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",
64 [FAULT_BLKADDR] = "invalid blkaddr",
67 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
70 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
73 atomic_set(&ffi->inject_ops, 0);
74 ffi->inject_rate = rate;
78 ffi->inject_type = type;
81 memset(ffi, 0, sizeof(struct f2fs_fault_info));
85 /* f2fs-wide shrinker description */
86 static struct shrinker f2fs_shrinker_info = {
87 .scan_objects = f2fs_shrink_scan,
88 .count_objects = f2fs_shrink_count,
89 .seeks = DEFAULT_SEEKS,
94 Opt_disable_roll_forward,
105 Opt_disable_ext_identify,
108 Opt_inline_xattr_size,
146 Opt_test_dummy_encryption,
148 Opt_checkpoint_disable,
149 Opt_checkpoint_disable_cap,
150 Opt_checkpoint_disable_cap_perc,
151 Opt_checkpoint_enable,
152 Opt_checkpoint_merge,
153 Opt_nocheckpoint_merge,
154 Opt_compress_algorithm,
155 Opt_compress_log_size,
156 Opt_compress_extension,
157 Opt_nocompress_extension,
166 Opt_age_extent_cache,
170 static match_table_t f2fs_tokens = {
171 {Opt_gc_background, "background_gc=%s"},
172 {Opt_disable_roll_forward, "disable_roll_forward"},
173 {Opt_norecovery, "norecovery"},
174 {Opt_discard, "discard"},
175 {Opt_nodiscard, "nodiscard"},
176 {Opt_noheap, "no_heap"},
178 {Opt_user_xattr, "user_xattr"},
179 {Opt_nouser_xattr, "nouser_xattr"},
181 {Opt_noacl, "noacl"},
182 {Opt_active_logs, "active_logs=%u"},
183 {Opt_disable_ext_identify, "disable_ext_identify"},
184 {Opt_inline_xattr, "inline_xattr"},
185 {Opt_noinline_xattr, "noinline_xattr"},
186 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
187 {Opt_inline_data, "inline_data"},
188 {Opt_inline_dentry, "inline_dentry"},
189 {Opt_noinline_dentry, "noinline_dentry"},
190 {Opt_flush_merge, "flush_merge"},
191 {Opt_noflush_merge, "noflush_merge"},
192 {Opt_barrier, "barrier"},
193 {Opt_nobarrier, "nobarrier"},
194 {Opt_fastboot, "fastboot"},
195 {Opt_extent_cache, "extent_cache"},
196 {Opt_noextent_cache, "noextent_cache"},
197 {Opt_noinline_data, "noinline_data"},
198 {Opt_data_flush, "data_flush"},
199 {Opt_reserve_root, "reserve_root=%u"},
200 {Opt_resgid, "resgid=%u"},
201 {Opt_resuid, "resuid=%u"},
202 {Opt_mode, "mode=%s"},
203 {Opt_io_size_bits, "io_bits=%u"},
204 {Opt_fault_injection, "fault_injection=%u"},
205 {Opt_fault_type, "fault_type=%u"},
206 {Opt_lazytime, "lazytime"},
207 {Opt_nolazytime, "nolazytime"},
208 {Opt_quota, "quota"},
209 {Opt_noquota, "noquota"},
210 {Opt_usrquota, "usrquota"},
211 {Opt_grpquota, "grpquota"},
212 {Opt_prjquota, "prjquota"},
213 {Opt_usrjquota, "usrjquota=%s"},
214 {Opt_grpjquota, "grpjquota=%s"},
215 {Opt_prjjquota, "prjjquota=%s"},
216 {Opt_offusrjquota, "usrjquota="},
217 {Opt_offgrpjquota, "grpjquota="},
218 {Opt_offprjjquota, "prjjquota="},
219 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
220 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
221 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
222 {Opt_alloc, "alloc_mode=%s"},
223 {Opt_fsync, "fsync_mode=%s"},
224 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
225 {Opt_test_dummy_encryption, "test_dummy_encryption"},
226 {Opt_inlinecrypt, "inlinecrypt"},
227 {Opt_checkpoint_disable, "checkpoint=disable"},
228 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
229 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
230 {Opt_checkpoint_enable, "checkpoint=enable"},
231 {Opt_checkpoint_merge, "checkpoint_merge"},
232 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
233 {Opt_compress_algorithm, "compress_algorithm=%s"},
234 {Opt_compress_log_size, "compress_log_size=%u"},
235 {Opt_compress_extension, "compress_extension=%s"},
236 {Opt_nocompress_extension, "nocompress_extension=%s"},
237 {Opt_compress_chksum, "compress_chksum"},
238 {Opt_compress_mode, "compress_mode=%s"},
239 {Opt_compress_cache, "compress_cache"},
241 {Opt_gc_merge, "gc_merge"},
242 {Opt_nogc_merge, "nogc_merge"},
243 {Opt_discard_unit, "discard_unit=%s"},
244 {Opt_memory_mode, "memory=%s"},
245 {Opt_age_extent_cache, "age_extent_cache"},
249 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
251 struct va_format vaf;
257 level = printk_get_level(fmt);
258 vaf.fmt = printk_skip_level(fmt);
260 printk("%c%cF2FS-fs (%s): %pV\n",
261 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
266 #if IS_ENABLED(CONFIG_UNICODE)
267 static const struct f2fs_sb_encodings {
270 unsigned int version;
271 } f2fs_sb_encoding_map[] = {
272 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
275 static const struct f2fs_sb_encodings *
276 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
278 __u16 magic = le16_to_cpu(sb->s_encoding);
281 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
282 if (magic == f2fs_sb_encoding_map[i].magic)
283 return &f2fs_sb_encoding_map[i];
288 struct kmem_cache *f2fs_cf_name_slab;
289 static int __init f2fs_create_casefold_cache(void)
291 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
293 return f2fs_cf_name_slab ? 0 : -ENOMEM;
296 static void f2fs_destroy_casefold_cache(void)
298 kmem_cache_destroy(f2fs_cf_name_slab);
301 static int __init f2fs_create_casefold_cache(void) { return 0; }
302 static void f2fs_destroy_casefold_cache(void) { }
305 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
307 block_t limit = min((sbi->user_block_count >> 3),
308 sbi->user_block_count - sbi->reserved_blocks);
311 if (test_opt(sbi, RESERVE_ROOT) &&
312 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
313 F2FS_OPTION(sbi).root_reserved_blocks = limit;
314 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
315 F2FS_OPTION(sbi).root_reserved_blocks);
317 if (!test_opt(sbi, RESERVE_ROOT) &&
318 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
319 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
320 !gid_eq(F2FS_OPTION(sbi).s_resgid,
321 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
322 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
323 from_kuid_munged(&init_user_ns,
324 F2FS_OPTION(sbi).s_resuid),
325 from_kgid_munged(&init_user_ns,
326 F2FS_OPTION(sbi).s_resgid));
329 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
331 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
332 unsigned int avg_vblocks;
333 unsigned int wanted_reserved_segments;
334 block_t avail_user_block_count;
336 if (!F2FS_IO_ALIGNED(sbi))
339 /* average valid block count in section in worst case */
340 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
343 * we need enough free space when migrating one section in worst case
345 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
346 reserved_segments(sbi);
347 wanted_reserved_segments -= reserved_segments(sbi);
349 avail_user_block_count = sbi->user_block_count -
350 sbi->current_reserved_blocks -
351 F2FS_OPTION(sbi).root_reserved_blocks;
353 if (wanted_reserved_segments * sbi->blocks_per_seg >
354 avail_user_block_count) {
355 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
356 wanted_reserved_segments,
357 avail_user_block_count >> sbi->log_blocks_per_seg);
361 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
363 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
364 wanted_reserved_segments);
369 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
371 if (!F2FS_OPTION(sbi).unusable_cap_perc)
374 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
375 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
377 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
378 F2FS_OPTION(sbi).unusable_cap_perc;
380 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
381 F2FS_OPTION(sbi).unusable_cap,
382 F2FS_OPTION(sbi).unusable_cap_perc);
385 static void init_once(void *foo)
387 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
389 inode_init_once(&fi->vfs_inode);
393 static const char * const quotatypes[] = INITQFNAMES;
394 #define QTYPE2NAME(t) (quotatypes[t])
395 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
398 struct f2fs_sb_info *sbi = F2FS_SB(sb);
402 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
403 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
406 if (f2fs_sb_has_quota_ino(sbi)) {
407 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
411 qname = match_strdup(args);
413 f2fs_err(sbi, "Not enough memory for storing quotafile name");
416 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
417 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
420 f2fs_err(sbi, "%s quota file already specified",
424 if (strchr(qname, '/')) {
425 f2fs_err(sbi, "quotafile must be on filesystem root");
428 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
436 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
438 struct f2fs_sb_info *sbi = F2FS_SB(sb);
440 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
441 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
444 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
445 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
449 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
452 * We do the test below only for project quotas. 'usrquota' and
453 * 'grpquota' mount options are allowed even without quota feature
454 * to support legacy quotas in quota files.
456 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
457 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
460 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
461 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
462 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
463 if (test_opt(sbi, USRQUOTA) &&
464 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
465 clear_opt(sbi, USRQUOTA);
467 if (test_opt(sbi, GRPQUOTA) &&
468 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
469 clear_opt(sbi, GRPQUOTA);
471 if (test_opt(sbi, PRJQUOTA) &&
472 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
473 clear_opt(sbi, PRJQUOTA);
475 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
476 test_opt(sbi, PRJQUOTA)) {
477 f2fs_err(sbi, "old and new quota format mixing");
481 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
482 f2fs_err(sbi, "journaled quota format not specified");
487 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
488 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
489 F2FS_OPTION(sbi).s_jquota_fmt = 0;
495 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
497 const substring_t *arg,
500 struct f2fs_sb_info *sbi = F2FS_SB(sb);
501 struct fs_parameter param = {
502 .type = fs_value_is_string,
503 .string = arg->from ? arg->from : "",
505 struct fscrypt_dummy_policy *policy =
506 &F2FS_OPTION(sbi).dummy_enc_policy;
509 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
510 f2fs_warn(sbi, "test_dummy_encryption option not supported");
514 if (!f2fs_sb_has_encrypt(sbi)) {
515 f2fs_err(sbi, "Encrypt feature is off");
520 * This mount option is just for testing, and it's not worthwhile to
521 * implement the extra complexity (e.g. RCU protection) that would be
522 * needed to allow it to be set or changed during remount. We do allow
523 * it to be specified during remount, but only if there is no change.
525 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
526 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
530 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
534 "Can't change test_dummy_encryption on remount");
535 else if (err == -EINVAL)
536 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
539 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
543 err = fscrypt_add_test_dummy_key(sb, policy);
545 f2fs_warn(sbi, "Error adding test dummy encryption key [%d]",
549 f2fs_warn(sbi, "Test dummy encryption mode enabled");
553 #ifdef CONFIG_F2FS_FS_COMPRESSION
555 * 1. The same extension name cannot not appear in both compress and non-compress extension
557 * 2. If the compress extension specifies all files, the types specified by the non-compress
558 * extension will be treated as special cases and will not be compressed.
559 * 3. Don't allow the non-compress extension specifies all files.
561 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
563 unsigned char (*ext)[F2FS_EXTENSION_LEN];
564 unsigned char (*noext)[F2FS_EXTENSION_LEN];
565 int ext_cnt, noext_cnt, index = 0, no_index = 0;
567 ext = F2FS_OPTION(sbi).extensions;
568 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
569 noext = F2FS_OPTION(sbi).noextensions;
570 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
575 for (no_index = 0; no_index < noext_cnt; no_index++) {
576 if (!strcasecmp("*", noext[no_index])) {
577 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
580 for (index = 0; index < ext_cnt; index++) {
581 if (!strcasecmp(ext[index], noext[no_index])) {
582 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
591 #ifdef CONFIG_F2FS_FS_LZ4
592 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
594 #ifdef CONFIG_F2FS_FS_LZ4HC
598 if (strlen(str) == 3) {
599 F2FS_OPTION(sbi).compress_level = 0;
603 #ifdef CONFIG_F2FS_FS_LZ4HC
607 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
610 if (kstrtouint(str + 1, 10, &level))
613 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
614 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
618 F2FS_OPTION(sbi).compress_level = level;
621 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
627 #ifdef CONFIG_F2FS_FS_ZSTD
628 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
633 if (strlen(str) == len) {
634 F2FS_OPTION(sbi).compress_level = 0;
641 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
644 if (kstrtouint(str + 1, 10, &level))
647 if (!level || level > zstd_max_clevel()) {
648 f2fs_info(sbi, "invalid zstd compress level: %d", level);
652 F2FS_OPTION(sbi).compress_level = level;
658 static int parse_options(struct super_block *sb, char *options, bool is_remount)
660 struct f2fs_sb_info *sbi = F2FS_SB(sb);
661 substring_t args[MAX_OPT_ARGS];
662 #ifdef CONFIG_F2FS_FS_COMPRESSION
663 unsigned char (*ext)[F2FS_EXTENSION_LEN];
664 unsigned char (*noext)[F2FS_EXTENSION_LEN];
665 int ext_cnt, noext_cnt;
676 while ((p = strsep(&options, ",")) != NULL) {
682 * Initialize args struct so we know whether arg was
683 * found; some options take optional arguments.
