 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3368 if (segment_count != dev_seg_count) {
3369 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3370 segment_count, dev_seg_count);
3371 return -EFSCORRUPTED;
3374 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3375 !bdev_is_zoned(sbi->sb->s_bdev)) {
3376 f2fs_info(sbi, "Zoned block device path is missing");
3377 return -EFSCORRUPTED;
3381 if (secs_per_zone > total_sections || !secs_per_zone) {
3382 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3383 secs_per_zone, total_sections);
3384 return -EFSCORRUPTED;
3386 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3387 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3388 (le32_to_cpu(raw_super->extension_count) +
3389 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3390 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3391 le32_to_cpu(raw_super->extension_count),
3392 raw_super->hot_ext_count,
3393 F2FS_MAX_EXTENSION);
3394 return -EFSCORRUPTED;
3397 if (le32_to_cpu(raw_super->cp_payload) >=
3398 (blocks_per_seg - F2FS_CP_PACKS -
3399 NR_CURSEG_PERSIST_TYPE)) {
3400 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3401 le32_to_cpu(raw_super->cp_payload),
3402 blocks_per_seg - F2FS_CP_PACKS -
3403 NR_CURSEG_PERSIST_TYPE);
3404 return -EFSCORRUPTED;
3407 /* check reserved ino info */
3408 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3409 le32_to_cpu(raw_super->meta_ino) != 2 ||
3410 le32_to_cpu(raw_super->root_ino) != 3) {
3411 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3412 le32_to_cpu(raw_super->node_ino),
3413 le32_to_cpu(raw_super->meta_ino),
3414 le32_to_cpu(raw_super->root_ino));
3415 return -EFSCORRUPTED;
3418 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3419 if (sanity_check_area_boundary(sbi, bh))
3420 return -EFSCORRUPTED;
3425 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3427 unsigned int total, fsmeta;
3428 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3429 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3430 unsigned int ovp_segments, reserved_segments;
3431 unsigned int main_segs, blocks_per_seg;
3432 unsigned int sit_segs, nat_segs;
3433 unsigned int sit_bitmap_size, nat_bitmap_size;
3434 unsigned int log_blocks_per_seg;
3435 unsigned int segment_count_main;
3436 unsigned int cp_pack_start_sum, cp_payload;
3437 block_t user_block_count, valid_user_blocks;
3438 block_t avail_node_count, valid_node_count;
3439 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3442 total = le32_to_cpu(raw_super->segment_count);
3443 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3444 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3446 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3448 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3449 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3451 if (unlikely(fsmeta >= total))
3454 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3455 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3457 if (!f2fs_sb_has_readonly(sbi) &&
3458 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3459 ovp_segments == 0 || reserved_segments == 0)) {
3460 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3463 user_block_count = le64_to_cpu(ckpt->user_block_count);
3464 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3465 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3466 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3467 if (!user_block_count || user_block_count >=
3468 segment_count_main << log_blocks_per_seg) {
3469 f2fs_err(sbi, "Wrong user_block_count: %u",
3474 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3475 if (valid_user_blocks > user_block_count) {
3476 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3477 valid_user_blocks, user_block_count);
3481 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3482 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3483 if (valid_node_count > avail_node_count) {
3484 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3485 valid_node_count, avail_node_count);
3489 main_segs = le32_to_cpu(raw_super->segment_count_main);
3490 blocks_per_seg = sbi->blocks_per_seg;
3492 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3493 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3494 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3497 if (f2fs_sb_has_readonly(sbi))
3500 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3501 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3502 le32_to_cpu(ckpt->cur_node_segno[j])) {
3503 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3505 le32_to_cpu(ckpt->cur_node_segno[i]));
3511 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3512 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3513 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3516 if (f2fs_sb_has_readonly(sbi))
3519 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3520 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3521 le32_to_cpu(ckpt->cur_data_segno[j])) {
3522 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3524 le32_to_cpu(ckpt->cur_data_segno[i]));
3529 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3530 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3531 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3532 le32_to_cpu(ckpt->cur_data_segno[j])) {
3533 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3535 le32_to_cpu(ckpt->cur_node_segno[i]));
3541 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3542 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3544 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3545 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3546 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3547 sit_bitmap_size, nat_bitmap_size);
