1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
8 #include <linux/module.h>
9 #include <linux/init.h>
11 #include <linux/statfs.h>
12 #include <linux/buffer_head.h>
13 #include <linux/backing-dev.h>
14 #include <linux/kthread.h>
15 #include <linux/parser.h>
16 #include <linux/mount.h>
17 #include <linux/seq_file.h>
18 #include <linux/proc_fs.h>
19 #include <linux/random.h>
20 #include <linux/exportfs.h>
21 #include <linux/blkdev.h>
22 #include <linux/quotaops.h>
23 #include <linux/f2fs_fs.h>
24 #include <linux/sysfs.h>
25 #include <linux/quota.h>
26 #include <linux/unicode.h>
27 #include <linux/part_stat.h>
28 #include <linux/zstd.h>
29 #include <linux/lz4.h>
38 #define CREATE_TRACE_POINTS
39 #include <trace/events/f2fs.h>
41 static struct kmem_cache *f2fs_inode_cachep;
43 #ifdef CONFIG_F2FS_FAULT_INJECTION
45 const char *f2fs_fault_name[FAULT_MAX] = {
46 [FAULT_KMALLOC] = "kmalloc",
47 [FAULT_KVMALLOC] = "kvmalloc",
48 [FAULT_PAGE_ALLOC] = "page alloc",
49 [FAULT_PAGE_GET] = "page get",
50 [FAULT_ALLOC_NID] = "alloc nid",
51 [FAULT_ORPHAN] = "orphan",
52 [FAULT_BLOCK] = "no more block",
53 [FAULT_DIR_DEPTH] = "too big dir depth",
54 [FAULT_EVICT_INODE] = "evict_inode fail",
55 [FAULT_TRUNCATE] = "truncate fail",
56 [FAULT_READ_IO] = "read IO error",
57 [FAULT_CHECKPOINT] = "checkpoint error",
58 [FAULT_DISCARD] = "discard error",
59 [FAULT_WRITE_IO] = "write IO error",
60 [FAULT_SLAB_ALLOC] = "slab alloc",
63 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
66 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
69 atomic_set(&ffi->inject_ops, 0);
70 ffi->inject_rate = rate;
74 ffi->inject_type = type;
77 memset(ffi, 0, sizeof(struct f2fs_fault_info));
81 /* f2fs-wide shrinker description */
82 static struct shrinker f2fs_shrinker_info = {
83 .scan_objects = f2fs_shrink_scan,
84 .count_objects = f2fs_shrink_count,
85 .seeks = DEFAULT_SEEKS,
90 Opt_disable_roll_forward,
101 Opt_disable_ext_identify,
104 Opt_inline_xattr_size,
142 Opt_test_dummy_encryption,
144 Opt_checkpoint_disable,
145 Opt_checkpoint_disable_cap,
146 Opt_checkpoint_disable_cap_perc,
147 Opt_checkpoint_enable,
148 Opt_checkpoint_merge,
149 Opt_nocheckpoint_merge,
150 Opt_compress_algorithm,
151 Opt_compress_log_size,
152 Opt_compress_extension,
153 Opt_nocompress_extension,
164 static match_table_t f2fs_tokens = {
165 {Opt_gc_background, "background_gc=%s"},
166 {Opt_disable_roll_forward, "disable_roll_forward"},
167 {Opt_norecovery, "norecovery"},
168 {Opt_discard, "discard"},
169 {Opt_nodiscard, "nodiscard"},
170 {Opt_noheap, "no_heap"},
172 {Opt_user_xattr, "user_xattr"},
173 {Opt_nouser_xattr, "nouser_xattr"},
175 {Opt_noacl, "noacl"},
176 {Opt_active_logs, "active_logs=%u"},
177 {Opt_disable_ext_identify, "disable_ext_identify"},
178 {Opt_inline_xattr, "inline_xattr"},
179 {Opt_noinline_xattr, "noinline_xattr"},
180 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
181 {Opt_inline_data, "inline_data"},
182 {Opt_inline_dentry, "inline_dentry"},
183 {Opt_noinline_dentry, "noinline_dentry"},
184 {Opt_flush_merge, "flush_merge"},
185 {Opt_noflush_merge, "noflush_merge"},
186 {Opt_nobarrier, "nobarrier"},
187 {Opt_fastboot, "fastboot"},
188 {Opt_extent_cache, "extent_cache"},
189 {Opt_noextent_cache, "noextent_cache"},
190 {Opt_noinline_data, "noinline_data"},
191 {Opt_data_flush, "data_flush"},
192 {Opt_reserve_root, "reserve_root=%u"},
193 {Opt_resgid, "resgid=%u"},
194 {Opt_resuid, "resuid=%u"},
195 {Opt_mode, "mode=%s"},
196 {Opt_io_size_bits, "io_bits=%u"},
197 {Opt_fault_injection, "fault_injection=%u"},
198 {Opt_fault_type, "fault_type=%u"},
199 {Opt_lazytime, "lazytime"},
200 {Opt_nolazytime, "nolazytime"},
201 {Opt_quota, "quota"},
202 {Opt_noquota, "noquota"},
203 {Opt_usrquota, "usrquota"},
204 {Opt_grpquota, "grpquota"},
205 {Opt_prjquota, "prjquota"},
206 {Opt_usrjquota, "usrjquota=%s"},
207 {Opt_grpjquota, "grpjquota=%s"},
208 {Opt_prjjquota, "prjjquota=%s"},
209 {Opt_offusrjquota, "usrjquota="},
210 {Opt_offgrpjquota, "grpjquota="},
211 {Opt_offprjjquota, "prjjquota="},
212 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
213 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
214 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
215 {Opt_whint, "whint_mode=%s"},
216 {Opt_alloc, "alloc_mode=%s"},
217 {Opt_fsync, "fsync_mode=%s"},
218 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
219 {Opt_test_dummy_encryption, "test_dummy_encryption"},
220 {Opt_inlinecrypt, "inlinecrypt"},
221 {Opt_checkpoint_disable, "checkpoint=disable"},
222 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
223 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
224 {Opt_checkpoint_enable, "checkpoint=enable"},
225 {Opt_checkpoint_merge, "checkpoint_merge"},
226 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
227 {Opt_compress_algorithm, "compress_algorithm=%s"},
228 {Opt_compress_log_size, "compress_log_size=%u"},
229 {Opt_compress_extension, "compress_extension=%s"},
230 {Opt_nocompress_extension, "nocompress_extension=%s"},
231 {Opt_compress_chksum, "compress_chksum"},
232 {Opt_compress_mode, "compress_mode=%s"},
233 {Opt_compress_cache, "compress_cache"},
235 {Opt_gc_merge, "gc_merge"},
236 {Opt_nogc_merge, "nogc_merge"},
237 {Opt_discard_unit, "discard_unit=%s"},
241 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
243 struct va_format vaf;
249 level = printk_get_level(fmt);
250 vaf.fmt = printk_skip_level(fmt);
252 printk("%c%cF2FS-fs (%s): %pV\n",
253 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
258 #ifdef CONFIG_UNICODE
259 static const struct f2fs_sb_encodings {
263 } f2fs_sb_encoding_map[] = {
264 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
267 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
268 const struct f2fs_sb_encodings **encoding,
271 __u16 magic = le16_to_cpu(sb->s_encoding);
274 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
275 if (magic == f2fs_sb_encoding_map[i].magic)
278 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
281 *encoding = &f2fs_sb_encoding_map[i];
282 *flags = le16_to_cpu(sb->s_encoding_flags);
287 struct kmem_cache *f2fs_cf_name_slab;
288 static int __init f2fs_create_casefold_cache(void)
290 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
292 if (!f2fs_cf_name_slab)
297 static void f2fs_destroy_casefold_cache(void)
299 kmem_cache_destroy(f2fs_cf_name_slab);
302 static int __init f2fs_create_casefold_cache(void) { return 0; }
303 static void f2fs_destroy_casefold_cache(void) { }
306 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
308 block_t limit = min((sbi->user_block_count << 1) / 1000,
309 sbi->user_block_count - sbi->reserved_blocks);
312 if (test_opt(sbi, RESERVE_ROOT) &&
313 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
314 F2FS_OPTION(sbi).root_reserved_blocks = limit;
315 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
316 F2FS_OPTION(sbi).root_reserved_blocks);
318 if (!test_opt(sbi, RESERVE_ROOT) &&
319 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
320 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
321 !gid_eq(F2FS_OPTION(sbi).s_resgid,
322 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
323 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
324 from_kuid_munged(&init_user_ns,
325 F2FS_OPTION(sbi).s_resuid),
326 from_kgid_munged(&init_user_ns,
327 F2FS_OPTION(sbi).s_resgid));
330 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
332 if (!F2FS_OPTION(sbi).unusable_cap_perc)
335 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
336 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
338 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
339 F2FS_OPTION(sbi).unusable_cap_perc;
341 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
342 F2FS_OPTION(sbi).unusable_cap,
343 F2FS_OPTION(sbi).unusable_cap_perc);
346 static void init_once(void *foo)
348 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
350 inode_init_once(&fi->vfs_inode);
354 static const char * const quotatypes[] = INITQFNAMES;
355 #define QTYPE2NAME(t) (quotatypes[t])
356 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
359 struct f2fs_sb_info *sbi = F2FS_SB(sb);
363 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
364 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
367 if (f2fs_sb_has_quota_ino(sbi)) {
368 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
372 qname = match_strdup(args);
374 f2fs_err(sbi, "Not enough memory for storing quotafile name");
377 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
378 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
381 f2fs_err(sbi, "%s quota file already specified",
385 if (strchr(qname, '/')) {
386 f2fs_err(sbi, "quotafile must be on filesystem root");
389 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
397 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
399 struct f2fs_sb_info *sbi = F2FS_SB(sb);
401 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
402 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
405 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
406 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
410 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
413 * We do the test below only for project quotas. 'usrquota' and
414 * 'grpquota' mount options are allowed even without quota feature
415 * to support legacy quotas in quota files.
417 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
418 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
421 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
422 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
423 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
424 if (test_opt(sbi, USRQUOTA) &&
425 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
426 clear_opt(sbi, USRQUOTA);
428 if (test_opt(sbi, GRPQUOTA) &&
429 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
430 clear_opt(sbi, GRPQUOTA);
432 if (test_opt(sbi, PRJQUOTA) &&
433 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
434 clear_opt(sbi, PRJQUOTA);
436 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
437 test_opt(sbi, PRJQUOTA)) {
438 f2fs_err(sbi, "old and new quota format mixing");
442 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
443 f2fs_err(sbi, "journaled quota format not specified");
448 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
449 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
450 F2FS_OPTION(sbi).s_jquota_fmt = 0;
456 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
458 const substring_t *arg,
461 struct f2fs_sb_info *sbi = F2FS_SB(sb);
462 #ifdef CONFIG_FS_ENCRYPTION
465 if (!f2fs_sb_has_encrypt(sbi)) {
466 f2fs_err(sbi, "Encrypt feature is off");
471 * This mount option is just for testing, and it's not worthwhile to
472 * implement the extra complexity (e.g. RCU protection) that would be
473 * needed to allow it to be set or changed during remount. We do allow
474 * it to be specified during remount, but only if there is no change.
476 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
477 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
480 err = fscrypt_set_test_dummy_encryption(
481 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
485 "Can't change test_dummy_encryption on remount");
486 else if (err == -EINVAL)
487 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
490 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
494 f2fs_warn(sbi, "Test dummy encryption mode enabled");
496 f2fs_warn(sbi, "Test dummy encryption mount option ignored");
501 #ifdef CONFIG_F2FS_FS_COMPRESSION
503 * 1. The same extension name cannot not appear in both compress and non-compress extension
505 * 2. If the compress extension specifies all files, the types specified by the non-compress
506 * extension will be treated as special cases and will not be compressed.
507 * 3. Don't allow the non-compress extension specifies all files.
509 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
511 unsigned char (*ext)[F2FS_EXTENSION_LEN];
512 unsigned char (*noext)[F2FS_EXTENSION_LEN];
513 int ext_cnt, noext_cnt, index = 0, no_index = 0;
515 ext = F2FS_OPTION(sbi).extensions;
516 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
517 noext = F2FS_OPTION(sbi).noextensions;
518 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
523 for (no_index = 0; no_index < noext_cnt; no_index++) {
524 if (!strcasecmp("*", noext[no_index])) {
525 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
528 for (index = 0; index < ext_cnt; index++) {
529 if (!strcasecmp(ext[index], noext[no_index])) {
530 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
539 #ifdef CONFIG_F2FS_FS_LZ4
540 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
542 #ifdef CONFIG_F2FS_FS_LZ4HC
546 if (strlen(str) == 3) {
547 F2FS_OPTION(sbi).compress_level = 0;
551 #ifdef CONFIG_F2FS_FS_LZ4HC
555 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
558 if (kstrtouint(str + 1, 10, &level))
561 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
562 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
566 F2FS_OPTION(sbi).compress_level = level;
569 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
575 #ifdef CONFIG_F2FS_FS_ZSTD
576 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
581 if (strlen(str) == len) {
582 F2FS_OPTION(sbi).compress_level = 0;
589 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
592 if (kstrtouint(str + 1, 10, &level))
595 if (!level || level > ZSTD_maxCLevel()) {
596 f2fs_info(sbi, "invalid zstd compress level: %d", level);
600 F2FS_OPTION(sbi).compress_level = level;
606 static int parse_options(struct super_block *sb, char *options, bool is_remount)
608 struct f2fs_sb_info *sbi = F2FS_SB(sb);
609 substring_t args[MAX_OPT_ARGS];
610 #ifdef CONFIG_F2FS_FS_COMPRESSION
611 unsigned char (*ext)[F2FS_EXTENSION_LEN];
612 unsigned char (*noext)[F2FS_EXTENSION_LEN];
613 int ext_cnt, noext_cnt;
624 while ((p = strsep(&options, ",")) != NULL) {
630 * Initialize args struct so we know whether arg was
631 * found; some options take optional arguments.
