1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
8 #include <linux/module.h>
9 #include <linux/init.h>
11 #include <linux/sched/mm.h>
12 #include <linux/statfs.h>
13 #include <linux/buffer_head.h>
14 #include <linux/kthread.h>
15 #include <linux/parser.h>
16 #include <linux/mount.h>
17 #include <linux/seq_file.h>
18 #include <linux/proc_fs.h>
19 #include <linux/random.h>
20 #include <linux/exportfs.h>
21 #include <linux/blkdev.h>
22 #include <linux/quotaops.h>
23 #include <linux/f2fs_fs.h>
24 #include <linux/sysfs.h>
25 #include <linux/quota.h>
26 #include <linux/unicode.h>
27 #include <linux/part_stat.h>
28 #include <linux/zstd.h>
29 #include <linux/lz4.h>
38 #define CREATE_TRACE_POINTS
39 #include <trace/events/f2fs.h>
41 static struct kmem_cache *f2fs_inode_cachep;
43 #ifdef CONFIG_F2FS_FAULT_INJECTION
45 const char *f2fs_fault_name[FAULT_MAX] = {
46 [FAULT_KMALLOC] = "kmalloc",
47 [FAULT_KVMALLOC] = "kvmalloc",
48 [FAULT_PAGE_ALLOC] = "page alloc",
49 [FAULT_PAGE_GET] = "page get",
50 [FAULT_ALLOC_NID] = "alloc nid",
51 [FAULT_ORPHAN] = "orphan",
52 [FAULT_BLOCK] = "no more block",
53 [FAULT_DIR_DEPTH] = "too big dir depth",
54 [FAULT_EVICT_INODE] = "evict_inode fail",
55 [FAULT_TRUNCATE] = "truncate fail",
56 [FAULT_READ_IO] = "read IO error",
57 [FAULT_CHECKPOINT] = "checkpoint error",
58 [FAULT_DISCARD] = "discard error",
59 [FAULT_WRITE_IO] = "write IO error",
60 [FAULT_SLAB_ALLOC] = "slab alloc",
61 [FAULT_DQUOT_INIT] = "dquot initialize",
62 [FAULT_LOCK_OP] = "lock_op",
65 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
68 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
71 atomic_set(&ffi->inject_ops, 0);
72 ffi->inject_rate = rate;
76 ffi->inject_type = type;
79 memset(ffi, 0, sizeof(struct f2fs_fault_info));
83 /* f2fs-wide shrinker description */
84 static struct shrinker f2fs_shrinker_info = {
85 .scan_objects = f2fs_shrink_scan,
86 .count_objects = f2fs_shrink_count,
87 .seeks = DEFAULT_SEEKS,
92 Opt_disable_roll_forward,
103 Opt_disable_ext_identify,
106 Opt_inline_xattr_size,
144 Opt_test_dummy_encryption,
146 Opt_checkpoint_disable,
147 Opt_checkpoint_disable_cap,
148 Opt_checkpoint_disable_cap_perc,
149 Opt_checkpoint_enable,
150 Opt_checkpoint_merge,
151 Opt_nocheckpoint_merge,
152 Opt_compress_algorithm,
153 Opt_compress_log_size,
154 Opt_compress_extension,
155 Opt_nocompress_extension,
166 static match_table_t f2fs_tokens = {
167 {Opt_gc_background, "background_gc=%s"},
168 {Opt_disable_roll_forward, "disable_roll_forward"},
169 {Opt_norecovery, "norecovery"},
170 {Opt_discard, "discard"},
171 {Opt_nodiscard, "nodiscard"},
172 {Opt_noheap, "no_heap"},
174 {Opt_user_xattr, "user_xattr"},
175 {Opt_nouser_xattr, "nouser_xattr"},
177 {Opt_noacl, "noacl"},
178 {Opt_active_logs, "active_logs=%u"},
179 {Opt_disable_ext_identify, "disable_ext_identify"},
180 {Opt_inline_xattr, "inline_xattr"},
181 {Opt_noinline_xattr, "noinline_xattr"},
182 {Opt_inline_xattr_size, "inline_xattr_size=%u"},
183 {Opt_inline_data, "inline_data"},
184 {Opt_inline_dentry, "inline_dentry"},
185 {Opt_noinline_dentry, "noinline_dentry"},
186 {Opt_flush_merge, "flush_merge"},
187 {Opt_noflush_merge, "noflush_merge"},
188 {Opt_nobarrier, "nobarrier"},
189 {Opt_fastboot, "fastboot"},
190 {Opt_extent_cache, "extent_cache"},
191 {Opt_noextent_cache, "noextent_cache"},
192 {Opt_noinline_data, "noinline_data"},
193 {Opt_data_flush, "data_flush"},
194 {Opt_reserve_root, "reserve_root=%u"},
195 {Opt_resgid, "resgid=%u"},
196 {Opt_resuid, "resuid=%u"},
197 {Opt_mode, "mode=%s"},
198 {Opt_io_size_bits, "io_bits=%u"},
199 {Opt_fault_injection, "fault_injection=%u"},
200 {Opt_fault_type, "fault_type=%u"},
201 {Opt_lazytime, "lazytime"},
202 {Opt_nolazytime, "nolazytime"},
203 {Opt_quota, "quota"},
204 {Opt_noquota, "noquota"},
205 {Opt_usrquota, "usrquota"},
206 {Opt_grpquota, "grpquota"},
207 {Opt_prjquota, "prjquota"},
208 {Opt_usrjquota, "usrjquota=%s"},
209 {Opt_grpjquota, "grpjquota=%s"},
210 {Opt_prjjquota, "prjjquota=%s"},