3551 cp_pack_start_sum = __start_sum_addr(sbi);
3552 cp_payload = __cp_payload(sbi);
3553 if (cp_pack_start_sum < cp_payload + 1 ||
3554 cp_pack_start_sum > blocks_per_seg - 1 -
3555 NR_CURSEG_PERSIST_TYPE) {
3556 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3561 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3562 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3563 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3564 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3565 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3566 le32_to_cpu(ckpt->checksum_offset));
3570 nat_blocks = nat_segs << log_blocks_per_seg;
3571 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3572 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3573 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3574 (cp_payload + F2FS_CP_PACKS +
3575 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3576 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3577 cp_payload, nat_bits_blocks);
3581 if (unlikely(f2fs_cp_error(sbi))) {
3582 f2fs_err(sbi, "A bug case: need to run fsck");
3588 static void init_sb_info(struct f2fs_sb_info *sbi)
3590 struct f2fs_super_block *raw_super = sbi->raw_super;
3593 sbi->log_sectors_per_block =
3594 le32_to_cpu(raw_super->log_sectors_per_block);
3595 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3596 sbi->blocksize = 1 << sbi->log_blocksize;
3597 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3598 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3599 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3600 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3601 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3602 sbi->total_node_count =
3603 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3604 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3605 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3606 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3607 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3608 sbi->cur_victim_sec = NULL_SECNO;
3609 sbi->gc_mode = GC_NORMAL;
3610 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3611 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3612 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3613 sbi->migration_granularity = sbi->segs_per_sec;
3614 sbi->seq_file_ra_mul = MIN_RA_MUL;
3615 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3616 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3617 spin_lock_init(&sbi->gc_urgent_high_lock);
3618 atomic64_set(&sbi->current_atomic_write, 0);
3620 sbi->dir_level = DEF_DIR_LEVEL;
3621 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3622 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3623 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3624 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3625 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3626 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3627 DEF_UMOUNT_DISCARD_TIMEOUT;
3628 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3630 for (i = 0; i < NR_COUNT_TYPE; i++)
3631 atomic_set(&sbi->nr_pages[i], 0);
3633 for (i = 0; i < META; i++)
3634 atomic_set(&sbi->wb_sync_req[i], 0);
3636 INIT_LIST_HEAD(&sbi->s_list);
3637 mutex_init(&sbi->umount_mutex);
3638 init_f2fs_rwsem(&sbi->io_order_lock);
3639 spin_lock_init(&sbi->cp_lock);
3641 sbi->dirty_device = 0;
3642 spin_lock_init(&sbi->dev_lock);
3644 init_f2fs_rwsem(&sbi->sb_lock);
3645 init_f2fs_rwsem(&sbi->pin_sem);
3648 static int init_percpu_info(struct f2fs_sb_info *sbi)
3652 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3656 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3658 goto err_valid_block;
3660 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3663 goto err_node_block;
3667 percpu_counter_destroy(&sbi->rf_node_block_count);
3669 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3673 #ifdef CONFIG_BLK_DEV_ZONED
3675 struct f2fs_report_zones_args {
3676 struct f2fs_sb_info *sbi;
3677 struct f2fs_dev_info *dev;
3680 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3683 struct f2fs_report_zones_args *rz_args = data;
3684 block_t unusable_blocks = (zone->len - zone->capacity) >>
3685 F2FS_LOG_SECTORS_PER_BLOCK;
3687 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3690 set_bit(idx, rz_args->dev->blkz_seq);
3691 if (!rz_args->sbi->unusable_blocks_per_sec) {
3692 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3695 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3696 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3702 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3704 struct block_device *bdev = FDEV(devi).bdev;
3705 sector_t nr_sectors = bdev_nr_sectors(bdev);
3706 struct f2fs_report_zones_args rep_zone_arg;
3710 if (!f2fs_sb_has_blkzoned(sbi))
3713 zone_sectors = bdev_zone_sectors(bdev);
3714 if (!is_power_of_2(zone_sectors)) {
3715 f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n");
3719 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3720 SECTOR_TO_BLOCK(zone_sectors))
3722 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3723 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3724 __ilog2_u32(sbi->blocks_per_blkz))
3726 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3727 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3728 sbi->log_blocks_per_blkz;
3729 if (nr_sectors & (zone_sectors - 1))
3730 FDEV(devi).nr_blkz++;
3732 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3733 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3734 * sizeof(unsigned long),
3736 if (!FDEV(devi).blkz_seq)
3739 rep_zone_arg.sbi = sbi;
3740 rep_zone_arg.dev = &FDEV(devi);
3742 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3751 * Read f2fs raw super block.