633 args[0].to = args[0].from = NULL;
634 token = match_token(p, f2fs_tokens, args);
637 case Opt_gc_background:
638 name = match_strdup(&args[0]);
642 if (!strcmp(name, "on")) {
643 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
644 } else if (!strcmp(name, "off")) {
645 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
646 } else if (!strcmp(name, "sync")) {
647 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
654 case Opt_disable_roll_forward:
655 set_opt(sbi, DISABLE_ROLL_FORWARD);
658 /* this option mounts f2fs with ro */
659 set_opt(sbi, NORECOVERY);
660 if (!f2fs_readonly(sb))
664 if (!f2fs_hw_support_discard(sbi)) {
665 f2fs_warn(sbi, "device does not support discard");
668 set_opt(sbi, DISCARD);
671 if (f2fs_hw_should_discard(sbi)) {
672 f2fs_warn(sbi, "discard is required for zoned block devices");
675 clear_opt(sbi, DISCARD);
678 set_opt(sbi, NOHEAP);
681 clear_opt(sbi, NOHEAP);
683 #ifdef CONFIG_F2FS_FS_XATTR
685 set_opt(sbi, XATTR_USER);
687 case Opt_nouser_xattr:
688 clear_opt(sbi, XATTR_USER);
690 case Opt_inline_xattr:
691 set_opt(sbi, INLINE_XATTR);
693 case Opt_noinline_xattr:
694 clear_opt(sbi, INLINE_XATTR);
696 case Opt_inline_xattr_size:
697 if (args->from && match_int(args, &arg))
699 set_opt(sbi, INLINE_XATTR_SIZE);
700 F2FS_OPTION(sbi).inline_xattr_size = arg;
704 f2fs_info(sbi, "user_xattr options not supported");
706 case Opt_nouser_xattr:
707 f2fs_info(sbi, "nouser_xattr options not supported");
709 case Opt_inline_xattr:
710 f2fs_info(sbi, "inline_xattr options not supported");
712 case Opt_noinline_xattr:
713 f2fs_info(sbi, "noinline_xattr options not supported");
716 #ifdef CONFIG_F2FS_FS_POSIX_ACL
718 set_opt(sbi, POSIX_ACL);
721 clear_opt(sbi, POSIX_ACL);
725 f2fs_info(sbi, "acl options not supported");
728 f2fs_info(sbi, "noacl options not supported");
731 case Opt_active_logs:
732 if (args->from && match_int(args, &arg))
734 if (arg != 2 && arg != 4 &&
735 arg != NR_CURSEG_PERSIST_TYPE)
737 F2FS_OPTION(sbi).active_logs = arg;
739 case Opt_disable_ext_identify:
740 set_opt(sbi, DISABLE_EXT_IDENTIFY);
742 case Opt_inline_data:
743 set_opt(sbi, INLINE_DATA);
745 case Opt_inline_dentry:
746 set_opt(sbi, INLINE_DENTRY);
748 case Opt_noinline_dentry:
749 clear_opt(sbi, INLINE_DENTRY);
751 case Opt_flush_merge:
752 set_opt(sbi, FLUSH_MERGE);
754 case Opt_noflush_merge:
755 clear_opt(sbi, FLUSH_MERGE);
758 set_opt(sbi, NOBARRIER);
761 set_opt(sbi, FASTBOOT);
763 case Opt_extent_cache:
764 set_opt(sbi, EXTENT_CACHE);
766 case Opt_noextent_cache:
767 clear_opt(sbi, EXTENT_CACHE);
769 case Opt_noinline_data:
770 clear_opt(sbi, INLINE_DATA);
773 set_opt(sbi, DATA_FLUSH);
775 case Opt_reserve_root:
776 if (args->from && match_int(args, &arg))
778 if (test_opt(sbi, RESERVE_ROOT)) {
779 f2fs_info(sbi, "Preserve previous reserve_root=%u",
780 F2FS_OPTION(sbi).root_reserved_blocks);
782 F2FS_OPTION(sbi).root_reserved_blocks = arg;
783 set_opt(sbi, RESERVE_ROOT);
787 if (args->from && match_int(args, &arg))
789 uid = make_kuid(current_user_ns(), arg);
790 if (!uid_valid(uid)) {
791 f2fs_err(sbi, "Invalid uid value %d", arg);
794 F2FS_OPTION(sbi).s_resuid = uid;
797 if (args->from && match_int(args, &arg))
799 gid = make_kgid(current_user_ns(), arg);
800 if (!gid_valid(gid)) {
801 f2fs_err(sbi, "Invalid gid value %d", arg);
804 F2FS_OPTION(sbi).s_resgid = gid;
807 name = match_strdup(&args[0]);
811 if (!strcmp(name, "adaptive")) {
812 if (f2fs_sb_has_blkzoned(sbi)) {
813 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
817 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
818 } else if (!strcmp(name, "lfs")) {
819 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
820 } else if (!strcmp(name, "fragment:segment")) {
821 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
822 } else if (!strcmp(name, "fragment:block")) {
823 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
830 case Opt_io_size_bits:
831 if (args->from && match_int(args, &arg))
833 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
834 f2fs_warn(sbi, "Not support %d, larger than %d",
835 1 << arg, BIO_MAX_VECS);
838 F2FS_OPTION(sbi).write_io_size_bits = arg;
840 #ifdef CONFIG_F2FS_FAULT_INJECTION
841 case Opt_fault_injection:
842 if (args->from && match_int(args, &arg))
844 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
845 set_opt(sbi, FAULT_INJECTION);
849 if (args->from && match_int(args, &arg))
851 f2fs_build_fault_attr(sbi, 0, arg);
852 set_opt(sbi, FAULT_INJECTION);
855 case Opt_fault_injection:
856 f2fs_info(sbi, "fault_injection options not supported");
860 f2fs_info(sbi, "fault_type options not supported");
864 sb->s_flags |= SB_LAZYTIME;
867 sb->s_flags &= ~SB_LAZYTIME;
872 set_opt(sbi, USRQUOTA);
875 set_opt(sbi, GRPQUOTA);
878 set_opt(sbi, PRJQUOTA);
881 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
886 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
891 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
895 case Opt_offusrjquota:
896 ret = f2fs_clear_qf_name(sb, USRQUOTA);
900 case Opt_offgrpjquota:
901 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
905 case Opt_offprjjquota:
906 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
910 case Opt_jqfmt_vfsold:
911 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
913 case Opt_jqfmt_vfsv0:
914 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
916 case Opt_jqfmt_vfsv1:
917 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
920 clear_opt(sbi, QUOTA);
921 clear_opt(sbi, USRQUOTA);
922 clear_opt(sbi, GRPQUOTA);
923 clear_opt(sbi, PRJQUOTA);
933 case Opt_offusrjquota:
934 case Opt_offgrpjquota:
935 case Opt_offprjjquota:
936 case Opt_jqfmt_vfsold:
937 case Opt_jqfmt_vfsv0:
938 case Opt_jqfmt_vfsv1:
940 f2fs_info(sbi, "quota operations not supported");
944 name = match_strdup(&args[0]);
947 if (!strcmp(name, "user-based")) {
948 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
949 } else if (!strcmp(name, "off")) {
950 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
951 } else if (!strcmp(name, "fs-based")) {
952 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
960 name = match_strdup(&args[0]);
964 if (!strcmp(name, "default")) {
965 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
966 } else if (!strcmp(name, "reuse")) {
967 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
975 name = match_strdup(&args[0]);
978 if (!strcmp(name, "posix")) {
979 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
980 } else if (!strcmp(name, "strict")) {
981 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
982 } else if (!strcmp(name, "nobarrier")) {
983 F2FS_OPTION(sbi).fsync_mode =
984 FSYNC_MODE_NOBARRIER;
991 case Opt_test_dummy_encryption:
992 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
997 case Opt_inlinecrypt:
998 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
999 sb->s_flags |= SB_INLINECRYPT;
1001 f2fs_info(sbi, "inline encryption not supported");
1004 case Opt_checkpoint_disable_cap_perc:
1005 if (args->from && match_int(args, &arg))
1007 if (arg < 0 || arg > 100)
1009 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1010 set_opt(sbi, DISABLE_CHECKPOINT);
1012 case Opt_checkpoint_disable_cap:
1013 if (args->from && match_int(args, &arg))
1015 F2FS_OPTION(sbi).unusable_cap = arg;
1016 set_opt(sbi, DISABLE_CHECKPOINT);
1018 case Opt_checkpoint_disable:
1019 set_opt(sbi, DISABLE_CHECKPOINT);
1021 case Opt_checkpoint_enable:
1022 clear_opt(sbi, DISABLE_CHECKPOINT);
1024 case Opt_checkpoint_merge:
1025 set_opt(sbi, MERGE_CHECKPOINT);
1027 case Opt_nocheckpoint_merge:
1028 clear_opt(sbi, MERGE_CHECKPOINT);
1030 #ifdef CONFIG_F2FS_FS_COMPRESSION
1031 case Opt_compress_algorithm:
1032 if (!f2fs_sb_has_compression(sbi)) {
1033 f2fs_info(sbi, "Image doesn't support compression");
1036 name = match_strdup(&args[0]);
1039 if (!strcmp(name, "lzo")) {
1040 #ifdef CONFIG_F2FS_FS_LZO
1041 F2FS_OPTION(sbi).compress_level = 0;
1042 F2FS_OPTION(sbi).compress_algorithm =
1045 f2fs_info(sbi, "kernel doesn't support lzo compression");
1047 } else if (!strncmp(name, "lz4", 3)) {
1048 #ifdef CONFIG_F2FS_FS_LZ4
1049 ret = f2fs_set_lz4hc_level(sbi, name);
1054 F2FS_OPTION(sbi).compress_algorithm =
1057 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1059 } else if (!strncmp(name, "zstd", 4)) {
1060 #ifdef CONFIG_F2FS_FS_ZSTD
1061 ret = f2fs_set_zstd_level(sbi, name);
1066 F2FS_OPTION(sbi).compress_algorithm =
1069 f2fs_info(sbi, "kernel doesn't support zstd compression");
1071 } else if (!strcmp(name, "lzo-rle")) {
1072 #ifdef CONFIG_F2FS_FS_LZORLE
1073 F2FS_OPTION(sbi).compress_level = 0;
1074 F2FS_OPTION(sbi).compress_algorithm =
1077 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1085 case Opt_compress_log_size:
1086 if (!f2fs_sb_has_compression(sbi)) {
1087 f2fs_info(sbi, "Image doesn't support compression");
1090 if (args->from && match_int(args, &arg))
1092 if (arg < MIN_COMPRESS_LOG_SIZE ||
1093 arg > MAX_COMPRESS_LOG_SIZE) {
1095 "Compress cluster log size is out of range");
1098 F2FS_OPTION(sbi).compress_log_size = arg;
1100 case Opt_compress_extension:
1101 if (!f2fs_sb_has_compression(sbi)) {
1102 f2fs_info(sbi, "Image doesn't support compression");
1105 name = match_strdup(&args[0]);
1109 ext = F2FS_OPTION(sbi).extensions;
1110 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1112 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1113 ext_cnt >= COMPRESS_EXT_NUM) {
1115 "invalid extension length/number");
1120 strcpy(ext[ext_cnt], name);
1121 F2FS_OPTION(sbi).compress_ext_cnt++;
1124 case Opt_nocompress_extension:
1125 if (!f2fs_sb_has_compression(sbi)) {
1126 f2fs_info(sbi, "Image doesn't support compression");
1129 name = match_strdup(&args[0]);
1133 noext = F2FS_OPTION(sbi).noextensions;
1134 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1136 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1137 noext_cnt >= COMPRESS_EXT_NUM) {
1139 "invalid extension length/number");
1144 strcpy(noext[noext_cnt], name);
1145 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1148 case Opt_compress_chksum:
1149 F2FS_OPTION(sbi).compress_chksum = true;
1151 case Opt_compress_mode:
1152 name = match_strdup(&args[0]);
1155 if (!strcmp(name, "fs")) {
1156 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1157 } else if (!strcmp(name, "user")) {
1158 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1165 case Opt_compress_cache:
1166 set_opt(sbi, COMPRESS_CACHE);
1169 case Opt_compress_algorithm:
1170 case Opt_compress_log_size:
1171 case Opt_compress_extension:
1172 case Opt_nocompress_extension:
1173 case Opt_compress_chksum:
1174 case Opt_compress_mode:
1175 case Opt_compress_cache:
1176 f2fs_info(sbi, "compression options not supported");
1183 set_opt(sbi, GC_MERGE);
1185 case Opt_nogc_merge:
1186 clear_opt(sbi, GC_MERGE);
1188 case Opt_discard_unit:
1189 name = match_strdup(&args[0]);
1192 if (!strcmp(name, "block")) {
1193 F2FS_OPTION(sbi).discard_unit =
1195 } else if (!strcmp(name, "segment")) {
1196 F2FS_OPTION(sbi).discard_unit =
1197 DISCARD_UNIT_SEGMENT;
1198 } else if (!strcmp(name, "section")) {
1199 F2FS_OPTION(sbi).discard_unit =
1200 DISCARD_UNIT_SECTION;
1208 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1215 if (f2fs_check_quota_options(sbi))
1218 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1219 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1222 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1223 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1227 #ifndef CONFIG_UNICODE
1228 if (f2fs_sb_has_casefold(sbi)) {
1230 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1235 * The BLKZONED feature indicates that the drive was formatted with
1236 * zone alignment optimization. This is optional for host-aware
1237 * devices, but mandatory for host-managed zoned block devices.