685 args[0].to = args[0].from = NULL;
686 token = match_token(p, f2fs_tokens, args);
689 case Opt_gc_background:
690 name = match_strdup(&args[0]);
694 if (!strcmp(name, "on")) {
695 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
696 } else if (!strcmp(name, "off")) {
697 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
698 } else if (!strcmp(name, "sync")) {
699 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
706 case Opt_disable_roll_forward:
707 set_opt(sbi, DISABLE_ROLL_FORWARD);
710 /* this option mounts f2fs with ro */
711 set_opt(sbi, NORECOVERY);
712 if (!f2fs_readonly(sb))
716 if (!f2fs_hw_support_discard(sbi)) {
717 f2fs_warn(sbi, "device does not support discard");
720 set_opt(sbi, DISCARD);
723 if (f2fs_hw_should_discard(sbi)) {
724 f2fs_warn(sbi, "discard is required for zoned block devices");
727 clear_opt(sbi, DISCARD);
730 set_opt(sbi, NOHEAP);
733 clear_opt(sbi, NOHEAP);
735 #ifdef CONFIG_F2FS_FS_XATTR
737 set_opt(sbi, XATTR_USER);
739 case Opt_nouser_xattr:
740 clear_opt(sbi, XATTR_USER);
742 case Opt_inline_xattr:
743 set_opt(sbi, INLINE_XATTR);
745 case Opt_noinline_xattr:
746 clear_opt(sbi, INLINE_XATTR);
748 case Opt_inline_xattr_size:
749 if (args->from && match_int(args, &arg))
751 set_opt(sbi, INLINE_XATTR_SIZE);
752 F2FS_OPTION(sbi).inline_xattr_size = arg;
756 f2fs_info(sbi, "user_xattr options not supported");
758 case Opt_nouser_xattr:
759 f2fs_info(sbi, "nouser_xattr options not supported");
761 case Opt_inline_xattr:
762 f2fs_info(sbi, "inline_xattr options not supported");
764 case Opt_noinline_xattr:
765 f2fs_info(sbi, "noinline_xattr options not supported");
768 #ifdef CONFIG_F2FS_FS_POSIX_ACL
770 set_opt(sbi, POSIX_ACL);
773 clear_opt(sbi, POSIX_ACL);
777 f2fs_info(sbi, "acl options not supported");
780 f2fs_info(sbi, "noacl options not supported");
783 case Opt_active_logs:
784 if (args->from && match_int(args, &arg))
786 if (arg != 2 && arg != 4 &&
787 arg != NR_CURSEG_PERSIST_TYPE)
789 F2FS_OPTION(sbi).active_logs = arg;
791 case Opt_disable_ext_identify:
792 set_opt(sbi, DISABLE_EXT_IDENTIFY);
794 case Opt_inline_data:
795 set_opt(sbi, INLINE_DATA);
797 case Opt_inline_dentry:
798 set_opt(sbi, INLINE_DENTRY);
800 case Opt_noinline_dentry:
801 clear_opt(sbi, INLINE_DENTRY);
803 case Opt_flush_merge:
804 set_opt(sbi, FLUSH_MERGE);
806 case Opt_noflush_merge:
807 clear_opt(sbi, FLUSH_MERGE);
810 set_opt(sbi, NOBARRIER);
813 clear_opt(sbi, NOBARRIER);
816 set_opt(sbi, FASTBOOT);
818 case Opt_extent_cache:
819 set_opt(sbi, READ_EXTENT_CACHE);
821 case Opt_noextent_cache:
822 clear_opt(sbi, READ_EXTENT_CACHE);
824 case Opt_noinline_data:
825 clear_opt(sbi, INLINE_DATA);
828 set_opt(sbi, DATA_FLUSH);
830 case Opt_reserve_root:
831 if (args->from && match_int(args, &arg))
833 if (test_opt(sbi, RESERVE_ROOT)) {
834 f2fs_info(sbi, "Preserve previous reserve_root=%u",
835 F2FS_OPTION(sbi).root_reserved_blocks);
837 F2FS_OPTION(sbi).root_reserved_blocks = arg;
838 set_opt(sbi, RESERVE_ROOT);
842 if (args->from && match_int(args, &arg))
844 uid = make_kuid(current_user_ns(), arg);
845 if (!uid_valid(uid)) {
846 f2fs_err(sbi, "Invalid uid value %d", arg);
849 F2FS_OPTION(sbi).s_resuid = uid;
852 if (args->from && match_int(args, &arg))
854 gid = make_kgid(current_user_ns(), arg);
855 if (!gid_valid(gid)) {
856 f2fs_err(sbi, "Invalid gid value %d", arg);
859 F2FS_OPTION(sbi).s_resgid = gid;
862 name = match_strdup(&args[0]);
866 if (!strcmp(name, "adaptive")) {
867 if (f2fs_sb_has_blkzoned(sbi)) {
868 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
872 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
873 } else if (!strcmp(name, "lfs")) {
874 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
875 } else if (!strcmp(name, "fragment:segment")) {
876 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
877 } else if (!strcmp(name, "fragment:block")) {
878 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
885 case Opt_io_size_bits:
886 if (args->from && match_int(args, &arg))
888 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
889 f2fs_warn(sbi, "Not support %d, larger than %d",
890 1 << arg, BIO_MAX_VECS);
893 F2FS_OPTION(sbi).write_io_size_bits = arg;
895 #ifdef CONFIG_F2FS_FAULT_INJECTION
896 case Opt_fault_injection:
897 if (args->from && match_int(args, &arg))
899 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
900 set_opt(sbi, FAULT_INJECTION);
904 if (args->from && match_int(args, &arg))
906 f2fs_build_fault_attr(sbi, 0, arg);
907 set_opt(sbi, FAULT_INJECTION);
910 case Opt_fault_injection:
911 f2fs_info(sbi, "fault_injection options not supported");
915 f2fs_info(sbi, "fault_type options not supported");
919 sb->s_flags |= SB_LAZYTIME;
922 sb->s_flags &= ~SB_LAZYTIME;
927 set_opt(sbi, USRQUOTA);
930 set_opt(sbi, GRPQUOTA);
933 set_opt(sbi, PRJQUOTA);
936 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
941 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
946 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
950 case Opt_offusrjquota:
951 ret = f2fs_clear_qf_name(sb, USRQUOTA);
955 case Opt_offgrpjquota:
956 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
960 case Opt_offprjjquota:
961 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
965 case Opt_jqfmt_vfsold:
966 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
968 case Opt_jqfmt_vfsv0:
969 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
971 case Opt_jqfmt_vfsv1:
972 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
975 clear_opt(sbi, QUOTA);
976 clear_opt(sbi, USRQUOTA);
977 clear_opt(sbi, GRPQUOTA);
978 clear_opt(sbi, PRJQUOTA);
988 case Opt_offusrjquota:
989 case Opt_offgrpjquota:
990 case Opt_offprjjquota:
991 case Opt_jqfmt_vfsold:
992 case Opt_jqfmt_vfsv0:
993 case Opt_jqfmt_vfsv1:
995 f2fs_info(sbi, "quota operations not supported");
999 name = match_strdup(&args[0]);
1003 if (!strcmp(name, "default")) {
1004 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1005 } else if (!strcmp(name, "reuse")) {
1006 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1014 name = match_strdup(&args[0]);
1017 if (!strcmp(name, "posix")) {
1018 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1019 } else if (!strcmp(name, "strict")) {
1020 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1021 } else if (!strcmp(name, "nobarrier")) {
1022 F2FS_OPTION(sbi).fsync_mode =
1023 FSYNC_MODE_NOBARRIER;
1030 case Opt_test_dummy_encryption:
1031 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1036 case Opt_inlinecrypt:
1037 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1038 sb->s_flags |= SB_INLINECRYPT;
1040 f2fs_info(sbi, "inline encryption not supported");
1043 case Opt_checkpoint_disable_cap_perc:
1044 if (args->from && match_int(args, &arg))
1046 if (arg < 0 || arg > 100)
1048 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1049 set_opt(sbi, DISABLE_CHECKPOINT);
1051 case Opt_checkpoint_disable_cap:
1052 if (args->from && match_int(args, &arg))
1054 F2FS_OPTION(sbi).unusable_cap = arg;
1055 set_opt(sbi, DISABLE_CHECKPOINT);
1057 case Opt_checkpoint_disable:
1058 set_opt(sbi, DISABLE_CHECKPOINT);
1060 case Opt_checkpoint_enable:
1061 clear_opt(sbi, DISABLE_CHECKPOINT);
1063 case Opt_checkpoint_merge:
1064 set_opt(sbi, MERGE_CHECKPOINT);
1066 case Opt_nocheckpoint_merge:
1067 clear_opt(sbi, MERGE_CHECKPOINT);
1069 #ifdef CONFIG_F2FS_FS_COMPRESSION
1070 case Opt_compress_algorithm:
1071 if (!f2fs_sb_has_compression(sbi)) {
1072 f2fs_info(sbi, "Image doesn't support compression");
1075 name = match_strdup(&args[0]);
1078 if (!strcmp(name, "lzo")) {
1079 #ifdef CONFIG_F2FS_FS_LZO
1080 F2FS_OPTION(sbi).compress_level = 0;
1081 F2FS_OPTION(sbi).compress_algorithm =
1084 f2fs_info(sbi, "kernel doesn't support lzo compression");
1086 } else if (!strncmp(name, "lz4", 3)) {
1087 #ifdef CONFIG_F2FS_FS_LZ4
1088 ret = f2fs_set_lz4hc_level(sbi, name);
1093 F2FS_OPTION(sbi).compress_algorithm =
1096 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1098 } else if (!strncmp(name, "zstd", 4)) {
1099 #ifdef CONFIG_F2FS_FS_ZSTD
1100 ret = f2fs_set_zstd_level(sbi, name);
1105 F2FS_OPTION(sbi).compress_algorithm =
1108 f2fs_info(sbi, "kernel doesn't support zstd compression");
1110 } else if (!strcmp(name, "lzo-rle")) {
1111 #ifdef CONFIG_F2FS_FS_LZORLE
1112 F2FS_OPTION(sbi).compress_level = 0;
1113 F2FS_OPTION(sbi).compress_algorithm =
1116 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1124 case Opt_compress_log_size:
1125 if (!f2fs_sb_has_compression(sbi)) {
1126 f2fs_info(sbi, "Image doesn't support compression");
1129 if (args->from && match_int(args, &arg))
1131 if (arg < MIN_COMPRESS_LOG_SIZE ||
1132 arg > MAX_COMPRESS_LOG_SIZE) {
1134 "Compress cluster log size is out of range");
1137 F2FS_OPTION(sbi).compress_log_size = arg;
1139 case Opt_compress_extension:
1140 if (!f2fs_sb_has_compression(sbi)) {
1141 f2fs_info(sbi, "Image doesn't support compression");
1144 name = match_strdup(&args[0]);
1148 ext = F2FS_OPTION(sbi).extensions;
1149 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1151 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1152 ext_cnt >= COMPRESS_EXT_NUM) {
1154 "invalid extension length/number");
1159 strcpy(ext[ext_cnt], name);
1160 F2FS_OPTION(sbi).compress_ext_cnt++;
1163 case Opt_nocompress_extension:
1164 if (!f2fs_sb_has_compression(sbi)) {
1165 f2fs_info(sbi, "Image doesn't support compression");
1168 name = match_strdup(&args[0]);
1172 noext = F2FS_OPTION(sbi).noextensions;
1173 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1175 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1176 noext_cnt >= COMPRESS_EXT_NUM) {
1178 "invalid extension length/number");
1183 strcpy(noext[noext_cnt], name);
1184 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1187 case Opt_compress_chksum:
1188 F2FS_OPTION(sbi).compress_chksum = true;
1190 case Opt_compress_mode:
1191 name = match_strdup(&args[0]);
1194 if (!strcmp(name, "fs")) {
1195 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1196 } else if (!strcmp(name, "user")) {
1197 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1204 case Opt_compress_cache:
1205 set_opt(sbi, COMPRESS_CACHE);
1208 case Opt_compress_algorithm:
1209 case Opt_compress_log_size:
1210 case Opt_compress_extension:
1211 case Opt_nocompress_extension:
1212 case Opt_compress_chksum:
1213 case Opt_compress_mode:
1214 case Opt_compress_cache:
1215 f2fs_info(sbi, "compression options not supported");
1222 set_opt(sbi, GC_MERGE);
1224 case Opt_nogc_merge:
1225 clear_opt(sbi, GC_MERGE);
1227 case Opt_discard_unit:
1228 name = match_strdup(&args[0]);
1231 if (!strcmp(name, "block")) {
1232 F2FS_OPTION(sbi).discard_unit =
1234 } else if (!strcmp(name, "segment")) {
1235 F2FS_OPTION(sbi).discard_unit =
1236 DISCARD_UNIT_SEGMENT;
1237 } else if (!strcmp(name, "section")) {
1238 F2FS_OPTION(sbi).discard_unit =
1239 DISCARD_UNIT_SECTION;
1246 case Opt_memory_mode:
1247 name = match_strdup(&args[0]);
1250 if (!strcmp(name, "normal")) {
1251 F2FS_OPTION(sbi).memory_mode =
1253 } else if (!strcmp(name, "low")) {
1254 F2FS_OPTION(sbi).memory_mode =
1262 case Opt_age_extent_cache:
1263 set_opt(sbi, AGE_EXTENT_CACHE);
1266 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1273 if (f2fs_check_quota_options(sbi))
1276 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1277 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1280 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1281 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1285 #if !IS_ENABLED(CONFIG_UNICODE)
1286 if (f2fs_sb_has_casefold(sbi)) {
1288 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1293 * The BLKZONED feature indicates that the drive was formatted with
1294 * zone alignment optimization. This is optional for host-aware
1295 * devices, but mandatory for host-managed zoned block devices.
