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;
88 static int __init f2fs_init_shrinker(void)
90 f2fs_shrinker_info = shrinker_alloc(0, "f2fs-shrinker");
91 if (!f2fs_shrinker_info)
94 f2fs_shrinker_info->count_objects = f2fs_shrink_count;
95 f2fs_shrinker_info->scan_objects = f2fs_shrink_scan;
97 shrinker_register(f2fs_shrinker_info);
102 static void f2fs_exit_shrinker(void)
104 shrinker_free(f2fs_shrinker_info);
109 Opt_disable_roll_forward,
120 Opt_disable_ext_identify,
123 Opt_inline_xattr_size,
161 Opt_test_dummy_encryption,
163 Opt_checkpoint_disable,
164 Opt_checkpoint_disable_cap,
165 Opt_checkpoint_disable_cap_perc,
166 Opt_checkpoint_enable,
167 Opt_checkpoint_merge,
168 Opt_nocheckpoint_merge,
169 Opt_compress_algorithm,
170 Opt_compress_log_size,
171 Opt_compress_extension,
172 Opt_nocompress_extension,
181 Opt_age_extent_cache,
186 static match_table_t f2fs_tokens = {
187 {Opt_gc_background, "background_gc=%s"},
188 {Opt_disable_roll_forward, "disable_roll_forward"},
189 {Opt_norecovery, "norecovery"},
190 {Opt_discard, "discard"},
191 {Opt_nodiscard, "nodiscard"},
192 {Opt_noheap, "no_heap"},
194 {Opt_user_xattr, "user_xattr"},
195 {Opt_nouser_xattr, "nouser_xattr"},
197 {Opt_noacl, "noacl"},
198 {Opt_active_logs, "active_logs=%u"},
199 {Opt_disable_ext_identify, "disable_ext_identify"},
200 {Opt_inline_xattr, "inline_xattr"},
201 {Opt_noinline_xattr, "noinline_xattr"},
202 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
203 {Opt_inline_data, "inline_data"},
204 {Opt_inline_dentry, "inline_dentry"},
205 {Opt_noinline_dentry, "noinline_dentry"},
206 {Opt_flush_merge, "flush_merge"},
207 {Opt_noflush_merge, "noflush_merge"},
208 {Opt_barrier, "barrier"},
209 {Opt_nobarrier, "nobarrier"},
210 {Opt_fastboot, "fastboot"},
211 {Opt_extent_cache, "extent_cache"},
212 {Opt_noextent_cache, "noextent_cache"},
213 {Opt_noinline_data, "noinline_data"},
214 {Opt_data_flush, "data_flush"},
215 {Opt_reserve_root, "reserve_root=%u"},
216 {Opt_resgid, "resgid=%u"},
217 {Opt_resuid, "resuid=%u"},
218 {Opt_mode, "mode=%s"},
219 {Opt_io_size_bits, "io_bits=%u"},
220 {Opt_fault_injection, "fault_injection=%u"},
221 {Opt_fault_type, "fault_type=%u"},
222 {Opt_lazytime, "lazytime"},
223 {Opt_nolazytime, "nolazytime"},
224 {Opt_quota, "quota"},
225 {Opt_noquota, "noquota"},
226 {Opt_usrquota, "usrquota"},
227 {Opt_grpquota, "grpquota"},
228 {Opt_prjquota, "prjquota"},
229 {Opt_usrjquota, "usrjquota=%s"},
230 {Opt_grpjquota, "grpjquota=%s"},
231 {Opt_prjjquota, "prjjquota=%s"},
232 {Opt_offusrjquota, "usrjquota="},
233 {Opt_offgrpjquota, "grpjquota="},
234 {Opt_offprjjquota, "prjjquota="},
235 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
236 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
237 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
238 {Opt_alloc, "alloc_mode=%s"},
239 {Opt_fsync, "fsync_mode=%s"},
240 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
241 {Opt_test_dummy_encryption, "test_dummy_encryption"},
242 {Opt_inlinecrypt, "inlinecrypt"},
243 {Opt_checkpoint_disable, "checkpoint=disable"},
244 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
245 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
246 {Opt_checkpoint_enable, "checkpoint=enable"},
247 {Opt_checkpoint_merge, "checkpoint_merge"},
248 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
249 {Opt_compress_algorithm, "compress_algorithm=%s"},
250 {Opt_compress_log_size, "compress_log_size=%u"},
251 {Opt_compress_extension, "compress_extension=%s"},
252 {Opt_nocompress_extension, "nocompress_extension=%s"},
253 {Opt_compress_chksum, "compress_chksum"},
254 {Opt_compress_mode, "compress_mode=%s"},
255 {Opt_compress_cache, "compress_cache"},
257 {Opt_gc_merge, "gc_merge"},
258 {Opt_nogc_merge, "nogc_merge"},
259 {Opt_discard_unit, "discard_unit=%s"},
260 {Opt_memory_mode, "memory=%s"},
261 {Opt_age_extent_cache, "age_extent_cache"},
262 {Opt_errors, "errors=%s"},
266 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
268 struct va_format vaf;
274 level = printk_get_level(fmt);
275 vaf.fmt = printk_skip_level(fmt);
277 printk("%c%cF2FS-fs (%s): %pV\n",
278 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
283 #if IS_ENABLED(CONFIG_UNICODE)
284 static const struct f2fs_sb_encodings {
287 unsigned int version;
288 } f2fs_sb_encoding_map[] = {
289 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
292 static const struct f2fs_sb_encodings *
293 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
295 __u16 magic = le16_to_cpu(sb->s_encoding);
298 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
299 if (magic == f2fs_sb_encoding_map[i].magic)
300 return &f2fs_sb_encoding_map[i];
305 struct kmem_cache *f2fs_cf_name_slab;
306 static int __init f2fs_create_casefold_cache(void)
308 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
310 return f2fs_cf_name_slab ? 0 : -ENOMEM;
313 static void f2fs_destroy_casefold_cache(void)
315 kmem_cache_destroy(f2fs_cf_name_slab);
318 static int __init f2fs_create_casefold_cache(void) { return 0; }
319 static void f2fs_destroy_casefold_cache(void) { }
322 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
324 block_t limit = min((sbi->user_block_count >> 3),
325 sbi->user_block_count - sbi->reserved_blocks);
328 if (test_opt(sbi, RESERVE_ROOT) &&
329 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
330 F2FS_OPTION(sbi).root_reserved_blocks = limit;
331 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
332 F2FS_OPTION(sbi).root_reserved_blocks);
334 if (!test_opt(sbi, RESERVE_ROOT) &&
335 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
336 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
337 !gid_eq(F2FS_OPTION(sbi).s_resgid,
338 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
339 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
340 from_kuid_munged(&init_user_ns,
341 F2FS_OPTION(sbi).s_resuid),
342 from_kgid_munged(&init_user_ns,
343 F2FS_OPTION(sbi).s_resgid));
346 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
348 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
349 unsigned int avg_vblocks;
350 unsigned int wanted_reserved_segments;
351 block_t avail_user_block_count;
353 if (!F2FS_IO_ALIGNED(sbi))
356 /* average valid block count in section in worst case */
357 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
360 * we need enough free space when migrating one section in worst case
362 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
363 reserved_segments(sbi);
364 wanted_reserved_segments -= reserved_segments(sbi);
366 avail_user_block_count = sbi->user_block_count -
367 sbi->current_reserved_blocks -
368 F2FS_OPTION(sbi).root_reserved_blocks;
370 if (wanted_reserved_segments * sbi->blocks_per_seg >
371 avail_user_block_count) {
372 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
373 wanted_reserved_segments,
374 avail_user_block_count >> sbi->log_blocks_per_seg);
378 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
380 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
381 wanted_reserved_segments);
386 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
388 if (!F2FS_OPTION(sbi).unusable_cap_perc)
391 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
392 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
394 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
395 F2FS_OPTION(sbi).unusable_cap_perc;
397 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
398 F2FS_OPTION(sbi).unusable_cap,
399 F2FS_OPTION(sbi).unusable_cap_perc);
402 static void init_once(void *foo)
404 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
406 inode_init_once(&fi->vfs_inode);
410 static const char * const quotatypes[] = INITQFNAMES;
411 #define QTYPE2NAME(t) (quotatypes[t])
412 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
415 struct f2fs_sb_info *sbi = F2FS_SB(sb);
419 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
420 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
423 if (f2fs_sb_has_quota_ino(sbi)) {
424 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
428 qname = match_strdup(args);
430 f2fs_err(sbi, "Not enough memory for storing quotafile name");
433 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
434 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
437 f2fs_err(sbi, "%s quota file already specified",
441 if (strchr(qname, '/')) {
442 f2fs_err(sbi, "quotafile must be on filesystem root");
445 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
453 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
455 struct f2fs_sb_info *sbi = F2FS_SB(sb);
457 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
458 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
461 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
462 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
466 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
469 * We do the test below only for project quotas. 'usrquota' and
470 * 'grpquota' mount options are allowed even without quota feature
471 * to support legacy quotas in quota files.
473 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
474 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
477 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
478 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
479 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
480 if (test_opt(sbi, USRQUOTA) &&
481 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
482 clear_opt(sbi, USRQUOTA);
484 if (test_opt(sbi, GRPQUOTA) &&
485 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
486 clear_opt(sbi, GRPQUOTA);
488 if (test_opt(sbi, PRJQUOTA) &&
489 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
490 clear_opt(sbi, PRJQUOTA);
492 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
493 test_opt(sbi, PRJQUOTA)) {
494 f2fs_err(sbi, "old and new quota format mixing");
498 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
499 f2fs_err(sbi, "journaled quota format not specified");
504 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
505 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
506 F2FS_OPTION(sbi).s_jquota_fmt = 0;
512 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
514 const substring_t *arg,
517 struct f2fs_sb_info *sbi = F2FS_SB(sb);
518 struct fs_parameter param = {
519 .type = fs_value_is_string,
520 .string = arg->from ? arg->from : "",
522 struct fscrypt_dummy_policy *policy =
523 &F2FS_OPTION(sbi).dummy_enc_policy;
526 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
527 f2fs_warn(sbi, "test_dummy_encryption option not supported");
531 if (!f2fs_sb_has_encrypt(sbi)) {
532 f2fs_err(sbi, "Encrypt feature is off");
537 * This mount option is just for testing, and it's not worthwhile to
538 * implement the extra complexity (e.g. RCU protection) that would be
539 * needed to allow it to be set or changed during remount. We do allow
540 * it to be specified during remount, but only if there is no change.
542 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
543 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
547 err = fscrypt_parse_test_dummy_encryption(¶m, policy);
551 "Can't change test_dummy_encryption on remount");
552 else if (err == -EINVAL)
553 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
556 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
560 f2fs_warn(sbi, "Test dummy encryption mode enabled");
564 #ifdef CONFIG_F2FS_FS_COMPRESSION
565 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
566 const char *new_ext, bool is_ext)
568 unsigned char (*ext)[F2FS_EXTENSION_LEN];
573 ext = F2FS_OPTION(sbi).extensions;
574 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
576 ext = F2FS_OPTION(sbi).noextensions;
577 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
580 for (i = 0; i < ext_cnt; i++) {
581 if (!strcasecmp(new_ext, ext[i]))
589 * 1. The same extension name cannot not appear in both compress and non-compress extension
591 * 2. If the compress extension specifies all files, the types specified by the non-compress
592 * extension will be treated as special cases and will not be compressed.
593 * 3. Don't allow the non-compress extension specifies all files.
595 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
597 unsigned char (*ext)[F2FS_EXTENSION_LEN];
598 unsigned char (*noext)[F2FS_EXTENSION_LEN];
599 int ext_cnt, noext_cnt, index = 0, no_index = 0;
601 ext = F2FS_OPTION(sbi).extensions;
602 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
603 noext = F2FS_OPTION(sbi).noextensions;
604 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
609 for (no_index = 0; no_index < noext_cnt; no_index++) {
610 if (!strcasecmp("*", noext[no_index])) {
611 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
614 for (index = 0; index < ext_cnt; index++) {
615 if (!strcasecmp(ext[index], noext[no_index])) {
616 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
625 #ifdef CONFIG_F2FS_FS_LZ4
626 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
628 #ifdef CONFIG_F2FS_FS_LZ4HC
631 if (strlen(str) == 3) {
632 F2FS_OPTION(sbi).compress_level = 0;
639 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
642 if (kstrtouint(str + 1, 10, &level))
645 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
646 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
650 F2FS_OPTION(sbi).compress_level = level;
653 if (strlen(str) == 3) {
654 F2FS_OPTION(sbi).compress_level = 0;
657 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
663 #ifdef CONFIG_F2FS_FS_ZSTD
664 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
669 if (strlen(str) == len) {
670 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
677 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
680 if (kstrtoint(str + 1, 10, &level))
683 /* f2fs does not support negative compress level now */
685 f2fs_info(sbi, "do not support negative compress level: %d", level);
689 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
690 f2fs_info(sbi, "invalid zstd compress level: %d", level);
694 F2FS_OPTION(sbi).compress_level = level;
700 static int parse_options(struct super_block *sb, char *options, bool is_remount)
702 struct f2fs_sb_info *sbi = F2FS_SB(sb);
703 substring_t args[MAX_OPT_ARGS];
704 #ifdef CONFIG_F2FS_FS_COMPRESSION
705 unsigned char (*ext)[F2FS_EXTENSION_LEN];
706 unsigned char (*noext)[F2FS_EXTENSION_LEN];
707 int ext_cnt, noext_cnt;
718 while ((p = strsep(&options, ",")) != NULL) {
724 * Initialize args struct so we know whether arg was
725 * found; some options take optional arguments.