1239 #ifndef CONFIG_BLK_DEV_ZONED
1240 if (f2fs_sb_has_blkzoned(sbi)) {
1241 f2fs_err(sbi, "Zoned block device support is not enabled");
1245 if (f2fs_sb_has_blkzoned(sbi)) {
1246 if (F2FS_OPTION(sbi).discard_unit !=
1247 DISCARD_UNIT_SECTION) {
1248 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1249 F2FS_OPTION(sbi).discard_unit =
1250 DISCARD_UNIT_SECTION;
1254 #ifdef CONFIG_F2FS_FS_COMPRESSION
1255 if (f2fs_test_compress_extension(sbi)) {
1256 f2fs_err(sbi, "invalid compress or nocompress extension");
1261 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1262 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1263 F2FS_IO_SIZE_KB(sbi));
1267 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1268 int min_size, max_size;
1270 if (!f2fs_sb_has_extra_attr(sbi) ||
1271 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1272 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1275 if (!test_opt(sbi, INLINE_XATTR)) {
1276 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1280 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1281 max_size = MAX_INLINE_XATTR_SIZE;
1283 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1284 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1285 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1286 min_size, max_size);
1291 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1292 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1296 /* Not pass down write hints if the number of active logs is lesser
1297 * than NR_CURSEG_PERSIST_TYPE.
1299 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE)
1300 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1302 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1303 f2fs_err(sbi, "Allow to mount readonly mode only");
1309 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1311 struct f2fs_inode_info *fi;
1313 fi = f2fs_kmem_cache_alloc(f2fs_inode_cachep,
1314 GFP_F2FS_ZERO, false, F2FS_SB(sb));
1318 init_once((void *) fi);
1320 /* Initialize f2fs-specific inode info */
1321 atomic_set(&fi->dirty_pages, 0);
1322 atomic_set(&fi->i_compr_blocks, 0);
1323 init_rwsem(&fi->i_sem);
1324 spin_lock_init(&fi->i_size_lock);
1325 INIT_LIST_HEAD(&fi->dirty_list);
1326 INIT_LIST_HEAD(&fi->gdirty_list);
1327 INIT_LIST_HEAD(&fi->inmem_ilist);
1328 INIT_LIST_HEAD(&fi->inmem_pages);
1329 mutex_init(&fi->inmem_lock);
1330 init_rwsem(&fi->i_gc_rwsem[READ]);
1331 init_rwsem(&fi->i_gc_rwsem[WRITE]);
1332 init_rwsem(&fi->i_xattr_sem);
1334 /* Will be used by directory only */
1335 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1337 return &fi->vfs_inode;
1340 static int f2fs_drop_inode(struct inode *inode)
1342 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1346 * during filesystem shutdown, if checkpoint is disabled,
1347 * drop useless meta/node dirty pages.
1349 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1350 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1351 inode->i_ino == F2FS_META_INO(sbi)) {
1352 trace_f2fs_drop_inode(inode, 1);
1358 * This is to avoid a deadlock condition like below.
1359 * writeback_single_inode(inode)
1360 * - f2fs_write_data_page
1361 * - f2fs_gc -> iput -> evict
1362 * - inode_wait_for_writeback(inode)
1364 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1365 if (!inode->i_nlink && !is_bad_inode(inode)) {
1366 /* to avoid evict_inode call simultaneously */
1367 atomic_inc(&inode->i_count);
1368 spin_unlock(&inode->i_lock);
1370 /* some remained atomic pages should discarded */
1371 if (f2fs_is_atomic_file(inode))
1372 f2fs_drop_inmem_pages(inode);
1374 /* should remain fi->extent_tree for writepage */
1375 f2fs_destroy_extent_node(inode);
1377 sb_start_intwrite(inode->i_sb);
1378 f2fs_i_size_write(inode, 0);
1380 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1381 inode, NULL, 0, DATA);
1382 truncate_inode_pages_final(inode->i_mapping);
1384 if (F2FS_HAS_BLOCKS(inode))
1385 f2fs_truncate(inode);
1387 sb_end_intwrite(inode->i_sb);
1389 spin_lock(&inode->i_lock);
1390 atomic_dec(&inode->i_count);
1392 trace_f2fs_drop_inode(inode, 0);
1395 ret = generic_drop_inode(inode);
1397 ret = fscrypt_drop_inode(inode);
1398 trace_f2fs_drop_inode(inode, ret);
1402 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1404 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1407 spin_lock(&sbi->inode_lock[DIRTY_META]);
1408 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1411 set_inode_flag(inode, FI_DIRTY_INODE);
1412 stat_inc_dirty_inode(sbi, DIRTY_META);
1414 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1415 list_add_tail(&F2FS_I(inode)->gdirty_list,
1416 &sbi->inode_list[DIRTY_META]);
1417 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1419 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1423 void f2fs_inode_synced(struct inode *inode)
1425 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1427 spin_lock(&sbi->inode_lock[DIRTY_META]);
1428 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1429 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1432 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1433 list_del_init(&F2FS_I(inode)->gdirty_list);
1434 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1436 clear_inode_flag(inode, FI_DIRTY_INODE);
1437 clear_inode_flag(inode, FI_AUTO_RECOVER);
1438 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1439 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1443 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1445 * We should call set_dirty_inode to write the dirty inode through write_inode.
1447 static void f2fs_dirty_inode(struct inode *inode, int flags)
1449 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1451 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1452 inode->i_ino == F2FS_META_INO(sbi))
1455 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1456 clear_inode_flag(inode, FI_AUTO_RECOVER);
1458 f2fs_inode_dirtied(inode, false);
1461 static void f2fs_free_inode(struct inode *inode)
1463 fscrypt_free_inode(inode);
1464 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1467 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1469 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1470 percpu_counter_destroy(&sbi->total_valid_inode_count);
1473 static void destroy_device_list(struct f2fs_sb_info *sbi)
1477 for (i = 0; i < sbi->s_ndevs; i++) {
1478 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1479 #ifdef CONFIG_BLK_DEV_ZONED
1480 kvfree(FDEV(i).blkz_seq);
1481 kfree(FDEV(i).zone_capacity_blocks);
1487 static void f2fs_put_super(struct super_block *sb)
1489 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1493 /* unregister procfs/sysfs entries in advance to avoid race case */
1494 f2fs_unregister_sysfs(sbi);
1496 f2fs_quota_off_umount(sb);
1498 /* prevent remaining shrinker jobs */
1499 mutex_lock(&sbi->umount_mutex);
1502 * flush all issued checkpoints and stop checkpoint issue thread.
1503 * after then, all checkpoints should be done by each process context.
1505 f2fs_stop_ckpt_thread(sbi);
1508 * We don't need to do checkpoint when superblock is clean.
1509 * But, the previous checkpoint was not done by umount, it needs to do
1510 * clean checkpoint again.
1512 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1513 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1514 struct cp_control cpc = {
1515 .reason = CP_UMOUNT,
1517 f2fs_write_checkpoint(sbi, &cpc);
1520 /* be sure to wait for any on-going discard commands */
1521 dropped = f2fs_issue_discard_timeout(sbi);
1523 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1524 !sbi->discard_blks && !dropped) {
1525 struct cp_control cpc = {
1526 .reason = CP_UMOUNT | CP_TRIMMED,
1528 f2fs_write_checkpoint(sbi, &cpc);
1532 * normally superblock is clean, so we need to release this.
1533 * In addition, EIO will skip do checkpoint, we need this as well.
1535 f2fs_release_ino_entry(sbi, true);
1537 f2fs_leave_shrinker(sbi);
1538 mutex_unlock(&sbi->umount_mutex);
1540 /* our cp_error case, we can wait for any writeback page */
1541 f2fs_flush_merged_writes(sbi);
1543 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1545 f2fs_bug_on(sbi, sbi->fsync_node_num);
1547 f2fs_destroy_compress_inode(sbi);
1549 iput(sbi->node_inode);
1550 sbi->node_inode = NULL;
1552 iput(sbi->meta_inode);
1553 sbi->meta_inode = NULL;
1556 * iput() can update stat information, if f2fs_write_checkpoint()
1557 * above failed with error.
1559 f2fs_destroy_stats(sbi);
1561 /* destroy f2fs internal modules */
1562 f2fs_destroy_node_manager(sbi);
1563 f2fs_destroy_segment_manager(sbi);
1565 f2fs_destroy_post_read_wq(sbi);
1569 sb->s_fs_info = NULL;
1570 if (sbi->s_chksum_driver)
1571 crypto_free_shash(sbi->s_chksum_driver);
1572 kfree(sbi->raw_super);
1574 destroy_device_list(sbi);
1575 f2fs_destroy_page_array_cache(sbi);
1576 f2fs_destroy_xattr_caches(sbi);
1577 mempool_destroy(sbi->write_io_dummy);
1579 for (i = 0; i < MAXQUOTAS; i++)
1580 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1582 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1583 destroy_percpu_info(sbi);
1584 f2fs_destroy_iostat(sbi);
1585 for (i = 0; i < NR_PAGE_TYPE; i++)
1586 kvfree(sbi->write_io[i]);
1587 #ifdef CONFIG_UNICODE
1588 utf8_unload(sb->s_encoding);
1593 int f2fs_sync_fs(struct super_block *sb, int sync)
1595 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1598 if (unlikely(f2fs_cp_error(sbi)))
1600 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1603 trace_f2fs_sync_fs(sb, sync);
1605 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1609 err = f2fs_issue_checkpoint(sbi);
1614 static int f2fs_freeze(struct super_block *sb)
1616 if (f2fs_readonly(sb))
1619 /* IO error happened before */
1620 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1623 /* must be clean, since sync_filesystem() was already called */
1624 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1627 /* ensure no checkpoint required */
1628 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list))
1633 static int f2fs_unfreeze(struct super_block *sb)
1639 static int f2fs_statfs_project(struct super_block *sb,
1640 kprojid_t projid, struct kstatfs *buf)
1643 struct dquot *dquot;
1647 qid = make_kqid_projid(projid);
1648 dquot = dqget(sb, qid);
1650 return PTR_ERR(dquot);
1651 spin_lock(&dquot->dq_dqb_lock);
1653 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1654 dquot->dq_dqb.dqb_bhardlimit);
1656 limit >>= sb->s_blocksize_bits;
1658 if (limit && buf->f_blocks > limit) {
1659 curblock = (dquot->dq_dqb.dqb_curspace +
1660 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1661 buf->f_blocks = limit;
1662 buf->f_bfree = buf->f_bavail =
1663 (buf->f_blocks > curblock) ?