1297 if (f2fs_sb_has_blkzoned(sbi)) {
1298 #ifdef CONFIG_BLK_DEV_ZONED
1299 if (F2FS_OPTION(sbi).discard_unit !=
1300 DISCARD_UNIT_SECTION) {
1301 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1302 F2FS_OPTION(sbi).discard_unit =
1303 DISCARD_UNIT_SECTION;
1306 f2fs_err(sbi, "Zoned block device support is not enabled");
1311 #ifdef CONFIG_F2FS_FS_COMPRESSION
1312 if (f2fs_test_compress_extension(sbi)) {
1313 f2fs_err(sbi, "invalid compress or nocompress extension");
1318 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1319 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1320 F2FS_IO_SIZE_KB(sbi));
1324 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1325 int min_size, max_size;
1327 if (!f2fs_sb_has_extra_attr(sbi) ||
1328 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1329 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1332 if (!test_opt(sbi, INLINE_XATTR)) {
1333 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1337 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1338 max_size = MAX_INLINE_XATTR_SIZE;
1340 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1341 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1342 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1343 min_size, max_size);
1348 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1349 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1353 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1354 f2fs_err(sbi, "LFS not compatible with ATGC");
1358 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1359 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1363 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1364 f2fs_err(sbi, "Allow to mount readonly mode only");
1370 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1372 struct f2fs_inode_info *fi;
1374 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) {
1375 f2fs_show_injection_info(F2FS_SB(sb), FAULT_SLAB_ALLOC);
1379 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1383 init_once((void *) fi);
1385 /* Initialize f2fs-specific inode info */
1386 atomic_set(&fi->dirty_pages, 0);
1387 atomic_set(&fi->i_compr_blocks, 0);
1388 init_f2fs_rwsem(&fi->i_sem);
1389 spin_lock_init(&fi->i_size_lock);
1390 INIT_LIST_HEAD(&fi->dirty_list);
1391 INIT_LIST_HEAD(&fi->gdirty_list);
1392 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1393 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1394 init_f2fs_rwsem(&fi->i_xattr_sem);
1396 /* Will be used by directory only */
1397 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1399 return &fi->vfs_inode;
1402 static int f2fs_drop_inode(struct inode *inode)
1404 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1408 * during filesystem shutdown, if checkpoint is disabled,
1409 * drop useless meta/node dirty pages.
1411 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1412 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1413 inode->i_ino == F2FS_META_INO(sbi)) {
1414 trace_f2fs_drop_inode(inode, 1);
1420 * This is to avoid a deadlock condition like below.
1421 * writeback_single_inode(inode)
1422 * - f2fs_write_data_page
1423 * - f2fs_gc -> iput -> evict
1424 * - inode_wait_for_writeback(inode)
1426 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1427 if (!inode->i_nlink && !is_bad_inode(inode)) {
1428 /* to avoid evict_inode call simultaneously */
1429 atomic_inc(&inode->i_count);
1430 spin_unlock(&inode->i_lock);
1432 f2fs_abort_atomic_write(inode, true);
1434 /* should remain fi->extent_tree for writepage */
1435 f2fs_destroy_extent_node(inode);
1437 sb_start_intwrite(inode->i_sb);
1438 f2fs_i_size_write(inode, 0);
1440 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1441 inode, NULL, 0, DATA);
1442 truncate_inode_pages_final(inode->i_mapping);
1444 if (F2FS_HAS_BLOCKS(inode))
1445 f2fs_truncate(inode);
1447 sb_end_intwrite(inode->i_sb);
1449 spin_lock(&inode->i_lock);
1450 atomic_dec(&inode->i_count);
1452 trace_f2fs_drop_inode(inode, 0);
1455 ret = generic_drop_inode(inode);
1457 ret = fscrypt_drop_inode(inode);
1458 trace_f2fs_drop_inode(inode, ret);
1462 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1464 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1467 spin_lock(&sbi->inode_lock[DIRTY_META]);
1468 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1471 set_inode_flag(inode, FI_DIRTY_INODE);
1472 stat_inc_dirty_inode(sbi, DIRTY_META);
1474 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1475 list_add_tail(&F2FS_I(inode)->gdirty_list,
1476 &sbi->inode_list[DIRTY_META]);
1477 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1479 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1483 void f2fs_inode_synced(struct inode *inode)
1485 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1487 spin_lock(&sbi->inode_lock[DIRTY_META]);
1488 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1489 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1492 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1493 list_del_init(&F2FS_I(inode)->gdirty_list);
1494 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1496 clear_inode_flag(inode, FI_DIRTY_INODE);
1497 clear_inode_flag(inode, FI_AUTO_RECOVER);
1498 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1499 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1503 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1505 * We should call set_dirty_inode to write the dirty inode through write_inode.
1507 static void f2fs_dirty_inode(struct inode *inode, int flags)
1509 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1511 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1512 inode->i_ino == F2FS_META_INO(sbi))
1515 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1516 clear_inode_flag(inode, FI_AUTO_RECOVER);
1518 f2fs_inode_dirtied(inode, false);
1521 static void f2fs_free_inode(struct inode *inode)
1523 fscrypt_free_inode(inode);
1524 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1527 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1529 percpu_counter_destroy(&sbi->total_valid_inode_count);
1530 percpu_counter_destroy(&sbi->rf_node_block_count);
1531 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1534 static void destroy_device_list(struct f2fs_sb_info *sbi)
1538 for (i = 0; i < sbi->s_ndevs; i++) {
1539 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1540 #ifdef CONFIG_BLK_DEV_ZONED
1541 kvfree(FDEV(i).blkz_seq);
1547 static void f2fs_put_super(struct super_block *sb)
1549 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1553 /* unregister procfs/sysfs entries in advance to avoid race case */
1554 f2fs_unregister_sysfs(sbi);
1556 f2fs_quota_off_umount(sb);
1558 /* prevent remaining shrinker jobs */
1559 mutex_lock(&sbi->umount_mutex);
1562 * flush all issued checkpoints and stop checkpoint issue thread.
1563 * after then, all checkpoints should be done by each process context.
1565 f2fs_stop_ckpt_thread(sbi);
1568 * We don't need to do checkpoint when superblock is clean.
1569 * But, the previous checkpoint was not done by umount, it needs to do
1570 * clean checkpoint again.
1572 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1573 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1574 struct cp_control cpc = {
1575 .reason = CP_UMOUNT,
1577 f2fs_write_checkpoint(sbi, &cpc);
1580 /* be sure to wait for any on-going discard commands */
1581 dropped = f2fs_issue_discard_timeout(sbi);
1583 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && !dropped) {
1584 struct cp_control cpc = {
1585 .reason = CP_UMOUNT | CP_TRIMMED,
1587 f2fs_write_checkpoint(sbi, &cpc);
1591 * normally superblock is clean, so we need to release this.
1592 * In addition, EIO will skip do checkpoint, we need this as well.
1594 f2fs_release_ino_entry(sbi, true);
1596 f2fs_leave_shrinker(sbi);
1597 mutex_unlock(&sbi->umount_mutex);
1599 /* our cp_error case, we can wait for any writeback page */
1600 f2fs_flush_merged_writes(sbi);
1602 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1604 f2fs_bug_on(sbi, sbi->fsync_node_num);
1606 f2fs_destroy_compress_inode(sbi);
1608 iput(sbi->node_inode);
1609 sbi->node_inode = NULL;
1611 iput(sbi->meta_inode);
1612 sbi->meta_inode = NULL;
1615 * iput() can update stat information, if f2fs_write_checkpoint()
1616 * above failed with error.
1618 f2fs_destroy_stats(sbi);
1620 /* destroy f2fs internal modules */
1621 f2fs_destroy_node_manager(sbi);
1622 f2fs_destroy_segment_manager(sbi);
1624 f2fs_destroy_post_read_wq(sbi);
1628 sb->s_fs_info = NULL;
1629 if (sbi->s_chksum_driver)
1630 crypto_free_shash(sbi->s_chksum_driver);
1631 kfree(sbi->raw_super);
1633 destroy_device_list(sbi);
1634 f2fs_destroy_page_array_cache(sbi);
1635 f2fs_destroy_xattr_caches(sbi);
1636 mempool_destroy(sbi->write_io_dummy);
1638 for (i = 0; i < MAXQUOTAS; i++)
1639 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1641 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1642 destroy_percpu_info(sbi);
1643 f2fs_destroy_iostat(sbi);
1644 for (i = 0; i < NR_PAGE_TYPE; i++)
1645 kvfree(sbi->write_io[i]);
1646 #if IS_ENABLED(CONFIG_UNICODE)
1647 utf8_unload(sb->s_encoding);
1652 int f2fs_sync_fs(struct super_block *sb, int sync)
1654 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1657 if (unlikely(f2fs_cp_error(sbi)))
1659 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1662 trace_f2fs_sync_fs(sb, sync);
1664 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1668 err = f2fs_issue_checkpoint(sbi);
1673 static int f2fs_freeze(struct super_block *sb)
1675 if (f2fs_readonly(sb))
1678 /* IO error happened before */
1679 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1682 /* must be clean, since sync_filesystem() was already called */
1683 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1686 /* Let's flush checkpoints and stop the thread. */
1687 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1689 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1690 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1694 static int f2fs_unfreeze(struct super_block *sb)
1696 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1701 static int f2fs_statfs_project(struct super_block *sb,
1702 kprojid_t projid, struct kstatfs *buf)
1705 struct dquot *dquot;
1709 qid = make_kqid_projid(projid);
1710 dquot = dqget(sb, qid);
1712 return PTR_ERR(dquot);
1713 spin_lock(&dquot->dq_dqb_lock);
1715 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1716 dquot->dq_dqb.dqb_bhardlimit);
1718 limit >>= sb->s_blocksize_bits;
1720 if (limit && buf->f_blocks > limit) {
1721 curblock = (dquot->dq_dqb.dqb_curspace +
1722 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1723 buf->f_blocks = limit;
1724 buf->f_bfree = buf->f_bavail =
1725 (buf->f_blocks > curblock) ?