727 args[0].to = args[0].from = NULL;
728 token = match_token(p, f2fs_tokens, args);
731 case Opt_gc_background:
732 name = match_strdup(&args[0]);
736 if (!strcmp(name, "on")) {
737 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
738 } else if (!strcmp(name, "off")) {
739 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
740 } else if (!strcmp(name, "sync")) {
741 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
748 case Opt_disable_roll_forward:
749 set_opt(sbi, DISABLE_ROLL_FORWARD);
752 /* this option mounts f2fs with ro */
753 set_opt(sbi, NORECOVERY);
754 if (!f2fs_readonly(sb))
758 if (!f2fs_hw_support_discard(sbi)) {
759 f2fs_warn(sbi, "device does not support discard");
762 set_opt(sbi, DISCARD);
765 if (f2fs_hw_should_discard(sbi)) {
766 f2fs_warn(sbi, "discard is required for zoned block devices");
769 clear_opt(sbi, DISCARD);
772 set_opt(sbi, NOHEAP);
775 clear_opt(sbi, NOHEAP);
777 #ifdef CONFIG_F2FS_FS_XATTR
779 set_opt(sbi, XATTR_USER);
781 case Opt_nouser_xattr:
782 clear_opt(sbi, XATTR_USER);
784 case Opt_inline_xattr:
785 set_opt(sbi, INLINE_XATTR);
787 case Opt_noinline_xattr:
788 clear_opt(sbi, INLINE_XATTR);
790 case Opt_inline_xattr_size:
791 if (args->from && match_int(args, &arg))
793 set_opt(sbi, INLINE_XATTR_SIZE);
794 F2FS_OPTION(sbi).inline_xattr_size = arg;
798 f2fs_info(sbi, "user_xattr options not supported");
800 case Opt_nouser_xattr:
801 f2fs_info(sbi, "nouser_xattr options not supported");
803 case Opt_inline_xattr:
804 f2fs_info(sbi, "inline_xattr options not supported");
806 case Opt_noinline_xattr:
807 f2fs_info(sbi, "noinline_xattr options not supported");
810 #ifdef CONFIG_F2FS_FS_POSIX_ACL
812 set_opt(sbi, POSIX_ACL);
815 clear_opt(sbi, POSIX_ACL);
819 f2fs_info(sbi, "acl options not supported");
822 f2fs_info(sbi, "noacl options not supported");
825 case Opt_active_logs:
826 if (args->from && match_int(args, &arg))
828 if (arg != 2 && arg != 4 &&
829 arg != NR_CURSEG_PERSIST_TYPE)
831 F2FS_OPTION(sbi).active_logs = arg;
833 case Opt_disable_ext_identify:
834 set_opt(sbi, DISABLE_EXT_IDENTIFY);
836 case Opt_inline_data:
837 set_opt(sbi, INLINE_DATA);
839 case Opt_inline_dentry:
840 set_opt(sbi, INLINE_DENTRY);
842 case Opt_noinline_dentry:
843 clear_opt(sbi, INLINE_DENTRY);
845 case Opt_flush_merge:
846 set_opt(sbi, FLUSH_MERGE);
848 case Opt_noflush_merge:
849 clear_opt(sbi, FLUSH_MERGE);
852 set_opt(sbi, NOBARRIER);
855 clear_opt(sbi, NOBARRIER);
858 set_opt(sbi, FASTBOOT);
860 case Opt_extent_cache:
861 set_opt(sbi, READ_EXTENT_CACHE);
863 case Opt_noextent_cache:
864 clear_opt(sbi, READ_EXTENT_CACHE);
866 case Opt_noinline_data:
867 clear_opt(sbi, INLINE_DATA);
870 set_opt(sbi, DATA_FLUSH);
872 case Opt_reserve_root:
873 if (args->from && match_int(args, &arg))
875 if (test_opt(sbi, RESERVE_ROOT)) {
876 f2fs_info(sbi, "Preserve previous reserve_root=%u",
877 F2FS_OPTION(sbi).root_reserved_blocks);
879 F2FS_OPTION(sbi).root_reserved_blocks = arg;
880 set_opt(sbi, RESERVE_ROOT);
884 if (args->from && match_int(args, &arg))
886 uid = make_kuid(current_user_ns(), arg);
887 if (!uid_valid(uid)) {
888 f2fs_err(sbi, "Invalid uid value %d", arg);
891 F2FS_OPTION(sbi).s_resuid = uid;
894 if (args->from && match_int(args, &arg))
896 gid = make_kgid(current_user_ns(), arg);
897 if (!gid_valid(gid)) {
898 f2fs_err(sbi, "Invalid gid value %d", arg);
901 F2FS_OPTION(sbi).s_resgid = gid;
904 name = match_strdup(&args[0]);
908 if (!strcmp(name, "adaptive")) {
909 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
910 } else if (!strcmp(name, "lfs")) {
911 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
912 } else if (!strcmp(name, "fragment:segment")) {
913 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
914 } else if (!strcmp(name, "fragment:block")) {
915 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
922 case Opt_io_size_bits:
923 if (args->from && match_int(args, &arg))
925 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
926 f2fs_warn(sbi, "Not support %ld, larger than %d",
927 BIT(arg), BIO_MAX_VECS);
930 F2FS_OPTION(sbi).write_io_size_bits = arg;
932 #ifdef CONFIG_F2FS_FAULT_INJECTION
933 case Opt_fault_injection:
934 if (args->from && match_int(args, &arg))
936 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
937 set_opt(sbi, FAULT_INJECTION);
941 if (args->from && match_int(args, &arg))
943 f2fs_build_fault_attr(sbi, 0, arg);
944 set_opt(sbi, FAULT_INJECTION);
947 case Opt_fault_injection:
948 f2fs_info(sbi, "fault_injection options not supported");
952 f2fs_info(sbi, "fault_type options not supported");
956 sb->s_flags |= SB_LAZYTIME;
959 sb->s_flags &= ~SB_LAZYTIME;
964 set_opt(sbi, USRQUOTA);
967 set_opt(sbi, GRPQUOTA);
970 set_opt(sbi, PRJQUOTA);
973 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
978 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
983 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
987 case Opt_offusrjquota:
988 ret = f2fs_clear_qf_name(sb, USRQUOTA);
992 case Opt_offgrpjquota:
993 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
997 case Opt_offprjjquota:
998 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
1002 case Opt_jqfmt_vfsold:
1003 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
1005 case Opt_jqfmt_vfsv0:
1006 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
1008 case Opt_jqfmt_vfsv1:
1009 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
1012 clear_opt(sbi, QUOTA);
1013 clear_opt(sbi, USRQUOTA);
1014 clear_opt(sbi, GRPQUOTA);
1015 clear_opt(sbi, PRJQUOTA);
1025 case Opt_offusrjquota:
1026 case Opt_offgrpjquota:
1027 case Opt_offprjjquota:
1028 case Opt_jqfmt_vfsold:
1029 case Opt_jqfmt_vfsv0:
1030 case Opt_jqfmt_vfsv1:
1032 f2fs_info(sbi, "quota operations not supported");
1036 name = match_strdup(&args[0]);
1040 if (!strcmp(name, "default")) {
1041 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1042 } else if (!strcmp(name, "reuse")) {
1043 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1051 name = match_strdup(&args[0]);
1054 if (!strcmp(name, "posix")) {
1055 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1056 } else if (!strcmp(name, "strict")) {
1057 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1058 } else if (!strcmp(name, "nobarrier")) {
1059 F2FS_OPTION(sbi).fsync_mode =
1060 FSYNC_MODE_NOBARRIER;
1067 case Opt_test_dummy_encryption:
1068 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1073 case Opt_inlinecrypt:
1074 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1075 sb->s_flags |= SB_INLINECRYPT;
1077 f2fs_info(sbi, "inline encryption not supported");
1080 case Opt_checkpoint_disable_cap_perc:
1081 if (args->from && match_int(args, &arg))
1083 if (arg < 0 || arg > 100)
1085 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1086 set_opt(sbi, DISABLE_CHECKPOINT);
1088 case Opt_checkpoint_disable_cap:
1089 if (args->from && match_int(args, &arg))
1091 F2FS_OPTION(sbi).unusable_cap = arg;
1092 set_opt(sbi, DISABLE_CHECKPOINT);
1094 case Opt_checkpoint_disable:
1095 set_opt(sbi, DISABLE_CHECKPOINT);
1097 case Opt_checkpoint_enable:
1098 clear_opt(sbi, DISABLE_CHECKPOINT);
1100 case Opt_checkpoint_merge:
1101 set_opt(sbi, MERGE_CHECKPOINT);
1103 case Opt_nocheckpoint_merge:
1104 clear_opt(sbi, MERGE_CHECKPOINT);
1106 #ifdef CONFIG_F2FS_FS_COMPRESSION
1107 case Opt_compress_algorithm:
1108 if (!f2fs_sb_has_compression(sbi)) {
1109 f2fs_info(sbi, "Image doesn't support compression");
1112 name = match_strdup(&args[0]);
1115 if (!strcmp(name, "lzo")) {
1116 #ifdef CONFIG_F2FS_FS_LZO
1117 F2FS_OPTION(sbi).compress_level = 0;
1118 F2FS_OPTION(sbi).compress_algorithm =
1121 f2fs_info(sbi, "kernel doesn't support lzo compression");
1123 } else if (!strncmp(name, "lz4", 3)) {
1124 #ifdef CONFIG_F2FS_FS_LZ4
1125 ret = f2fs_set_lz4hc_level(sbi, name);
1130 F2FS_OPTION(sbi).compress_algorithm =
1133 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1135 } else if (!strncmp(name, "zstd", 4)) {
1136 #ifdef CONFIG_F2FS_FS_ZSTD
1137 ret = f2fs_set_zstd_level(sbi, name);
1142 F2FS_OPTION(sbi).compress_algorithm =
1145 f2fs_info(sbi, "kernel doesn't support zstd compression");
1147 } else if (!strcmp(name, "lzo-rle")) {
1148 #ifdef CONFIG_F2FS_FS_LZORLE
1149 F2FS_OPTION(sbi).compress_level = 0;
1150 F2FS_OPTION(sbi).compress_algorithm =
1153 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1161 case Opt_compress_log_size:
1162 if (!f2fs_sb_has_compression(sbi)) {
1163 f2fs_info(sbi, "Image doesn't support compression");
1166 if (args->from && match_int(args, &arg))
1168 if (arg < MIN_COMPRESS_LOG_SIZE ||
1169 arg > MAX_COMPRESS_LOG_SIZE) {
1171 "Compress cluster log size is out of range");
1174 F2FS_OPTION(sbi).compress_log_size = arg;
1176 case Opt_compress_extension:
1177 if (!f2fs_sb_has_compression(sbi)) {
1178 f2fs_info(sbi, "Image doesn't support compression");
1181 name = match_strdup(&args[0]);
1185 ext = F2FS_OPTION(sbi).extensions;
1186 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1188 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1189 ext_cnt >= COMPRESS_EXT_NUM) {
1191 "invalid extension length/number");
1196 if (is_compress_extension_exist(sbi, name, true)) {
1201 strcpy(ext[ext_cnt], name);
1202 F2FS_OPTION(sbi).compress_ext_cnt++;
1205 case Opt_nocompress_extension:
1206 if (!f2fs_sb_has_compression(sbi)) {
1207 f2fs_info(sbi, "Image doesn't support compression");
1210 name = match_strdup(&args[0]);
1214 noext = F2FS_OPTION(sbi).noextensions;
1215 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1217 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1218 noext_cnt >= COMPRESS_EXT_NUM) {
1220 "invalid extension length/number");
1225 if (is_compress_extension_exist(sbi, name, false)) {
1230 strcpy(noext[noext_cnt], name);
1231 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1234 case Opt_compress_chksum:
1235 if (!f2fs_sb_has_compression(sbi)) {
1236 f2fs_info(sbi, "Image doesn't support compression");
1239 F2FS_OPTION(sbi).compress_chksum = true;
1241 case Opt_compress_mode:
1242 if (!f2fs_sb_has_compression(sbi)) {
1243 f2fs_info(sbi, "Image doesn't support compression");
1246 name = match_strdup(&args[0]);
1249 if (!strcmp(name, "fs")) {
1250 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1251 } else if (!strcmp(name, "user")) {
1252 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1259 case Opt_compress_cache:
1260 if (!f2fs_sb_has_compression(sbi)) {
1261 f2fs_info(sbi, "Image doesn't support compression");
1264 set_opt(sbi, COMPRESS_CACHE);
1267 case Opt_compress_algorithm:
1268 case Opt_compress_log_size:
1269 case Opt_compress_extension:
1270 case Opt_nocompress_extension:
1271 case Opt_compress_chksum:
1272 case Opt_compress_mode:
1273 case Opt_compress_cache:
1274 f2fs_info(sbi, "compression options not supported");
1281 set_opt(sbi, GC_MERGE);
1283 case Opt_nogc_merge:
1284 clear_opt(sbi, GC_MERGE);
1286 case Opt_discard_unit:
1287 name = match_strdup(&args[0]);
1290 if (!strcmp(name, "block")) {
1291 F2FS_OPTION(sbi).discard_unit =
1293 } else if (!strcmp(name, "segment")) {
1294 F2FS_OPTION(sbi).discard_unit =
1295 DISCARD_UNIT_SEGMENT;
1296 } else if (!strcmp(name, "section")) {
1297 F2FS_OPTION(sbi).discard_unit =
1298 DISCARD_UNIT_SECTION;
1305 case Opt_memory_mode:
1306 name = match_strdup(&args[0]);
1309 if (!strcmp(name, "normal")) {
1310 F2FS_OPTION(sbi).memory_mode =
1312 } else if (!strcmp(name, "low")) {
1313 F2FS_OPTION(sbi).memory_mode =
1321 case Opt_age_extent_cache:
1322 set_opt(sbi, AGE_EXTENT_CACHE);
1325 name = match_strdup(&args[0]);
1328 if (!strcmp(name, "remount-ro")) {
1329 F2FS_OPTION(sbi).errors =
1330 MOUNT_ERRORS_READONLY;
1331 } else if (!strcmp(name, "continue")) {
1332 F2FS_OPTION(sbi).errors =
1333 MOUNT_ERRORS_CONTINUE;
1334 } else if (!strcmp(name, "panic")) {
1335 F2FS_OPTION(sbi).errors =
1344 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1351 if (f2fs_check_quota_options(sbi))
1354 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1355 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1358 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1359 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1363 #if !IS_ENABLED(CONFIG_UNICODE)
1364 if (f2fs_sb_has_casefold(sbi)) {
1366 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1371 * The BLKZONED feature indicates that the drive was formatted with
1372 * zone alignment optimization. This is optional for host-aware
1373 * devices, but mandatory for host-managed zoned block devices.