1664 (buf->f_blocks - curblock) : 0;
1667 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1668 dquot->dq_dqb.dqb_ihardlimit);
1670 if (limit && buf->f_files > limit) {
1671 buf->f_files = limit;
1673 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1674 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1677 spin_unlock(&dquot->dq_dqb_lock);
1683 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1685 struct super_block *sb = dentry->d_sb;
1686 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1687 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1688 block_t total_count, user_block_count, start_count;
1689 u64 avail_node_count;
1691 total_count = le64_to_cpu(sbi->raw_super->block_count);
1692 user_block_count = sbi->user_block_count;
1693 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1694 buf->f_type = F2FS_SUPER_MAGIC;
1695 buf->f_bsize = sbi->blocksize;
1697 buf->f_blocks = total_count - start_count;
1698 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1699 sbi->current_reserved_blocks;
1701 spin_lock(&sbi->stat_lock);
1702 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1705 buf->f_bfree -= sbi->unusable_block_count;
1706 spin_unlock(&sbi->stat_lock);
1708 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1709 buf->f_bavail = buf->f_bfree -
1710 F2FS_OPTION(sbi).root_reserved_blocks;
1714 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1716 if (avail_node_count > user_block_count) {
1717 buf->f_files = user_block_count;
1718 buf->f_ffree = buf->f_bavail;
1720 buf->f_files = avail_node_count;
1721 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1725 buf->f_namelen = F2FS_NAME_LEN;
1726 buf->f_fsid = u64_to_fsid(id);
1729 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1730 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1731 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1737 static inline void f2fs_show_quota_options(struct seq_file *seq,
1738 struct super_block *sb)
1741 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1743 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1746 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1757 seq_printf(seq, ",jqfmt=%s", fmtname);
1760 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1761 seq_show_option(seq, "usrjquota",
1762 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1764 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1765 seq_show_option(seq, "grpjquota",
1766 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1768 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1769 seq_show_option(seq, "prjjquota",
1770 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1774 #ifdef CONFIG_F2FS_FS_COMPRESSION
1775 static inline void f2fs_show_compress_options(struct seq_file *seq,
1776 struct super_block *sb)
1778 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1782 if (!f2fs_sb_has_compression(sbi))
1785 switch (F2FS_OPTION(sbi).compress_algorithm) {
1795 case COMPRESS_LZORLE:
1796 algtype = "lzo-rle";
1799 seq_printf(seq, ",compress_algorithm=%s", algtype);
1801 if (F2FS_OPTION(sbi).compress_level)
1802 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1804 seq_printf(seq, ",compress_log_size=%u",
1805 F2FS_OPTION(sbi).compress_log_size);
1807 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1808 seq_printf(seq, ",compress_extension=%s",
1809 F2FS_OPTION(sbi).extensions[i]);
1812 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1813 seq_printf(seq, ",nocompress_extension=%s",
1814 F2FS_OPTION(sbi).noextensions[i]);
1817 if (F2FS_OPTION(sbi).compress_chksum)
1818 seq_puts(seq, ",compress_chksum");
1820 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1821 seq_printf(seq, ",compress_mode=%s", "fs");
1822 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1823 seq_printf(seq, ",compress_mode=%s", "user");
1825 if (test_opt(sbi, COMPRESS_CACHE))
1826 seq_puts(seq, ",compress_cache");
1830 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1832 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1834 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1835 seq_printf(seq, ",background_gc=%s", "sync");
1836 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1837 seq_printf(seq, ",background_gc=%s", "on");
1838 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1839 seq_printf(seq, ",background_gc=%s", "off");
1841 if (test_opt(sbi, GC_MERGE))
1842 seq_puts(seq, ",gc_merge");
1844 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1845 seq_puts(seq, ",disable_roll_forward");
1846 if (test_opt(sbi, NORECOVERY))
1847 seq_puts(seq, ",norecovery");
1848 if (test_opt(sbi, DISCARD))
1849 seq_puts(seq, ",discard");
1851 seq_puts(seq, ",nodiscard");
1852 if (test_opt(sbi, NOHEAP))
1853 seq_puts(seq, ",no_heap");
1855 seq_puts(seq, ",heap");
1856 #ifdef CONFIG_F2FS_FS_XATTR
1857 if (test_opt(sbi, XATTR_USER))
1858 seq_puts(seq, ",user_xattr");
1860 seq_puts(seq, ",nouser_xattr");
1861 if (test_opt(sbi, INLINE_XATTR))
1862 seq_puts(seq, ",inline_xattr");
1864 seq_puts(seq, ",noinline_xattr");
1865 if (test_opt(sbi, INLINE_XATTR_SIZE))
1866 seq_printf(seq, ",inline_xattr_size=%u",
1867 F2FS_OPTION(sbi).inline_xattr_size);
1869 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1870 if (test_opt(sbi, POSIX_ACL))
1871 seq_puts(seq, ",acl");
1873 seq_puts(seq, ",noacl");
1875 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1876 seq_puts(seq, ",disable_ext_identify");
1877 if (test_opt(sbi, INLINE_DATA))
1878 seq_puts(seq, ",inline_data");
1880 seq_puts(seq, ",noinline_data");
1881 if (test_opt(sbi, INLINE_DENTRY))
1882 seq_puts(seq, ",inline_dentry");
1884 seq_puts(seq, ",noinline_dentry");
1885 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1886 seq_puts(seq, ",flush_merge");
1887 if (test_opt(sbi, NOBARRIER))
1888 seq_puts(seq, ",nobarrier");
1889 if (test_opt(sbi, FASTBOOT))
1890 seq_puts(seq, ",fastboot");
1891 if (test_opt(sbi, EXTENT_CACHE))
1892 seq_puts(seq, ",extent_cache");
1894 seq_puts(seq, ",noextent_cache");
1895 if (test_opt(sbi, DATA_FLUSH))
1896 seq_puts(seq, ",data_flush");
1898 seq_puts(seq, ",mode=");
1899 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1900 seq_puts(seq, "adaptive");
1901 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1902 seq_puts(seq, "lfs");
1903 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1904 seq_puts(seq, "fragment:segment");
1905 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1906 seq_puts(seq, "fragment:block");
1907 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1908 if (test_opt(sbi, RESERVE_ROOT))
1909 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1910 F2FS_OPTION(sbi).root_reserved_blocks,
1911 from_kuid_munged(&init_user_ns,
1912 F2FS_OPTION(sbi).s_resuid),
1913 from_kgid_munged(&init_user_ns,
1914 F2FS_OPTION(sbi).s_resgid));
1915 if (F2FS_IO_SIZE_BITS(sbi))
1916 seq_printf(seq, ",io_bits=%u",
1917 F2FS_OPTION(sbi).write_io_size_bits);
1918 #ifdef CONFIG_F2FS_FAULT_INJECTION
1919 if (test_opt(sbi, FAULT_INJECTION)) {
1920 seq_printf(seq, ",fault_injection=%u",
1921 F2FS_OPTION(sbi).fault_info.inject_rate);
1922 seq_printf(seq, ",fault_type=%u",
1923 F2FS_OPTION(sbi).fault_info.inject_type);
1927 if (test_opt(sbi, QUOTA))
1928 seq_puts(seq, ",quota");
1929 if (test_opt(sbi, USRQUOTA))
1930 seq_puts(seq, ",usrquota");
1931 if (test_opt(sbi, GRPQUOTA))
1932 seq_puts(seq, ",grpquota");
1933 if (test_opt(sbi, PRJQUOTA))
1934 seq_puts(seq, ",prjquota");
1936 f2fs_show_quota_options(seq, sbi->sb);
1937 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1938 seq_printf(seq, ",whint_mode=%s", "user-based");
1939 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1940 seq_printf(seq, ",whint_mode=%s", "fs-based");
1942 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1944 if (sbi->sb->s_flags & SB_INLINECRYPT)
1945 seq_puts(seq, ",inlinecrypt");
1947 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1948 seq_printf(seq, ",alloc_mode=%s", "default");
1949 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1950 seq_printf(seq, ",alloc_mode=%s", "reuse");
1952 if (test_opt(sbi, DISABLE_CHECKPOINT))
1953 seq_printf(seq, ",checkpoint=disable:%u",
1954 F2FS_OPTION(sbi).unusable_cap);
1955 if (test_opt(sbi, MERGE_CHECKPOINT))
1956 seq_puts(seq, ",checkpoint_merge");
1958 seq_puts(seq, ",nocheckpoint_merge");
1959 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1960 seq_printf(seq, ",fsync_mode=%s", "posix");
1961 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1962 seq_printf(seq, ",fsync_mode=%s", "strict");
1963 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1964 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1966 #ifdef CONFIG_F2FS_FS_COMPRESSION
1967 f2fs_show_compress_options(seq, sbi->sb);
1970 if (test_opt(sbi, ATGC))
1971 seq_puts(seq, ",atgc");
1973 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1974 seq_printf(seq, ",discard_unit=%s", "block");
1975 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1976 seq_printf(seq, ",discard_unit=%s", "segment");
1977 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1978 seq_printf(seq, ",discard_unit=%s", "section");
1983 static void default_options(struct f2fs_sb_info *sbi)
1985 /* init some FS parameters */
1986 if (f2fs_sb_has_readonly(sbi))
1987 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
1989 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
1991 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1992 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1993 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1994 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1995 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1996 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1997 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1998 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1999 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2000 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2001 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2003 sbi->sb->s_flags &= ~SB_INLINECRYPT;
2005 set_opt(sbi, INLINE_XATTR);
2006 set_opt(sbi, INLINE_DATA);
2007 set_opt(sbi, INLINE_DENTRY);
2008 set_opt(sbi, EXTENT_CACHE);
2009 set_opt(sbi, NOHEAP);
2010 clear_opt(sbi, DISABLE_CHECKPOINT);
2011 set_opt(sbi, MERGE_CHECKPOINT);
2012 F2FS_OPTION(sbi).unusable_cap = 0;
2013 sbi->sb->s_flags |= SB_LAZYTIME;
2014 set_opt(sbi, FLUSH_MERGE);
2015 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2016 set_opt(sbi, DISCARD);
2017 if (f2fs_sb_has_blkzoned(sbi)) {
2018 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2019 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2021 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2022 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2025 #ifdef CONFIG_F2FS_FS_XATTR
2026 set_opt(sbi, XATTR_USER);
2028 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2029 set_opt(sbi, POSIX_ACL);
2032 f2fs_build_fault_attr(sbi, 0, 0);
2036 static int f2fs_enable_quotas(struct super_block *sb);
2039 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2041 unsigned int s_flags = sbi->sb->s_flags;
2042 struct cp_control cpc;
2047 if (s_flags & SB_RDONLY) {
2048 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2051 sbi->sb->s_flags |= SB_ACTIVE;
2053 f2fs_update_time(sbi, DISABLE_TIME);
2055 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2056 down_write(&sbi->gc_lock);
2057 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
2058 if (err == -ENODATA) {
2062 if (err && err != -EAGAIN)
2066 ret = sync_filesystem(sbi->sb);
2068 err = ret ? ret : err;
2072 unusable = f2fs_get_unusable_blocks(sbi);
2073 if (f2fs_disable_cp_again(sbi, unusable)) {
2078 down_write(&sbi->gc_lock);
2079 cpc.reason = CP_PAUSE;
2080 set_sbi_flag(sbi, SBI_CP_DISABLED);
2081 err = f2fs_write_checkpoint(sbi, &cpc);
2085 spin_lock(&sbi->stat_lock);
2086 sbi->unusable_block_count = unusable;
2087 spin_unlock(&sbi->stat_lock);
2090 up_write(&sbi->gc_lock);
2092 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2096 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2098 int retry = DEFAULT_RETRY_IO_COUNT;
2100 /* we should flush all the data to keep data consistency */
2102 sync_inodes_sb(sbi->sb);
2104 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
2105 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2107 if (unlikely(retry < 0))
2108 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2110 down_write(&sbi->gc_lock);
2111 f2fs_dirty_to_prefree(sbi);
2113 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2114 set_sbi_flag(sbi, SBI_IS_DIRTY);
2115 up_write(&sbi->gc_lock);
2117 f2fs_sync_fs(sbi->sb, 1);
2120 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2122 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2123 struct f2fs_mount_info org_mount_opt;
2124 unsigned long old_sb_flags;
2126 bool need_restart_gc = false, need_stop_gc = false;
2127 bool need_restart_ckpt = false, need_stop_ckpt = false;
2128 bool need_restart_flush = false, need_stop_flush = false;
2129 bool need_restart_discard = false, need_stop_discard = false;
2130 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
2131 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2132 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2133 bool no_atgc = !test_opt(sbi, ATGC);
2134 bool no_discard = !test_opt(sbi, DISCARD);
2135 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2136 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2137 struct discard_cmd_control *dcc;
2143 * Save the old mount options in case we
2144 * need to restore them.
2146 org_mount_opt = sbi->mount_opt;
2147 old_sb_flags = sb->s_flags;
2150 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2151 for (i = 0; i < MAXQUOTAS; i++) {
2152 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2153 org_mount_opt.s_qf_names[i] =
2154 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2156 if (!org_mount_opt.s_qf_names[i]) {
2157 for (j = 0; j < i; j++)
2158 kfree(org_mount_opt.s_qf_names[j]);
2162 org_mount_opt.s_qf_names[i] = NULL;
2167 /* recover superblocks we couldn't write due to previous RO mount */
2168 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2169 err = f2fs_commit_super(sbi, false);
2170 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2173 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2176 default_options(sbi);
2178 /* parse mount options */
2179 err = parse_options(sb, data, true);
2184 * Previous and new state of filesystem is RO,
2185 * so skip checking GC and FLUSH_MERGE conditions.
2187 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2190 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2196 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2197 err = dquot_suspend(sb, -1);
2200 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2201 /* dquot_resume needs RW */
2202 sb->s_flags &= ~SB_RDONLY;
2203 if (sb_any_quota_suspended(sb)) {
2204 dquot_resume(sb, -1);
2205 } else if (f2fs_sb_has_quota_ino(sbi)) {
2206 err = f2fs_enable_quotas(sb);
2212 /* disallow enable atgc dynamically */
2213 if (no_atgc == !!test_opt(sbi, ATGC)) {
2215 f2fs_warn(sbi, "switch atgc option is not allowed");
2219 /* disallow enable/disable extent_cache dynamically */
2220 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
2222 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2226 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2228 f2fs_warn(sbi, "switch io_bits option is not allowed");
2232 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2234 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2238 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2240 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2244 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2246 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2251 * We stop the GC thread if FS is mounted as RO
2252 * or if background_gc = off is passed in mount
2253 * option. Also sync the filesystem.