1726 (buf->f_blocks - curblock) : 0;
1729 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1730 dquot->dq_dqb.dqb_ihardlimit);
1732 if (limit && buf->f_files > limit) {
1733 buf->f_files = limit;
1735 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1736 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1739 spin_unlock(&dquot->dq_dqb_lock);
1745 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1747 struct super_block *sb = dentry->d_sb;
1748 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1749 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1750 block_t total_count, user_block_count, start_count;
1751 u64 avail_node_count;
1752 unsigned int total_valid_node_count;
1754 total_count = le64_to_cpu(sbi->raw_super->block_count);
1755 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1756 buf->f_type = F2FS_SUPER_MAGIC;
1757 buf->f_bsize = sbi->blocksize;
1759 buf->f_blocks = total_count - start_count;
1761 spin_lock(&sbi->stat_lock);
1763 user_block_count = sbi->user_block_count;
1764 total_valid_node_count = valid_node_count(sbi);
1765 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1766 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1767 sbi->current_reserved_blocks;
1769 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1772 buf->f_bfree -= sbi->unusable_block_count;
1773 spin_unlock(&sbi->stat_lock);
1775 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1776 buf->f_bavail = buf->f_bfree -
1777 F2FS_OPTION(sbi).root_reserved_blocks;
1781 if (avail_node_count > user_block_count) {
1782 buf->f_files = user_block_count;
1783 buf->f_ffree = buf->f_bavail;
1785 buf->f_files = avail_node_count;
1786 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1790 buf->f_namelen = F2FS_NAME_LEN;
1791 buf->f_fsid = u64_to_fsid(id);
1794 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1795 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1796 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1802 static inline void f2fs_show_quota_options(struct seq_file *seq,
1803 struct super_block *sb)
1806 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1808 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1811 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1822 seq_printf(seq, ",jqfmt=%s", fmtname);
1825 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1826 seq_show_option(seq, "usrjquota",
1827 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1829 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1830 seq_show_option(seq, "grpjquota",
1831 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1833 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1834 seq_show_option(seq, "prjjquota",
1835 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1839 #ifdef CONFIG_F2FS_FS_COMPRESSION
1840 static inline void f2fs_show_compress_options(struct seq_file *seq,
1841 struct super_block *sb)
1843 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1847 if (!f2fs_sb_has_compression(sbi))
1850 switch (F2FS_OPTION(sbi).compress_algorithm) {
1860 case COMPRESS_LZORLE:
1861 algtype = "lzo-rle";
1864 seq_printf(seq, ",compress_algorithm=%s", algtype);
1866 if (F2FS_OPTION(sbi).compress_level)
1867 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1869 seq_printf(seq, ",compress_log_size=%u",
1870 F2FS_OPTION(sbi).compress_log_size);
1872 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1873 seq_printf(seq, ",compress_extension=%s",
1874 F2FS_OPTION(sbi).extensions[i]);
1877 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1878 seq_printf(seq, ",nocompress_extension=%s",
1879 F2FS_OPTION(sbi).noextensions[i]);
1882 if (F2FS_OPTION(sbi).compress_chksum)
1883 seq_puts(seq, ",compress_chksum");
1885 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1886 seq_printf(seq, ",compress_mode=%s", "fs");
1887 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1888 seq_printf(seq, ",compress_mode=%s", "user");
1890 if (test_opt(sbi, COMPRESS_CACHE))
1891 seq_puts(seq, ",compress_cache");
1895 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1897 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1899 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1900 seq_printf(seq, ",background_gc=%s", "sync");
1901 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1902 seq_printf(seq, ",background_gc=%s", "on");
1903 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1904 seq_printf(seq, ",background_gc=%s", "off");
1906 if (test_opt(sbi, GC_MERGE))
1907 seq_puts(seq, ",gc_merge");
1909 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1910 seq_puts(seq, ",disable_roll_forward");
1911 if (test_opt(sbi, NORECOVERY))
1912 seq_puts(seq, ",norecovery");
1913 if (test_opt(sbi, DISCARD))
1914 seq_puts(seq, ",discard");
1916 seq_puts(seq, ",nodiscard");
1917 if (test_opt(sbi, NOHEAP))
1918 seq_puts(seq, ",no_heap");
1920 seq_puts(seq, ",heap");
1921 #ifdef CONFIG_F2FS_FS_XATTR
1922 if (test_opt(sbi, XATTR_USER))
1923 seq_puts(seq, ",user_xattr");
1925 seq_puts(seq, ",nouser_xattr");
1926 if (test_opt(sbi, INLINE_XATTR))
1927 seq_puts(seq, ",inline_xattr");
1929 seq_puts(seq, ",noinline_xattr");
1930 if (test_opt(sbi, INLINE_XATTR_SIZE))
1931 seq_printf(seq, ",inline_xattr_size=%u",
1932 F2FS_OPTION(sbi).inline_xattr_size);
1934 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1935 if (test_opt(sbi, POSIX_ACL))
1936 seq_puts(seq, ",acl");
1938 seq_puts(seq, ",noacl");
1940 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1941 seq_puts(seq, ",disable_ext_identify");
1942 if (test_opt(sbi, INLINE_DATA))
1943 seq_puts(seq, ",inline_data");
1945 seq_puts(seq, ",noinline_data");
1946 if (test_opt(sbi, INLINE_DENTRY))
1947 seq_puts(seq, ",inline_dentry");
1949 seq_puts(seq, ",noinline_dentry");
1950 if (test_opt(sbi, FLUSH_MERGE))
1951 seq_puts(seq, ",flush_merge");
1953 seq_puts(seq, ",noflush_merge");
1954 if (test_opt(sbi, NOBARRIER))
1955 seq_puts(seq, ",nobarrier");
1957 seq_puts(seq, ",barrier");
1958 if (test_opt(sbi, FASTBOOT))
1959 seq_puts(seq, ",fastboot");
1960 if (test_opt(sbi, READ_EXTENT_CACHE))
1961 seq_puts(seq, ",extent_cache");
1963 seq_puts(seq, ",noextent_cache");
1964 if (test_opt(sbi, AGE_EXTENT_CACHE))
1965 seq_puts(seq, ",age_extent_cache");
1966 if (test_opt(sbi, DATA_FLUSH))
1967 seq_puts(seq, ",data_flush");
1969 seq_puts(seq, ",mode=");
1970 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1971 seq_puts(seq, "adaptive");
1972 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1973 seq_puts(seq, "lfs");
1974 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1975 seq_puts(seq, "fragment:segment");
1976 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1977 seq_puts(seq, "fragment:block");
1978 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1979 if (test_opt(sbi, RESERVE_ROOT))
1980 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1981 F2FS_OPTION(sbi).root_reserved_blocks,
1982 from_kuid_munged(&init_user_ns,
1983 F2FS_OPTION(sbi).s_resuid),
1984 from_kgid_munged(&init_user_ns,
1985 F2FS_OPTION(sbi).s_resgid));
1986 if (F2FS_IO_SIZE_BITS(sbi))
1987 seq_printf(seq, ",io_bits=%u",
1988 F2FS_OPTION(sbi).write_io_size_bits);
1989 #ifdef CONFIG_F2FS_FAULT_INJECTION
1990 if (test_opt(sbi, FAULT_INJECTION)) {
1991 seq_printf(seq, ",fault_injection=%u",
1992 F2FS_OPTION(sbi).fault_info.inject_rate);
1993 seq_printf(seq, ",fault_type=%u",
1994 F2FS_OPTION(sbi).fault_info.inject_type);
1998 if (test_opt(sbi, QUOTA))
1999 seq_puts(seq, ",quota");
2000 if (test_opt(sbi, USRQUOTA))
2001 seq_puts(seq, ",usrquota");
2002 if (test_opt(sbi, GRPQUOTA))
2003 seq_puts(seq, ",grpquota");
2004 if (test_opt(sbi, PRJQUOTA))
2005 seq_puts(seq, ",prjquota");
2007 f2fs_show_quota_options(seq, sbi->sb);
2009 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2011 if (sbi->sb->s_flags & SB_INLINECRYPT)
2012 seq_puts(seq, ",inlinecrypt");
2014 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2015 seq_printf(seq, ",alloc_mode=%s", "default");
2016 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2017 seq_printf(seq, ",alloc_mode=%s", "reuse");
2019 if (test_opt(sbi, DISABLE_CHECKPOINT))
2020 seq_printf(seq, ",checkpoint=disable:%u",
2021 F2FS_OPTION(sbi).unusable_cap);
2022 if (test_opt(sbi, MERGE_CHECKPOINT))
2023 seq_puts(seq, ",checkpoint_merge");
2025 seq_puts(seq, ",nocheckpoint_merge");
2026 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2027 seq_printf(seq, ",fsync_mode=%s", "posix");
2028 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2029 seq_printf(seq, ",fsync_mode=%s", "strict");
2030 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2031 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2033 #ifdef CONFIG_F2FS_FS_COMPRESSION
2034 f2fs_show_compress_options(seq, sbi->sb);
2037 if (test_opt(sbi, ATGC))
2038 seq_puts(seq, ",atgc");
2040 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2041 seq_printf(seq, ",discard_unit=%s", "block");
2042 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2043 seq_printf(seq, ",discard_unit=%s", "segment");
2044 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2045 seq_printf(seq, ",discard_unit=%s", "section");
2047 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2048 seq_printf(seq, ",memory=%s", "normal");
2049 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2050 seq_printf(seq, ",memory=%s", "low");
2055 static void default_options(struct f2fs_sb_info *sbi)
2057 /* init some FS parameters */
2058 if (f2fs_sb_has_readonly(sbi))
2059 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2061 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2063 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2064 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2065 SMALL_VOLUME_SEGMENTS)
2066 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2068 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2069 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2070 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2071 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2072 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2073 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2074 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2075 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2076 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2077 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2079 sbi->sb->s_flags &= ~SB_INLINECRYPT;
2081 set_opt(sbi, INLINE_XATTR);
2082 set_opt(sbi, INLINE_DATA);
2083 set_opt(sbi, INLINE_DENTRY);
2084 set_opt(sbi, READ_EXTENT_CACHE);
2085 set_opt(sbi, NOHEAP);
2086 clear_opt(sbi, DISABLE_CHECKPOINT);
2087 set_opt(sbi, MERGE_CHECKPOINT);
2088 F2FS_OPTION(sbi).unusable_cap = 0;
2089 sbi->sb->s_flags |= SB_LAZYTIME;
2090 if (!f2fs_is_readonly(sbi))
2091 set_opt(sbi, FLUSH_MERGE);
2092 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2093 set_opt(sbi, DISCARD);
2094 if (f2fs_sb_has_blkzoned(sbi)) {
2095 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2096 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2098 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2099 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2102 #ifdef CONFIG_F2FS_FS_XATTR
2103 set_opt(sbi, XATTR_USER);
2105 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2106 set_opt(sbi, POSIX_ACL);
2109 f2fs_build_fault_attr(sbi, 0, 0);
2113 static int f2fs_enable_quotas(struct super_block *sb);
2116 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2118 unsigned int s_flags = sbi->sb->s_flags;
2119 struct cp_control cpc;
2120 unsigned int gc_mode = sbi->gc_mode;
2125 if (s_flags & SB_RDONLY) {
2126 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2129 sbi->sb->s_flags |= SB_ACTIVE;
2131 /* check if we need more GC first */
2132 unusable = f2fs_get_unusable_blocks(sbi);
2133 if (!f2fs_disable_cp_again(sbi, unusable))
2136 f2fs_update_time(sbi, DISABLE_TIME);
2138 sbi->gc_mode = GC_URGENT_HIGH;
2140 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2141 struct f2fs_gc_control gc_control = {
2142 .victim_segno = NULL_SEGNO,
2143 .init_gc_type = FG_GC,
2144 .should_migrate_blocks = false,
2145 .err_gc_skipped = true,
2146 .nr_free_secs = 1 };
2148 f2fs_down_write(&sbi->gc_lock);
2149 err = f2fs_gc(sbi, &gc_control);
2150 if (err == -ENODATA) {
2154 if (err && err != -EAGAIN)
2158 ret = sync_filesystem(sbi->sb);
2160 err = ret ? ret : err;
2164 unusable = f2fs_get_unusable_blocks(sbi);
2165 if (f2fs_disable_cp_again(sbi, unusable)) {
2171 f2fs_down_write(&sbi->gc_lock);
2172 cpc.reason = CP_PAUSE;
2173 set_sbi_flag(sbi, SBI_CP_DISABLED);
2174 err = f2fs_write_checkpoint(sbi, &cpc);
2178 spin_lock(&sbi->stat_lock);
2179 sbi->unusable_block_count = unusable;
2180 spin_unlock(&sbi->stat_lock);
2183 f2fs_up_write(&sbi->gc_lock);
2185 sbi->gc_mode = gc_mode;
2186 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2190 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2192 int retry = DEFAULT_RETRY_IO_COUNT;
2194 /* we should flush all the data to keep data consistency */
2196 sync_inodes_sb(sbi->sb);
2197 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2198 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2200 if (unlikely(retry < 0))
2201 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2203 f2fs_down_write(&sbi->gc_lock);
2204 f2fs_dirty_to_prefree(sbi);
2206 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2207 set_sbi_flag(sbi, SBI_IS_DIRTY);
2208 f2fs_up_write(&sbi->gc_lock);
2210 f2fs_sync_fs(sbi->sb, 1);
2212 /* Let's ensure there's no pending checkpoint anymore */
2213 f2fs_flush_ckpt_thread(sbi);
2216 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2218 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2219 struct f2fs_mount_info org_mount_opt;
2220 unsigned long old_sb_flags;
2222 bool need_restart_gc = false, need_stop_gc = false;
2223 bool need_restart_ckpt = false, need_stop_ckpt = false;
2224 bool need_restart_flush = false, need_stop_flush = false;
2225 bool need_restart_discard = false, need_stop_discard = false;
2226 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2227 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2228 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2229 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2230 bool no_atgc = !test_opt(sbi, ATGC);
2231 bool no_discard = !test_opt(sbi, DISCARD);
2232 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2233 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2239 * Save the old mount options in case we
2240 * need to restore them.
2242 org_mount_opt = sbi->mount_opt;
2243 old_sb_flags = sb->s_flags;
2246 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2247 for (i = 0; i < MAXQUOTAS; i++) {
2248 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2249 org_mount_opt.s_qf_names[i] =
2250 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2252 if (!org_mount_opt.s_qf_names[i]) {
2253 for (j = 0; j < i; j++)
2254 kfree(org_mount_opt.s_qf_names[j]);
2258 org_mount_opt.s_qf_names[i] = NULL;
2263 /* recover superblocks we couldn't write due to previous RO mount */
2264 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2265 err = f2fs_commit_super(sbi, false);
2266 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2269 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2272 default_options(sbi);
2274 /* parse mount options */
2275 err = parse_options(sb, data, true);
2280 * Previous and new state of filesystem is RO,
2281 * so skip checking GC and FLUSH_MERGE conditions.
2283 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2286 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2292 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2293 err = dquot_suspend(sb, -1);
2296 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2297 /* dquot_resume needs RW */
2298 sb->s_flags &= ~SB_RDONLY;
2299 if (sb_any_quota_suspended(sb)) {
2300 dquot_resume(sb, -1);
2301 } else if (f2fs_sb_has_quota_ino(sbi)) {
2302 err = f2fs_enable_quotas(sb);
2308 /* disallow enable atgc dynamically */
2309 if (no_atgc == !!test_opt(sbi, ATGC)) {
2311 f2fs_warn(sbi, "switch atgc option is not allowed");
2315 /* disallow enable/disable extent_cache dynamically */
2316 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2318 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2321 /* disallow enable/disable age extent_cache dynamically */
2322 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2324 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2328 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2330 f2fs_warn(sbi, "switch io_bits option is not allowed");
2334 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2336 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2340 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2342 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2346 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2348 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2353 * We stop the GC thread if FS is mounted as RO
2354 * or if background_gc = off is passed in mount
2355 * option. Also sync the filesystem.
2357 if ((*flags & SB_RDONLY) ||
2358 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2359 !test_opt(sbi, GC_MERGE))) {
2360 if (sbi->gc_thread) {
2361 f2fs_stop_gc_thread(sbi);
2362 need_restart_gc = true;
2364 } else if (!sbi->gc_thread) {
2365 err = f2fs_start_gc_thread(sbi);
2368 need_stop_gc = true;
2371 if (*flags & SB_RDONLY) {
2374 set_sbi_flag(sbi, SBI_IS_DIRTY);
2375 set_sbi_flag(sbi, SBI_IS_CLOSE);
2376 f2fs_sync_fs(sb, 1);
2377 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2380 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2381 !test_opt(sbi, MERGE_CHECKPOINT)) {
2382 f2fs_stop_ckpt_thread(sbi);
2383 need_restart_ckpt = true;
2385 /* Flush if the prevous checkpoint, if exists. */
2386 f2fs_flush_ckpt_thread(sbi);
2388 err = f2fs_start_ckpt_thread(sbi);
2391 "Failed to start F2FS issue_checkpoint_thread (%d)",
2395 need_stop_ckpt = true;
2399 * We stop issue flush thread if FS is mounted as RO
2400 * or if flush_merge is not passed in mount option.