3752 * Because we have two copies of super block, so read both of them
3753 * to get the first valid one. If any one of them is broken, we pass
3754 * them recovery flag back to the caller.
3756 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3757 struct f2fs_super_block **raw_super,
3758 int *valid_super_block, int *recovery)
3760 struct super_block *sb = sbi->sb;
3762 struct buffer_head *bh;
3763 struct f2fs_super_block *super;
3766 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3770 for (block = 0; block < 2; block++) {
3771 bh = sb_bread(sb, block);
3773 f2fs_err(sbi, "Unable to read %dth superblock",
3780 /* sanity checking of raw super */
3781 err = sanity_check_raw_super(sbi, bh);
3783 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3791 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3793 *valid_super_block = block;
3799 /* No valid superblock */
3808 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3810 struct buffer_head *bh;
3814 if ((recover && f2fs_readonly(sbi->sb)) ||
3815 bdev_read_only(sbi->sb->s_bdev)) {
3816 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3820 /* we should update superblock crc here */
3821 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3822 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3823 offsetof(struct f2fs_super_block, crc));
3824 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3827 /* write back-up superblock first */
3828 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3831 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3834 /* if we are in recovery path, skip writing valid superblock */
3838 /* write current valid superblock */
3839 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3842 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3847 void f2fs_handle_stop(struct f2fs_sb_info *sbi, unsigned char reason)
3849 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3852 f2fs_down_write(&sbi->sb_lock);
3854 if (raw_super->s_stop_reason[reason] < ((1 << BITS_PER_BYTE) - 1))
3855 raw_super->s_stop_reason[reason]++;
3857 err = f2fs_commit_super(sbi, false);
3859 f2fs_err(sbi, "f2fs_commit_super fails to record reason:%u err:%d",
3861 f2fs_up_write(&sbi->sb_lock);
3864 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
3866 spin_lock(&sbi->error_lock);
3867 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
3868 set_bit(flag, (unsigned long *)sbi->errors);
3869 sbi->error_dirty = true;
3871 spin_unlock(&sbi->error_lock);
3874 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
3876 bool need_update = false;
3878 spin_lock(&sbi->error_lock);
3879 if (sbi->error_dirty) {
3880 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
3882 sbi->error_dirty = false;
3885 spin_unlock(&sbi->error_lock);
3890 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
3894 f2fs_save_errors(sbi, error);
3896 f2fs_down_write(&sbi->sb_lock);
3898 if (!f2fs_update_errors(sbi))
3901 err = f2fs_commit_super(sbi, false);
3903 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
3906 f2fs_up_write(&sbi->sb_lock);
3909 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3911 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3912 unsigned int max_devices = MAX_DEVICES;
3913 unsigned int logical_blksize;
3916 /* Initialize single device information */
3917 if (!RDEV(0).path[0]) {
3918 if (!bdev_is_zoned(sbi->sb->s_bdev))
3924 * Initialize multiple devices information, or single
3925 * zoned block device information.