2255 if ((*flags & SB_RDONLY) ||
2256 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2257 !test_opt(sbi, GC_MERGE))) {
2258 if (sbi->gc_thread) {
2259 f2fs_stop_gc_thread(sbi);
2260 need_restart_gc = true;
2262 } else if (!sbi->gc_thread) {
2263 err = f2fs_start_gc_thread(sbi);
2266 need_stop_gc = true;
2269 if (*flags & SB_RDONLY ||
2270 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
2273 set_sbi_flag(sbi, SBI_IS_DIRTY);
2274 set_sbi_flag(sbi, SBI_IS_CLOSE);
2275 f2fs_sync_fs(sb, 1);
2276 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2279 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2280 !test_opt(sbi, MERGE_CHECKPOINT)) {
2281 f2fs_stop_ckpt_thread(sbi);
2282 need_restart_ckpt = true;
2284 err = f2fs_start_ckpt_thread(sbi);
2287 "Failed to start F2FS issue_checkpoint_thread (%d)",
2291 need_stop_ckpt = true;
2295 * We stop issue flush thread if FS is mounted as RO
2296 * or if flush_merge is not passed in mount option.
2298 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2299 clear_opt(sbi, FLUSH_MERGE);
2300 f2fs_destroy_flush_cmd_control(sbi, false);
2301 need_restart_flush = true;
2303 err = f2fs_create_flush_cmd_control(sbi);
2306 need_stop_flush = true;
2309 if (no_discard == !!test_opt(sbi, DISCARD)) {
2310 if (test_opt(sbi, DISCARD)) {
2311 err = f2fs_start_discard_thread(sbi);
2314 need_stop_discard = true;
2316 dcc = SM_I(sbi)->dcc_info;
2317 f2fs_stop_discard_thread(sbi);
2318 if (atomic_read(&dcc->discard_cmd_cnt))
2319 f2fs_issue_discard_timeout(sbi);
2320 need_restart_discard = true;
2324 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2325 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2326 err = f2fs_disable_checkpoint(sbi);
2328 goto restore_discard;
2330 f2fs_enable_checkpoint(sbi);
2336 /* Release old quota file names */
2337 for (i = 0; i < MAXQUOTAS; i++)
2338 kfree(org_mount_opt.s_qf_names[i]);
2340 /* Update the POSIXACL Flag */
2341 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2342 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2344 limit_reserve_root(sbi);
2345 adjust_unusable_cap_perc(sbi);
2346 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2349 if (need_restart_discard) {
2350 if (f2fs_start_discard_thread(sbi))
2351 f2fs_warn(sbi, "discard has been stopped");
2352 } else if (need_stop_discard) {
2353 f2fs_stop_discard_thread(sbi);
2356 if (need_restart_flush) {
2357 if (f2fs_create_flush_cmd_control(sbi))
2358 f2fs_warn(sbi, "background flush thread has stopped");
2359 } else if (need_stop_flush) {
2360 clear_opt(sbi, FLUSH_MERGE);
2361 f2fs_destroy_flush_cmd_control(sbi, false);
2364 if (need_restart_ckpt) {
2365 if (f2fs_start_ckpt_thread(sbi))
2366 f2fs_warn(sbi, "background ckpt thread has stopped");
2367 } else if (need_stop_ckpt) {
2368 f2fs_stop_ckpt_thread(sbi);
2371 if (need_restart_gc) {
2372 if (f2fs_start_gc_thread(sbi))
2373 f2fs_warn(sbi, "background gc thread has stopped");
2374 } else if (need_stop_gc) {
2375 f2fs_stop_gc_thread(sbi);
2379 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2380 for (i = 0; i < MAXQUOTAS; i++) {
2381 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2382 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2385 sbi->mount_opt = org_mount_opt;
2386 sb->s_flags = old_sb_flags;
2391 /* Read data from quotafile */
2392 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2393 size_t len, loff_t off)
2395 struct inode *inode = sb_dqopt(sb)->files[type];
2396 struct address_space *mapping = inode->i_mapping;
2397 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2398 int offset = off & (sb->s_blocksize - 1);
2401 loff_t i_size = i_size_read(inode);
2408 if (off + len > i_size)
2411 while (toread > 0) {
2412 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2414 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2416 if (PTR_ERR(page) == -ENOMEM) {
2417 congestion_wait(BLK_RW_ASYNC,
2418 DEFAULT_IO_TIMEOUT);
2421 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2422 return PTR_ERR(page);
2427 if (unlikely(page->mapping != mapping)) {
2428 f2fs_put_page(page, 1);
2431 if (unlikely(!PageUptodate(page))) {
2432 f2fs_put_page(page, 1);
2433 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2437 kaddr = kmap_atomic(page);
2438 memcpy(data, kaddr + offset, tocopy);
2439 kunmap_atomic(kaddr);
2440 f2fs_put_page(page, 1);
2450 /* Write to quotafile */
2451 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2452 const char *data, size_t len, loff_t off)
2454 struct inode *inode = sb_dqopt(sb)->files[type];
2455 struct address_space *mapping = inode->i_mapping;
2456 const struct address_space_operations *a_ops = mapping->a_ops;
2457 int offset = off & (sb->s_blocksize - 1);
2458 size_t towrite = len;
2460 void *fsdata = NULL;
2465 while (towrite > 0) {
2466 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2469 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2471 if (unlikely(err)) {
2472 if (err == -ENOMEM) {
2473 congestion_wait(BLK_RW_ASYNC,
2474 DEFAULT_IO_TIMEOUT);
2477 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2481 kaddr = kmap_atomic(page);
2482 memcpy(kaddr + offset, data, tocopy);
2483 kunmap_atomic(kaddr);
2484 flush_dcache_page(page);
2486 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2497 inode->i_mtime = inode->i_ctime = current_time(inode);
2498 f2fs_mark_inode_dirty_sync(inode, false);
2499 return len - towrite;
2502 static struct dquot **f2fs_get_dquots(struct inode *inode)
2504 return F2FS_I(inode)->i_dquot;
2507 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2509 return &F2FS_I(inode)->i_reserved_quota;
2512 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2514 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2515 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2519 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2520 F2FS_OPTION(sbi).s_jquota_fmt, type);
2523 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2528 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2529 err = f2fs_enable_quotas(sbi->sb);
2531 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2537 for (i = 0; i < MAXQUOTAS; i++) {
2538 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2539 err = f2fs_quota_on_mount(sbi, i);
2544 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2551 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2554 struct inode *qf_inode;
2555 unsigned long qf_inum;
2558 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2560 qf_inum = f2fs_qf_ino(sb, type);
2564 qf_inode = f2fs_iget(sb, qf_inum);
2565 if (IS_ERR(qf_inode)) {
2566 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2567 return PTR_ERR(qf_inode);
2570 /* Don't account quota for quota files to avoid recursion */
2571 qf_inode->i_flags |= S_NOQUOTA;
2572 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2577 static int f2fs_enable_quotas(struct super_block *sb)
2579 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2581 unsigned long qf_inum;
2582 bool quota_mopt[MAXQUOTAS] = {
2583 test_opt(sbi, USRQUOTA),
2584 test_opt(sbi, GRPQUOTA),
2585 test_opt(sbi, PRJQUOTA),
2588 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2589 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2593 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2595 for (type = 0; type < MAXQUOTAS; type++) {
2596 qf_inum = f2fs_qf_ino(sb, type);
2598 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2599 DQUOT_USAGE_ENABLED |
2600 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2602 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2604 for (type--; type >= 0; type--)
2605 dquot_quota_off(sb, type);
2606 set_sbi_flag(F2FS_SB(sb),
2607 SBI_QUOTA_NEED_REPAIR);
2615 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2617 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2618 struct address_space *mapping = dqopt->files[type]->i_mapping;
2621 ret = dquot_writeback_dquots(sbi->sb, type);
2625 ret = filemap_fdatawrite(mapping);
2629 /* if we are using journalled quota */
2630 if (is_journalled_quota(sbi))
2633 ret = filemap_fdatawait(mapping);
2635 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2638 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2642 int f2fs_quota_sync(struct super_block *sb, int type)
2644 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2645 struct quota_info *dqopt = sb_dqopt(sb);
2650 * Now when everything is written we can discard the pagecache so
2651 * that userspace sees the changes.
2653 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2655 if (type != -1 && cnt != type)
2658 if (!sb_has_quota_active(sb, type))
2661 inode_lock(dqopt->files[cnt]);
2666 * down_read(quota_sem)
2667 * dquot_writeback_dquots()
2670 * down_read(quota_sem)
2673 down_read(&sbi->quota_sem);
2675 ret = f2fs_quota_sync_file(sbi, cnt);
2677 up_read(&sbi->quota_sem);
2678 f2fs_unlock_op(sbi);
2680 inode_unlock(dqopt->files[cnt]);
2688 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2689 const struct path *path)
2691 struct inode *inode;
2694 /* if quota sysfile exists, deny enabling quota with specific file */
2695 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2696 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2700 err = f2fs_quota_sync(sb, type);
2704 err = dquot_quota_on(sb, type, format_id, path);
2708 inode = d_inode(path->dentry);
2711 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2712 f2fs_set_inode_flags(inode);
2713 inode_unlock(inode);
2714 f2fs_mark_inode_dirty_sync(inode, false);
2719 static int __f2fs_quota_off(struct super_block *sb, int type)
2721 struct inode *inode = sb_dqopt(sb)->files[type];
2724 if (!inode || !igrab(inode))
2725 return dquot_quota_off(sb, type);
2727 err = f2fs_quota_sync(sb, type);
2731 err = dquot_quota_off(sb, type);
2732 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2736 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2737 f2fs_set_inode_flags(inode);
2738 inode_unlock(inode);
2739 f2fs_mark_inode_dirty_sync(inode, false);
2745 static int f2fs_quota_off(struct super_block *sb, int type)
2747 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2750 err = __f2fs_quota_off(sb, type);
2753 * quotactl can shutdown journalled quota, result in inconsistence
2754 * between quota record and fs data by following updates, tag the
2755 * flag to let fsck be aware of it.
2757 if (is_journalled_quota(sbi))
2758 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2762 void f2fs_quota_off_umount(struct super_block *sb)
2767 for (type = 0; type < MAXQUOTAS; type++) {
2768 err = __f2fs_quota_off(sb, type);
2770 int ret = dquot_quota_off(sb, type);
2772 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2774 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2778 * In case of checkpoint=disable, we must flush quota blocks.
2779 * This can cause NULL exception for node_inode in end_io, since
2780 * put_super already dropped it.
2782 sync_filesystem(sb);
2785 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2787 struct quota_info *dqopt = sb_dqopt(sb);
2790 for (type = 0; type < MAXQUOTAS; type++) {
2791 if (!dqopt->files[type])
2793 f2fs_inode_synced(dqopt->files[type]);
2797 static int f2fs_dquot_commit(struct dquot *dquot)
2799 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2802 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2803 ret = dquot_commit(dquot);
2805 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2806 up_read(&sbi->quota_sem);
2810 static int f2fs_dquot_acquire(struct dquot *dquot)
2812 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2815 down_read(&sbi->quota_sem);
2816 ret = dquot_acquire(dquot);
2818 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2819 up_read(&sbi->quota_sem);
2823 static int f2fs_dquot_release(struct dquot *dquot)
2825 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2826 int ret = dquot_release(dquot);
2829 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2833 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2835 struct super_block *sb = dquot->dq_sb;
2836 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2837 int ret = dquot_mark_dquot_dirty(dquot);
2839 /* if we are using journalled quota */
2840 if (is_journalled_quota(sbi))
2841 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2846 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2848 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2849 int ret = dquot_commit_info(sb, type);
2852 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2856 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2858 *projid = F2FS_I(inode)->i_projid;
2862 static const struct dquot_operations f2fs_quota_operations = {
2863 .get_reserved_space = f2fs_get_reserved_space,
2864 .write_dquot = f2fs_dquot_commit,
2865 .acquire_dquot = f2fs_dquot_acquire,
2866 .release_dquot = f2fs_dquot_release,
2867 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2868 .write_info = f2fs_dquot_commit_info,
2869 .alloc_dquot = dquot_alloc,
2870 .destroy_dquot = dquot_destroy,
2871 .get_projid = f2fs_get_projid,
2872 .get_next_id = dquot_get_next_id,
2875 static const struct quotactl_ops f2fs_quotactl_ops = {
2876 .quota_on = f2fs_quota_on,
2877 .quota_off = f2fs_quota_off,
2878 .quota_sync = f2fs_quota_sync,
2879 .get_state = dquot_get_state,
2880 .set_info = dquot_set_dqinfo,
2881 .get_dqblk = dquot_get_dqblk,
2882 .set_dqblk = dquot_set_dqblk,
2883 .get_nextdqblk = dquot_get_next_dqblk,
2886 int f2fs_quota_sync(struct super_block *sb, int type)
2891 void f2fs_quota_off_umount(struct super_block *sb)
2896 static const struct super_operations f2fs_sops = {
2897 .alloc_inode = f2fs_alloc_inode,
2898 .free_inode = f2fs_free_inode,
2899 .drop_inode = f2fs_drop_inode,
2900 .write_inode = f2fs_write_inode,
2901 .dirty_inode = f2fs_dirty_inode,
2902 .show_options = f2fs_show_options,
2904 .quota_read = f2fs_quota_read,
2905 .quota_write = f2fs_quota_write,
2906 .get_dquots = f2fs_get_dquots,
2908 .evict_inode = f2fs_evict_inode,
2909 .put_super = f2fs_put_super,
2910 .sync_fs = f2fs_sync_fs,
2911 .freeze_fs = f2fs_freeze,
2912 .unfreeze_fs = f2fs_unfreeze,
2913 .statfs = f2fs_statfs,
2914 .remount_fs = f2fs_remount,
2917 #ifdef CONFIG_FS_ENCRYPTION
2918 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2920 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2921 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2925 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2928 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2931 * Encrypting the root directory is not allowed because fsck
2932 * expects lost+found directory to exist and remain unencrypted
2933 * if LOST_FOUND feature is enabled.