2402 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2403 clear_opt(sbi, FLUSH_MERGE);
2404 f2fs_destroy_flush_cmd_control(sbi, false);
2405 need_restart_flush = true;
2407 err = f2fs_create_flush_cmd_control(sbi);
2410 need_stop_flush = true;
2413 if (no_discard == !!test_opt(sbi, DISCARD)) {
2414 if (test_opt(sbi, DISCARD)) {
2415 err = f2fs_start_discard_thread(sbi);
2418 need_stop_discard = true;
2420 f2fs_stop_discard_thread(sbi);
2421 f2fs_issue_discard_timeout(sbi);
2422 need_restart_discard = true;
2426 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2427 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2428 err = f2fs_disable_checkpoint(sbi);
2430 goto restore_discard;
2432 f2fs_enable_checkpoint(sbi);
2438 /* Release old quota file names */
2439 for (i = 0; i < MAXQUOTAS; i++)
2440 kfree(org_mount_opt.s_qf_names[i]);
2442 /* Update the POSIXACL Flag */
2443 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2444 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2446 limit_reserve_root(sbi);
2447 adjust_unusable_cap_perc(sbi);
2448 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2451 if (need_restart_discard) {
2452 if (f2fs_start_discard_thread(sbi))
2453 f2fs_warn(sbi, "discard has been stopped");
2454 } else if (need_stop_discard) {
2455 f2fs_stop_discard_thread(sbi);
2458 if (need_restart_flush) {
2459 if (f2fs_create_flush_cmd_control(sbi))
2460 f2fs_warn(sbi, "background flush thread has stopped");
2461 } else if (need_stop_flush) {
2462 clear_opt(sbi, FLUSH_MERGE);
2463 f2fs_destroy_flush_cmd_control(sbi, false);
2466 if (need_restart_ckpt) {
2467 if (f2fs_start_ckpt_thread(sbi))
2468 f2fs_warn(sbi, "background ckpt thread has stopped");
2469 } else if (need_stop_ckpt) {
2470 f2fs_stop_ckpt_thread(sbi);
2473 if (need_restart_gc) {
2474 if (f2fs_start_gc_thread(sbi))
2475 f2fs_warn(sbi, "background gc thread has stopped");
2476 } else if (need_stop_gc) {
2477 f2fs_stop_gc_thread(sbi);
2481 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2482 for (i = 0; i < MAXQUOTAS; i++) {
2483 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2484 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2487 sbi->mount_opt = org_mount_opt;
2488 sb->s_flags = old_sb_flags;
2493 /* Read data from quotafile */
2494 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2495 size_t len, loff_t off)
2497 struct inode *inode = sb_dqopt(sb)->files[type];
2498 struct address_space *mapping = inode->i_mapping;
2499 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2500 int offset = off & (sb->s_blocksize - 1);
2503 loff_t i_size = i_size_read(inode);
2509 if (off + len > i_size)
2512 while (toread > 0) {
2513 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2515 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2517 if (PTR_ERR(page) == -ENOMEM) {
2518 memalloc_retry_wait(GFP_NOFS);
2521 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2522 return PTR_ERR(page);
2527 if (unlikely(page->mapping != mapping)) {
2528 f2fs_put_page(page, 1);
2531 if (unlikely(!PageUptodate(page))) {
2532 f2fs_put_page(page, 1);
2533 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2537 memcpy_from_page(data, page, offset, tocopy);
2538 f2fs_put_page(page, 1);
2548 /* Write to quotafile */
2549 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2550 const char *data, size_t len, loff_t off)
2552 struct inode *inode = sb_dqopt(sb)->files[type];
2553 struct address_space *mapping = inode->i_mapping;
2554 const struct address_space_operations *a_ops = mapping->a_ops;
2555 int offset = off & (sb->s_blocksize - 1);
2556 size_t towrite = len;
2558 void *fsdata = NULL;
2562 while (towrite > 0) {
2563 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2566 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2568 if (unlikely(err)) {
2569 if (err == -ENOMEM) {
2570 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2573 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2577 memcpy_to_page(page, offset, data, tocopy);
2579 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2590 inode->i_mtime = inode->i_ctime = current_time(inode);
2591 f2fs_mark_inode_dirty_sync(inode, false);
2592 return len - towrite;
2595 int f2fs_dquot_initialize(struct inode *inode)
2597 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) {
2598 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT);
2602 return dquot_initialize(inode);
2605 static struct dquot **f2fs_get_dquots(struct inode *inode)
2607 return F2FS_I(inode)->i_dquot;
2610 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2612 return &F2FS_I(inode)->i_reserved_quota;
2615 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2617 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2618 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2622 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2623 F2FS_OPTION(sbi).s_jquota_fmt, type);
2626 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2631 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2632 err = f2fs_enable_quotas(sbi->sb);
2634 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2640 for (i = 0; i < MAXQUOTAS; i++) {
2641 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2642 err = f2fs_quota_on_mount(sbi, i);
2647 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2654 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2657 struct inode *qf_inode;
2658 unsigned long qf_inum;
2661 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2663 qf_inum = f2fs_qf_ino(sb, type);
2667 qf_inode = f2fs_iget(sb, qf_inum);
2668 if (IS_ERR(qf_inode)) {
2669 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2670 return PTR_ERR(qf_inode);
2673 /* Don't account quota for quota files to avoid recursion */
2674 qf_inode->i_flags |= S_NOQUOTA;
2675 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2680 static int f2fs_enable_quotas(struct super_block *sb)
2682 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2684 unsigned long qf_inum;
2685 bool quota_mopt[MAXQUOTAS] = {
2686 test_opt(sbi, USRQUOTA),
2687 test_opt(sbi, GRPQUOTA),
2688 test_opt(sbi, PRJQUOTA),
2691 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2692 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2696 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2698 for (type = 0; type < MAXQUOTAS; type++) {
2699 qf_inum = f2fs_qf_ino(sb, type);
2701 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2702 DQUOT_USAGE_ENABLED |
2703 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2705 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2707 for (type--; type >= 0; type--)
2708 dquot_quota_off(sb, type);
2709 set_sbi_flag(F2FS_SB(sb),
2710 SBI_QUOTA_NEED_REPAIR);
2718 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2720 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2721 struct address_space *mapping = dqopt->files[type]->i_mapping;
2724 ret = dquot_writeback_dquots(sbi->sb, type);
2728 ret = filemap_fdatawrite(mapping);
2732 /* if we are using journalled quota */
2733 if (is_journalled_quota(sbi))
2736 ret = filemap_fdatawait(mapping);
2738 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2741 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2745 int f2fs_quota_sync(struct super_block *sb, int type)
2747 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2748 struct quota_info *dqopt = sb_dqopt(sb);
2753 * Now when everything is written we can discard the pagecache so
2754 * that userspace sees the changes.
2756 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2758 if (type != -1 && cnt != type)
2761 if (!sb_has_quota_active(sb, cnt))
2764 if (!f2fs_sb_has_quota_ino(sbi))
2765 inode_lock(dqopt->files[cnt]);
2770 * f2fs_down_read(quota_sem)
2771 * dquot_writeback_dquots()
2774 * f2fs_down_read(quota_sem)
2777 f2fs_down_read(&sbi->quota_sem);
2779 ret = f2fs_quota_sync_file(sbi, cnt);
2781 f2fs_up_read(&sbi->quota_sem);
2782 f2fs_unlock_op(sbi);
2784 if (!f2fs_sb_has_quota_ino(sbi))
2785 inode_unlock(dqopt->files[cnt]);
2793 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2794 const struct path *path)
2796 struct inode *inode;
2799 /* if quota sysfile exists, deny enabling quota with specific file */
2800 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2801 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2805 err = f2fs_quota_sync(sb, type);
2809 err = dquot_quota_on(sb, type, format_id, path);
2813 inode = d_inode(path->dentry);
2816 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2817 f2fs_set_inode_flags(inode);
2818 inode_unlock(inode);
2819 f2fs_mark_inode_dirty_sync(inode, false);
2824 static int __f2fs_quota_off(struct super_block *sb, int type)
2826 struct inode *inode = sb_dqopt(sb)->files[type];
2829 if (!inode || !igrab(inode))
2830 return dquot_quota_off(sb, type);
2832 err = f2fs_quota_sync(sb, type);
2836 err = dquot_quota_off(sb, type);
2837 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2841 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2842 f2fs_set_inode_flags(inode);
2843 inode_unlock(inode);
2844 f2fs_mark_inode_dirty_sync(inode, false);
2850 static int f2fs_quota_off(struct super_block *sb, int type)
2852 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2855 err = __f2fs_quota_off(sb, type);
2858 * quotactl can shutdown journalled quota, result in inconsistence
2859 * between quota record and fs data by following updates, tag the
2860 * flag to let fsck be aware of it.
2862 if (is_journalled_quota(sbi))
2863 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2867 void f2fs_quota_off_umount(struct super_block *sb)
2872 for (type = 0; type < MAXQUOTAS; type++) {
2873 err = __f2fs_quota_off(sb, type);
2875 int ret = dquot_quota_off(sb, type);
2877 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2879 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2883 * In case of checkpoint=disable, we must flush quota blocks.
2884 * This can cause NULL exception for node_inode in end_io, since
2885 * put_super already dropped it.
2887 sync_filesystem(sb);
2890 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2892 struct quota_info *dqopt = sb_dqopt(sb);
2895 for (type = 0; type < MAXQUOTAS; type++) {
2896 if (!dqopt->files[type])
2898 f2fs_inode_synced(dqopt->files[type]);
2902 static int f2fs_dquot_commit(struct dquot *dquot)
2904 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2907 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2908 ret = dquot_commit(dquot);
2910 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2911 f2fs_up_read(&sbi->quota_sem);
2915 static int f2fs_dquot_acquire(struct dquot *dquot)
2917 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2920 f2fs_down_read(&sbi->quota_sem);
2921 ret = dquot_acquire(dquot);
2923 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2924 f2fs_up_read(&sbi->quota_sem);
2928 static int f2fs_dquot_release(struct dquot *dquot)
2930 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2931 int ret = dquot_release(dquot);
2934 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2938 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2940 struct super_block *sb = dquot->dq_sb;
2941 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2942 int ret = dquot_mark_dquot_dirty(dquot);
2944 /* if we are using journalled quota */
2945 if (is_journalled_quota(sbi))
2946 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2951 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2953 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2954 int ret = dquot_commit_info(sb, type);
2957 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2961 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2963 *projid = F2FS_I(inode)->i_projid;
2967 static const struct dquot_operations f2fs_quota_operations = {
2968 .get_reserved_space = f2fs_get_reserved_space,
2969 .write_dquot = f2fs_dquot_commit,
2970 .acquire_dquot = f2fs_dquot_acquire,
2971 .release_dquot = f2fs_dquot_release,
2972 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2973 .write_info = f2fs_dquot_commit_info,
2974 .alloc_dquot = dquot_alloc,
2975 .destroy_dquot = dquot_destroy,
2976 .get_projid = f2fs_get_projid,
2977 .get_next_id = dquot_get_next_id,
2980 static const struct quotactl_ops f2fs_quotactl_ops = {
2981 .quota_on = f2fs_quota_on,
2982 .quota_off = f2fs_quota_off,
2983 .quota_sync = f2fs_quota_sync,
2984 .get_state = dquot_get_state,
2985 .set_info = dquot_set_dqinfo,
2986 .get_dqblk = dquot_get_dqblk,
2987 .set_dqblk = dquot_set_dqblk,
2988 .get_nextdqblk = dquot_get_next_dqblk,
2991 int f2fs_dquot_initialize(struct inode *inode)
2996 int f2fs_quota_sync(struct super_block *sb, int type)
3001 void f2fs_quota_off_umount(struct super_block *sb)
3006 static const struct super_operations f2fs_sops = {
3007 .alloc_inode = f2fs_alloc_inode,
3008 .free_inode = f2fs_free_inode,
3009 .drop_inode = f2fs_drop_inode,
3010 .write_inode = f2fs_write_inode,
3011 .dirty_inode = f2fs_dirty_inode,
3012 .show_options = f2fs_show_options,
3014 .quota_read = f2fs_quota_read,
3015 .quota_write = f2fs_quota_write,
3016 .get_dquots = f2fs_get_dquots,
3018 .evict_inode = f2fs_evict_inode,
3019 .put_super = f2fs_put_super,
3020 .sync_fs = f2fs_sync_fs,
3021 .freeze_fs = f2fs_freeze,
3022 .unfreeze_fs = f2fs_unfreeze,
3023 .statfs = f2fs_statfs,
3024 .remount_fs = f2fs_remount,
3027 #ifdef CONFIG_FS_ENCRYPTION
3028 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3030 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3031 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3035 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3038 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3041 * Encrypting the root directory is not allowed because fsck
3042 * expects lost+found directory to exist and remain unencrypted
3043 * if LOST_FOUND feature is enabled.