211 {Opt_offusrjquota, "usrjquota="},
212 {Opt_offgrpjquota, "grpjquota="},
213 {Opt_offprjjquota, "prjjquota="},
214 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
215 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
216 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
217 {Opt_whint, "whint_mode=%s"},
218 {Opt_alloc, "alloc_mode=%s"},
219 {Opt_fsync, "fsync_mode=%s"},
220 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
221 {Opt_test_dummy_encryption, "test_dummy_encryption"},
222 {Opt_inlinecrypt, "inlinecrypt"},
223 {Opt_checkpoint_disable, "checkpoint=disable"},
224 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
225 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
226 {Opt_checkpoint_enable, "checkpoint=enable"},
227 {Opt_checkpoint_merge, "checkpoint_merge"},
228 {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
229 {Opt_compress_algorithm, "compress_algorithm=%s"},
230 {Opt_compress_log_size, "compress_log_size=%u"},
231 {Opt_compress_extension, "compress_extension=%s"},
232 {Opt_nocompress_extension, "nocompress_extension=%s"},
233 {Opt_compress_chksum, "compress_chksum"},
234 {Opt_compress_mode, "compress_mode=%s"},
235 {Opt_compress_cache, "compress_cache"},
237 {Opt_gc_merge, "gc_merge"},
238 {Opt_nogc_merge, "nogc_merge"},
239 {Opt_discard_unit, "discard_unit=%s"},
243 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
245 struct va_format vaf;
251 level = printk_get_level(fmt);
252 vaf.fmt = printk_skip_level(fmt);
254 printk("%c%cF2FS-fs (%s): %pV\n",
255 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
260 #if IS_ENABLED(CONFIG_UNICODE)
261 static const struct f2fs_sb_encodings {
264 unsigned int version;
265 } f2fs_sb_encoding_map[] = {
266 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
269 static const struct f2fs_sb_encodings *
270 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
272 __u16 magic = le16_to_cpu(sb->s_encoding);
275 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
276 if (magic == f2fs_sb_encoding_map[i].magic)
277 return &f2fs_sb_encoding_map[i];
282 struct kmem_cache *f2fs_cf_name_slab;
283 static int __init f2fs_create_casefold_cache(void)
285 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
287 if (!f2fs_cf_name_slab)
292 static void f2fs_destroy_casefold_cache(void)
294 kmem_cache_destroy(f2fs_cf_name_slab);
297 static int __init f2fs_create_casefold_cache(void) { return 0; }
298 static void f2fs_destroy_casefold_cache(void) { }
301 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
303 block_t limit = min((sbi->user_block_count << 1) / 1000,
304 sbi->user_block_count - sbi->reserved_blocks);
307 if (test_opt(sbi, RESERVE_ROOT) &&
308 F2FS_OPTION(sbi).root_reserved_blocks > limit) {
309 F2FS_OPTION(sbi).root_reserved_blocks = limit;
310 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
311 F2FS_OPTION(sbi).root_reserved_blocks);
313 if (!test_opt(sbi, RESERVE_ROOT) &&
314 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
315 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
316 !gid_eq(F2FS_OPTION(sbi).s_resgid,
317 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
318 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
319 from_kuid_munged(&init_user_ns,
320 F2FS_OPTION(sbi).s_resuid),
321 from_kgid_munged(&init_user_ns,
322 F2FS_OPTION(sbi).s_resgid));
325 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
327 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
328 unsigned int avg_vblocks;
329 unsigned int wanted_reserved_segments;
330 block_t avail_user_block_count;
332 if (!F2FS_IO_ALIGNED(sbi))
335 /* average valid block count in section in worst case */
336 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
339 * we need enough free space when migrating one section in worst case
341 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
342 reserved_segments(sbi);
343 wanted_reserved_segments -= reserved_segments(sbi);
345 avail_user_block_count = sbi->user_block_count -
346 sbi->current_reserved_blocks -