1375 if (f2fs_sb_has_blkzoned(sbi)) {
1376 #ifdef CONFIG_BLK_DEV_ZONED
1377 if (F2FS_OPTION(sbi).discard_unit !=
1378 DISCARD_UNIT_SECTION) {
1379 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1380 F2FS_OPTION(sbi).discard_unit =
1381 DISCARD_UNIT_SECTION;
1384 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1385 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1389 f2fs_err(sbi, "Zoned block device support is not enabled");
1394 #ifdef CONFIG_F2FS_FS_COMPRESSION
1395 if (f2fs_test_compress_extension(sbi)) {
1396 f2fs_err(sbi, "invalid compress or nocompress extension");
1401 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1402 f2fs_err(sbi, "Should set mode=lfs with %luKB-sized IO",
1403 F2FS_IO_SIZE_KB(sbi));
1407 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1408 int min_size, max_size;
1410 if (!f2fs_sb_has_extra_attr(sbi) ||
1411 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1412 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1415 if (!test_opt(sbi, INLINE_XATTR)) {
1416 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1420 min_size = MIN_INLINE_XATTR_SIZE;
1421 max_size = MAX_INLINE_XATTR_SIZE;
1423 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1424 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1425 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1426 min_size, max_size);
1431 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1432 f2fs_err(sbi, "LFS is not compatible with ATGC");
1436 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1437 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1441 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1442 f2fs_err(sbi, "Allow to mount readonly mode only");
1448 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1450 struct f2fs_inode_info *fi;
1452 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1455 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1459 init_once((void *) fi);
1461 /* Initialize f2fs-specific inode info */
1462 atomic_set(&fi->dirty_pages, 0);
1463 atomic_set(&fi->i_compr_blocks, 0);
1464 init_f2fs_rwsem(&fi->i_sem);
1465 spin_lock_init(&fi->i_size_lock);
1466 INIT_LIST_HEAD(&fi->dirty_list);
1467 INIT_LIST_HEAD(&fi->gdirty_list);
1468 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1469 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1470 init_f2fs_rwsem(&fi->i_xattr_sem);
1472 /* Will be used by directory only */
1473 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1475 return &fi->vfs_inode;
1478 static int f2fs_drop_inode(struct inode *inode)
1480 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1484 * during filesystem shutdown, if checkpoint is disabled,
1485 * drop useless meta/node dirty pages.
1487 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1488 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1489 inode->i_ino == F2FS_META_INO(sbi)) {
1490 trace_f2fs_drop_inode(inode, 1);
1496 * This is to avoid a deadlock condition like below.
1497 * writeback_single_inode(inode)
1498 * - f2fs_write_data_page
1499 * - f2fs_gc -> iput -> evict
1500 * - inode_wait_for_writeback(inode)
1502 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1503 if (!inode->i_nlink && !is_bad_inode(inode)) {
1504 /* to avoid evict_inode call simultaneously */
1505 atomic_inc(&inode->i_count);
1506 spin_unlock(&inode->i_lock);
1508 /* should remain fi->extent_tree for writepage */
1509 f2fs_destroy_extent_node(inode);
1511 sb_start_intwrite(inode->i_sb);
1512 f2fs_i_size_write(inode, 0);
1514 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1515 inode, NULL, 0, DATA);
1516 truncate_inode_pages_final(inode->i_mapping);
1518 if (F2FS_HAS_BLOCKS(inode))
1519 f2fs_truncate(inode);
1521 sb_end_intwrite(inode->i_sb);
1523 spin_lock(&inode->i_lock);
1524 atomic_dec(&inode->i_count);
1526 trace_f2fs_drop_inode(inode, 0);
1529 ret = generic_drop_inode(inode);
1531 ret = fscrypt_drop_inode(inode);
1532 trace_f2fs_drop_inode(inode, ret);
1536 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1538 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1541 spin_lock(&sbi->inode_lock[DIRTY_META]);
1542 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1545 set_inode_flag(inode, FI_DIRTY_INODE);
1546 stat_inc_dirty_inode(sbi, DIRTY_META);
1548 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1549 list_add_tail(&F2FS_I(inode)->gdirty_list,
1550 &sbi->inode_list[DIRTY_META]);
1551 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1553 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1557 void f2fs_inode_synced(struct inode *inode)
1559 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1561 spin_lock(&sbi->inode_lock[DIRTY_META]);
1562 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1563 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1566 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1567 list_del_init(&F2FS_I(inode)->gdirty_list);
1568 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1570 clear_inode_flag(inode, FI_DIRTY_INODE);
1571 clear_inode_flag(inode, FI_AUTO_RECOVER);
1572 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1573 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1577 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1579 * We should call set_dirty_inode to write the dirty inode through write_inode.
1581 static void f2fs_dirty_inode(struct inode *inode, int flags)
1583 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1585 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1586 inode->i_ino == F2FS_META_INO(sbi))
1589 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1590 clear_inode_flag(inode, FI_AUTO_RECOVER);
1592 f2fs_inode_dirtied(inode, false);
1595 static void f2fs_free_inode(struct inode *inode)
1597 fscrypt_free_inode(inode);
1598 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1601 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1603 percpu_counter_destroy(&sbi->total_valid_inode_count);
1604 percpu_counter_destroy(&sbi->rf_node_block_count);
1605 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1608 static void destroy_device_list(struct f2fs_sb_info *sbi)
1612 for (i = 0; i < sbi->s_ndevs; i++) {
1614 bdev_release(FDEV(i).bdev_handle);
1615 #ifdef CONFIG_BLK_DEV_ZONED
1616 kvfree(FDEV(i).blkz_seq);
1622 static void f2fs_put_super(struct super_block *sb)
1624 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1629 /* unregister procfs/sysfs entries in advance to avoid race case */
1630 f2fs_unregister_sysfs(sbi);
1632 f2fs_quota_off_umount(sb);
1634 /* prevent remaining shrinker jobs */
1635 mutex_lock(&sbi->umount_mutex);
1638 * flush all issued checkpoints and stop checkpoint issue thread.
1639 * after then, all checkpoints should be done by each process context.
1641 f2fs_stop_ckpt_thread(sbi);
1644 * We don't need to do checkpoint when superblock is clean.
1645 * But, the previous checkpoint was not done by umount, it needs to do
1646 * clean checkpoint again.
1648 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1649 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1650 struct cp_control cpc = {
1651 .reason = CP_UMOUNT,
1653 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1654 err = f2fs_write_checkpoint(sbi, &cpc);
1657 /* be sure to wait for any on-going discard commands */
1658 done = f2fs_issue_discard_timeout(sbi);
1659 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1660 struct cp_control cpc = {
1661 .reason = CP_UMOUNT | CP_TRIMMED,
1663 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1664 err = f2fs_write_checkpoint(sbi, &cpc);
1668 * normally superblock is clean, so we need to release this.
1669 * In addition, EIO will skip do checkpoint, we need this as well.
1671 f2fs_release_ino_entry(sbi, true);
1673 f2fs_leave_shrinker(sbi);
1674 mutex_unlock(&sbi->umount_mutex);
1676 /* our cp_error case, we can wait for any writeback page */
1677 f2fs_flush_merged_writes(sbi);
1679 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1681 if (err || f2fs_cp_error(sbi)) {
1682 truncate_inode_pages_final(NODE_MAPPING(sbi));
1683 truncate_inode_pages_final(META_MAPPING(sbi));
1686 for (i = 0; i < NR_COUNT_TYPE; i++) {
1687 if (!get_pages(sbi, i))
1689 f2fs_err(sbi, "detect filesystem reference count leak during "
1690 "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1691 f2fs_bug_on(sbi, 1);
1694 f2fs_bug_on(sbi, sbi->fsync_node_num);
1696 f2fs_destroy_compress_inode(sbi);
1698 iput(sbi->node_inode);
1699 sbi->node_inode = NULL;
1701 iput(sbi->meta_inode);
1702 sbi->meta_inode = NULL;
1705 * iput() can update stat information, if f2fs_write_checkpoint()
1706 * above failed with error.
1708 f2fs_destroy_stats(sbi);
1710 /* destroy f2fs internal modules */
1711 f2fs_destroy_node_manager(sbi);
1712 f2fs_destroy_segment_manager(sbi);
1714 /* flush s_error_work before sbi destroy */
1715 flush_work(&sbi->s_error_work);
1717 f2fs_destroy_post_read_wq(sbi);
1721 if (sbi->s_chksum_driver)
1722 crypto_free_shash(sbi->s_chksum_driver);
1723 kfree(sbi->raw_super);
1725 f2fs_destroy_page_array_cache(sbi);
1726 f2fs_destroy_xattr_caches(sbi);
1727 mempool_destroy(sbi->write_io_dummy);
1729 for (i = 0; i < MAXQUOTAS; i++)
1730 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1732 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1733 destroy_percpu_info(sbi);
1734 f2fs_destroy_iostat(sbi);
1735 for (i = 0; i < NR_PAGE_TYPE; i++)
1736 kvfree(sbi->write_io[i]);
1737 #if IS_ENABLED(CONFIG_UNICODE)
1738 utf8_unload(sb->s_encoding);
1742 int f2fs_sync_fs(struct super_block *sb, int sync)
1744 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1747 if (unlikely(f2fs_cp_error(sbi)))
1749 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1752 trace_f2fs_sync_fs(sb, sync);
1754 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1758 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1759 err = f2fs_issue_checkpoint(sbi);
1765 static int f2fs_freeze(struct super_block *sb)
1767 if (f2fs_readonly(sb))
1770 /* IO error happened before */
1771 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1774 /* must be clean, since sync_filesystem() was already called */
1775 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1778 /* Let's flush checkpoints and stop the thread. */
1779 f2fs_flush_ckpt_thread(F2FS_SB(sb));
1781 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1782 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1786 static int f2fs_unfreeze(struct super_block *sb)
1788 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1793 static int f2fs_statfs_project(struct super_block *sb,
1794 kprojid_t projid, struct kstatfs *buf)
1797 struct dquot *dquot;
1801 qid = make_kqid_projid(projid);
1802 dquot = dqget(sb, qid);
1804 return PTR_ERR(dquot);
1805 spin_lock(&dquot->dq_dqb_lock);
1807 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1808 dquot->dq_dqb.dqb_bhardlimit);
1810 limit >>= sb->s_blocksize_bits;
1812 if (limit && buf->f_blocks > limit) {
1813 curblock = (dquot->dq_dqb.dqb_curspace +
1814 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1815 buf->f_blocks = limit;
1816 buf->f_bfree = buf->f_bavail =
1817 (buf->f_blocks > curblock) ?