3046 if (f2fs_sb_has_lost_found(sbi) &&
3047 inode->i_ino == F2FS_ROOT_INO(sbi))
3050 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3051 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3052 ctx, len, fs_data, XATTR_CREATE);
3055 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3057 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3060 static bool f2fs_has_stable_inodes(struct super_block *sb)
3065 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3066 int *ino_bits_ret, int *lblk_bits_ret)
3068 *ino_bits_ret = 8 * sizeof(nid_t);
3069 *lblk_bits_ret = 8 * sizeof(block_t);
3072 static struct block_device **f2fs_get_devices(struct super_block *sb,
3073 unsigned int *num_devs)
3075 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3076 struct block_device **devs;
3079 if (!f2fs_is_multi_device(sbi))
3082 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3084 return ERR_PTR(-ENOMEM);
3086 for (i = 0; i < sbi->s_ndevs; i++)
3087 devs[i] = FDEV(i).bdev;
3088 *num_devs = sbi->s_ndevs;
3092 static const struct fscrypt_operations f2fs_cryptops = {
3093 .key_prefix = "f2fs:",
3094 .get_context = f2fs_get_context,
3095 .set_context = f2fs_set_context,
3096 .get_dummy_policy = f2fs_get_dummy_policy,
3097 .empty_dir = f2fs_empty_dir,
3098 .has_stable_inodes = f2fs_has_stable_inodes,
3099 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
3100 .get_devices = f2fs_get_devices,
3104 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3105 u64 ino, u32 generation)
3107 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3108 struct inode *inode;
3110 if (f2fs_check_nid_range(sbi, ino))
3111 return ERR_PTR(-ESTALE);
3114 * f2fs_iget isn't quite right if the inode is currently unallocated!
3115 * However f2fs_iget currently does appropriate checks to handle stale
3116 * inodes so everything is OK.
3118 inode = f2fs_iget(sb, ino);
3120 return ERR_CAST(inode);
3121 if (unlikely(generation && inode->i_generation != generation)) {
3122 /* we didn't find the right inode.. */
3124 return ERR_PTR(-ESTALE);
3129 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3130 int fh_len, int fh_type)
3132 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3133 f2fs_nfs_get_inode);
3136 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3137 int fh_len, int fh_type)
3139 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3140 f2fs_nfs_get_inode);
3143 static const struct export_operations f2fs_export_ops = {
3144 .fh_to_dentry = f2fs_fh_to_dentry,
3145 .fh_to_parent = f2fs_fh_to_parent,
3146 .get_parent = f2fs_get_parent,
3149 loff_t max_file_blocks(struct inode *inode)
3155 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3156 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3157 * space in inode.i_addr, it will be more safe to reassign
3161 if (inode && f2fs_compressed_file(inode))
3162 leaf_count = ADDRS_PER_BLOCK(inode);
3164 leaf_count = DEF_ADDRS_PER_BLOCK;
3166 /* two direct node blocks */
3167 result += (leaf_count * 2);
3169 /* two indirect node blocks */
3170 leaf_count *= NIDS_PER_BLOCK;
3171 result += (leaf_count * 2);
3173 /* one double indirect node block */
3174 leaf_count *= NIDS_PER_BLOCK;
3175 result += leaf_count;
3180 static int __f2fs_commit_super(struct buffer_head *bh,
3181 struct f2fs_super_block *super)
3185 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3186 set_buffer_dirty(bh);
3189 /* it's rare case, we can do fua all the time */
3190 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3193 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3194 struct buffer_head *bh)
3196 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3197 (bh->b_data + F2FS_SUPER_OFFSET);
3198 struct super_block *sb = sbi->sb;
3199 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3200 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3201 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3202 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3203 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3204 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3205 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3206 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3207 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3208 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3209 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3210 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3211 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3212 u64 main_end_blkaddr = main_blkaddr +
3213 (segment_count_main << log_blocks_per_seg);
3214 u64 seg_end_blkaddr = segment0_blkaddr +
3215 (segment_count << log_blocks_per_seg);
3217 if (segment0_blkaddr != cp_blkaddr) {
3218 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3219 segment0_blkaddr, cp_blkaddr);
3223 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3225 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3226 cp_blkaddr, sit_blkaddr,
3227 segment_count_ckpt << log_blocks_per_seg);
3231 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3233 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3234 sit_blkaddr, nat_blkaddr,
3235 segment_count_sit << log_blocks_per_seg);
3239 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3241 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3242 nat_blkaddr, ssa_blkaddr,
3243 segment_count_nat << log_blocks_per_seg);
3247 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3249 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3250 ssa_blkaddr, main_blkaddr,
3251 segment_count_ssa << log_blocks_per_seg);
3255 if (main_end_blkaddr > seg_end_blkaddr) {
3256 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3257 main_blkaddr, seg_end_blkaddr,
3258 segment_count_main << log_blocks_per_seg);
3260 } else if (main_end_blkaddr < seg_end_blkaddr) {
3264 /* fix in-memory information all the time */
3265 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3266 segment0_blkaddr) >> log_blocks_per_seg);
3268 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3269 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3272 err = __f2fs_commit_super(bh, NULL);
3273 res = err ? "failed" : "done";
3275 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3276 res, main_blkaddr, seg_end_blkaddr,
3277 segment_count_main << log_blocks_per_seg);
3284 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3285 struct buffer_head *bh)
3287 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3288 block_t total_sections, blocks_per_seg;
3289 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3290 (bh->b_data + F2FS_SUPER_OFFSET);
3291 size_t crc_offset = 0;
3294 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3295 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3296 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3300 /* Check checksum_offset and crc in superblock */
3301 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3302 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3304 offsetof(struct f2fs_super_block, crc)) {
3305 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3307 return -EFSCORRUPTED;
3309 crc = le32_to_cpu(raw_super->crc);
3310 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3311 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3312 return -EFSCORRUPTED;
3316 /* Currently, support only 4KB block size */
3317 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3318 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3319 le32_to_cpu(raw_super->log_blocksize),
3321 return -EFSCORRUPTED;
3324 /* check log blocks per segment */
3325 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3326 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3327 le32_to_cpu(raw_super->log_blocks_per_seg));
3328 return -EFSCORRUPTED;
3331 /* Currently, support 512/1024/2048/4096 bytes sector size */
3332 if (le32_to_cpu(raw_super->log_sectorsize) >
3333 F2FS_MAX_LOG_SECTOR_SIZE ||
3334 le32_to_cpu(raw_super->log_sectorsize) <
3335 F2FS_MIN_LOG_SECTOR_SIZE) {
3336 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3337 le32_to_cpu(raw_super->log_sectorsize));
3338 return -EFSCORRUPTED;
3340 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3341 le32_to_cpu(raw_super->log_sectorsize) !=
3342 F2FS_MAX_LOG_SECTOR_SIZE) {
3343 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3344 le32_to_cpu(raw_super->log_sectors_per_block),
3345 le32_to_cpu(raw_super->log_sectorsize));
3346 return -EFSCORRUPTED;
3349 segment_count = le32_to_cpu(raw_super->segment_count);
3350 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3351 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3352 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3353 total_sections = le32_to_cpu(raw_super->section_count);
3355 /* blocks_per_seg should be 512, given the above check */
3356 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3358 if (segment_count > F2FS_MAX_SEGMENT ||
3359 segment_count < F2FS_MIN_SEGMENTS) {
3360 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3361 return -EFSCORRUPTED;
3364 if (total_sections > segment_count_main || total_sections < 1 ||
3365 segs_per_sec > segment_count || !segs_per_sec) {
3366 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3367 segment_count, total_sections, segs_per_sec);
3368 return -EFSCORRUPTED;
3371 if (segment_count_main != total_sections * segs_per_sec) {
3372 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3373 segment_count_main, total_sections, segs_per_sec);
3374 return -EFSCORRUPTED;
3377 if ((segment_count / segs_per_sec) < total_sections) {
3378 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3379 segment_count, segs_per_sec, total_sections);
3380 return -EFSCORRUPTED;
3383 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3384 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3385 segment_count, le64_to_cpu(raw_super->block_count));
3386 return -EFSCORRUPTED;
3389 if (RDEV(0).path[0]) {
3390 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3393 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3394 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3397 if (segment_count != dev_seg_count) {
3398 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3399 segment_count, dev_seg_count);
3400 return -EFSCORRUPTED;
3403 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3404 !bdev_is_zoned(sbi->sb->s_bdev)) {
3405 f2fs_info(sbi, "Zoned block device path is missing");
3406 return -EFSCORRUPTED;
3410 if (secs_per_zone > total_sections || !secs_per_zone) {
3411 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3412 secs_per_zone, total_sections);
3413 return -EFSCORRUPTED;
3415 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3416 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3417 (le32_to_cpu(raw_super->extension_count) +
3418 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3419 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3420 le32_to_cpu(raw_super->extension_count),
3421 raw_super->hot_ext_count,
3422 F2FS_MAX_EXTENSION);
3423 return -EFSCORRUPTED;
3426 if (le32_to_cpu(raw_super->cp_payload) >=
3427 (blocks_per_seg - F2FS_CP_PACKS -
3428 NR_CURSEG_PERSIST_TYPE)) {
3429 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3430 le32_to_cpu(raw_super->cp_payload),
3431 blocks_per_seg - F2FS_CP_PACKS -
3432 NR_CURSEG_PERSIST_TYPE);
3433 return -EFSCORRUPTED;
3436 /* check reserved ino info */
3437 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3438 le32_to_cpu(raw_super->meta_ino) != 2 ||
3439 le32_to_cpu(raw_super->root_ino) != 3) {
3440 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3441 le32_to_cpu(raw_super->node_ino),
3442 le32_to_cpu(raw_super->meta_ino),
3443 le32_to_cpu(raw_super->root_ino));
3444 return -EFSCORRUPTED;
3447 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3448 if (sanity_check_area_boundary(sbi, bh))
3449 return -EFSCORRUPTED;
3454 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3456 unsigned int total, fsmeta;
3457 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3458 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3459 unsigned int ovp_segments, reserved_segments;
3460 unsigned int main_segs, blocks_per_seg;
3461 unsigned int sit_segs, nat_segs;
3462 unsigned int sit_bitmap_size, nat_bitmap_size;
3463 unsigned int log_blocks_per_seg;
3464 unsigned int segment_count_main;
3465 unsigned int cp_pack_start_sum, cp_payload;
3466 block_t user_block_count, valid_user_blocks;
3467 block_t avail_node_count, valid_node_count;
3468 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3471 total = le32_to_cpu(raw_super->segment_count);
3472 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3473 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3475 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3477 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3478 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3480 if (unlikely(fsmeta >= total))
3483 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3484 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3486 if (!f2fs_sb_has_readonly(sbi) &&
3487 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3488 ovp_segments == 0 || reserved_segments == 0)) {
3489 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3492 user_block_count = le64_to_cpu(ckpt->user_block_count);
3493 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3494 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3495 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3496 if (!user_block_count || user_block_count >=
3497 segment_count_main << log_blocks_per_seg) {
3498 f2fs_err(sbi, "Wrong user_block_count: %u",
3503 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3504 if (valid_user_blocks > user_block_count) {
3505 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3506 valid_user_blocks, user_block_count);
3510 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3511 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3512 if (valid_node_count > avail_node_count) {
3513 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3514 valid_node_count, avail_node_count);
3518 main_segs = le32_to_cpu(raw_super->segment_count_main);
3519 blocks_per_seg = sbi->blocks_per_seg;
3521 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3522 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3523 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3526 if (f2fs_sb_has_readonly(sbi))
3529 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3530 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3531 le32_to_cpu(ckpt->cur_node_segno[j])) {
3532 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3534 le32_to_cpu(ckpt->cur_node_segno[i]));
3540 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3541 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3542 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3545 if (f2fs_sb_has_readonly(sbi))
3548 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3549 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3550 le32_to_cpu(ckpt->cur_data_segno[j])) {
3551 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3553 le32_to_cpu(ckpt->cur_data_segno[i]));
3558 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3559 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3560 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3561 le32_to_cpu(ckpt->cur_data_segno[j])) {
3562 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3564 le32_to_cpu(ckpt->cur_node_segno[i]));
3570 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3571 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3573 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3574 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3575 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3576 sit_bitmap_size, nat_bitmap_size);
3580 cp_pack_start_sum = __start_sum_addr(sbi);
3581 cp_payload = __cp_payload(sbi);
3582 if (cp_pack_start_sum < cp_payload + 1 ||
3583 cp_pack_start_sum > blocks_per_seg - 1 -
3584 NR_CURSEG_PERSIST_TYPE) {
3585 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3590 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3591 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3592 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3593 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3594 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3595 le32_to_cpu(ckpt->checksum_offset));
3599 nat_blocks = nat_segs << log_blocks_per_seg;
3600 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3601 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3602 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3603 (cp_payload + F2FS_CP_PACKS +
3604 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3605 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3606 cp_payload, nat_bits_blocks);
3610 if (unlikely(f2fs_cp_error(sbi))) {
3611 f2fs_err(sbi, "A bug case: need to run fsck");
3617 static void init_sb_info(struct f2fs_sb_info *sbi)
3619 struct f2fs_super_block *raw_super = sbi->raw_super;
3622 sbi->log_sectors_per_block =
3623 le32_to_cpu(raw_super->log_sectors_per_block);
3624 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3625 sbi->blocksize = 1 << sbi->log_blocksize;
3626 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3627 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3628 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3629 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3630 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3631 sbi->total_node_count =
3632 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3633 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3634 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3635 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3636 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3637 sbi->cur_victim_sec = NULL_SECNO;
3638 sbi->gc_mode = GC_NORMAL;
3639 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3640 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3641 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3642 sbi->migration_granularity = sbi->segs_per_sec;
3643 sbi->seq_file_ra_mul = MIN_RA_MUL;
3644 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3645 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3646 spin_lock_init(&sbi->gc_remaining_trials_lock);
3647 atomic64_set(&sbi->current_atomic_write, 0);
3649 sbi->dir_level = DEF_DIR_LEVEL;
3650 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3651 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3652 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3653 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3654 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3655 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3656 DEF_UMOUNT_DISCARD_TIMEOUT;
3657 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3659 for (i = 0; i < NR_COUNT_TYPE; i++)
3660 atomic_set(&sbi->nr_pages[i], 0);
3662 for (i = 0; i < META; i++)
3663 atomic_set(&sbi->wb_sync_req[i], 0);
3665 INIT_LIST_HEAD(&sbi->s_list);
3666 mutex_init(&sbi->umount_mutex);
3667 init_f2fs_rwsem(&sbi->io_order_lock);
3668 spin_lock_init(&sbi->cp_lock);
3670 sbi->dirty_device = 0;
3671 spin_lock_init(&sbi->dev_lock);
3673 init_f2fs_rwsem(&sbi->sb_lock);
3674 init_f2fs_rwsem(&sbi->pin_sem);
3677 static int init_percpu_info(struct f2fs_sb_info *sbi)
3681 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3685 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3687 goto err_valid_block;
3689 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3692 goto err_node_block;
3696 percpu_counter_destroy(&sbi->rf_node_block_count);
3698 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3702 #ifdef CONFIG_BLK_DEV_ZONED
3704 struct f2fs_report_zones_args {
3705 struct f2fs_sb_info *sbi;
3706 struct f2fs_dev_info *dev;
3709 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3712 struct f2fs_report_zones_args *rz_args = data;
3713 block_t unusable_blocks = (zone->len - zone->capacity) >>
3714 F2FS_LOG_SECTORS_PER_BLOCK;
3716 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3719 set_bit(idx, rz_args->dev->blkz_seq);
3720 if (!rz_args->sbi->unusable_blocks_per_sec) {
3721 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3724 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3725 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3731 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3733 struct block_device *bdev = FDEV(devi).bdev;
3734 sector_t nr_sectors = bdev_nr_sectors(bdev);
3735 struct f2fs_report_zones_args rep_zone_arg;
3739 if (!f2fs_sb_has_blkzoned(sbi))
3742 zone_sectors = bdev_zone_sectors(bdev);
3743 if (!is_power_of_2(zone_sectors)) {
3744 f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n");
3748 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3749 SECTOR_TO_BLOCK(zone_sectors))
3751 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3752 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3753 __ilog2_u32(sbi->blocks_per_blkz))
3755 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3756 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3757 sbi->log_blocks_per_blkz;
3758 if (nr_sectors & (zone_sectors - 1))
3759 FDEV(devi).nr_blkz++;
3761 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3762 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3763 * sizeof(unsigned long),
3765 if (!FDEV(devi).blkz_seq)
3768 rep_zone_arg.sbi = sbi;
3769 rep_zone_arg.dev = &FDEV(devi);
3771 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3780 * Read f2fs raw super block.