347 F2FS_OPTION(sbi).root_reserved_blocks;
349 if (wanted_reserved_segments * sbi->blocks_per_seg >
350 avail_user_block_count) {
351 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
352 wanted_reserved_segments,
353 avail_user_block_count >> sbi->log_blocks_per_seg);
357 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
359 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
360 wanted_reserved_segments);
365 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
367 if (!F2FS_OPTION(sbi).unusable_cap_perc)
370 if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
371 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
373 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
374 F2FS_OPTION(sbi).unusable_cap_perc;
376 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
377 F2FS_OPTION(sbi).unusable_cap,
378 F2FS_OPTION(sbi).unusable_cap_perc);
381 static void init_once(void *foo)
383 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
385 inode_init_once(&fi->vfs_inode);
389 static const char * const quotatypes[] = INITQFNAMES;
390 #define QTYPE2NAME(t) (quotatypes[t])
391 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
394 struct f2fs_sb_info *sbi = F2FS_SB(sb);
398 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
399 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
402 if (f2fs_sb_has_quota_ino(sbi)) {
403 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
407 qname = match_strdup(args);
409 f2fs_err(sbi, "Not enough memory for storing quotafile name");
412 if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
413 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
416 f2fs_err(sbi, "%s quota file already specified",
420 if (strchr(qname, '/')) {
421 f2fs_err(sbi, "quotafile must be on filesystem root");
424 F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
432 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
434 struct f2fs_sb_info *sbi = F2FS_SB(sb);
436 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
437 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
440 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
441 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
445 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
448 * We do the test below only for project quotas. 'usrquota' and
449 * 'grpquota' mount options are allowed even without quota feature
450 * to support legacy quotas in quota files.
452 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
453 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
456 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
457 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
458 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
459 if (test_opt(sbi, USRQUOTA) &&
460 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
461 clear_opt(sbi, USRQUOTA);
463 if (test_opt(sbi, GRPQUOTA) &&
464 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
465 clear_opt(sbi, GRPQUOTA);
467 if (test_opt(sbi, PRJQUOTA) &&
468 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
469 clear_opt(sbi, PRJQUOTA);
471 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
472 test_opt(sbi, PRJQUOTA)) {
473 f2fs_err(sbi, "old and new quota format mixing");
477 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
478 f2fs_err(sbi, "journaled quota format not specified");
483 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
484 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
485 F2FS_OPTION(sbi).s_jquota_fmt = 0;
491 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
493 const substring_t *arg,
496 struct f2fs_sb_info *sbi = F2FS_SB(sb);
497 #ifdef CONFIG_FS_ENCRYPTION
500 if (!f2fs_sb_has_encrypt(sbi)) {
501 f2fs_err(sbi, "Encrypt feature is off");
506 * This mount option is just for testing, and it's not worthwhile to
507 * implement the extra complexity (e.g. RCU protection) that would be
508 * needed to allow it to be set or changed during remount. We do allow
509 * it to be specified during remount, but only if there is no change.