2936 if (f2fs_sb_has_lost_found(sbi) &&
2937 inode->i_ino == F2FS_ROOT_INO(sbi))
2940 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2941 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2942 ctx, len, fs_data, XATTR_CREATE);
2945 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
2947 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
2950 static bool f2fs_has_stable_inodes(struct super_block *sb)
2955 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2956 int *ino_bits_ret, int *lblk_bits_ret)
2958 *ino_bits_ret = 8 * sizeof(nid_t);
2959 *lblk_bits_ret = 8 * sizeof(block_t);
2962 static int f2fs_get_num_devices(struct super_block *sb)
2964 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2966 if (f2fs_is_multi_device(sbi))
2967 return sbi->s_ndevs;
2971 static void f2fs_get_devices(struct super_block *sb,
2972 struct request_queue **devs)
2974 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2977 for (i = 0; i < sbi->s_ndevs; i++)
2978 devs[i] = bdev_get_queue(FDEV(i).bdev);
2981 static const struct fscrypt_operations f2fs_cryptops = {
2982 .key_prefix = "f2fs:",
2983 .get_context = f2fs_get_context,
2984 .set_context = f2fs_set_context,
2985 .get_dummy_policy = f2fs_get_dummy_policy,
2986 .empty_dir = f2fs_empty_dir,
2987 .max_namelen = F2FS_NAME_LEN,
2988 .has_stable_inodes = f2fs_has_stable_inodes,
2989 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
2990 .get_num_devices = f2fs_get_num_devices,
2991 .get_devices = f2fs_get_devices,
2995 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2996 u64 ino, u32 generation)
2998 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2999 struct inode *inode;
3001 if (f2fs_check_nid_range(sbi, ino))
3002 return ERR_PTR(-ESTALE);
3005 * f2fs_iget isn't quite right if the inode is currently unallocated!
3006 * However f2fs_iget currently does appropriate checks to handle stale
3007 * inodes so everything is OK.
3009 inode = f2fs_iget(sb, ino);
3011 return ERR_CAST(inode);
3012 if (unlikely(generation && inode->i_generation != generation)) {
3013 /* we didn't find the right inode.. */
3015 return ERR_PTR(-ESTALE);
3020 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3021 int fh_len, int fh_type)
3023 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3024 f2fs_nfs_get_inode);
3027 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3028 int fh_len, int fh_type)
3030 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3031 f2fs_nfs_get_inode);
3034 static const struct export_operations f2fs_export_ops = {
3035 .fh_to_dentry = f2fs_fh_to_dentry,
3036 .fh_to_parent = f2fs_fh_to_parent,
3037 .get_parent = f2fs_get_parent,
3040 loff_t max_file_blocks(struct inode *inode)
3046 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3047 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3048 * space in inode.i_addr, it will be more safe to reassign
3052 if (inode && f2fs_compressed_file(inode))
3053 leaf_count = ADDRS_PER_BLOCK(inode);
3055 leaf_count = DEF_ADDRS_PER_BLOCK;
3057 /* two direct node blocks */
3058 result += (leaf_count * 2);
3060 /* two indirect node blocks */
3061 leaf_count *= NIDS_PER_BLOCK;
3062 result += (leaf_count * 2);
3064 /* one double indirect node block */
3065 leaf_count *= NIDS_PER_BLOCK;
3066 result += leaf_count;
3071 static int __f2fs_commit_super(struct buffer_head *bh,
3072 struct f2fs_super_block *super)
3076 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3077 set_buffer_dirty(bh);
3080 /* it's rare case, we can do fua all the time */
3081 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3084 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3085 struct buffer_head *bh)
3087 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3088 (bh->b_data + F2FS_SUPER_OFFSET);
3089 struct super_block *sb = sbi->sb;
3090 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3091 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3092 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3093 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3094 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3095 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3096 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3097 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3098 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3099 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3100 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3101 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3102 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3103 u64 main_end_blkaddr = main_blkaddr +
3104 (segment_count_main << log_blocks_per_seg);
3105 u64 seg_end_blkaddr = segment0_blkaddr +
3106 (segment_count << log_blocks_per_seg);
3108 if (segment0_blkaddr != cp_blkaddr) {
3109 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3110 segment0_blkaddr, cp_blkaddr);
3114 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3116 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3117 cp_blkaddr, sit_blkaddr,
3118 segment_count_ckpt << log_blocks_per_seg);
3122 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3124 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3125 sit_blkaddr, nat_blkaddr,
3126 segment_count_sit << log_blocks_per_seg);
3130 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3132 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3133 nat_blkaddr, ssa_blkaddr,
3134 segment_count_nat << log_blocks_per_seg);
3138 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3140 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3141 ssa_blkaddr, main_blkaddr,
3142 segment_count_ssa << log_blocks_per_seg);
3146 if (main_end_blkaddr > seg_end_blkaddr) {
3147 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3148 main_blkaddr, seg_end_blkaddr,
3149 segment_count_main << log_blocks_per_seg);
3151 } else if (main_end_blkaddr < seg_end_blkaddr) {
3155 /* fix in-memory information all the time */
3156 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3157 segment0_blkaddr) >> log_blocks_per_seg);
3159 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3160 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3163 err = __f2fs_commit_super(bh, NULL);
3164 res = err ? "failed" : "done";
3166 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3167 res, main_blkaddr, seg_end_blkaddr,
3168 segment_count_main << log_blocks_per_seg);
3175 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3176 struct buffer_head *bh)
3178 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3179 block_t total_sections, blocks_per_seg;
3180 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3181 (bh->b_data + F2FS_SUPER_OFFSET);
3182 size_t crc_offset = 0;
3185 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3186 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3187 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3191 /* Check checksum_offset and crc in superblock */
3192 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3193 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3195 offsetof(struct f2fs_super_block, crc)) {
3196 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3198 return -EFSCORRUPTED;
3200 crc = le32_to_cpu(raw_super->crc);
3201 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3202 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3203 return -EFSCORRUPTED;
3207 /* Currently, support only 4KB block size */
3208 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3209 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3210 le32_to_cpu(raw_super->log_blocksize),
3212 return -EFSCORRUPTED;
3215 /* check log blocks per segment */
3216 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3217 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3218 le32_to_cpu(raw_super->log_blocks_per_seg));
3219 return -EFSCORRUPTED;
3222 /* Currently, support 512/1024/2048/4096 bytes sector size */
3223 if (le32_to_cpu(raw_super->log_sectorsize) >
3224 F2FS_MAX_LOG_SECTOR_SIZE ||
3225 le32_to_cpu(raw_super->log_sectorsize) <
3226 F2FS_MIN_LOG_SECTOR_SIZE) {
3227 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3228 le32_to_cpu(raw_super->log_sectorsize));
3229 return -EFSCORRUPTED;
3231 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3232 le32_to_cpu(raw_super->log_sectorsize) !=
3233 F2FS_MAX_LOG_SECTOR_SIZE) {
3234 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3235 le32_to_cpu(raw_super->log_sectors_per_block),
3236 le32_to_cpu(raw_super->log_sectorsize));
3237 return -EFSCORRUPTED;
3240 segment_count = le32_to_cpu(raw_super->segment_count);
3241 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3242 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3243 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3244 total_sections = le32_to_cpu(raw_super->section_count);
3246 /* blocks_per_seg should be 512, given the above check */
3247 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3249 if (segment_count > F2FS_MAX_SEGMENT ||
3250 segment_count < F2FS_MIN_SEGMENTS) {
3251 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3252 return -EFSCORRUPTED;
3255 if (total_sections > segment_count_main || total_sections < 1 ||
3256 segs_per_sec > segment_count || !segs_per_sec) {
3257 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3258 segment_count, total_sections, segs_per_sec);
3259 return -EFSCORRUPTED;
3262 if (segment_count_main != total_sections * segs_per_sec) {
3263 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3264 segment_count_main, total_sections, segs_per_sec);
3265 return -EFSCORRUPTED;
3268 if ((segment_count / segs_per_sec) < total_sections) {
3269 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3270 segment_count, segs_per_sec, total_sections);
3271 return -EFSCORRUPTED;
3274 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3275 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3276 segment_count, le64_to_cpu(raw_super->block_count));
3277 return -EFSCORRUPTED;
3280 if (RDEV(0).path[0]) {
3281 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3284 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3285 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3288 if (segment_count != dev_seg_count) {
3289 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3290 segment_count, dev_seg_count);
3291 return -EFSCORRUPTED;
3294 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3295 !bdev_is_zoned(sbi->sb->s_bdev)) {
3296 f2fs_info(sbi, "Zoned block device path is missing");
3297 return -EFSCORRUPTED;
3301 if (secs_per_zone > total_sections || !secs_per_zone) {
3302 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3303 secs_per_zone, total_sections);
3304 return -EFSCORRUPTED;
3306 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3307 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3308 (le32_to_cpu(raw_super->extension_count) +
3309 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3310 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3311 le32_to_cpu(raw_super->extension_count),
3312 raw_super->hot_ext_count,
3313 F2FS_MAX_EXTENSION);
3314 return -EFSCORRUPTED;
3317 if (le32_to_cpu(raw_super->cp_payload) >=
3318 (blocks_per_seg - F2FS_CP_PACKS -
3319 NR_CURSEG_PERSIST_TYPE)) {
3320 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3321 le32_to_cpu(raw_super->cp_payload),
3322 blocks_per_seg - F2FS_CP_PACKS -
3323 NR_CURSEG_PERSIST_TYPE);
3324 return -EFSCORRUPTED;
3327 /* check reserved ino info */
3328 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3329 le32_to_cpu(raw_super->meta_ino) != 2 ||
3330 le32_to_cpu(raw_super->root_ino) != 3) {
3331 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3332 le32_to_cpu(raw_super->node_ino),
3333 le32_to_cpu(raw_super->meta_ino),
3334 le32_to_cpu(raw_super->root_ino));
3335 return -EFSCORRUPTED;
3338 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3339 if (sanity_check_area_boundary(sbi, bh))
3340 return -EFSCORRUPTED;
3345 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3347 unsigned int total, fsmeta;
3348 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3349 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3350 unsigned int ovp_segments, reserved_segments;
3351 unsigned int main_segs, blocks_per_seg;
3352 unsigned int sit_segs, nat_segs;
3353 unsigned int sit_bitmap_size, nat_bitmap_size;
3354 unsigned int log_blocks_per_seg;
3355 unsigned int segment_count_main;
3356 unsigned int cp_pack_start_sum, cp_payload;
3357 block_t user_block_count, valid_user_blocks;
3358 block_t avail_node_count, valid_node_count;
3359 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3362 total = le32_to_cpu(raw_super->segment_count);
3363 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3364 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3366 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3368 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3369 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3371 if (unlikely(fsmeta >= total))
3374 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3375 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3377 if (!f2fs_sb_has_readonly(sbi) &&
3378 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3379 ovp_segments == 0 || reserved_segments == 0)) {
3380 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3383 user_block_count = le64_to_cpu(ckpt->user_block_count);
3384 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3385 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3386 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3387 if (!user_block_count || user_block_count >=
3388 segment_count_main << log_blocks_per_seg) {
3389 f2fs_err(sbi, "Wrong user_block_count: %u",
3394 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3395 if (valid_user_blocks > user_block_count) {
3396 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3397 valid_user_blocks, user_block_count);
3401 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3402 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3403 if (valid_node_count > avail_node_count) {
3404 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3405 valid_node_count, avail_node_count);
3409 main_segs = le32_to_cpu(raw_super->segment_count_main);
3410 blocks_per_seg = sbi->blocks_per_seg;
3412 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3413 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3414 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3417 if (f2fs_sb_has_readonly(sbi))
3420 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3421 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3422 le32_to_cpu(ckpt->cur_node_segno[j])) {
3423 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3425 le32_to_cpu(ckpt->cur_node_segno[i]));
3431 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3432 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3433 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3436 if (f2fs_sb_has_readonly(sbi))
3439 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3440 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3441 le32_to_cpu(ckpt->cur_data_segno[j])) {
3442 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3444 le32_to_cpu(ckpt->cur_data_segno[i]));
3449 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3450 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3451 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3452 le32_to_cpu(ckpt->cur_data_segno[j])) {
3453 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3455 le32_to_cpu(ckpt->cur_node_segno[i]));
3461 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3462 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3464 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3465 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3466 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3467 sit_bitmap_size, nat_bitmap_size);
3471 cp_pack_start_sum = __start_sum_addr(sbi);
3472 cp_payload = __cp_payload(sbi);
3473 if (cp_pack_start_sum < cp_payload + 1 ||
3474 cp_pack_start_sum > blocks_per_seg - 1 -
3475 NR_CURSEG_PERSIST_TYPE) {
3476 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3481 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3482 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3483 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3484 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3485 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3486 le32_to_cpu(ckpt->checksum_offset));
3490 nat_blocks = nat_segs << log_blocks_per_seg;
3491 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3492 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3493 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3494 (cp_payload + F2FS_CP_PACKS +
3495 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3496 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3497 cp_payload, nat_bits_blocks);
3498 return -EFSCORRUPTED;
3501 if (unlikely(f2fs_cp_error(sbi))) {
3502 f2fs_err(sbi, "A bug case: need to run fsck");
3508 static void init_sb_info(struct f2fs_sb_info *sbi)
3510 struct f2fs_super_block *raw_super = sbi->raw_super;
3513 sbi->log_sectors_per_block =
3514 le32_to_cpu(raw_super->log_sectors_per_block);
3515 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3516 sbi->blocksize = 1 << sbi->log_blocksize;
3517 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3518 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3519 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3520 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3521 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3522 sbi->total_node_count =
3523 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3524 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3525 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3526 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3527 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3528 sbi->cur_victim_sec = NULL_SECNO;
3529 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3530 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3531 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3532 sbi->migration_granularity = sbi->segs_per_sec;
3533 sbi->seq_file_ra_mul = MIN_RA_MUL;
3534 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3535 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3537 sbi->dir_level = DEF_DIR_LEVEL;
3538 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3539 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3540 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3541 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3542 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3543 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3544 DEF_UMOUNT_DISCARD_TIMEOUT;
3545 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3547 for (i = 0; i < NR_COUNT_TYPE; i++)
3548 atomic_set(&sbi->nr_pages[i], 0);
3550 for (i = 0; i < META; i++)
3551 atomic_set(&sbi->wb_sync_req[i], 0);
3553 INIT_LIST_HEAD(&sbi->s_list);
3554 mutex_init(&sbi->umount_mutex);
3555 init_rwsem(&sbi->io_order_lock);
3556 spin_lock_init(&sbi->cp_lock);
3558 sbi->dirty_device = 0;
3559 spin_lock_init(&sbi->dev_lock);
3561 init_rwsem(&sbi->sb_lock);
3562 init_rwsem(&sbi->pin_sem);
3565 static int init_percpu_info(struct f2fs_sb_info *sbi)
3569 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3573 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3576 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3581 #ifdef CONFIG_BLK_DEV_ZONED
3583 struct f2fs_report_zones_args {
3584 struct f2fs_dev_info *dev;
3585 bool zone_cap_mismatch;
3588 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3591 struct f2fs_report_zones_args *rz_args = data;
3593 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3596 set_bit(idx, rz_args->dev->blkz_seq);
3597 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >>
3598 F2FS_LOG_SECTORS_PER_BLOCK;
3599 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch)
3600 rz_args->zone_cap_mismatch = true;
3605 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3607 struct block_device *bdev = FDEV(devi).bdev;
3608 sector_t nr_sectors = bdev_nr_sectors(bdev);
3609 struct f2fs_report_zones_args rep_zone_arg;
3612 if (!f2fs_sb_has_blkzoned(sbi))
3615 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3616 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3618 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3619 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3620 __ilog2_u32(sbi->blocks_per_blkz))
3622 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3623 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3624 sbi->log_blocks_per_blkz;
3625 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3626 FDEV(devi).nr_blkz++;
3628 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3629 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3630 * sizeof(unsigned long),
3632 if (!FDEV(devi).blkz_seq)
3635 /* Get block zones type and zone-capacity */
3636 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi,
3637 FDEV(devi).nr_blkz * sizeof(block_t),
3639 if (!FDEV(devi).zone_capacity_blocks)
3642 rep_zone_arg.dev = &FDEV(devi);
3643 rep_zone_arg.zone_cap_mismatch = false;
3645 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3650 if (!rep_zone_arg.zone_cap_mismatch) {
3651 kfree(FDEV(devi).zone_capacity_blocks);
3652 FDEV(devi).zone_capacity_blocks = NULL;
3660 * Read f2fs raw super block.