3781 * Because we have two copies of super block, so read both of them
3782 * to get the first valid one. If any one of them is broken, we pass
3783 * them recovery flag back to the caller.
3785 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3786 struct f2fs_super_block **raw_super,
3787 int *valid_super_block, int *recovery)
3789 struct super_block *sb = sbi->sb;
3791 struct buffer_head *bh;
3792 struct f2fs_super_block *super;
3795 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3799 for (block = 0; block < 2; block++) {
3800 bh = sb_bread(sb, block);
3802 f2fs_err(sbi, "Unable to read %dth superblock",
3809 /* sanity checking of raw super */
3810 err = sanity_check_raw_super(sbi, bh);
3812 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3820 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3822 *valid_super_block = block;
3828 /* No valid superblock */
3837 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3839 struct buffer_head *bh;
3843 if ((recover && f2fs_readonly(sbi->sb)) ||
3844 bdev_read_only(sbi->sb->s_bdev)) {
3845 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3849 /* we should update superblock crc here */
3850 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3851 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3852 offsetof(struct f2fs_super_block, crc));
3853 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3856 /* write back-up superblock first */
3857 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3860 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3863 /* if we are in recovery path, skip writing valid superblock */
3867 /* write current valid superblock */
3868 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3871 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3876 void f2fs_handle_stop(struct f2fs_sb_info *sbi, unsigned char reason)
3878 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3881 f2fs_down_write(&sbi->sb_lock);
3883 if (raw_super->s_stop_reason[reason] < ((1 << BITS_PER_BYTE) - 1))
3884 raw_super->s_stop_reason[reason]++;
3886 err = f2fs_commit_super(sbi, false);
3888 f2fs_err(sbi, "f2fs_commit_super fails to record reason:%u err:%d",
3890 f2fs_up_write(&sbi->sb_lock);
3893 static void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
3895 spin_lock(&sbi->error_lock);
3896 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
3897 set_bit(flag, (unsigned long *)sbi->errors);
3898 sbi->error_dirty = true;
3900 spin_unlock(&sbi->error_lock);
3903 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
3905 bool need_update = false;
3907 spin_lock(&sbi->error_lock);
3908 if (sbi->error_dirty) {
3909 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
3911 sbi->error_dirty = false;
3914 spin_unlock(&sbi->error_lock);
3919 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
3923 f2fs_save_errors(sbi, error);
3925 f2fs_down_write(&sbi->sb_lock);
3927 if (!f2fs_update_errors(sbi))
3930 err = f2fs_commit_super(sbi, false);
3932 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
3935 f2fs_up_write(&sbi->sb_lock);
3938 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3940 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3941 unsigned int max_devices = MAX_DEVICES;
3942 unsigned int logical_blksize;
3945 /* Initialize single device information */
3946 if (!RDEV(0).path[0]) {
3947 if (!bdev_is_zoned(sbi->sb->s_bdev))
3953 * Initialize multiple devices information, or single
3954 * zoned block device information.
3956 sbi->devs = f2fs_kzalloc(sbi,
3957 array_size(max_devices,
3958 sizeof(struct f2fs_dev_info)),
3963 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3964 sbi->aligned_blksize = true;
3966 for (i = 0; i < max_devices; i++) {
3968 if (i > 0 && !RDEV(i).path[0])
3971 if (max_devices == 1) {
3972 /* Single zoned block device mount */
3974 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3975 sbi->sb->s_mode, sbi->sb->s_type);
3977 /* Multi-device mount */
3978 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3979 FDEV(i).total_segments =
3980 le32_to_cpu(RDEV(i).total_segments);
3982 FDEV(i).start_blk = 0;
3983 FDEV(i).end_blk = FDEV(i).start_blk +
3984 (FDEV(i).total_segments <<
3985 sbi->log_blocks_per_seg) - 1 +
3986 le32_to_cpu(raw_super->segment0_blkaddr);
3988 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3989 FDEV(i).end_blk = FDEV(i).start_blk +
3990 (FDEV(i).total_segments <<
3991 sbi->log_blocks_per_seg) - 1;
3993 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3994 sbi->sb->s_mode, sbi->sb->s_type);
3996 if (IS_ERR(FDEV(i).bdev))
3997 return PTR_ERR(FDEV(i).bdev);
3999 /* to release errored devices */
4000 sbi->s_ndevs = i + 1;
4002 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4003 sbi->aligned_blksize = false;
4005 #ifdef CONFIG_BLK_DEV_ZONED
4006 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
4007 !f2fs_sb_has_blkzoned(sbi)) {
4008 f2fs_err(sbi, "Zoned block device feature not enabled");
4011 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
4012 if (init_blkz_info(sbi, i)) {
4013 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4016 if (max_devices == 1)
4018 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
4020 FDEV(i).total_segments,
4021 FDEV(i).start_blk, FDEV(i).end_blk,
4022 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
4023 "Host-aware" : "Host-managed");
4027 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4029 FDEV(i).total_segments,
4030 FDEV(i).start_blk, FDEV(i).end_blk);
4033 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
4037 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4039 #if IS_ENABLED(CONFIG_UNICODE)
4040 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4041 const struct f2fs_sb_encodings *encoding_info;
4042 struct unicode_map *encoding;
4043 __u16 encoding_flags;
4045 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4046 if (!encoding_info) {
4048 "Encoding requested by superblock is unknown");
4052 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4053 encoding = utf8_load(encoding_info->version);
4054 if (IS_ERR(encoding)) {
4056 "can't mount with superblock charset: %s-%u.%u.%u "
4057 "not supported by the kernel. flags: 0x%x.",
4058 encoding_info->name,
4059 unicode_major(encoding_info->version),
4060 unicode_minor(encoding_info->version),
4061 unicode_rev(encoding_info->version),
4063 return PTR_ERR(encoding);
4065 f2fs_info(sbi, "Using encoding defined by superblock: "
4066 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4067 unicode_major(encoding_info->version),
4068 unicode_minor(encoding_info->version),
4069 unicode_rev(encoding_info->version),
4072 sbi->sb->s_encoding = encoding;
4073 sbi->sb->s_encoding_flags = encoding_flags;
4076 if (f2fs_sb_has_casefold(sbi)) {
4077 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4084 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4086 /* adjust parameters according to the volume size */
4087 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4088 if (f2fs_block_unit_discard(sbi))
4089 SM_I(sbi)->dcc_info->discard_granularity =
4090 MIN_DISCARD_GRANULARITY;
4091 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4092 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4095 sbi->readdir_ra = true;
4098 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4100 struct f2fs_sb_info *sbi;
4101 struct f2fs_super_block *raw_super;
4104 bool skip_recovery = false, need_fsck = false;
4105 char *options = NULL;
4106 int recovery, i, valid_super_block;
4107 struct curseg_info *seg_i;
4113 valid_super_block = -1;
4116 /* allocate memory for f2fs-specific super block info */
4117 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4123 /* initialize locks within allocated memory */
4124 init_f2fs_rwsem(&sbi->gc_lock);
4125 mutex_init(&sbi->writepages);
4126 init_f2fs_rwsem(&sbi->cp_global_sem);
4127 init_f2fs_rwsem(&sbi->node_write);
4128 init_f2fs_rwsem(&sbi->node_change);
4129 spin_lock_init(&sbi->stat_lock);
4130 init_f2fs_rwsem(&sbi->cp_rwsem);
4131 init_f2fs_rwsem(&sbi->quota_sem);
4132 init_waitqueue_head(&sbi->cp_wait);
4133 spin_lock_init(&sbi->error_lock);
4135 for (i = 0; i < NR_INODE_TYPE; i++) {
4136 INIT_LIST_HEAD(&sbi->inode_list[i]);
4137 spin_lock_init(&sbi->inode_lock[i]);
4139 mutex_init(&sbi->flush_lock);
4141 /* Load the checksum driver */
4142 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4143 if (IS_ERR(sbi->s_chksum_driver)) {
4144 f2fs_err(sbi, "Cannot load crc32 driver.");
4145 err = PTR_ERR(sbi->s_chksum_driver);
4146 sbi->s_chksum_driver = NULL;
4150 /* set a block size */
4151 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4152 f2fs_err(sbi, "unable to set blocksize");
4156 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4161 sb->s_fs_info = sbi;
4162 sbi->raw_super = raw_super;
4164 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4166 /* precompute checksum seed for metadata */
4167 if (f2fs_sb_has_inode_chksum(sbi))
4168 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4169 sizeof(raw_super->uuid));
4171 default_options(sbi);
4172 /* parse mount options */
4173 options = kstrdup((const char *)data, GFP_KERNEL);
4174 if (data && !options) {
4179 err = parse_options(sb, options, false);
4183 sb->s_maxbytes = max_file_blocks(NULL) <<
4184 le32_to_cpu(raw_super->log_blocksize);
4185 sb->s_max_links = F2FS_LINK_MAX;
4187 err = f2fs_setup_casefold(sbi);
4192 sb->dq_op = &f2fs_quota_operations;
4193 sb->s_qcop = &f2fs_quotactl_ops;
4194 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4196 if (f2fs_sb_has_quota_ino(sbi)) {
4197 for (i = 0; i < MAXQUOTAS; i++) {
4198 if (f2fs_qf_ino(sbi->sb, i))
4199 sbi->nquota_files++;
4204 sb->s_op = &f2fs_sops;
4205 #ifdef CONFIG_FS_ENCRYPTION
4206 sb->s_cop = &f2fs_cryptops;
4208 #ifdef CONFIG_FS_VERITY
4209 sb->s_vop = &f2fs_verityops;
4211 sb->s_xattr = f2fs_xattr_handlers;
4212 sb->s_export_op = &f2fs_export_ops;
4213 sb->s_magic = F2FS_SUPER_MAGIC;
4214 sb->s_time_gran = 1;
4215 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4216 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4217 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4218 sb->s_iflags |= SB_I_CGROUPWB;
4220 /* init f2fs-specific super block info */
4221 sbi->valid_super_block = valid_super_block;
4223 /* disallow all the data/node/meta page writes */
4224 set_sbi_flag(sbi, SBI_POR_DOING);
4226 err = f2fs_init_write_merge_io(sbi);
4232 err = f2fs_init_iostat(sbi);
4236 err = init_percpu_info(sbi);
4240 if (F2FS_IO_ALIGNED(sbi)) {
4241 sbi->write_io_dummy =
4242 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4243 if (!sbi->write_io_dummy) {
4249 /* init per sbi slab cache */
4250 err = f2fs_init_xattr_caches(sbi);
4253 err = f2fs_init_page_array_cache(sbi);
4255 goto free_xattr_cache;
4257 /* get an inode for meta space */
4258 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4259 if (IS_ERR(sbi->meta_inode)) {
4260 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4261 err = PTR_ERR(sbi->meta_inode);
4262 goto free_page_array_cache;
4265 err = f2fs_get_valid_checkpoint(sbi);
4267 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4268 goto free_meta_inode;
4271 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4272 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4273 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4274 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4275 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4278 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4279 set_sbi_flag(sbi, SBI_NEED_FSCK);
4281 /* Initialize device list */
4282 err = f2fs_scan_devices(sbi);
4284 f2fs_err(sbi, "Failed to find devices");
4288 err = f2fs_init_post_read_wq(sbi);
4290 f2fs_err(sbi, "Failed to initialize post read workqueue");
4294 sbi->total_valid_node_count =
4295 le32_to_cpu(sbi->ckpt->valid_node_count);
4296 percpu_counter_set(&sbi->total_valid_inode_count,
4297 le32_to_cpu(sbi->ckpt->valid_inode_count));
4298 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4299 sbi->total_valid_block_count =
4300 le64_to_cpu(sbi->ckpt->valid_block_count);
4301 sbi->last_valid_block_count = sbi->total_valid_block_count;
4302 sbi->reserved_blocks = 0;
4303 sbi->current_reserved_blocks = 0;
4304 limit_reserve_root(sbi);
4305 adjust_unusable_cap_perc(sbi);
4307 f2fs_init_extent_cache_info(sbi);
4309 f2fs_init_ino_entry_info(sbi);
4311 f2fs_init_fsync_node_info(sbi);
4313 /* setup checkpoint request control and start checkpoint issue thread */
4314 f2fs_init_ckpt_req_control(sbi);
4315 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4316 test_opt(sbi, MERGE_CHECKPOINT)) {
4317 err = f2fs_start_ckpt_thread(sbi);
4320 "Failed to start F2FS issue_checkpoint_thread (%d)",
4322 goto stop_ckpt_thread;
4326 /* setup f2fs internal modules */
4327 err = f2fs_build_segment_manager(sbi);
4329 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4333 err = f2fs_build_node_manager(sbi);
4335 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4340 err = adjust_reserved_segment(sbi);
4344 /* For write statistics */
4345 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4347 /* Read accumulated write IO statistics if exists */
4348 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4349 if (__exist_node_summaries(sbi))
4350 sbi->kbytes_written =
4351 le64_to_cpu(seg_i->journal->info.kbytes_written);
4353 f2fs_build_gc_manager(sbi);
4355 err = f2fs_build_stats(sbi);
4359 /* get an inode for node space */
4360 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4361 if (IS_ERR(sbi->node_inode)) {
4362 f2fs_err(sbi, "Failed to read node inode");
4363 err = PTR_ERR(sbi->node_inode);
4367 /* read root inode and dentry */
4368 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4370 f2fs_err(sbi, "Failed to read root inode");
4371 err = PTR_ERR(root);
4372 goto free_node_inode;
4374 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4375 !root->i_size || !root->i_nlink) {
4378 goto free_node_inode;
4381 sb->s_root = d_make_root(root); /* allocate root dentry */
4384 goto free_node_inode;
4387 err = f2fs_init_compress_inode(sbi);
4389 goto free_root_inode;
4391 err = f2fs_register_sysfs(sbi);
4393 goto free_compress_inode;
4396 /* Enable quota usage during mount */
4397 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4398 err = f2fs_enable_quotas(sb);
4400 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4403 /* if there are any orphan inodes, free them */
4404 err = f2fs_recover_orphan_inodes(sbi);
4408 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4409 goto reset_checkpoint;
4411 /* recover fsynced data */
4412 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4413 !test_opt(sbi, NORECOVERY)) {
4415 * mount should be failed, when device has readonly mode, and
4416 * previous checkpoint was not done by clean system shutdown.