1818 (buf->f_blocks - curblock) : 0;
1821 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1822 dquot->dq_dqb.dqb_ihardlimit);
1824 if (limit && buf->f_files > limit) {
1825 buf->f_files = limit;
1827 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1828 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1831 spin_unlock(&dquot->dq_dqb_lock);
1837 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1839 struct super_block *sb = dentry->d_sb;
1840 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1841 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1842 block_t total_count, user_block_count, start_count;
1843 u64 avail_node_count;
1844 unsigned int total_valid_node_count;
1846 total_count = le64_to_cpu(sbi->raw_super->block_count);
1847 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1848 buf->f_type = F2FS_SUPER_MAGIC;
1849 buf->f_bsize = sbi->blocksize;
1851 buf->f_blocks = total_count - start_count;
1853 spin_lock(&sbi->stat_lock);
1855 user_block_count = sbi->user_block_count;
1856 total_valid_node_count = valid_node_count(sbi);
1857 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1858 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1859 sbi->current_reserved_blocks;
1861 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1864 buf->f_bfree -= sbi->unusable_block_count;
1865 spin_unlock(&sbi->stat_lock);
1867 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1868 buf->f_bavail = buf->f_bfree -
1869 F2FS_OPTION(sbi).root_reserved_blocks;
1873 if (avail_node_count > user_block_count) {
1874 buf->f_files = user_block_count;
1875 buf->f_ffree = buf->f_bavail;
1877 buf->f_files = avail_node_count;
1878 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1882 buf->f_namelen = F2FS_NAME_LEN;
1883 buf->f_fsid = u64_to_fsid(id);
1886 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1887 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1888 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1894 static inline void f2fs_show_quota_options(struct seq_file *seq,
1895 struct super_block *sb)
1898 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1900 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1903 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1914 seq_printf(seq, ",jqfmt=%s", fmtname);
1917 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1918 seq_show_option(seq, "usrjquota",
1919 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1921 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1922 seq_show_option(seq, "grpjquota",
1923 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1925 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1926 seq_show_option(seq, "prjjquota",
1927 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1931 #ifdef CONFIG_F2FS_FS_COMPRESSION
1932 static inline void f2fs_show_compress_options(struct seq_file *seq,
1933 struct super_block *sb)
1935 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1939 if (!f2fs_sb_has_compression(sbi))
1942 switch (F2FS_OPTION(sbi).compress_algorithm) {
1952 case COMPRESS_LZORLE:
1953 algtype = "lzo-rle";
1956 seq_printf(seq, ",compress_algorithm=%s", algtype);
1958 if (F2FS_OPTION(sbi).compress_level)
1959 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1961 seq_printf(seq, ",compress_log_size=%u",
1962 F2FS_OPTION(sbi).compress_log_size);
1964 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1965 seq_printf(seq, ",compress_extension=%s",
1966 F2FS_OPTION(sbi).extensions[i]);
1969 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1970 seq_printf(seq, ",nocompress_extension=%s",
1971 F2FS_OPTION(sbi).noextensions[i]);
1974 if (F2FS_OPTION(sbi).compress_chksum)
1975 seq_puts(seq, ",compress_chksum");
1977 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1978 seq_printf(seq, ",compress_mode=%s", "fs");
1979 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1980 seq_printf(seq, ",compress_mode=%s", "user");
1982 if (test_opt(sbi, COMPRESS_CACHE))
1983 seq_puts(seq, ",compress_cache");
1987 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1989 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1991 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1992 seq_printf(seq, ",background_gc=%s", "sync");
1993 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1994 seq_printf(seq, ",background_gc=%s", "on");
1995 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1996 seq_printf(seq, ",background_gc=%s", "off");
1998 if (test_opt(sbi, GC_MERGE))
1999 seq_puts(seq, ",gc_merge");
2001 seq_puts(seq, ",nogc_merge");
2003 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2004 seq_puts(seq, ",disable_roll_forward");
2005 if (test_opt(sbi, NORECOVERY))
2006 seq_puts(seq, ",norecovery");
2007 if (test_opt(sbi, DISCARD)) {
2008 seq_puts(seq, ",discard");
2009 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2010 seq_printf(seq, ",discard_unit=%s", "block");
2011 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2012 seq_printf(seq, ",discard_unit=%s", "segment");
2013 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2014 seq_printf(seq, ",discard_unit=%s", "section");
2016 seq_puts(seq, ",nodiscard");
2018 if (test_opt(sbi, NOHEAP))
2019 seq_puts(seq, ",no_heap");
2021 seq_puts(seq, ",heap");
2022 #ifdef CONFIG_F2FS_FS_XATTR
2023 if (test_opt(sbi, XATTR_USER))
2024 seq_puts(seq, ",user_xattr");
2026 seq_puts(seq, ",nouser_xattr");
2027 if (test_opt(sbi, INLINE_XATTR))
2028 seq_puts(seq, ",inline_xattr");
2030 seq_puts(seq, ",noinline_xattr");
2031 if (test_opt(sbi, INLINE_XATTR_SIZE))
2032 seq_printf(seq, ",inline_xattr_size=%u",
2033 F2FS_OPTION(sbi).inline_xattr_size);
2035 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2036 if (test_opt(sbi, POSIX_ACL))
2037 seq_puts(seq, ",acl");
2039 seq_puts(seq, ",noacl");
2041 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
2042 seq_puts(seq, ",disable_ext_identify");
2043 if (test_opt(sbi, INLINE_DATA))
2044 seq_puts(seq, ",inline_data");
2046 seq_puts(seq, ",noinline_data");
2047 if (test_opt(sbi, INLINE_DENTRY))
2048 seq_puts(seq, ",inline_dentry");
2050 seq_puts(seq, ",noinline_dentry");
2051 if (test_opt(sbi, FLUSH_MERGE))
2052 seq_puts(seq, ",flush_merge");
2054 seq_puts(seq, ",noflush_merge");
2055 if (test_opt(sbi, NOBARRIER))
2056 seq_puts(seq, ",nobarrier");
2058 seq_puts(seq, ",barrier");
2059 if (test_opt(sbi, FASTBOOT))
2060 seq_puts(seq, ",fastboot");
2061 if (test_opt(sbi, READ_EXTENT_CACHE))
2062 seq_puts(seq, ",extent_cache");
2064 seq_puts(seq, ",noextent_cache");
2065 if (test_opt(sbi, AGE_EXTENT_CACHE))
2066 seq_puts(seq, ",age_extent_cache");
2067 if (test_opt(sbi, DATA_FLUSH))
2068 seq_puts(seq, ",data_flush");
2070 seq_puts(seq, ",mode=");
2071 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2072 seq_puts(seq, "adaptive");
2073 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2074 seq_puts(seq, "lfs");
2075 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2076 seq_puts(seq, "fragment:segment");
2077 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2078 seq_puts(seq, "fragment:block");
2079 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2080 if (test_opt(sbi, RESERVE_ROOT))
2081 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2082 F2FS_OPTION(sbi).root_reserved_blocks,
2083 from_kuid_munged(&init_user_ns,
2084 F2FS_OPTION(sbi).s_resuid),
2085 from_kgid_munged(&init_user_ns,
2086 F2FS_OPTION(sbi).s_resgid));
2087 if (F2FS_IO_SIZE_BITS(sbi))
2088 seq_printf(seq, ",io_bits=%u",
2089 F2FS_OPTION(sbi).write_io_size_bits);
2090 #ifdef CONFIG_F2FS_FAULT_INJECTION
2091 if (test_opt(sbi, FAULT_INJECTION)) {
2092 seq_printf(seq, ",fault_injection=%u",
2093 F2FS_OPTION(sbi).fault_info.inject_rate);
2094 seq_printf(seq, ",fault_type=%u",
2095 F2FS_OPTION(sbi).fault_info.inject_type);
2099 if (test_opt(sbi, QUOTA))
2100 seq_puts(seq, ",quota");
2101 if (test_opt(sbi, USRQUOTA))
2102 seq_puts(seq, ",usrquota");
2103 if (test_opt(sbi, GRPQUOTA))
2104 seq_puts(seq, ",grpquota");
2105 if (test_opt(sbi, PRJQUOTA))
2106 seq_puts(seq, ",prjquota");
2108 f2fs_show_quota_options(seq, sbi->sb);
2110 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2112 if (sbi->sb->s_flags & SB_INLINECRYPT)
2113 seq_puts(seq, ",inlinecrypt");
2115 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2116 seq_printf(seq, ",alloc_mode=%s", "default");
2117 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2118 seq_printf(seq, ",alloc_mode=%s", "reuse");
2120 if (test_opt(sbi, DISABLE_CHECKPOINT))
2121 seq_printf(seq, ",checkpoint=disable:%u",
2122 F2FS_OPTION(sbi).unusable_cap);
2123 if (test_opt(sbi, MERGE_CHECKPOINT))
2124 seq_puts(seq, ",checkpoint_merge");
2126 seq_puts(seq, ",nocheckpoint_merge");
2127 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2128 seq_printf(seq, ",fsync_mode=%s", "posix");
2129 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2130 seq_printf(seq, ",fsync_mode=%s", "strict");
2131 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2132 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2134 #ifdef CONFIG_F2FS_FS_COMPRESSION
2135 f2fs_show_compress_options(seq, sbi->sb);
2138 if (test_opt(sbi, ATGC))
2139 seq_puts(seq, ",atgc");
2141 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2142 seq_printf(seq, ",memory=%s", "normal");
2143 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2144 seq_printf(seq, ",memory=%s", "low");
2146 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2147 seq_printf(seq, ",errors=%s", "remount-ro");
2148 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2149 seq_printf(seq, ",errors=%s", "continue");
2150 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2151 seq_printf(seq, ",errors=%s", "panic");
2156 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2158 /* init some FS parameters */
2160 set_opt(sbi, READ_EXTENT_CACHE);
2161 clear_opt(sbi, DISABLE_CHECKPOINT);
2163 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2164 set_opt(sbi, DISCARD);
2166 if (f2fs_sb_has_blkzoned(sbi))
2167 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2169 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2172 if (f2fs_sb_has_readonly(sbi))
2173 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2175 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2177 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2178 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2179 SMALL_VOLUME_SEGMENTS)
2180 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2182 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2183 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2184 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2185 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2186 if (f2fs_sb_has_compression(sbi)) {
2187 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2188 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2189 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2190 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2192 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2193 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2194 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2196 sbi->sb->s_flags &= ~SB_INLINECRYPT;
2198 set_opt(sbi, INLINE_XATTR);
2199 set_opt(sbi, INLINE_DATA);
2200 set_opt(sbi, INLINE_DENTRY);
2201 set_opt(sbi, NOHEAP);
2202 set_opt(sbi, MERGE_CHECKPOINT);
2203 F2FS_OPTION(sbi).unusable_cap = 0;
2204 sbi->sb->s_flags |= SB_LAZYTIME;
2205 if (!f2fs_is_readonly(sbi))
2206 set_opt(sbi, FLUSH_MERGE);
2207 if (f2fs_sb_has_blkzoned(sbi))
2208 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2210 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2212 #ifdef CONFIG_F2FS_FS_XATTR
2213 set_opt(sbi, XATTR_USER);
2215 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2216 set_opt(sbi, POSIX_ACL);
2219 f2fs_build_fault_attr(sbi, 0, 0);
2223 static int f2fs_enable_quotas(struct super_block *sb);
2226 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2228 unsigned int s_flags = sbi->sb->s_flags;
2229 struct cp_control cpc;
2230 unsigned int gc_mode = sbi->gc_mode;
2235 if (s_flags & SB_RDONLY) {
2236 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2239 sbi->sb->s_flags |= SB_ACTIVE;
2241 /* check if we need more GC first */
2242 unusable = f2fs_get_unusable_blocks(sbi);
2243 if (!f2fs_disable_cp_again(sbi, unusable))
2246 f2fs_update_time(sbi, DISABLE_TIME);
2248 sbi->gc_mode = GC_URGENT_HIGH;
2250 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2251 struct f2fs_gc_control gc_control = {
2252 .victim_segno = NULL_SEGNO,
2253 .init_gc_type = FG_GC,
2254 .should_migrate_blocks = false,
2255 .err_gc_skipped = true,
2256 .nr_free_secs = 1 };
2258 f2fs_down_write(&sbi->gc_lock);
2259 stat_inc_gc_call_count(sbi, FOREGROUND);
2260 err = f2fs_gc(sbi, &gc_control);
2261 if (err == -ENODATA) {
2265 if (err && err != -EAGAIN)
2269 ret = sync_filesystem(sbi->sb);
2271 err = ret ? ret : err;
2275 unusable = f2fs_get_unusable_blocks(sbi);
2276 if (f2fs_disable_cp_again(sbi, unusable)) {
2282 f2fs_down_write(&sbi->gc_lock);
2283 cpc.reason = CP_PAUSE;
2284 set_sbi_flag(sbi, SBI_CP_DISABLED);
2285 stat_inc_cp_call_count(sbi, TOTAL_CALL);
2286 err = f2fs_write_checkpoint(sbi, &cpc);
2290 spin_lock(&sbi->stat_lock);
2291 sbi->unusable_block_count = unusable;
2292 spin_unlock(&sbi->stat_lock);
2295 f2fs_up_write(&sbi->gc_lock);
2297 sbi->gc_mode = gc_mode;
2298 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2302 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2304 int retry = DEFAULT_RETRY_IO_COUNT;
2306 /* we should flush all the data to keep data consistency */
2308 sync_inodes_sb(sbi->sb);
2309 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2310 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2312 if (unlikely(retry < 0))
2313 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2315 f2fs_down_write(&sbi->gc_lock);
2316 f2fs_dirty_to_prefree(sbi);
2318 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2319 set_sbi_flag(sbi, SBI_IS_DIRTY);
2320 f2fs_up_write(&sbi->gc_lock);
2322 f2fs_sync_fs(sbi->sb, 1);
2324 /* Let's ensure there's no pending checkpoint anymore */
2325 f2fs_flush_ckpt_thread(sbi);
2328 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2330 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2331 struct f2fs_mount_info org_mount_opt;
2332 unsigned long old_sb_flags;
2334 bool need_restart_gc = false, need_stop_gc = false;
2335 bool need_restart_flush = false, need_stop_flush = false;
2336 bool need_restart_discard = false, need_stop_discard = false;
2337 bool need_enable_checkpoint = false, need_disable_checkpoint = false;
2338 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2339 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2340 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2341 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2342 bool no_atgc = !test_opt(sbi, ATGC);
2343 bool no_discard = !test_opt(sbi, DISCARD);
2344 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2345 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2351 * Save the old mount options in case we
2352 * need to restore them.
2354 org_mount_opt = sbi->mount_opt;
2355 old_sb_flags = sb->s_flags;
2358 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2359 for (i = 0; i < MAXQUOTAS; i++) {
2360 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2361 org_mount_opt.s_qf_names[i] =
2362 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2364 if (!org_mount_opt.s_qf_names[i]) {
2365 for (j = 0; j < i; j++)
2366 kfree(org_mount_opt.s_qf_names[j]);
2370 org_mount_opt.s_qf_names[i] = NULL;
2375 /* recover superblocks we couldn't write due to previous RO mount */
2376 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2377 err = f2fs_commit_super(sbi, false);
2378 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2381 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2384 default_options(sbi, true);
2386 /* parse mount options */
2387 err = parse_options(sb, data, true);
2391 /* flush outstanding errors before changing fs state */
2392 flush_work(&sbi->s_error_work);
2395 * Previous and new state of filesystem is RO,
2396 * so skip checking GC and FLUSH_MERGE conditions.
2398 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2401 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2407 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2408 err = dquot_suspend(sb, -1);
2411 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2412 /* dquot_resume needs RW */
2413 sb->s_flags &= ~SB_RDONLY;
2414 if (sb_any_quota_suspended(sb)) {
2415 dquot_resume(sb, -1);
2416 } else if (f2fs_sb_has_quota_ino(sbi)) {
2417 err = f2fs_enable_quotas(sb);
2423 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2425 f2fs_warn(sbi, "LFS is not compatible with IPU");
2429 /* disallow enable atgc dynamically */
2430 if (no_atgc == !!test_opt(sbi, ATGC)) {
2432 f2fs_warn(sbi, "switch atgc option is not allowed");
2436 /* disallow enable/disable extent_cache dynamically */
2437 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2439 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2442 /* disallow enable/disable age extent_cache dynamically */
2443 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2445 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2449 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2451 f2fs_warn(sbi, "switch io_bits option is not allowed");
2455 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2457 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2461 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2463 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2467 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2469 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2474 * We stop the GC thread if FS is mounted as RO
2475 * or if background_gc = off is passed in mount
2476 * option. Also sync the filesystem.
2478 if ((*flags & SB_RDONLY) ||
2479 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2480 !test_opt(sbi, GC_MERGE))) {
2481 if (sbi->gc_thread) {
2482 f2fs_stop_gc_thread(sbi);
2483 need_restart_gc = true;
2485 } else if (!sbi->gc_thread) {
2486 err = f2fs_start_gc_thread(sbi);
2489 need_stop_gc = true;
2492 if (*flags & SB_RDONLY) {
2495 set_sbi_flag(sbi, SBI_IS_DIRTY);
2496 set_sbi_flag(sbi, SBI_IS_CLOSE);
2497 f2fs_sync_fs(sb, 1);
2498 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2502 * We stop issue flush thread if FS is mounted as RO
2503 * or if flush_merge is not passed in mount option.
2505 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2506 clear_opt(sbi, FLUSH_MERGE);
2507 f2fs_destroy_flush_cmd_control(sbi, false);
2508 need_restart_flush = true;
2510 err = f2fs_create_flush_cmd_control(sbi);
2513 need_stop_flush = true;
2516 if (no_discard == !!test_opt(sbi, DISCARD)) {
2517 if (test_opt(sbi, DISCARD)) {
2518 err = f2fs_start_discard_thread(sbi);
2521 need_stop_discard = true;
2523 f2fs_stop_discard_thread(sbi);
2524 f2fs_issue_discard_timeout(sbi);
2525 need_restart_discard = true;
2529 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2530 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2531 err = f2fs_disable_checkpoint(sbi);
2533 goto restore_discard;
2534 need_enable_checkpoint = true;
2536 f2fs_enable_checkpoint(sbi);
2537 need_disable_checkpoint = true;
2542 * Place this routine at the end, since a new checkpoint would be
2543 * triggered while remount and we need to take care of it before
2544 * returning from remount.