3661 * Because we have two copies of super block, so read both of them
3662 * to get the first valid one. If any one of them is broken, we pass
3663 * them recovery flag back to the caller.
3665 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3666 struct f2fs_super_block **raw_super,
3667 int *valid_super_block, int *recovery)
3669 struct super_block *sb = sbi->sb;
3671 struct buffer_head *bh;
3672 struct f2fs_super_block *super;
3675 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3679 for (block = 0; block < 2; block++) {
3680 bh = sb_bread(sb, block);
3682 f2fs_err(sbi, "Unable to read %dth superblock",
3689 /* sanity checking of raw super */
3690 err = sanity_check_raw_super(sbi, bh);
3692 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3700 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3702 *valid_super_block = block;
3708 /* No valid superblock */
3717 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3719 struct buffer_head *bh;
3723 if ((recover && f2fs_readonly(sbi->sb)) ||
3724 bdev_read_only(sbi->sb->s_bdev)) {
3725 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3729 /* we should update superblock crc here */
3730 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3731 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3732 offsetof(struct f2fs_super_block, crc));
3733 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3736 /* write back-up superblock first */
3737 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3740 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3743 /* if we are in recovery path, skip writing valid superblock */
3747 /* write current valid superblock */
3748 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3751 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3756 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3758 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3759 unsigned int max_devices = MAX_DEVICES;
3760 unsigned int logical_blksize;
3763 /* Initialize single device information */
3764 if (!RDEV(0).path[0]) {
3765 if (!bdev_is_zoned(sbi->sb->s_bdev))
3771 * Initialize multiple devices information, or single
3772 * zoned block device information.
3774 sbi->devs = f2fs_kzalloc(sbi,
3775 array_size(max_devices,
3776 sizeof(struct f2fs_dev_info)),
3781 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3782 sbi->aligned_blksize = true;
3784 for (i = 0; i < max_devices; i++) {
3786 if (i > 0 && !RDEV(i).path[0])
3789 if (max_devices == 1) {
3790 /* Single zoned block device mount */
3792 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3793 sbi->sb->s_mode, sbi->sb->s_type);
3795 /* Multi-device mount */
3796 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3797 FDEV(i).total_segments =
3798 le32_to_cpu(RDEV(i).total_segments);
3800 FDEV(i).start_blk = 0;
3801 FDEV(i).end_blk = FDEV(i).start_blk +
3802 (FDEV(i).total_segments <<
3803 sbi->log_blocks_per_seg) - 1 +
3804 le32_to_cpu(raw_super->segment0_blkaddr);
3806 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3807 FDEV(i).end_blk = FDEV(i).start_blk +
3808 (FDEV(i).total_segments <<
3809 sbi->log_blocks_per_seg) - 1;
3811 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3812 sbi->sb->s_mode, sbi->sb->s_type);
3814 if (IS_ERR(FDEV(i).bdev))
3815 return PTR_ERR(FDEV(i).bdev);
3817 /* to release errored devices */
3818 sbi->s_ndevs = i + 1;
3820 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
3821 sbi->aligned_blksize = false;
3823 #ifdef CONFIG_BLK_DEV_ZONED
3824 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3825 !f2fs_sb_has_blkzoned(sbi)) {
3826 f2fs_err(sbi, "Zoned block device feature not enabled");
3829 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3830 if (init_blkz_info(sbi, i)) {
3831 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3834 if (max_devices == 1)
3836 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3838 FDEV(i).total_segments,
3839 FDEV(i).start_blk, FDEV(i).end_blk,
3840 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3841 "Host-aware" : "Host-managed");
3845 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3847 FDEV(i).total_segments,
3848 FDEV(i).start_blk, FDEV(i).end_blk);
3851 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3855 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3857 #ifdef CONFIG_UNICODE
3858 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3859 const struct f2fs_sb_encodings *encoding_info;
3860 struct unicode_map *encoding;
3861 __u16 encoding_flags;
3863 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3866 "Encoding requested by superblock is unknown");
3870 encoding = utf8_load(encoding_info->version);
3871 if (IS_ERR(encoding)) {
3873 "can't mount with superblock charset: %s-%s "
3874 "not supported by the kernel. flags: 0x%x.",
3875 encoding_info->name, encoding_info->version,
3877 return PTR_ERR(encoding);
3879 f2fs_info(sbi, "Using encoding defined by superblock: "
3880 "%s-%s with flags 0x%hx", encoding_info->name,
3881 encoding_info->version?:"\b", encoding_flags);
3883 sbi->sb->s_encoding = encoding;
3884 sbi->sb->s_encoding_flags = encoding_flags;
3887 if (f2fs_sb_has_casefold(sbi)) {
3888 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3895 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3897 struct f2fs_sm_info *sm_i = SM_I(sbi);
3899 /* adjust parameters according to the volume size */
3900 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3901 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3902 if (f2fs_block_unit_discard(sbi))
3903 sm_i->dcc_info->discard_granularity = 1;
3904 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3907 sbi->readdir_ra = 1;
3910 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3912 struct f2fs_sb_info *sbi;
3913 struct f2fs_super_block *raw_super;
3916 bool skip_recovery = false, need_fsck = false;
3917 char *options = NULL;
3918 int recovery, i, valid_super_block;
3919 struct curseg_info *seg_i;
3925 valid_super_block = -1;
3928 /* allocate memory for f2fs-specific super block info */
3929 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3935 /* Load the checksum driver */
3936 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3937 if (IS_ERR(sbi->s_chksum_driver)) {
3938 f2fs_err(sbi, "Cannot load crc32 driver.");
3939 err = PTR_ERR(sbi->s_chksum_driver);
3940 sbi->s_chksum_driver = NULL;
3944 /* set a block size */
3945 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3946 f2fs_err(sbi, "unable to set blocksize");
3950 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3955 sb->s_fs_info = sbi;
3956 sbi->raw_super = raw_super;
3958 /* precompute checksum seed for metadata */
3959 if (f2fs_sb_has_inode_chksum(sbi))
3960 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3961 sizeof(raw_super->uuid));
3963 default_options(sbi);
3964 /* parse mount options */
3965 options = kstrdup((const char *)data, GFP_KERNEL);
3966 if (data && !options) {
3971 err = parse_options(sb, options, false);
3975 sb->s_maxbytes = max_file_blocks(NULL) <<
3976 le32_to_cpu(raw_super->log_blocksize);
3977 sb->s_max_links = F2FS_LINK_MAX;
3979 err = f2fs_setup_casefold(sbi);
3984 sb->dq_op = &f2fs_quota_operations;
3985 sb->s_qcop = &f2fs_quotactl_ops;
3986 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3988 if (f2fs_sb_has_quota_ino(sbi)) {
3989 for (i = 0; i < MAXQUOTAS; i++) {
3990 if (f2fs_qf_ino(sbi->sb, i))
3991 sbi->nquota_files++;
3996 sb->s_op = &f2fs_sops;
3997 #ifdef CONFIG_FS_ENCRYPTION
3998 sb->s_cop = &f2fs_cryptops;
4000 #ifdef CONFIG_FS_VERITY
4001 sb->s_vop = &f2fs_verityops;
4003 sb->s_xattr = f2fs_xattr_handlers;
4004 sb->s_export_op = &f2fs_export_ops;
4005 sb->s_magic = F2FS_SUPER_MAGIC;
4006 sb->s_time_gran = 1;
4007 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4008 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4009 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4010 sb->s_iflags |= SB_I_CGROUPWB;
4012 /* init f2fs-specific super block info */
4013 sbi->valid_super_block = valid_super_block;
4014 init_rwsem(&sbi->gc_lock);
4015 mutex_init(&sbi->writepages);
4016 init_rwsem(&sbi->cp_global_sem);
4017 init_rwsem(&sbi->node_write);
4018 init_rwsem(&sbi->node_change);
4020 /* disallow all the data/node/meta page writes */
4021 set_sbi_flag(sbi, SBI_POR_DOING);
4022 spin_lock_init(&sbi->stat_lock);
4024 for (i = 0; i < NR_PAGE_TYPE; i++) {
4025 int n = (i == META) ? 1 : NR_TEMP_TYPE;
4031 sizeof(struct f2fs_bio_info)),
4033 if (!sbi->write_io[i]) {
4038 for (j = HOT; j < n; j++) {
4039 init_rwsem(&sbi->write_io[i][j].io_rwsem);
4040 sbi->write_io[i][j].sbi = sbi;
4041 sbi->write_io[i][j].bio = NULL;
4042 spin_lock_init(&sbi->write_io[i][j].io_lock);
4043 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
4044 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
4045 init_rwsem(&sbi->write_io[i][j].bio_list_lock);
4049 init_rwsem(&sbi->cp_rwsem);
4050 init_rwsem(&sbi->quota_sem);
4051 init_waitqueue_head(&sbi->cp_wait);
4054 err = f2fs_init_iostat(sbi);
4058 err = init_percpu_info(sbi);
4062 if (F2FS_IO_ALIGNED(sbi)) {
4063 sbi->write_io_dummy =
4064 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4065 if (!sbi->write_io_dummy) {
4071 /* init per sbi slab cache */
4072 err = f2fs_init_xattr_caches(sbi);
4075 err = f2fs_init_page_array_cache(sbi);
4077 goto free_xattr_cache;
4079 /* get an inode for meta space */
4080 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4081 if (IS_ERR(sbi->meta_inode)) {
4082 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4083 err = PTR_ERR(sbi->meta_inode);
4084 goto free_page_array_cache;
4087 err = f2fs_get_valid_checkpoint(sbi);
4089 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4090 goto free_meta_inode;
4093 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4094 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4095 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4096 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4097 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4100 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4101 set_sbi_flag(sbi, SBI_NEED_FSCK);
4103 /* Initialize device list */
4104 err = f2fs_scan_devices(sbi);
4106 f2fs_err(sbi, "Failed to find devices");
4110 err = f2fs_init_post_read_wq(sbi);
4112 f2fs_err(sbi, "Failed to initialize post read workqueue");
4116 sbi->total_valid_node_count =
4117 le32_to_cpu(sbi->ckpt->valid_node_count);
4118 percpu_counter_set(&sbi->total_valid_inode_count,
4119 le32_to_cpu(sbi->ckpt->valid_inode_count));
4120 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4121 sbi->total_valid_block_count =
4122 le64_to_cpu(sbi->ckpt->valid_block_count);
4123 sbi->last_valid_block_count = sbi->total_valid_block_count;
4124 sbi->reserved_blocks = 0;
4125 sbi->current_reserved_blocks = 0;
4126 limit_reserve_root(sbi);
4127 adjust_unusable_cap_perc(sbi);
4129 for (i = 0; i < NR_INODE_TYPE; i++) {
4130 INIT_LIST_HEAD(&sbi->inode_list[i]);
4131 spin_lock_init(&sbi->inode_lock[i]);
4133 mutex_init(&sbi->flush_lock);
4135 f2fs_init_extent_cache_info(sbi);
4137 f2fs_init_ino_entry_info(sbi);
4139 f2fs_init_fsync_node_info(sbi);
4141 /* setup checkpoint request control and start checkpoint issue thread */
4142 f2fs_init_ckpt_req_control(sbi);
4143 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4144 test_opt(sbi, MERGE_CHECKPOINT)) {
4145 err = f2fs_start_ckpt_thread(sbi);
4148 "Failed to start F2FS issue_checkpoint_thread (%d)",
4150 goto stop_ckpt_thread;
4154 /* setup f2fs internal modules */
4155 err = f2fs_build_segment_manager(sbi);
4157 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4161 err = f2fs_build_node_manager(sbi);
4163 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4168 /* For write statistics */
4169 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4171 /* Read accumulated write IO statistics if exists */
4172 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4173 if (__exist_node_summaries(sbi))
4174 sbi->kbytes_written =
4175 le64_to_cpu(seg_i->journal->info.kbytes_written);
4177 f2fs_build_gc_manager(sbi);
4179 err = f2fs_build_stats(sbi);
4183 /* get an inode for node space */
4184 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4185 if (IS_ERR(sbi->node_inode)) {
4186 f2fs_err(sbi, "Failed to read node inode");
4187 err = PTR_ERR(sbi->node_inode);
4191 /* read root inode and dentry */
4192 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4194 f2fs_err(sbi, "Failed to read root inode");
4195 err = PTR_ERR(root);
4196 goto free_node_inode;
4198 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4199 !root->i_size || !root->i_nlink) {
4202 goto free_node_inode;
4205 sb->s_root = d_make_root(root); /* allocate root dentry */
4208 goto free_node_inode;
4211 err = f2fs_init_compress_inode(sbi);
4213 goto free_root_inode;
4215 err = f2fs_register_sysfs(sbi);
4217 goto free_compress_inode;
4220 /* Enable quota usage during mount */
4221 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4222 err = f2fs_enable_quotas(sb);
4224 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4227 /* if there are any orphan inodes, free them */
4228 err = f2fs_recover_orphan_inodes(sbi);
4232 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4233 goto reset_checkpoint;
4235 /* recover fsynced data */
4236 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4237 !test_opt(sbi, NORECOVERY)) {
4239 * mount should be failed, when device has readonly mode, and
4240 * previous checkpoint was not done by clean system shutdown.