4418 if (f2fs_hw_is_readonly(sbi)) {
4419 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4420 err = f2fs_recover_fsync_data(sbi, true);
4423 f2fs_err(sbi, "Need to recover fsync data, but "
4424 "write access unavailable, please try "
4425 "mount w/ disable_roll_forward or norecovery");
4430 f2fs_info(sbi, "write access unavailable, skipping recovery");
4431 goto reset_checkpoint;
4435 set_sbi_flag(sbi, SBI_NEED_FSCK);
4438 goto reset_checkpoint;
4440 err = f2fs_recover_fsync_data(sbi, false);
4443 skip_recovery = true;
4445 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4450 err = f2fs_recover_fsync_data(sbi, true);
4452 if (!f2fs_readonly(sb) && err > 0) {
4454 f2fs_err(sbi, "Need to recover fsync data");
4460 * If the f2fs is not readonly and fsync data recovery succeeds,
4461 * check zoned block devices' write pointer consistency.
4463 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4464 err = f2fs_check_write_pointer(sbi);
4470 f2fs_init_inmem_curseg(sbi);
4472 /* f2fs_recover_fsync_data() cleared this already */
4473 clear_sbi_flag(sbi, SBI_POR_DOING);
4475 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4476 err = f2fs_disable_checkpoint(sbi);
4478 goto sync_free_meta;
4479 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4480 f2fs_enable_checkpoint(sbi);
4484 * If filesystem is not mounted as read-only then
4485 * do start the gc_thread.
4487 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4488 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4489 /* After POR, we can run background GC thread.*/
4490 err = f2fs_start_gc_thread(sbi);
4492 goto sync_free_meta;
4496 /* recover broken superblock */
4498 err = f2fs_commit_super(sbi, true);
4499 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4500 sbi->valid_super_block ? 1 : 2, err);
4503 f2fs_join_shrinker(sbi);
4505 f2fs_tuning_parameters(sbi);
4507 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4508 cur_cp_version(F2FS_CKPT(sbi)));
4509 f2fs_update_time(sbi, CP_TIME);
4510 f2fs_update_time(sbi, REQ_TIME);
4511 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4515 /* safe to flush all the data */
4516 sync_filesystem(sbi->sb);
4521 f2fs_truncate_quota_inode_pages(sb);
4522 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4523 f2fs_quota_off_umount(sbi->sb);
4526 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4527 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4528 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4529 * falls into an infinite loop in f2fs_sync_meta_pages().
4531 truncate_inode_pages_final(META_MAPPING(sbi));
4532 /* evict some inodes being cached by GC */
4534 f2fs_unregister_sysfs(sbi);
4535 free_compress_inode:
4536 f2fs_destroy_compress_inode(sbi);
4541 f2fs_release_ino_entry(sbi, true);
4542 truncate_inode_pages_final(NODE_MAPPING(sbi));
4543 iput(sbi->node_inode);
4544 sbi->node_inode = NULL;
4546 f2fs_destroy_stats(sbi);
4548 /* stop discard thread before destroying node manager */
4549 f2fs_stop_discard_thread(sbi);
4550 f2fs_destroy_node_manager(sbi);
4552 f2fs_destroy_segment_manager(sbi);
4554 f2fs_stop_ckpt_thread(sbi);
4555 f2fs_destroy_post_read_wq(sbi);
4557 destroy_device_list(sbi);
4560 make_bad_inode(sbi->meta_inode);
4561 iput(sbi->meta_inode);
4562 sbi->meta_inode = NULL;
4563 free_page_array_cache:
4564 f2fs_destroy_page_array_cache(sbi);
4566 f2fs_destroy_xattr_caches(sbi);
4568 mempool_destroy(sbi->write_io_dummy);
4570 destroy_percpu_info(sbi);
4572 f2fs_destroy_iostat(sbi);
4574 for (i = 0; i < NR_PAGE_TYPE; i++)
4575 kvfree(sbi->write_io[i]);
4577 #if IS_ENABLED(CONFIG_UNICODE)
4578 utf8_unload(sb->s_encoding);
4579 sb->s_encoding = NULL;
4583 for (i = 0; i < MAXQUOTAS; i++)
4584 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4586 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4591 if (sbi->s_chksum_driver)
4592 crypto_free_shash(sbi->s_chksum_driver);
4595 /* give only one another chance */
4596 if (retry_cnt > 0 && skip_recovery) {
4598 shrink_dcache_sb(sb);
4604 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4605 const char *dev_name, void *data)
4607 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4610 static void kill_f2fs_super(struct super_block *sb)
4613 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4615 set_sbi_flag(sbi, SBI_IS_CLOSE);
4616 f2fs_stop_gc_thread(sbi);
4617 f2fs_stop_discard_thread(sbi);
4619 #ifdef CONFIG_F2FS_FS_COMPRESSION
4621 * latter evict_inode() can bypass checking and invalidating
4622 * compress inode cache.
4624 if (test_opt(sbi, COMPRESS_CACHE))
4625 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4628 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4629 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4630 struct cp_control cpc = {
4631 .reason = CP_UMOUNT,
4633 f2fs_write_checkpoint(sbi, &cpc);
4636 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4637 sb->s_flags &= ~SB_RDONLY;
4639 kill_block_super(sb);
4642 static struct file_system_type f2fs_fs_type = {
4643 .owner = THIS_MODULE,
4645 .mount = f2fs_mount,
4646 .kill_sb = kill_f2fs_super,
4647 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4649 MODULE_ALIAS_FS("f2fs");
4651 static int __init init_inodecache(void)
4653 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4654 sizeof(struct f2fs_inode_info), 0,
4655 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4656 return f2fs_inode_cachep ? 0 : -ENOMEM;
4659 static void destroy_inodecache(void)
4662 * Make sure all delayed rcu free inodes are flushed before we
4666 kmem_cache_destroy(f2fs_inode_cachep);
4669 static int __init init_f2fs_fs(void)
4673 if (PAGE_SIZE != F2FS_BLKSIZE) {
4674 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4675 PAGE_SIZE, F2FS_BLKSIZE);
4679 err = init_inodecache();
4682 err = f2fs_create_node_manager_caches();
4684 goto free_inodecache;
4685 err = f2fs_create_segment_manager_caches();
4687 goto free_node_manager_caches;
4688 err = f2fs_create_checkpoint_caches();
4690 goto free_segment_manager_caches;
4691 err = f2fs_create_recovery_cache();
4693 goto free_checkpoint_caches;
4694 err = f2fs_create_extent_cache();
4696 goto free_recovery_cache;
4697 err = f2fs_create_garbage_collection_cache();
4699 goto free_extent_cache;
4700 err = f2fs_init_sysfs();
4702 goto free_garbage_collection_cache;
4703 err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4706 err = register_filesystem(&f2fs_fs_type);
4709 f2fs_create_root_stats();
4710 err = f2fs_init_post_read_processing();
4712 goto free_root_stats;
4713 err = f2fs_init_iostat_processing();
4715 goto free_post_read;
4716 err = f2fs_init_bio_entry_cache();
4719 err = f2fs_init_bioset();
4721 goto free_bio_entry_cache;
4722 err = f2fs_init_compress_mempool();
4725 err = f2fs_init_compress_cache();
4727 goto free_compress_mempool;
4728 err = f2fs_create_casefold_cache();
4730 goto free_compress_cache;
4732 free_compress_cache:
4733 f2fs_destroy_compress_cache();
4734 free_compress_mempool:
4735 f2fs_destroy_compress_mempool();
4737 f2fs_destroy_bioset();
4738 free_bio_entry_cache:
4739 f2fs_destroy_bio_entry_cache();
4741 f2fs_destroy_iostat_processing();
4743 f2fs_destroy_post_read_processing();
4745 f2fs_destroy_root_stats();
4746 unregister_filesystem(&f2fs_fs_type);
4748 unregister_shrinker(&f2fs_shrinker_info);
4751 free_garbage_collection_cache:
4752 f2fs_destroy_garbage_collection_cache();
4754 f2fs_destroy_extent_cache();
4755 free_recovery_cache:
4756 f2fs_destroy_recovery_cache();
4757 free_checkpoint_caches:
4758 f2fs_destroy_checkpoint_caches();
4759 free_segment_manager_caches:
4760 f2fs_destroy_segment_manager_caches();
4761 free_node_manager_caches:
4762 f2fs_destroy_node_manager_caches();
4764 destroy_inodecache();
4769 static void __exit exit_f2fs_fs(void)
4771 f2fs_destroy_casefold_cache();
4772 f2fs_destroy_compress_cache();
4773 f2fs_destroy_compress_mempool();
4774 f2fs_destroy_bioset();
4775 f2fs_destroy_bio_entry_cache();
4776 f2fs_destroy_iostat_processing();
4777 f2fs_destroy_post_read_processing();
4778 f2fs_destroy_root_stats();
4779 unregister_filesystem(&f2fs_fs_type);
4780 unregister_shrinker(&f2fs_shrinker_info);
4782 f2fs_destroy_garbage_collection_cache();
4783 f2fs_destroy_extent_cache();
4784 f2fs_destroy_recovery_cache();
4785 f2fs_destroy_checkpoint_caches();
4786 f2fs_destroy_segment_manager_caches();
4787 f2fs_destroy_node_manager_caches();
4788 destroy_inodecache();
4791 module_init(init_f2fs_fs)
4792 module_exit(exit_f2fs_fs)
4794 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4795 MODULE_DESCRIPTION("Flash Friendly File System");
4796 MODULE_LICENSE("GPL");
4797 MODULE_SOFTDEP("pre: crc32");