2546 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2547 !test_opt(sbi, MERGE_CHECKPOINT)) {
2548 f2fs_stop_ckpt_thread(sbi);
2550 /* Flush if the prevous checkpoint, if exists. */
2551 f2fs_flush_ckpt_thread(sbi);
2553 err = f2fs_start_ckpt_thread(sbi);
2556 "Failed to start F2FS issue_checkpoint_thread (%d)",
2558 goto restore_checkpoint;
2564 /* Release old quota file names */
2565 for (i = 0; i < MAXQUOTAS; i++)
2566 kfree(org_mount_opt.s_qf_names[i]);
2568 /* Update the POSIXACL Flag */
2569 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2570 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2572 limit_reserve_root(sbi);
2573 adjust_unusable_cap_perc(sbi);
2574 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2577 if (need_enable_checkpoint) {
2578 f2fs_enable_checkpoint(sbi);
2579 } else if (need_disable_checkpoint) {
2580 if (f2fs_disable_checkpoint(sbi))
2581 f2fs_warn(sbi, "checkpoint has not been disabled");
2584 if (need_restart_discard) {
2585 if (f2fs_start_discard_thread(sbi))
2586 f2fs_warn(sbi, "discard has been stopped");
2587 } else if (need_stop_discard) {
2588 f2fs_stop_discard_thread(sbi);
2591 if (need_restart_flush) {
2592 if (f2fs_create_flush_cmd_control(sbi))
2593 f2fs_warn(sbi, "background flush thread has stopped");
2594 } else if (need_stop_flush) {
2595 clear_opt(sbi, FLUSH_MERGE);
2596 f2fs_destroy_flush_cmd_control(sbi, false);
2599 if (need_restart_gc) {
2600 if (f2fs_start_gc_thread(sbi))
2601 f2fs_warn(sbi, "background gc thread has stopped");
2602 } else if (need_stop_gc) {
2603 f2fs_stop_gc_thread(sbi);
2607 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2608 for (i = 0; i < MAXQUOTAS; i++) {
2609 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2610 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2613 sbi->mount_opt = org_mount_opt;
2614 sb->s_flags = old_sb_flags;
2619 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2621 /* need to recovery orphan */
2622 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2624 /* need to recovery data */
2625 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2627 if (test_opt(sbi, NORECOVERY))
2629 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2632 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2634 bool readonly = f2fs_readonly(sbi->sb);
2636 if (!f2fs_need_recovery(sbi))
2639 /* it doesn't need to check f2fs_sb_has_readonly() */
2640 if (f2fs_hw_is_readonly(sbi))
2644 sbi->sb->s_flags &= ~SB_RDONLY;
2645 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2649 * Turn on quotas which were not enabled for read-only mounts if
2650 * filesystem has quota feature, so that they are updated correctly.
2652 return f2fs_enable_quota_files(sbi, readonly);
2655 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2659 f2fs_quota_off_umount(sbi->sb);
2661 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2662 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2663 sbi->sb->s_flags |= SB_RDONLY;
2667 /* Read data from quotafile */
2668 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2669 size_t len, loff_t off)
2671 struct inode *inode = sb_dqopt(sb)->files[type];
2672 struct address_space *mapping = inode->i_mapping;
2673 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2674 int offset = off & (sb->s_blocksize - 1);
2677 loff_t i_size = i_size_read(inode);
2683 if (off + len > i_size)
2686 while (toread > 0) {
2687 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2689 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2691 if (PTR_ERR(page) == -ENOMEM) {
2692 memalloc_retry_wait(GFP_NOFS);
2695 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2696 return PTR_ERR(page);
2701 if (unlikely(page->mapping != mapping)) {
2702 f2fs_put_page(page, 1);
2705 if (unlikely(!PageUptodate(page))) {
2706 f2fs_put_page(page, 1);
2707 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2711 memcpy_from_page(data, page, offset, tocopy);
2712 f2fs_put_page(page, 1);
2722 /* Write to quotafile */
2723 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2724 const char *data, size_t len, loff_t off)
2726 struct inode *inode = sb_dqopt(sb)->files[type];
2727 struct address_space *mapping = inode->i_mapping;
2728 const struct address_space_operations *a_ops = mapping->a_ops;
2729 int offset = off & (sb->s_blocksize - 1);
2730 size_t towrite = len;
2732 void *fsdata = NULL;
2736 while (towrite > 0) {
2737 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2740 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2742 if (unlikely(err)) {
2743 if (err == -ENOMEM) {
2744 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2747 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2751 memcpy_to_page(page, offset, data, tocopy);
2753 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2764 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2765 f2fs_mark_inode_dirty_sync(inode, false);
2766 return len - towrite;
2769 int f2fs_dquot_initialize(struct inode *inode)
2771 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2774 return dquot_initialize(inode);
2777 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2779 return F2FS_I(inode)->i_dquot;
2782 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2784 return &F2FS_I(inode)->i_reserved_quota;
2787 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2789 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2790 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2794 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2795 F2FS_OPTION(sbi).s_jquota_fmt, type);
2798 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2803 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2804 err = f2fs_enable_quotas(sbi->sb);
2806 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2812 for (i = 0; i < MAXQUOTAS; i++) {
2813 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2814 err = f2fs_quota_on_mount(sbi, i);
2819 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2826 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2829 struct inode *qf_inode;
2830 unsigned long qf_inum;
2831 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2834 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2836 qf_inum = f2fs_qf_ino(sb, type);
2840 qf_inode = f2fs_iget(sb, qf_inum);
2841 if (IS_ERR(qf_inode)) {
2842 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2843 return PTR_ERR(qf_inode);
2846 /* Don't account quota for quota files to avoid recursion */
2847 inode_lock(qf_inode);
2848 qf_inode->i_flags |= S_NOQUOTA;
2850 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2851 F2FS_I(qf_inode)->i_flags |= qf_flag;
2852 f2fs_set_inode_flags(qf_inode);
2854 inode_unlock(qf_inode);
2856 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2861 static int f2fs_enable_quotas(struct super_block *sb)
2863 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2865 unsigned long qf_inum;
2866 bool quota_mopt[MAXQUOTAS] = {
2867 test_opt(sbi, USRQUOTA),
2868 test_opt(sbi, GRPQUOTA),
2869 test_opt(sbi, PRJQUOTA),
2872 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2873 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2877 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2879 for (type = 0; type < MAXQUOTAS; type++) {
2880 qf_inum = f2fs_qf_ino(sb, type);
2882 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2883 DQUOT_USAGE_ENABLED |
2884 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2886 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2888 for (type--; type >= 0; type--)
2889 dquot_quota_off(sb, type);
2890 set_sbi_flag(F2FS_SB(sb),
2891 SBI_QUOTA_NEED_REPAIR);
2899 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2901 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2902 struct address_space *mapping = dqopt->files[type]->i_mapping;
2905 ret = dquot_writeback_dquots(sbi->sb, type);
2909 ret = filemap_fdatawrite(mapping);
2913 /* if we are using journalled quota */
2914 if (is_journalled_quota(sbi))
2917 ret = filemap_fdatawait(mapping);
2919 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2922 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2926 int f2fs_quota_sync(struct super_block *sb, int type)
2928 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2929 struct quota_info *dqopt = sb_dqopt(sb);
2934 * Now when everything is written we can discard the pagecache so
2935 * that userspace sees the changes.
2937 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2939 if (type != -1 && cnt != type)
2942 if (!sb_has_quota_active(sb, cnt))
2945 if (!f2fs_sb_has_quota_ino(sbi))
2946 inode_lock(dqopt->files[cnt]);
2951 * f2fs_down_read(quota_sem)
2952 * dquot_writeback_dquots()
2955 * f2fs_down_read(quota_sem)
2958 f2fs_down_read(&sbi->quota_sem);
2960 ret = f2fs_quota_sync_file(sbi, cnt);
2962 f2fs_up_read(&sbi->quota_sem);
2963 f2fs_unlock_op(sbi);
2965 if (!f2fs_sb_has_quota_ino(sbi))
2966 inode_unlock(dqopt->files[cnt]);
2974 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2975 const struct path *path)
2977 struct inode *inode;
2980 /* if quota sysfile exists, deny enabling quota with specific file */
2981 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2982 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2986 if (path->dentry->d_sb != sb)
2989 err = f2fs_quota_sync(sb, type);
2993 inode = d_inode(path->dentry);
2995 err = filemap_fdatawrite(inode->i_mapping);
2999 err = filemap_fdatawait(inode->i_mapping);
3003 err = dquot_quota_on(sb, type, format_id, path);
3008 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
3009 f2fs_set_inode_flags(inode);
3010 inode_unlock(inode);
3011 f2fs_mark_inode_dirty_sync(inode, false);
3016 static int __f2fs_quota_off(struct super_block *sb, int type)
3018 struct inode *inode = sb_dqopt(sb)->files[type];
3021 if (!inode || !igrab(inode))
3022 return dquot_quota_off(sb, type);
3024 err = f2fs_quota_sync(sb, type);
3028 err = dquot_quota_off(sb, type);
3029 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
3033 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
3034 f2fs_set_inode_flags(inode);
3035 inode_unlock(inode);
3036 f2fs_mark_inode_dirty_sync(inode, false);
3042 static int f2fs_quota_off(struct super_block *sb, int type)
3044 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3047 err = __f2fs_quota_off(sb, type);
3050 * quotactl can shutdown journalled quota, result in inconsistence
3051 * between quota record and fs data by following updates, tag the
3052 * flag to let fsck be aware of it.
3054 if (is_journalled_quota(sbi))
3055 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3059 void f2fs_quota_off_umount(struct super_block *sb)
3064 for (type = 0; type < MAXQUOTAS; type++) {
3065 err = __f2fs_quota_off(sb, type);
3067 int ret = dquot_quota_off(sb, type);
3069 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3071 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3075 * In case of checkpoint=disable, we must flush quota blocks.
3076 * This can cause NULL exception for node_inode in end_io, since
3077 * put_super already dropped it.
3079 sync_filesystem(sb);
3082 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3084 struct quota_info *dqopt = sb_dqopt(sb);
3087 for (type = 0; type < MAXQUOTAS; type++) {
3088 if (!dqopt->files[type])
3090 f2fs_inode_synced(dqopt->files[type]);
3094 static int f2fs_dquot_commit(struct dquot *dquot)
3096 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3099 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3100 ret = dquot_commit(dquot);
3102 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3103 f2fs_up_read(&sbi->quota_sem);
3107 static int f2fs_dquot_acquire(struct dquot *dquot)
3109 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3112 f2fs_down_read(&sbi->quota_sem);
3113 ret = dquot_acquire(dquot);
3115 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3116 f2fs_up_read(&sbi->quota_sem);
3120 static int f2fs_dquot_release(struct dquot *dquot)
3122 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3123 int ret = dquot_release(dquot);
3126 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3130 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3132 struct super_block *sb = dquot->dq_sb;
3133 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3134 int ret = dquot_mark_dquot_dirty(dquot);
3136 /* if we are using journalled quota */
3137 if (is_journalled_quota(sbi))
3138 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3143 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3145 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3146 int ret = dquot_commit_info(sb, type);
3149 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3153 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3155 *projid = F2FS_I(inode)->i_projid;
3159 static const struct dquot_operations f2fs_quota_operations = {
3160 .get_reserved_space = f2fs_get_reserved_space,
3161 .write_dquot = f2fs_dquot_commit,
3162 .acquire_dquot = f2fs_dquot_acquire,
3163 .release_dquot = f2fs_dquot_release,
3164 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
3165 .write_info = f2fs_dquot_commit_info,
3166 .alloc_dquot = dquot_alloc,
3167 .destroy_dquot = dquot_destroy,
3168 .get_projid = f2fs_get_projid,
3169 .get_next_id = dquot_get_next_id,
3172 static const struct quotactl_ops f2fs_quotactl_ops = {
3173 .quota_on = f2fs_quota_on,
3174 .quota_off = f2fs_quota_off,
3175 .quota_sync = f2fs_quota_sync,
3176 .get_state = dquot_get_state,
3177 .set_info = dquot_set_dqinfo,
3178 .get_dqblk = dquot_get_dqblk,
3179 .set_dqblk = dquot_set_dqblk,
3180 .get_nextdqblk = dquot_get_next_dqblk,
3183 int f2fs_dquot_initialize(struct inode *inode)
3188 int f2fs_quota_sync(struct super_block *sb, int type)
3193 void f2fs_quota_off_umount(struct super_block *sb)
3198 static const struct super_operations f2fs_sops = {
3199 .alloc_inode = f2fs_alloc_inode,
3200 .free_inode = f2fs_free_inode,
3201 .drop_inode = f2fs_drop_inode,
3202 .write_inode = f2fs_write_inode,
3203 .dirty_inode = f2fs_dirty_inode,
3204 .show_options = f2fs_show_options,
3206 .quota_read = f2fs_quota_read,
3207 .quota_write = f2fs_quota_write,
3208 .get_dquots = f2fs_get_dquots,
3210 .evict_inode = f2fs_evict_inode,
3211 .put_super = f2fs_put_super,
3212 .sync_fs = f2fs_sync_fs,
3213 .freeze_fs = f2fs_freeze,
3214 .unfreeze_fs = f2fs_unfreeze,
3215 .statfs = f2fs_statfs,
3216 .remount_fs = f2fs_remount,
3219 #ifdef CONFIG_FS_ENCRYPTION
3220 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3222 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3223 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3227 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3230 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3233 * Encrypting the root directory is not allowed because fsck
3234 * expects lost+found directory to exist and remain unencrypted
3235 * if LOST_FOUND feature is enabled.