3927 sbi->devs = f2fs_kzalloc(sbi,
3928 array_size(max_devices,
3929 sizeof(struct f2fs_dev_info)),
3934 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3935 sbi->aligned_blksize = true;
3937 for (i = 0; i < max_devices; i++) {
3939 if (i > 0 && !RDEV(i).path[0])
3942 if (max_devices == 1) {
3943 /* Single zoned block device mount */
3945 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3946 sbi->sb->s_mode, sbi->sb->s_type);
3948 /* Multi-device mount */
3949 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3950 FDEV(i).total_segments =
3951 le32_to_cpu(RDEV(i).total_segments);
3953 FDEV(i).start_blk = 0;
3954 FDEV(i).end_blk = FDEV(i).start_blk +
3955 (FDEV(i).total_segments <<
3956 sbi->log_blocks_per_seg) - 1 +
3957 le32_to_cpu(raw_super->segment0_blkaddr);
3959 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3960 FDEV(i).end_blk = FDEV(i).start_blk +
3961 (FDEV(i).total_segments <<
3962 sbi->log_blocks_per_seg) - 1;
3964 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3965 sbi->sb->s_mode, sbi->sb->s_type);
3967 if (IS_ERR(FDEV(i).bdev))
3968 return PTR_ERR(FDEV(i).bdev);
3970 /* to release errored devices */
3971 sbi->s_ndevs = i + 1;
3973 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
3974 sbi->aligned_blksize = false;
3976 #ifdef CONFIG_BLK_DEV_ZONED
3977 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3978 !f2fs_sb_has_blkzoned(sbi)) {
3979 f2fs_err(sbi, "Zoned block device feature not enabled");
3982 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3983 if (init_blkz_info(sbi, i)) {
3984 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3987 if (max_devices == 1)
3989 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3991 FDEV(i).total_segments,
3992 FDEV(i).start_blk, FDEV(i).end_blk,
3993 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3994 "Host-aware" : "Host-managed");
3998 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4000 FDEV(i).total_segments,
4001 FDEV(i).start_blk, FDEV(i).end_blk);
4004 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
4008 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4010 #if IS_ENABLED(CONFIG_UNICODE)
4011 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4012 const struct f2fs_sb_encodings *encoding_info;
4013 struct unicode_map *encoding;
4014 __u16 encoding_flags;
4016 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4017 if (!encoding_info) {
4019 "Encoding requested by superblock is unknown");
4023 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4024 encoding = utf8_load(encoding_info->version);
4025 if (IS_ERR(encoding)) {
4027 "can't mount with superblock charset: %s-%u.%u.%u "
4028 "not supported by the kernel. flags: 0x%x.",
4029 encoding_info->name,
4030 unicode_major(encoding_info->version),
4031 unicode_minor(encoding_info->version),
4032 unicode_rev(encoding_info->version),
4034 return PTR_ERR(encoding);
4036 f2fs_info(sbi, "Using encoding defined by superblock: "
4037 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4038 unicode_major(encoding_info->version),
4039 unicode_minor(encoding_info->version),
4040 unicode_rev(encoding_info->version),
4043 sbi->sb->s_encoding = encoding;
4044 sbi->sb->s_encoding_flags = encoding_flags;
4047 if (f2fs_sb_has_casefold(sbi)) {
4048 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4055 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4057 struct f2fs_sm_info *sm_i = SM_I(sbi);
4059 /* adjust parameters according to the volume size */
4060 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
4061 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
4062 if (f2fs_block_unit_discard(sbi))
4063 sm_i->dcc_info->discard_granularity = 1;
4064 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE |
4065 1 << F2FS_IPU_HONOR_OPU_WRITE;
4068 sbi->readdir_ra = 1;
4071 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4073 struct f2fs_sb_info *sbi;
4074 struct f2fs_super_block *raw_super;
4077 bool skip_recovery = false, need_fsck = false;
4078 char *options = NULL;
4079 int recovery, i, valid_super_block;
4080 struct curseg_info *seg_i;
4086 valid_super_block = -1;
4089 /* allocate memory for f2fs-specific super block info */
4090 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4096 /* initialize locks within allocated memory */