4242 if (f2fs_hw_is_readonly(sbi)) {
4243 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4244 err = f2fs_recover_fsync_data(sbi, true);
4247 f2fs_err(sbi, "Need to recover fsync data, but "
4248 "write access unavailable, please try "
4249 "mount w/ disable_roll_forward or norecovery");
4254 f2fs_info(sbi, "write access unavailable, skipping recovery");
4255 goto reset_checkpoint;
4259 set_sbi_flag(sbi, SBI_NEED_FSCK);
4262 goto reset_checkpoint;
4264 err = f2fs_recover_fsync_data(sbi, false);
4267 skip_recovery = true;
4269 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4274 err = f2fs_recover_fsync_data(sbi, true);
4276 if (!f2fs_readonly(sb) && err > 0) {
4278 f2fs_err(sbi, "Need to recover fsync data");
4284 * If the f2fs is not readonly and fsync data recovery succeeds,
4285 * check zoned block devices' write pointer consistency.
4287 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4288 err = f2fs_check_write_pointer(sbi);
4294 f2fs_init_inmem_curseg(sbi);
4296 /* f2fs_recover_fsync_data() cleared this already */
4297 clear_sbi_flag(sbi, SBI_POR_DOING);
4299 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4300 err = f2fs_disable_checkpoint(sbi);
4302 goto sync_free_meta;
4303 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4304 f2fs_enable_checkpoint(sbi);
4308 * If filesystem is not mounted as read-only then
4309 * do start the gc_thread.
4311 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4312 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4313 /* After POR, we can run background GC thread.*/
4314 err = f2fs_start_gc_thread(sbi);
4316 goto sync_free_meta;
4320 /* recover broken superblock */
4322 err = f2fs_commit_super(sbi, true);
4323 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4324 sbi->valid_super_block ? 1 : 2, err);
4327 f2fs_join_shrinker(sbi);
4329 f2fs_tuning_parameters(sbi);
4331 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4332 cur_cp_version(F2FS_CKPT(sbi)));
4333 f2fs_update_time(sbi, CP_TIME);
4334 f2fs_update_time(sbi, REQ_TIME);
4335 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4339 /* safe to flush all the data */
4340 sync_filesystem(sbi->sb);
4345 f2fs_truncate_quota_inode_pages(sb);
4346 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4347 f2fs_quota_off_umount(sbi->sb);
4350 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4351 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4352 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4353 * falls into an infinite loop in f2fs_sync_meta_pages().
4355 truncate_inode_pages_final(META_MAPPING(sbi));
4356 /* evict some inodes being cached by GC */
4358 f2fs_unregister_sysfs(sbi);
4359 free_compress_inode:
4360 f2fs_destroy_compress_inode(sbi);
4365 f2fs_release_ino_entry(sbi, true);
4366 truncate_inode_pages_final(NODE_MAPPING(sbi));
4367 iput(sbi->node_inode);
4368 sbi->node_inode = NULL;
4370 f2fs_destroy_stats(sbi);
4372 f2fs_destroy_node_manager(sbi);
4374 f2fs_destroy_segment_manager(sbi);
4375 f2fs_destroy_post_read_wq(sbi);
4377 f2fs_stop_ckpt_thread(sbi);
4379 destroy_device_list(sbi);
4382 make_bad_inode(sbi->meta_inode);
4383 iput(sbi->meta_inode);
4384 sbi->meta_inode = NULL;
4385 free_page_array_cache:
4386 f2fs_destroy_page_array_cache(sbi);
4388 f2fs_destroy_xattr_caches(sbi);
4390 mempool_destroy(sbi->write_io_dummy);
4392 destroy_percpu_info(sbi);
4394 f2fs_destroy_iostat(sbi);
4396 for (i = 0; i < NR_PAGE_TYPE; i++)
4397 kvfree(sbi->write_io[i]);
4399 #ifdef CONFIG_UNICODE
4400 utf8_unload(sb->s_encoding);
4401 sb->s_encoding = NULL;
4405 for (i = 0; i < MAXQUOTAS; i++)
4406 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4408 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4413 if (sbi->s_chksum_driver)
4414 crypto_free_shash(sbi->s_chksum_driver);
4417 /* give only one another chance */
4418 if (retry_cnt > 0 && skip_recovery) {
4420 shrink_dcache_sb(sb);
4426 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4427 const char *dev_name, void *data)
4429 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4432 static void kill_f2fs_super(struct super_block *sb)
4435 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4437 set_sbi_flag(sbi, SBI_IS_CLOSE);
4438 f2fs_stop_gc_thread(sbi);
4439 f2fs_stop_discard_thread(sbi);
4441 #ifdef CONFIG_F2FS_FS_COMPRESSION
4443 * latter evict_inode() can bypass checking and invalidating
4444 * compress inode cache.
4446 if (test_opt(sbi, COMPRESS_CACHE))
4447 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4450 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4451 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4452 struct cp_control cpc = {
4453 .reason = CP_UMOUNT,
4455 f2fs_write_checkpoint(sbi, &cpc);
4458 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4459 sb->s_flags &= ~SB_RDONLY;
4461 kill_block_super(sb);
4464 static struct file_system_type f2fs_fs_type = {
4465 .owner = THIS_MODULE,
4467 .mount = f2fs_mount,
4468 .kill_sb = kill_f2fs_super,
4469 .fs_flags = FS_REQUIRES_DEV,
4471 MODULE_ALIAS_FS("f2fs");
4473 static int __init init_inodecache(void)
4475 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4476 sizeof(struct f2fs_inode_info), 0,
4477 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4478 if (!f2fs_inode_cachep)
4483 static void destroy_inodecache(void)
4486 * Make sure all delayed rcu free inodes are flushed before we
4490 kmem_cache_destroy(f2fs_inode_cachep);
4493 static int __init init_f2fs_fs(void)
4497 if (PAGE_SIZE != F2FS_BLKSIZE) {
4498 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4499 PAGE_SIZE, F2FS_BLKSIZE);
4503 err = init_inodecache();
4506 err = f2fs_create_node_manager_caches();
4508 goto free_inodecache;
4509 err = f2fs_create_segment_manager_caches();
4511 goto free_node_manager_caches;
4512 err = f2fs_create_checkpoint_caches();
4514 goto free_segment_manager_caches;
4515 err = f2fs_create_recovery_cache();
4517 goto free_checkpoint_caches;
4518 err = f2fs_create_extent_cache();
4520 goto free_recovery_cache;
4521 err = f2fs_create_garbage_collection_cache();
4523 goto free_extent_cache;
4524 err = f2fs_init_sysfs();
4526 goto free_garbage_collection_cache;
4527 err = register_shrinker(&f2fs_shrinker_info);
4530 err = register_filesystem(&f2fs_fs_type);
4533 f2fs_create_root_stats();
4534 err = f2fs_init_post_read_processing();
4536 goto free_root_stats;
4537 err = f2fs_init_iostat_processing();
4539 goto free_post_read;
4540 err = f2fs_init_bio_entry_cache();
4543 err = f2fs_init_bioset();
4545 goto free_bio_enrty_cache;
4546 err = f2fs_init_compress_mempool();
4549 err = f2fs_init_compress_cache();
4551 goto free_compress_mempool;
4552 err = f2fs_create_casefold_cache();
4554 goto free_compress_cache;
4556 free_compress_cache:
4557 f2fs_destroy_compress_cache();
4558 free_compress_mempool:
4559 f2fs_destroy_compress_mempool();
4561 f2fs_destroy_bioset();
4562 free_bio_enrty_cache:
4563 f2fs_destroy_bio_entry_cache();
4565 f2fs_destroy_iostat_processing();
4567 f2fs_destroy_post_read_processing();
4569 f2fs_destroy_root_stats();
4570 unregister_filesystem(&f2fs_fs_type);
4572 unregister_shrinker(&f2fs_shrinker_info);
4575 free_garbage_collection_cache:
4576 f2fs_destroy_garbage_collection_cache();
4578 f2fs_destroy_extent_cache();
4579 free_recovery_cache:
4580 f2fs_destroy_recovery_cache();
4581 free_checkpoint_caches:
4582 f2fs_destroy_checkpoint_caches();
4583 free_segment_manager_caches:
4584 f2fs_destroy_segment_manager_caches();
4585 free_node_manager_caches:
4586 f2fs_destroy_node_manager_caches();
4588 destroy_inodecache();
4593 static void __exit exit_f2fs_fs(void)
4595 f2fs_destroy_casefold_cache();
4596 f2fs_destroy_compress_cache();
4597 f2fs_destroy_compress_mempool();
4598 f2fs_destroy_bioset();
4599 f2fs_destroy_bio_entry_cache();
4600 f2fs_destroy_iostat_processing();
4601 f2fs_destroy_post_read_processing();
4602 f2fs_destroy_root_stats();
4603 unregister_filesystem(&f2fs_fs_type);
4604 unregister_shrinker(&f2fs_shrinker_info);
4606 f2fs_destroy_garbage_collection_cache();
4607 f2fs_destroy_extent_cache();
4608 f2fs_destroy_recovery_cache();
4609 f2fs_destroy_checkpoint_caches();
4610 f2fs_destroy_segment_manager_caches();
4611 f2fs_destroy_node_manager_caches();
4612 destroy_inodecache();
4615 module_init(init_f2fs_fs)
4616 module_exit(exit_f2fs_fs)
4618 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4619 MODULE_DESCRIPTION("Flash Friendly File System");
4620 MODULE_LICENSE("GPL");
4621 MODULE_SOFTDEP("pre: crc32");