3238 if (f2fs_sb_has_lost_found(sbi) &&
3239 inode->i_ino == F2FS_ROOT_INO(sbi))
3242 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3243 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3244 ctx, len, fs_data, XATTR_CREATE);
3247 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3249 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3252 static bool f2fs_has_stable_inodes(struct super_block *sb)
3257 static struct block_device **f2fs_get_devices(struct super_block *sb,
3258 unsigned int *num_devs)
3260 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3261 struct block_device **devs;
3264 if (!f2fs_is_multi_device(sbi))
3267 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3269 return ERR_PTR(-ENOMEM);
3271 for (i = 0; i < sbi->s_ndevs; i++)
3272 devs[i] = FDEV(i).bdev;
3273 *num_devs = sbi->s_ndevs;
3277 static const struct fscrypt_operations f2fs_cryptops = {
3278 .needs_bounce_pages = 1,
3279 .has_32bit_inodes = 1,
3280 .supports_subblock_data_units = 1,
3281 .legacy_key_prefix = "f2fs:",
3282 .get_context = f2fs_get_context,
3283 .set_context = f2fs_set_context,
3284 .get_dummy_policy = f2fs_get_dummy_policy,
3285 .empty_dir = f2fs_empty_dir,
3286 .has_stable_inodes = f2fs_has_stable_inodes,
3287 .get_devices = f2fs_get_devices,
3291 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3292 u64 ino, u32 generation)
3294 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3295 struct inode *inode;
3297 if (f2fs_check_nid_range(sbi, ino))
3298 return ERR_PTR(-ESTALE);
3301 * f2fs_iget isn't quite right if the inode is currently unallocated!
3302 * However f2fs_iget currently does appropriate checks to handle stale
3303 * inodes so everything is OK.
3305 inode = f2fs_iget(sb, ino);
3307 return ERR_CAST(inode);
3308 if (unlikely(generation && inode->i_generation != generation)) {
3309 /* we didn't find the right inode.. */
3311 return ERR_PTR(-ESTALE);
3316 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3317 int fh_len, int fh_type)
3319 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3320 f2fs_nfs_get_inode);
3323 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3324 int fh_len, int fh_type)
3326 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3327 f2fs_nfs_get_inode);
3330 static const struct export_operations f2fs_export_ops = {
3331 .encode_fh = generic_encode_ino32_fh,
3332 .fh_to_dentry = f2fs_fh_to_dentry,
3333 .fh_to_parent = f2fs_fh_to_parent,
3334 .get_parent = f2fs_get_parent,
3337 loff_t max_file_blocks(struct inode *inode)
3343 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3344 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3345 * space in inode.i_addr, it will be more safe to reassign
3349 if (inode && f2fs_compressed_file(inode))
3350 leaf_count = ADDRS_PER_BLOCK(inode);
3352 leaf_count = DEF_ADDRS_PER_BLOCK;
3354 /* two direct node blocks */
3355 result += (leaf_count * 2);
3357 /* two indirect node blocks */
3358 leaf_count *= NIDS_PER_BLOCK;
3359 result += (leaf_count * 2);
3361 /* one double indirect node block */
3362 leaf_count *= NIDS_PER_BLOCK;
3363 result += leaf_count;
3366 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with
3367 * a 4K crypto data unit, we must restrict the max filesize to what can
3368 * fit within U32_MAX + 1 data units.
3371 result = min(result, (((loff_t)U32_MAX + 1) * 4096) >> F2FS_BLKSIZE_BITS);
3376 static int __f2fs_commit_super(struct buffer_head *bh,
3377 struct f2fs_super_block *super)
3381 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3382 set_buffer_dirty(bh);
3385 /* it's rare case, we can do fua all the time */
3386 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3389 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3390 struct buffer_head *bh)
3392 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3393 (bh->b_data + F2FS_SUPER_OFFSET);
3394 struct super_block *sb = sbi->sb;
3395 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3396 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3397 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3398 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3399 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3400 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3401 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3402 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3403 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3404 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3405 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3406 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3407 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3408 u64 main_end_blkaddr = main_blkaddr +
3409 (segment_count_main << log_blocks_per_seg);
3410 u64 seg_end_blkaddr = segment0_blkaddr +
3411 (segment_count << log_blocks_per_seg);
3413 if (segment0_blkaddr != cp_blkaddr) {
3414 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3415 segment0_blkaddr, cp_blkaddr);
3419 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3421 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3422 cp_blkaddr, sit_blkaddr,
3423 segment_count_ckpt << log_blocks_per_seg);
3427 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3429 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3430 sit_blkaddr, nat_blkaddr,
3431 segment_count_sit << log_blocks_per_seg);
3435 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3437 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3438 nat_blkaddr, ssa_blkaddr,
3439 segment_count_nat << log_blocks_per_seg);
3443 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3445 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3446 ssa_blkaddr, main_blkaddr,
3447 segment_count_ssa << log_blocks_per_seg);
3451 if (main_end_blkaddr > seg_end_blkaddr) {
3452 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3453 main_blkaddr, seg_end_blkaddr,
3454 segment_count_main << log_blocks_per_seg);
3456 } else if (main_end_blkaddr < seg_end_blkaddr) {
3460 /* fix in-memory information all the time */
3461 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3462 segment0_blkaddr) >> log_blocks_per_seg);
3464 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3465 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3468 err = __f2fs_commit_super(bh, NULL);
3469 res = err ? "failed" : "done";
3471 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3472 res, main_blkaddr, seg_end_blkaddr,
3473 segment_count_main << log_blocks_per_seg);
3480 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3481 struct buffer_head *bh)
3483 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3484 block_t total_sections, blocks_per_seg;
3485 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3486 (bh->b_data + F2FS_SUPER_OFFSET);
3487 size_t crc_offset = 0;
3490 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3491 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3492 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3496 /* Check checksum_offset and crc in superblock */
3497 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3498 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3500 offsetof(struct f2fs_super_block, crc)) {
3501 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3503 return -EFSCORRUPTED;
3505 crc = le32_to_cpu(raw_super->crc);
3506 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3507 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3508 return -EFSCORRUPTED;
3512 /* Currently, support only 4KB block size */
3513 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3514 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3515 le32_to_cpu(raw_super->log_blocksize),
3517 return -EFSCORRUPTED;
3520 /* check log blocks per segment */
3521 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3522 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3523 le32_to_cpu(raw_super->log_blocks_per_seg));
3524 return -EFSCORRUPTED;
3527 /* Currently, support 512/1024/2048/4096/16K bytes sector size */
3528 if (le32_to_cpu(raw_super->log_sectorsize) >
3529 F2FS_MAX_LOG_SECTOR_SIZE ||
3530 le32_to_cpu(raw_super->log_sectorsize) <
3531 F2FS_MIN_LOG_SECTOR_SIZE) {
3532 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3533 le32_to_cpu(raw_super->log_sectorsize));
3534 return -EFSCORRUPTED;
3536 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3537 le32_to_cpu(raw_super->log_sectorsize) !=
3538 F2FS_MAX_LOG_SECTOR_SIZE) {
3539 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3540 le32_to_cpu(raw_super->log_sectors_per_block),
3541 le32_to_cpu(raw_super->log_sectorsize));
3542 return -EFSCORRUPTED;
3545 segment_count = le32_to_cpu(raw_super->segment_count);
3546 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3547 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3548 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3549 total_sections = le32_to_cpu(raw_super->section_count);
3551 /* blocks_per_seg should be 512, given the above check */
3552 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3554 if (segment_count > F2FS_MAX_SEGMENT ||
3555 segment_count < F2FS_MIN_SEGMENTS) {
3556 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3557 return -EFSCORRUPTED;
3560 if (total_sections > segment_count_main || total_sections < 1 ||
3561 segs_per_sec > segment_count || !segs_per_sec) {
3562 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3563 segment_count, total_sections, segs_per_sec);
3564 return -EFSCORRUPTED;
3567 if (segment_count_main != total_sections * segs_per_sec) {
3568 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3569 segment_count_main, total_sections, segs_per_sec);
3570 return -EFSCORRUPTED;
3573 if ((segment_count / segs_per_sec) < total_sections) {
3574 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3575 segment_count, segs_per_sec, total_sections);
3576 return -EFSCORRUPTED;
3579 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3580 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3581 segment_count, le64_to_cpu(raw_super->block_count));
3582 return -EFSCORRUPTED;
3585 if (RDEV(0).path[0]) {
3586 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3589 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3590 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3593 if (segment_count != dev_seg_count) {
3594 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3595 segment_count, dev_seg_count);
3596 return -EFSCORRUPTED;
3599 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3600 !bdev_is_zoned(sbi->sb->s_bdev)) {
3601 f2fs_info(sbi, "Zoned block device path is missing");
3602 return -EFSCORRUPTED;
3606 if (secs_per_zone > total_sections || !secs_per_zone) {
3607 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3608 secs_per_zone, total_sections);
3609 return -EFSCORRUPTED;
3611 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3612 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3613 (le32_to_cpu(raw_super->extension_count) +
3614 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3615 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3616 le32_to_cpu(raw_super->extension_count),
3617 raw_super->hot_ext_count,
3618 F2FS_MAX_EXTENSION);
3619 return -EFSCORRUPTED;
3622 if (le32_to_cpu(raw_super->cp_payload) >=
3623 (blocks_per_seg - F2FS_CP_PACKS -
3624 NR_CURSEG_PERSIST_TYPE)) {
3625 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3626 le32_to_cpu(raw_super->cp_payload),
3627 blocks_per_seg - F2FS_CP_PACKS -
3628 NR_CURSEG_PERSIST_TYPE);
3629 return -EFSCORRUPTED;
3632 /* check reserved ino info */
3633 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3634 le32_to_cpu(raw_super->meta_ino) != 2 ||
3635 le32_to_cpu(raw_super->root_ino) != 3) {
3636 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3637 le32_to_cpu(raw_super->node_ino),
3638 le32_to_cpu(raw_super->meta_ino),
3639 le32_to_cpu(raw_super->root_ino));
3640 return -EFSCORRUPTED;
3643 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3644 if (sanity_check_area_boundary(sbi, bh))
3645 return -EFSCORRUPTED;
3650 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3652 unsigned int total, fsmeta;
3653 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3654 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3655 unsigned int ovp_segments, reserved_segments;
3656 unsigned int main_segs, blocks_per_seg;
3657 unsigned int sit_segs, nat_segs;
3658 unsigned int sit_bitmap_size, nat_bitmap_size;
3659 unsigned int log_blocks_per_seg;
3660 unsigned int segment_count_main;
3661 unsigned int cp_pack_start_sum, cp_payload;
3662 block_t user_block_count, valid_user_blocks;
3663 block_t avail_node_count, valid_node_count;
3664 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3667 total = le32_to_cpu(raw_super->segment_count);
3668 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3669 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3671 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3673 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3674 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3676 if (unlikely(fsmeta >= total))
3679 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3680 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3682 if (!f2fs_sb_has_readonly(sbi) &&
3683 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3684 ovp_segments == 0 || reserved_segments == 0)) {
3685 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3688 user_block_count = le64_to_cpu(ckpt->user_block_count);
3689 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3690 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3691 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3692 if (!user_block_count || user_block_count >=
3693 segment_count_main << log_blocks_per_seg) {
3694 f2fs_err(sbi, "Wrong user_block_count: %u",
3699 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3700 if (valid_user_blocks > user_block_count) {
3701 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3702 valid_user_blocks, user_block_count);
3706 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3707 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3708 if (valid_node_count > avail_node_count) {
3709 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3710 valid_node_count, avail_node_count);
3714 main_segs = le32_to_cpu(raw_super->segment_count_main);
3715 blocks_per_seg = sbi->blocks_per_seg;
3717 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3718 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3719 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3722 if (f2fs_sb_has_readonly(sbi))
3725 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3726 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3727 le32_to_cpu(ckpt->cur_node_segno[j])) {
3728 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3730 le32_to_cpu(ckpt->cur_node_segno[i]));
3736 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3737 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3738 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3741 if (f2fs_sb_has_readonly(sbi))
3744 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3745 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3746 le32_to_cpu(ckpt->cur_data_segno[j])) {
3747 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3749 le32_to_cpu(ckpt->cur_data_segno[i]));
3754 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3755 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3756 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3757 le32_to_cpu(ckpt->cur_data_segno[j])) {
3758 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3760 le32_to_cpu(ckpt->cur_node_segno[i]));
3766 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3767 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3769 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3770 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3771 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3772 sit_bitmap_size, nat_bitmap_size);
3776 cp_pack_start_sum = __start_sum_addr(sbi);
3777 cp_payload = __cp_payload(sbi);
3778 if (cp_pack_start_sum < cp_payload + 1 ||
3779 cp_pack_start_sum > blocks_per_seg - 1 -
3780 NR_CURSEG_PERSIST_TYPE) {
3781 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3786 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3787 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3788 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3789 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3790 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3791 le32_to_cpu(ckpt->checksum_offset));
3795 nat_blocks = nat_segs << log_blocks_per_seg;
3796 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3797 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3798 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3799 (cp_payload + F2FS_CP_PACKS +
3800 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3801 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3802 cp_payload, nat_bits_blocks);
3806 if (unlikely(f2fs_cp_error(sbi))) {
3807 f2fs_err(sbi, "A bug case: need to run fsck");
3813 static void init_sb_info(struct f2fs_sb_info *sbi)
3815 struct f2fs_super_block *raw_super = sbi->raw_super;
3818 sbi->log_sectors_per_block =
3819 le32_to_cpu(raw_super->log_sectors_per_block);
3820 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3821 sbi->blocksize = BIT(sbi->log_blocksize);
3822 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3823 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3824 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3825 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3826 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3827 sbi->total_node_count =
3828 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3829 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3830 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3831 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3832 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3833 sbi->cur_victim_sec = NULL_SECNO;
3834 sbi->gc_mode = GC_NORMAL;
3835 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3836 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3837 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3838 sbi->migration_granularity = sbi->segs_per_sec;
3839 sbi->seq_file_ra_mul = MIN_RA_MUL;
3840 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3841 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3842 spin_lock_init(&sbi->gc_remaining_trials_lock);
3843 atomic64_set(&sbi->current_atomic_write, 0);
3845 sbi->dir_level = DEF_DIR_LEVEL;
3846 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3847 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3848 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3849 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3850 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3851 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3852 DEF_UMOUNT_DISCARD_TIMEOUT;
3853 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3855 for (i = 0; i < NR_COUNT_TYPE; i++)
3856 atomic_set(&sbi->nr_pages[i], 0);
3858 for (i = 0; i < META; i++)
3859 atomic_set(&sbi->wb_sync_req[i], 0);
3861 INIT_LIST_HEAD(&sbi->s_list);
3862 mutex_init(&sbi->umount_mutex);
3863 init_f2fs_rwsem(&sbi->io_order_lock);
3864 spin_lock_init(&sbi->cp_lock);
3866 sbi->dirty_device = 0;
3867 spin_lock_init(&sbi->dev_lock);
3869 init_f2fs_rwsem(&sbi->sb_lock);
3870 init_f2fs_rwsem(&sbi->pin_sem);
3873 static int init_percpu_info(struct f2fs_sb_info *sbi)
3877 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3881 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3883 goto err_valid_block;
3885 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3888 goto err_node_block;
3892 percpu_counter_destroy(&sbi->rf_node_block_count);
3894 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3898 #ifdef CONFIG_BLK_DEV_ZONED
3900 struct f2fs_report_zones_args {
3901 struct f2fs_sb_info *sbi;
3902 struct f2fs_dev_info *dev;
3905 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3908 struct f2fs_report_zones_args *rz_args = data;
3909 block_t unusable_blocks = (zone->len - zone->capacity) >>
3910 F2FS_LOG_SECTORS_PER_BLOCK;
3912 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3915 set_bit(idx, rz_args->dev->blkz_seq);
3916 if (!rz_args->sbi->unusable_blocks_per_sec) {
3917 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3920 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3921 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3927 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3929 struct block_device *bdev = FDEV(devi).bdev;
3930 sector_t nr_sectors = bdev_nr_sectors(bdev);
3931 struct f2fs_report_zones_args rep_zone_arg;
3935 if (!f2fs_sb_has_blkzoned(sbi))
3938 zone_sectors = bdev_zone_sectors(bdev);
3939 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3940 SECTOR_TO_BLOCK(zone_sectors))
3942 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3943 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3944 sbi->blocks_per_blkz);
3945 if (nr_sectors & (zone_sectors - 1))
3946 FDEV(devi).nr_blkz++;
3948 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3949 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3950 * sizeof(unsigned long),
3952 if (!FDEV(devi).blkz_seq)
3955 rep_zone_arg.sbi = sbi;
3956 rep_zone_arg.dev = &FDEV(devi);
3958 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3967 * Read f2fs raw super block.