4097 init_f2fs_rwsem(&sbi->gc_lock);
4098 mutex_init(&sbi->writepages);
4099 init_f2fs_rwsem(&sbi->cp_global_sem);
4100 init_f2fs_rwsem(&sbi->node_write);
4101 init_f2fs_rwsem(&sbi->node_change);
4102 spin_lock_init(&sbi->stat_lock);
4103 init_f2fs_rwsem(&sbi->cp_rwsem);
4104 init_f2fs_rwsem(&sbi->quota_sem);
4105 init_waitqueue_head(&sbi->cp_wait);
4106 spin_lock_init(&sbi->error_lock);
4108 for (i = 0; i < NR_INODE_TYPE; i++) {
4109 INIT_LIST_HEAD(&sbi->inode_list[i]);
4110 spin_lock_init(&sbi->inode_lock[i]);
4112 mutex_init(&sbi->flush_lock);
4114 /* Load the checksum driver */
4115 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4116 if (IS_ERR(sbi->s_chksum_driver)) {
4117 f2fs_err(sbi, "Cannot load crc32 driver.");
4118 err = PTR_ERR(sbi->s_chksum_driver);
4119 sbi->s_chksum_driver = NULL;
4123 /* set a block size */
4124 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4125 f2fs_err(sbi, "unable to set blocksize");
4129 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4134 sb->s_fs_info = sbi;
4135 sbi->raw_super = raw_super;
4137 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4139 /* precompute checksum seed for metadata */
4140 if (f2fs_sb_has_inode_chksum(sbi))
4141 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4142 sizeof(raw_super->uuid));
4144 default_options(sbi);
4145 /* parse mount options */
4146 options = kstrdup((const char *)data, GFP_KERNEL);
4147 if (data && !options) {
4152 err = parse_options(sb, options, false);
4156 sb->s_maxbytes = max_file_blocks(NULL) <<
4157 le32_to_cpu(raw_super->log_blocksize);
4158 sb->s_max_links = F2FS_LINK_MAX;
4160 err = f2fs_setup_casefold(sbi);
4165 sb->dq_op = &f2fs_quota_operations;
4166 sb->s_qcop = &f2fs_quotactl_ops;
4167 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4169 if (f2fs_sb_has_quota_ino(sbi)) {
4170 for (i = 0; i < MAXQUOTAS; i++) {
4171 if (f2fs_qf_ino(sbi->sb, i))
4172 sbi->nquota_files++;
4177 sb->s_op = &f2fs_sops;
4178 #ifdef CONFIG_FS_ENCRYPTION
4179 sb->s_cop = &f2fs_cryptops;
4181 #ifdef CONFIG_FS_VERITY
4182 sb->s_vop = &f2fs_verityops;
4184 sb->s_xattr = f2fs_xattr_handlers;
4185 sb->s_export_op = &f2fs_export_ops;
4186 sb->s_magic = F2FS_SUPER_MAGIC;
4187 sb->s_time_gran = 1;
4188 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4189 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4190 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4191 sb->s_iflags |= SB_I_CGROUPWB;
4193 /* init f2fs-specific super block info */
4194 sbi->valid_super_block = valid_super_block;
4196 /* disallow all the data/node/meta page writes */
4197 set_sbi_flag(sbi, SBI_POR_DOING);
4199 err = f2fs_init_write_merge_io(sbi);
4205 err = f2fs_init_iostat(sbi);
4209 err = init_percpu_info(sbi);
4213 if (F2FS_IO_ALIGNED(sbi)) {
4214 sbi->write_io_dummy =
4215 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4216 if (!sbi->write_io_dummy) {
4222 /* init per sbi slab cache */
4223 err = f2fs_init_xattr_caches(sbi);
4226 err = f2fs_init_page_array_cache(sbi);
4228 goto free_xattr_cache;
4230 /* get an inode for meta space */
4231 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4232 if (IS_ERR(sbi->meta_inode)) {
4233 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4234 err = PTR_ERR(sbi->meta_inode);
4235 goto free_page_array_cache;
4238 err = f2fs_get_valid_checkpoint(sbi);
4240 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4241 goto free_meta_inode;
4244 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4245 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4246 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4247 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4248 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4251 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4252 set_sbi_flag(sbi, SBI_NEED_FSCK);
4254 /* Initialize device list */
4255 err = f2fs_scan_devices(sbi);
4257 f2fs_err(sbi, "Failed to find devices");
4261 err = f2fs_init_post_read_wq(sbi);
4263 f2fs_err(sbi, "Failed to initialize post read workqueue");
4267 sbi->total_valid_node_count =
4268 le32_to_cpu(sbi->ckpt->valid_node_count);
4269 percpu_counter_set(&sbi->total_valid_inode_count,
4270 le32_to_cpu(sbi->ckpt->valid_inode_count));