3968 * Because we have two copies of super block, so read both of them
3969 * to get the first valid one. If any one of them is broken, we pass
3970 * them recovery flag back to the caller.
3972 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3973 struct f2fs_super_block **raw_super,
3974 int *valid_super_block, int *recovery)
3976 struct super_block *sb = sbi->sb;
3978 struct buffer_head *bh;
3979 struct f2fs_super_block *super;
3982 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3986 for (block = 0; block < 2; block++) {
3987 bh = sb_bread(sb, block);
3989 f2fs_err(sbi, "Unable to read %dth superblock",
3996 /* sanity checking of raw super */
3997 err = sanity_check_raw_super(sbi, bh);
3999 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
4007 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
4009 *valid_super_block = block;
4015 /* No valid superblock */
4024 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
4026 struct buffer_head *bh;
4030 if ((recover && f2fs_readonly(sbi->sb)) ||
4031 f2fs_hw_is_readonly(sbi)) {
4032 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
4036 /* we should update superblock crc here */
4037 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
4038 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
4039 offsetof(struct f2fs_super_block, crc));
4040 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
4043 /* write back-up superblock first */
4044 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
4047 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4050 /* if we are in recovery path, skip writing valid superblock */
4054 /* write current valid superblock */
4055 bh = sb_bread(sbi->sb, sbi->valid_super_block);
4058 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4063 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4065 unsigned long flags;
4067 spin_lock_irqsave(&sbi->error_lock, flags);
4068 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4069 sbi->stop_reason[reason]++;
4070 spin_unlock_irqrestore(&sbi->error_lock, flags);
4073 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4075 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4076 unsigned long flags;
4079 f2fs_down_write(&sbi->sb_lock);
4081 spin_lock_irqsave(&sbi->error_lock, flags);
4082 if (sbi->error_dirty) {
4083 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4085 sbi->error_dirty = false;
4087 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4088 spin_unlock_irqrestore(&sbi->error_lock, flags);
4090 err = f2fs_commit_super(sbi, false);
4092 f2fs_up_write(&sbi->sb_lock);
4094 f2fs_err(sbi, "f2fs_commit_super fails to record err:%d", err);
4097 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4099 unsigned long flags;
4101 spin_lock_irqsave(&sbi->error_lock, flags);
4102 if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4103 set_bit(flag, (unsigned long *)sbi->errors);
4104 sbi->error_dirty = true;
4106 spin_unlock_irqrestore(&sbi->error_lock, flags);
4109 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4111 unsigned long flags;
4112 bool need_update = false;
4114 spin_lock_irqsave(&sbi->error_lock, flags);
4115 if (sbi->error_dirty) {
4116 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4118 sbi->error_dirty = false;
4121 spin_unlock_irqrestore(&sbi->error_lock, flags);
4126 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4130 f2fs_down_write(&sbi->sb_lock);
4132 if (!f2fs_update_errors(sbi))
4135 err = f2fs_commit_super(sbi, false);
4137 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
4140 f2fs_up_write(&sbi->sb_lock);
4143 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4145 f2fs_save_errors(sbi, error);
4146 f2fs_record_errors(sbi, error);
4149 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4151 f2fs_save_errors(sbi, error);
4153 if (!sbi->error_dirty)
4155 if (!test_bit(error, (unsigned long *)sbi->errors))
4157 schedule_work(&sbi->s_error_work);
4160 static bool system_going_down(void)
4162 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4163 || system_state == SYSTEM_RESTART;
4166 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason,
4169 struct super_block *sb = sbi->sb;
4170 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4171 bool continue_fs = !shutdown &&
4172 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4174 set_ckpt_flags(sbi, CP_ERROR_FLAG);
4176 if (!f2fs_hw_is_readonly(sbi)) {
4177 save_stop_reason(sbi, reason);
4179 if (irq_context && !shutdown)
4180 schedule_work(&sbi->s_error_work);
4182 f2fs_record_stop_reason(sbi);
4186 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4187 * could panic during 'reboot -f' as the underlying device got already
4190 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4191 !shutdown && !system_going_down() &&
4192 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4193 panic("F2FS-fs (device %s): panic forced after error\n",
4197 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4199 /* continue filesystem operators if errors=continue */
4200 if (continue_fs || f2fs_readonly(sb))
4203 f2fs_warn(sbi, "Remounting filesystem read-only");
4205 * Make sure updated value of ->s_mount_flags will be visible before
4209 sb->s_flags |= SB_RDONLY;
4212 static void f2fs_record_error_work(struct work_struct *work)
4214 struct f2fs_sb_info *sbi = container_of(work,
4215 struct f2fs_sb_info, s_error_work);
4217 f2fs_record_stop_reason(sbi);
4220 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4222 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4223 unsigned int max_devices = MAX_DEVICES;
4224 unsigned int logical_blksize;
4225 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4228 /* Initialize single device information */
4229 if (!RDEV(0).path[0]) {
4230 if (!bdev_is_zoned(sbi->sb->s_bdev))
4236 * Initialize multiple devices information, or single
4237 * zoned block device information.
4239 sbi->devs = f2fs_kzalloc(sbi,
4240 array_size(max_devices,
4241 sizeof(struct f2fs_dev_info)),
4246 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4247 sbi->aligned_blksize = true;
4249 for (i = 0; i < max_devices; i++) {
4251 FDEV(0).bdev_handle = sbi->sb->s_bdev_handle;
4252 else if (!RDEV(i).path[0])
4255 if (max_devices > 1) {
4256 /* Multi-device mount */
4257 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4258 FDEV(i).total_segments =
4259 le32_to_cpu(RDEV(i).total_segments);
4261 FDEV(i).start_blk = 0;
4262 FDEV(i).end_blk = FDEV(i).start_blk +
4263 (FDEV(i).total_segments <<
4264 sbi->log_blocks_per_seg) - 1 +
4265 le32_to_cpu(raw_super->segment0_blkaddr);
4267 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4268 FDEV(i).end_blk = FDEV(i).start_blk +
4269 (FDEV(i).total_segments <<
4270 sbi->log_blocks_per_seg) - 1;
4271 FDEV(i).bdev_handle = bdev_open_by_path(
4272 FDEV(i).path, mode, sbi->sb, NULL);
4275 if (IS_ERR(FDEV(i).bdev_handle))
4276 return PTR_ERR(FDEV(i).bdev_handle);
4278 FDEV(i).bdev = FDEV(i).bdev_handle->bdev;
4279 /* to release errored devices */
4280 sbi->s_ndevs = i + 1;
4282 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4283 sbi->aligned_blksize = false;
4285 #ifdef CONFIG_BLK_DEV_ZONED
4286 if (bdev_is_zoned(FDEV(i).bdev)) {
4287 if (!f2fs_sb_has_blkzoned(sbi)) {
4288 f2fs_err(sbi, "Zoned block device feature not enabled");
4291 if (init_blkz_info(sbi, i)) {
4292 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4295 if (max_devices == 1)
4297 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)",
4299 FDEV(i).total_segments,
4300 FDEV(i).start_blk, FDEV(i).end_blk);
4304 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4306 FDEV(i).total_segments,
4307 FDEV(i).start_blk, FDEV(i).end_blk);
4310 "IO Block Size: %8ld KB", F2FS_IO_SIZE_KB(sbi));
4314 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4316 #if IS_ENABLED(CONFIG_UNICODE)
4317 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4318 const struct f2fs_sb_encodings *encoding_info;
4319 struct unicode_map *encoding;
4320 __u16 encoding_flags;
4322 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4323 if (!encoding_info) {
4325 "Encoding requested by superblock is unknown");
4329 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4330 encoding = utf8_load(encoding_info->version);
4331 if (IS_ERR(encoding)) {
4333 "can't mount with superblock charset: %s-%u.%u.%u "
4334 "not supported by the kernel. flags: 0x%x.",
4335 encoding_info->name,
4336 unicode_major(encoding_info->version),
4337 unicode_minor(encoding_info->version),
4338 unicode_rev(encoding_info->version),
4340 return PTR_ERR(encoding);
4342 f2fs_info(sbi, "Using encoding defined by superblock: "
4343 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4344 unicode_major(encoding_info->version),
4345 unicode_minor(encoding_info->version),
4346 unicode_rev(encoding_info->version),
4349 sbi->sb->s_encoding = encoding;
4350 sbi->sb->s_encoding_flags = encoding_flags;
4353 if (f2fs_sb_has_casefold(sbi)) {
4354 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4361 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4363 /* adjust parameters according to the volume size */
4364 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4365 if (f2fs_block_unit_discard(sbi))
4366 SM_I(sbi)->dcc_info->discard_granularity =
4367 MIN_DISCARD_GRANULARITY;
4368 if (!f2fs_lfs_mode(sbi))
4369 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4370 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4373 sbi->readdir_ra = true;
4376 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4378 struct f2fs_sb_info *sbi;
4379 struct f2fs_super_block *raw_super;
4382 bool skip_recovery = false, need_fsck = false;
4383 char *options = NULL;
4384 int recovery, i, valid_super_block;
4385 struct curseg_info *seg_i;
4388 bool quota_enabled = false;
4394 valid_super_block = -1;
4397 /* allocate memory for f2fs-specific super block info */
4398 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4404 /* initialize locks within allocated memory */
4405 init_f2fs_rwsem(&sbi->gc_lock);
4406 mutex_init(&sbi->writepages);
4407 init_f2fs_rwsem(&sbi->cp_global_sem);
4408 init_f2fs_rwsem(&sbi->node_write);
4409 init_f2fs_rwsem(&sbi->node_change);
4410 spin_lock_init(&sbi->stat_lock);
4411 init_f2fs_rwsem(&sbi->cp_rwsem);
4412 init_f2fs_rwsem(&sbi->quota_sem);
4413 init_waitqueue_head(&sbi->cp_wait);
4414 spin_lock_init(&sbi->error_lock);
4416 for (i = 0; i < NR_INODE_TYPE; i++) {
4417 INIT_LIST_HEAD(&sbi->inode_list[i]);
4418 spin_lock_init(&sbi->inode_lock[i]);
4420 mutex_init(&sbi->flush_lock);
4422 /* Load the checksum driver */
4423 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4424 if (IS_ERR(sbi->s_chksum_driver)) {
4425 f2fs_err(sbi, "Cannot load crc32 driver.");
4426 err = PTR_ERR(sbi->s_chksum_driver);
4427 sbi->s_chksum_driver = NULL;
4431 /* set a block size */
4432 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4433 f2fs_err(sbi, "unable to set blocksize");
4437 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4442 sb->s_fs_info = sbi;
4443 sbi->raw_super = raw_super;
4445 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4446 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4447 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4449 /* precompute checksum seed for metadata */
4450 if (f2fs_sb_has_inode_chksum(sbi))
4451 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4452 sizeof(raw_super->uuid));
4454 default_options(sbi, false);
4455 /* parse mount options */
4456 options = kstrdup((const char *)data, GFP_KERNEL);
4457 if (data && !options) {
4462 err = parse_options(sb, options, false);
4466 sb->s_maxbytes = max_file_blocks(NULL) <<
4467 le32_to_cpu(raw_super->log_blocksize);
4468 sb->s_max_links = F2FS_LINK_MAX;
4470 err = f2fs_setup_casefold(sbi);
4475 sb->dq_op = &f2fs_quota_operations;
4476 sb->s_qcop = &f2fs_quotactl_ops;
4477 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4479 if (f2fs_sb_has_quota_ino(sbi)) {
4480 for (i = 0; i < MAXQUOTAS; i++) {
4481 if (f2fs_qf_ino(sbi->sb, i))
4482 sbi->nquota_files++;
4487 sb->s_op = &f2fs_sops;
4488 #ifdef CONFIG_FS_ENCRYPTION
4489 sb->s_cop = &f2fs_cryptops;
4491 #ifdef CONFIG_FS_VERITY
4492 sb->s_vop = &f2fs_verityops;
4494 sb->s_xattr = f2fs_xattr_handlers;
4495 sb->s_export_op = &f2fs_export_ops;
4496 sb->s_magic = F2FS_SUPER_MAGIC;
4497 sb->s_time_gran = 1;
4498 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4499 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4500 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4501 sb->s_iflags |= SB_I_CGROUPWB;
4503 /* init f2fs-specific super block info */
4504 sbi->valid_super_block = valid_super_block;
4506 /* disallow all the data/node/meta page writes */
4507 set_sbi_flag(sbi, SBI_POR_DOING);
4509 err = f2fs_init_write_merge_io(sbi);
4515 err = f2fs_init_iostat(sbi);
4519 err = init_percpu_info(sbi);
4523 if (F2FS_IO_ALIGNED(sbi)) {
4524 sbi->write_io_dummy =
4525 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4526 if (!sbi->write_io_dummy) {
4532 /* init per sbi slab cache */
4533 err = f2fs_init_xattr_caches(sbi);
4536 err = f2fs_init_page_array_cache(sbi);
4538 goto free_xattr_cache;
4540 /* get an inode for meta space */
4541 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4542 if (IS_ERR(sbi->meta_inode)) {
4543 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4544 err = PTR_ERR(sbi->meta_inode);
4545 goto free_page_array_cache;
4548 err = f2fs_get_valid_checkpoint(sbi);
4550 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4551 goto free_meta_inode;
4554 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4555 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4556 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4557 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4558 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4561 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4562 set_sbi_flag(sbi, SBI_NEED_FSCK);
4564 /* Initialize device list */
4565 err = f2fs_scan_devices(sbi);
4567 f2fs_err(sbi, "Failed to find devices");
4571 err = f2fs_init_post_read_wq(sbi);
4573 f2fs_err(sbi, "Failed to initialize post read workqueue");
4577 sbi->total_valid_node_count =
4578 le32_to_cpu(sbi->ckpt->valid_node_count);
4579 percpu_counter_set(&sbi->total_valid_inode_count,
4580 le32_to_cpu(sbi->ckpt->valid_inode_count));
4581 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4582 sbi->total_valid_block_count =
4583 le64_to_cpu(sbi->ckpt->valid_block_count);
4584 sbi->last_valid_block_count = sbi->total_valid_block_count;
4585 sbi->reserved_blocks = 0;
4586 sbi->current_reserved_blocks = 0;
4587 limit_reserve_root(sbi);
4588 adjust_unusable_cap_perc(sbi);
4590 f2fs_init_extent_cache_info(sbi);
4592 f2fs_init_ino_entry_info(sbi);
4594 f2fs_init_fsync_node_info(sbi);
4596 /* setup checkpoint request control and start checkpoint issue thread */
4597 f2fs_init_ckpt_req_control(sbi);
4598 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4599 test_opt(sbi, MERGE_CHECKPOINT)) {
4600 err = f2fs_start_ckpt_thread(sbi);
4603 "Failed to start F2FS issue_checkpoint_thread (%d)",
4605 goto stop_ckpt_thread;
4609 /* setup f2fs internal modules */
4610 err = f2fs_build_segment_manager(sbi);
4612 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4616 err = f2fs_build_node_manager(sbi);
4618 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4623 err = adjust_reserved_segment(sbi);
4627 /* For write statistics */
4628 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4630 /* Read accumulated write IO statistics if exists */
4631 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4632 if (__exist_node_summaries(sbi))
4633 sbi->kbytes_written =
4634 le64_to_cpu(seg_i->journal->info.kbytes_written);
4636 f2fs_build_gc_manager(sbi);
4638 err = f2fs_build_stats(sbi);
4642 /* get an inode for node space */
4643 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4644 if (IS_ERR(sbi->node_inode)) {
4645 f2fs_err(sbi, "Failed to read node inode");
4646 err = PTR_ERR(sbi->node_inode);
4650 /* read root inode and dentry */
4651 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4653 f2fs_err(sbi, "Failed to read root inode");
4654 err = PTR_ERR(root);
4655 goto free_node_inode;
4657 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4658 !root->i_size || !root->i_nlink) {
4661 goto free_node_inode;
4664 sb->s_root = d_make_root(root); /* allocate root dentry */
4667 goto free_node_inode;
4670 err = f2fs_init_compress_inode(sbi);
4672 goto free_root_inode;
4674 err = f2fs_register_sysfs(sbi);
4676 goto free_compress_inode;
4679 /* Enable quota usage during mount */
4680 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4681 err = f2fs_enable_quotas(sb);
4683 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4686 quota_enabled = f2fs_recover_quota_begin(sbi);
4688 /* if there are any orphan inodes, free them */
4689 err = f2fs_recover_orphan_inodes(sbi);
4693 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4694 goto reset_checkpoint;
4696 /* recover fsynced data */
4697 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4698 !test_opt(sbi, NORECOVERY)) {
4700 * mount should be failed, when device has readonly mode, and
4701 * previous checkpoint was not done by clean system shutdown.