4271 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4272 sbi->total_valid_block_count =
4273 le64_to_cpu(sbi->ckpt->valid_block_count);
4274 sbi->last_valid_block_count = sbi->total_valid_block_count;
4275 sbi->reserved_blocks = 0;
4276 sbi->current_reserved_blocks = 0;
4277 limit_reserve_root(sbi);
4278 adjust_unusable_cap_perc(sbi);
4280 f2fs_init_extent_cache_info(sbi);
4282 f2fs_init_ino_entry_info(sbi);
4284 f2fs_init_fsync_node_info(sbi);
4286 /* setup checkpoint request control and start checkpoint issue thread */
4287 f2fs_init_ckpt_req_control(sbi);
4288 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4289 test_opt(sbi, MERGE_CHECKPOINT)) {
4290 err = f2fs_start_ckpt_thread(sbi);
4293 "Failed to start F2FS issue_checkpoint_thread (%d)",
4295 goto stop_ckpt_thread;
4299 /* setup f2fs internal modules */
4300 err = f2fs_build_segment_manager(sbi);
4302 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4306 err = f2fs_build_node_manager(sbi);
4308 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4313 err = adjust_reserved_segment(sbi);
4317 /* For write statistics */
4318 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4320 /* Read accumulated write IO statistics if exists */
4321 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4322 if (__exist_node_summaries(sbi))
4323 sbi->kbytes_written =
4324 le64_to_cpu(seg_i->journal->info.kbytes_written);
4326 f2fs_build_gc_manager(sbi);
4328 err = f2fs_build_stats(sbi);
4332 /* get an inode for node space */
4333 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4334 if (IS_ERR(sbi->node_inode)) {
4335 f2fs_err(sbi, "Failed to read node inode");
4336 err = PTR_ERR(sbi->node_inode);
4340 /* read root inode and dentry */
4341 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4343 f2fs_err(sbi, "Failed to read root inode");
4344 err = PTR_ERR(root);
4345 goto free_node_inode;
4347 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4348 !root->i_size || !root->i_nlink) {
4351 goto free_node_inode;
4354 sb->s_root = d_make_root(root); /* allocate root dentry */
4357 goto free_node_inode;
4360 err = f2fs_init_compress_inode(sbi);
4362 goto free_root_inode;
4364 err = f2fs_register_sysfs(sbi);
4366 goto free_compress_inode;
4369 /* Enable quota usage during mount */
4370 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4371 err = f2fs_enable_quotas(sb);
4373 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4376 /* if there are any orphan inodes, free them */
4377 err = f2fs_recover_orphan_inodes(sbi);
4381 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4382 goto reset_checkpoint;
4384 /* recover fsynced data */
4385 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4386 !test_opt(sbi, NORECOVERY)) {
4388 * mount should be failed, when device has readonly mode, and
4389 * previous checkpoint was not done by clean system shutdown.
4391 if (f2fs_hw_is_readonly(sbi)) {
4392 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4393 err = f2fs_recover_fsync_data(sbi, true);
4396 f2fs_err(sbi, "Need to recover fsync data, but "
4397 "write access unavailable, please try "
4398 "mount w/ disable_roll_forward or norecovery");
4403 f2fs_info(sbi, "write access unavailable, skipping recovery");
4404 goto reset_checkpoint;
4408 set_sbi_flag(sbi, SBI_NEED_FSCK);
4411 goto reset_checkpoint;
4413 err = f2fs_recover_fsync_data(sbi, false);
4416 skip_recovery = true;
4418 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4423 err = f2fs_recover_fsync_data(sbi, true);
4425 if (!f2fs_readonly(sb) && err > 0) {
4427 f2fs_err(sbi, "Need to recover fsync data");
4433 * If the f2fs is not readonly and fsync data recovery succeeds,
4434 * check zoned block devices' write pointer consistency.
4436 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4437 err = f2fs_check_write_pointer(sbi);
4443 f2fs_init_inmem_curseg(sbi);
4445 /* f2fs_recover_fsync_data() cleared this already */
4446 clear_sbi_flag(sbi, SBI_POR_DOING);
4448 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4449 err = f2fs_disable_checkpoint(sbi);
4451 goto sync_free_meta;
4452 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4453 f2fs_enable_checkpoint(sbi);
4457 * If filesystem is not mounted as read-only then
4458 * do start the gc_thread.
4460 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4461 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4462 /* After POR, we can run background GC thread.*/
4463 err = f2fs_start_gc_thread(sbi);
4465 goto sync_free_meta;
4469 /* recover broken superblock */
4471 err = f2fs_commit_super(sbi, true);
4472 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4473 sbi->valid_super_block ? 1 : 2, err);
4476 f2fs_join_shrinker(sbi);
4478 f2fs_tuning_parameters(sbi);
4480 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4481 cur_cp_version(F2FS_CKPT(sbi)));
4482 f2fs_update_time(sbi, CP_TIME);
4483 f2fs_update_time(sbi, REQ_TIME);
4484 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4488 /* safe to flush all the data */
4489 sync_filesystem(sbi->sb);
4494 f2fs_truncate_quota_inode_pages(sb);
4495 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4496 f2fs_quota_off_umount(sbi->sb);
4499 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4500 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4501 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4502 * falls into an infinite loop in f2fs_sync_meta_pages().
4504 truncate_inode_pages_final(META_MAPPING(sbi));
4505 /* evict some inodes being cached by GC */
4507 f2fs_unregister_sysfs(sbi);
4508 free_compress_inode:
4509 f2fs_destroy_compress_inode(sbi);
4514 f2fs_release_ino_entry(sbi, true);
4515 truncate_inode_pages_final(NODE_MAPPING(sbi));
4516 iput(sbi->node_inode);
4517 sbi->node_inode = NULL;
4519 f2fs_destroy_stats(sbi);
4521 /* stop discard thread before destroying node manager */
4522 f2fs_stop_discard_thread(sbi);
4523 f2fs_destroy_node_manager(sbi);
4525 f2fs_destroy_segment_manager(sbi);
4527 f2fs_stop_ckpt_thread(sbi);
4528 f2fs_destroy_post_read_wq(sbi);
4530 destroy_device_list(sbi);
4533 make_bad_inode(sbi->meta_inode);
4534 iput(sbi->meta_inode);
4535 sbi->meta_inode = NULL;
4536 free_page_array_cache:
4537 f2fs_destroy_page_array_cache(sbi);
4539 f2fs_destroy_xattr_caches(sbi);
4541 mempool_destroy(sbi->write_io_dummy);
4543 destroy_percpu_info(sbi);
4545 f2fs_destroy_iostat(sbi);
4547 for (i = 0; i < NR_PAGE_TYPE; i++)
4548 kvfree(sbi->write_io[i]);
4550 #if IS_ENABLED(CONFIG_UNICODE)
4551 utf8_unload(sb->s_encoding);
4552 sb->s_encoding = NULL;
4556 for (i = 0; i < MAXQUOTAS; i++)
4557 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4559 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4564 if (sbi->s_chksum_driver)
4565 crypto_free_shash(sbi->s_chksum_driver);
4568 /* give only one another chance */
4569 if (retry_cnt > 0 && skip_recovery) {
4571 shrink_dcache_sb(sb);
4577 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4578 const char *dev_name, void *data)
4580 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4583 static void kill_f2fs_super(struct super_block *sb)
4586 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4588 set_sbi_flag(sbi, SBI_IS_CLOSE);
4589 f2fs_stop_gc_thread(sbi);
4590 f2fs_stop_discard_thread(sbi);
4592 #ifdef CONFIG_F2FS_FS_COMPRESSION
4594 * latter evict_inode() can bypass checking and invalidating
4595 * compress inode cache.