4703 if (f2fs_hw_is_readonly(sbi)) {
4704 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4705 err = f2fs_recover_fsync_data(sbi, true);
4708 f2fs_err(sbi, "Need to recover fsync data, but "
4709 "write access unavailable, please try "
4710 "mount w/ disable_roll_forward or norecovery");
4715 f2fs_info(sbi, "write access unavailable, skipping recovery");
4716 goto reset_checkpoint;
4720 set_sbi_flag(sbi, SBI_NEED_FSCK);
4723 goto reset_checkpoint;
4725 err = f2fs_recover_fsync_data(sbi, false);
4728 skip_recovery = true;
4730 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4735 err = f2fs_recover_fsync_data(sbi, true);
4737 if (!f2fs_readonly(sb) && err > 0) {
4739 f2fs_err(sbi, "Need to recover fsync data");
4745 f2fs_recover_quota_end(sbi, quota_enabled);
4749 * If the f2fs is not readonly and fsync data recovery succeeds,
4750 * check zoned block devices' write pointer consistency.
4752 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4753 err = f2fs_check_write_pointer(sbi);
4758 f2fs_init_inmem_curseg(sbi);
4760 /* f2fs_recover_fsync_data() cleared this already */
4761 clear_sbi_flag(sbi, SBI_POR_DOING);
4763 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4764 err = f2fs_disable_checkpoint(sbi);
4766 goto sync_free_meta;
4767 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4768 f2fs_enable_checkpoint(sbi);
4772 * If filesystem is not mounted as read-only then
4773 * do start the gc_thread.
4775 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4776 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4777 /* After POR, we can run background GC thread.*/
4778 err = f2fs_start_gc_thread(sbi);
4780 goto sync_free_meta;
4784 /* recover broken superblock */
4786 err = f2fs_commit_super(sbi, true);
4787 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4788 sbi->valid_super_block ? 1 : 2, err);
4791 f2fs_join_shrinker(sbi);
4793 f2fs_tuning_parameters(sbi);
4795 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4796 cur_cp_version(F2FS_CKPT(sbi)));
4797 f2fs_update_time(sbi, CP_TIME);
4798 f2fs_update_time(sbi, REQ_TIME);
4799 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4803 /* safe to flush all the data */
4804 sync_filesystem(sbi->sb);
4809 f2fs_truncate_quota_inode_pages(sb);
4810 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4811 f2fs_quota_off_umount(sbi->sb);
4814 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4815 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4816 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4817 * falls into an infinite loop in f2fs_sync_meta_pages().
4819 truncate_inode_pages_final(META_MAPPING(sbi));
4820 /* evict some inodes being cached by GC */
4822 f2fs_unregister_sysfs(sbi);
4823 free_compress_inode:
4824 f2fs_destroy_compress_inode(sbi);
4829 f2fs_release_ino_entry(sbi, true);
4830 truncate_inode_pages_final(NODE_MAPPING(sbi));
4831 iput(sbi->node_inode);
4832 sbi->node_inode = NULL;
4834 f2fs_destroy_stats(sbi);
4836 /* stop discard thread before destroying node manager */
4837 f2fs_stop_discard_thread(sbi);
4838 f2fs_destroy_node_manager(sbi);
4840 f2fs_destroy_segment_manager(sbi);
4842 f2fs_stop_ckpt_thread(sbi);
4843 /* flush s_error_work before sbi destroy */
4844 flush_work(&sbi->s_error_work);
4845 f2fs_destroy_post_read_wq(sbi);
4847 destroy_device_list(sbi);
4850 make_bad_inode(sbi->meta_inode);
4851 iput(sbi->meta_inode);
4852 sbi->meta_inode = NULL;
4853 free_page_array_cache:
4854 f2fs_destroy_page_array_cache(sbi);
4856 f2fs_destroy_xattr_caches(sbi);
4858 mempool_destroy(sbi->write_io_dummy);
4860 destroy_percpu_info(sbi);
4862 f2fs_destroy_iostat(sbi);
4864 for (i = 0; i < NR_PAGE_TYPE; i++)
4865 kvfree(sbi->write_io[i]);
4867 #if IS_ENABLED(CONFIG_UNICODE)
4868 utf8_unload(sb->s_encoding);
4869 sb->s_encoding = NULL;
4873 for (i = 0; i < MAXQUOTAS; i++)
4874 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4876 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4881 if (sbi->s_chksum_driver)
4882 crypto_free_shash(sbi->s_chksum_driver);
4884 sb->s_fs_info = NULL;
4886 /* give only one another chance */
4887 if (retry_cnt > 0 && skip_recovery) {
4889 shrink_dcache_sb(sb);
4895 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4896 const char *dev_name, void *data)
4898 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4901 static void kill_f2fs_super(struct super_block *sb)
4903 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4906 set_sbi_flag(sbi, SBI_IS_CLOSE);
4907 f2fs_stop_gc_thread(sbi);
4908 f2fs_stop_discard_thread(sbi);
4910 #ifdef CONFIG_F2FS_FS_COMPRESSION
4912 * latter evict_inode() can bypass checking and invalidating
4913 * compress inode cache.
4915 if (test_opt(sbi, COMPRESS_CACHE))
4916 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4919 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4920 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4921 struct cp_control cpc = {
4922 .reason = CP_UMOUNT,
4924 stat_inc_cp_call_count(sbi, TOTAL_CALL);
4925 f2fs_write_checkpoint(sbi, &cpc);
4928 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4929 sb->s_flags &= ~SB_RDONLY;
4931 kill_block_super(sb);
4932 /* Release block devices last, after fscrypt_destroy_keyring(). */
4934 destroy_device_list(sbi);
4936 sb->s_fs_info = NULL;
4940 static struct file_system_type f2fs_fs_type = {
4941 .owner = THIS_MODULE,
4943 .mount = f2fs_mount,
4944 .kill_sb = kill_f2fs_super,
4945 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4947 MODULE_ALIAS_FS("f2fs");
4949 static int __init init_inodecache(void)
4951 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4952 sizeof(struct f2fs_inode_info), 0,
4953 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4954 return f2fs_inode_cachep ? 0 : -ENOMEM;
4957 static void destroy_inodecache(void)
4960 * Make sure all delayed rcu free inodes are flushed before we
4964 kmem_cache_destroy(f2fs_inode_cachep);
4967 static int __init init_f2fs_fs(void)
4971 if (PAGE_SIZE != F2FS_BLKSIZE) {
4972 printk("F2FS not supported on PAGE_SIZE(%lu) != BLOCK_SIZE(%lu)\n",
4973 PAGE_SIZE, F2FS_BLKSIZE);
4977 err = init_inodecache();
4980 err = f2fs_create_node_manager_caches();
4982 goto free_inodecache;
4983 err = f2fs_create_segment_manager_caches();
4985 goto free_node_manager_caches;
4986 err = f2fs_create_checkpoint_caches();
4988 goto free_segment_manager_caches;
4989 err = f2fs_create_recovery_cache();
4991 goto free_checkpoint_caches;
4992 err = f2fs_create_extent_cache();
4994 goto free_recovery_cache;
4995 err = f2fs_create_garbage_collection_cache();
4997 goto free_extent_cache;
4998 err = f2fs_init_sysfs();
5000 goto free_garbage_collection_cache;
5001 err = f2fs_init_shrinker();
5004 err = register_filesystem(&f2fs_fs_type);
5007 f2fs_create_root_stats();
5008 err = f2fs_init_post_read_processing();
5010 goto free_root_stats;
5011 err = f2fs_init_iostat_processing();
5013 goto free_post_read;
5014 err = f2fs_init_bio_entry_cache();
5017 err = f2fs_init_bioset();
5019 goto free_bio_entry_cache;
5020 err = f2fs_init_compress_mempool();
5023 err = f2fs_init_compress_cache();
5025 goto free_compress_mempool;
5026 err = f2fs_create_casefold_cache();
5028 goto free_compress_cache;
5030 free_compress_cache:
5031 f2fs_destroy_compress_cache();
5032 free_compress_mempool:
5033 f2fs_destroy_compress_mempool();
5035 f2fs_destroy_bioset();
5036 free_bio_entry_cache:
5037 f2fs_destroy_bio_entry_cache();
5039 f2fs_destroy_iostat_processing();
5041 f2fs_destroy_post_read_processing();
5043 f2fs_destroy_root_stats();
5044 unregister_filesystem(&f2fs_fs_type);
5046 f2fs_exit_shrinker();
5049 free_garbage_collection_cache:
5050 f2fs_destroy_garbage_collection_cache();
5052 f2fs_destroy_extent_cache();
5053 free_recovery_cache:
5054 f2fs_destroy_recovery_cache();
5055 free_checkpoint_caches:
5056 f2fs_destroy_checkpoint_caches();
5057 free_segment_manager_caches:
5058 f2fs_destroy_segment_manager_caches();
5059 free_node_manager_caches:
5060 f2fs_destroy_node_manager_caches();
5062 destroy_inodecache();
5067 static void __exit exit_f2fs_fs(void)
5069 f2fs_destroy_casefold_cache();
5070 f2fs_destroy_compress_cache();
5071 f2fs_destroy_compress_mempool();
5072 f2fs_destroy_bioset();
5073 f2fs_destroy_bio_entry_cache();
5074 f2fs_destroy_iostat_processing();
5075 f2fs_destroy_post_read_processing();
5076 f2fs_destroy_root_stats();
5077 unregister_filesystem(&f2fs_fs_type);
5078 f2fs_exit_shrinker();
5080 f2fs_destroy_garbage_collection_cache();
5081 f2fs_destroy_extent_cache();
5082 f2fs_destroy_recovery_cache();
5083 f2fs_destroy_checkpoint_caches();
5084 f2fs_destroy_segment_manager_caches();
5085 f2fs_destroy_node_manager_caches();
5086 destroy_inodecache();
5089 module_init(init_f2fs_fs)
5090 module_exit(exit_f2fs_fs)
5092 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5093 MODULE_DESCRIPTION("Flash Friendly File System");
5094 MODULE_LICENSE("GPL");
5095 MODULE_SOFTDEP("pre: crc32");