4597 if (test_opt(sbi, COMPRESS_CACHE))
4598 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4601 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4602 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4603 struct cp_control cpc = {
4604 .reason = CP_UMOUNT,
4606 f2fs_write_checkpoint(sbi, &cpc);
4609 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4610 sb->s_flags &= ~SB_RDONLY;
4612 kill_block_super(sb);
4615 static struct file_system_type f2fs_fs_type = {
4616 .owner = THIS_MODULE,
4618 .mount = f2fs_mount,
4619 .kill_sb = kill_f2fs_super,
4620 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4622 MODULE_ALIAS_FS("f2fs");
4624 static int __init init_inodecache(void)
4626 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4627 sizeof(struct f2fs_inode_info), 0,
4628 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4629 if (!f2fs_inode_cachep)
4634 static void destroy_inodecache(void)
4637 * Make sure all delayed rcu free inodes are flushed before we
4641 kmem_cache_destroy(f2fs_inode_cachep);
4644 static int __init init_f2fs_fs(void)
4648 if (PAGE_SIZE != F2FS_BLKSIZE) {
4649 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4650 PAGE_SIZE, F2FS_BLKSIZE);
4654 err = init_inodecache();
4657 err = f2fs_create_node_manager_caches();
4659 goto free_inodecache;
4660 err = f2fs_create_segment_manager_caches();
4662 goto free_node_manager_caches;
4663 err = f2fs_create_checkpoint_caches();
4665 goto free_segment_manager_caches;
4666 err = f2fs_create_recovery_cache();
4668 goto free_checkpoint_caches;
4669 err = f2fs_create_extent_cache();
4671 goto free_recovery_cache;
4672 err = f2fs_create_garbage_collection_cache();
4674 goto free_extent_cache;
4675 err = f2fs_init_sysfs();
4677 goto free_garbage_collection_cache;
4678 err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4681 err = register_filesystem(&f2fs_fs_type);
4684 f2fs_create_root_stats();
4685 err = f2fs_init_post_read_processing();
4687 goto free_root_stats;
4688 err = f2fs_init_iostat_processing();
4690 goto free_post_read;
4691 err = f2fs_init_bio_entry_cache();
4694 err = f2fs_init_bioset();
4696 goto free_bio_enrty_cache;
4697 err = f2fs_init_compress_mempool();
4700 err = f2fs_init_compress_cache();
4702 goto free_compress_mempool;
4703 err = f2fs_create_casefold_cache();
4705 goto free_compress_cache;
4707 free_compress_cache:
4708 f2fs_destroy_compress_cache();
4709 free_compress_mempool:
4710 f2fs_destroy_compress_mempool();
4712 f2fs_destroy_bioset();
4713 free_bio_enrty_cache:
4714 f2fs_destroy_bio_entry_cache();
4716 f2fs_destroy_iostat_processing();
4718 f2fs_destroy_post_read_processing();
4720 f2fs_destroy_root_stats();
4721 unregister_filesystem(&f2fs_fs_type);
4723 unregister_shrinker(&f2fs_shrinker_info);
4726 free_garbage_collection_cache:
4727 f2fs_destroy_garbage_collection_cache();
4729 f2fs_destroy_extent_cache();
4730 free_recovery_cache:
4731 f2fs_destroy_recovery_cache();
4732 free_checkpoint_caches:
4733 f2fs_destroy_checkpoint_caches();
4734 free_segment_manager_caches:
4735 f2fs_destroy_segment_manager_caches();
4736 free_node_manager_caches:
4737 f2fs_destroy_node_manager_caches();
4739 destroy_inodecache();
4744 static void __exit exit_f2fs_fs(void)
4746 f2fs_destroy_casefold_cache();
4747 f2fs_destroy_compress_cache();
4748 f2fs_destroy_compress_mempool();
4749 f2fs_destroy_bioset();
4750 f2fs_destroy_bio_entry_cache();
4751 f2fs_destroy_iostat_processing();
4752 f2fs_destroy_post_read_processing();
4753 f2fs_destroy_root_stats();
4754 unregister_filesystem(&f2fs_fs_type);
4755 unregister_shrinker(&f2fs_shrinker_info);
4757 f2fs_destroy_garbage_collection_cache();
4758 f2fs_destroy_extent_cache();
4759 f2fs_destroy_recovery_cache();
4760 f2fs_destroy_checkpoint_caches();
4761 f2fs_destroy_segment_manager_caches();
4762 f2fs_destroy_node_manager_caches();
4763 destroy_inodecache();
4766 module_init(init_f2fs_fs)
4767 module_exit(exit_f2fs_fs)
4769 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4770 MODULE_DESCRIPTION("Flash Friendly File System");
4771 MODULE_LICENSE("GPL");
4772 MODULE_SOFTDEP("pre: crc32");