511 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
512 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
515 err = fscrypt_set_test_dummy_encryption(
516 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
520 "Can't change test_dummy_encryption on remount");
521 else if (err == -EINVAL)
522 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
525 f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
529 f2fs_warn(sbi, "Test dummy encryption mode enabled");
531 f2fs_warn(sbi, "Test dummy encryption mount option ignored");
536 #ifdef CONFIG_F2FS_FS_COMPRESSION
538 * 1. The same extension name cannot not appear in both compress and non-compress extension
540 * 2. If the compress extension specifies all files, the types specified by the non-compress
541 * extension will be treated as special cases and will not be compressed.
542 * 3. Don't allow the non-compress extension specifies all files.
544 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
546 unsigned char (*ext)[F2FS_EXTENSION_LEN];
547 unsigned char (*noext)[F2FS_EXTENSION_LEN];
548 int ext_cnt, noext_cnt, index = 0, no_index = 0;
550 ext = F2FS_OPTION(sbi).extensions;
551 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
552 noext = F2FS_OPTION(sbi).noextensions;
553 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
558 for (no_index = 0; no_index < noext_cnt; no_index++) {
559 if (!strcasecmp("*", noext[no_index])) {
560 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
563 for (index = 0; index < ext_cnt; index++) {
564 if (!strcasecmp(ext[index], noext[no_index])) {
565 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
574 #ifdef CONFIG_F2FS_FS_LZ4
575 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
577 #ifdef CONFIG_F2FS_FS_LZ4HC
581 if (strlen(str) == 3) {
582 F2FS_OPTION(sbi).compress_level = 0;
586 #ifdef CONFIG_F2FS_FS_LZ4HC
590 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
593 if (kstrtouint(str + 1, 10, &level))
596 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
597 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
601 F2FS_OPTION(sbi).compress_level = level;
604 f2fs_info(sbi, "kernel doesn't support lz4hc compression");
610 #ifdef CONFIG_F2FS_FS_ZSTD
611 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
616 if (strlen(str) == len) {
617 F2FS_OPTION(sbi).compress_level = 0;
624 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
627 if (kstrtouint(str + 1, 10, &level))
630 if (!level || level > zstd_max_clevel()) {
631 f2fs_info(sbi, "invalid zstd compress level: %d", level);
635 F2FS_OPTION(sbi).compress_level = level;
641 static int parse_options(struct super_block *sb, char *options, bool is_remount)
643 struct f2fs_sb_info *sbi = F2FS_SB(sb);
644 substring_t args[MAX_OPT_ARGS];
645 #ifdef CONFIG_F2FS_FS_COMPRESSION
646 unsigned char (*ext)[F2FS_EXTENSION_LEN];
647 unsigned char (*noext)[F2FS_EXTENSION_LEN];
648 int ext_cnt, noext_cnt;
659 while ((p = strsep(&options, ",")) != NULL) {
665 * Initialize args struct so we know whether arg was
666 * found; some options take optional arguments.
668 args[0].to = args[0].from = NULL;
669 token = match_token(p, f2fs_tokens, args);
672 case Opt_gc_background:
673 name = match_strdup(&args[0]);
677 if (!strcmp(name, "on")) {
678 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
679 } else if (!strcmp(name, "off")) {
680 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
681 } else if (!strcmp(name, "sync")) {
682 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
689 case Opt_disable_roll_forward:
690 set_opt(sbi, DISABLE_ROLL_FORWARD);
693 /* this option mounts f2fs with ro */
694 set_opt(sbi, NORECOVERY);
695 if (!f2fs_readonly(sb))
699 if (!f2fs_hw_support_discard(sbi)) {
700 f2fs_warn(sbi, "device does not support discard");
703 set_opt(sbi, DISCARD);
706 if (f2fs_hw_should_discard(sbi)) {
707 f2fs_warn(sbi, "discard is required for zoned block devices");
710 clear_opt(sbi, DISCARD);
713 set_opt(sbi, NOHEAP);
716 clear_opt(sbi, NOHEAP);
718 #ifdef CONFIG_F2FS_FS_XATTR
720 set_opt(sbi, XATTR_USER);
722 case Opt_nouser_xattr:
723 clear_opt(sbi, XATTR_USER);
725 case Opt_inline_xattr:
726 set_opt(sbi, INLINE_XATTR);
728 case Opt_noinline_xattr:
729 clear_opt(sbi, INLINE_XATTR);
731 case Opt_inline_xattr_size:
732 if (args->from && match_int(args, &arg))
734 set_opt(sbi, INLINE_XATTR_SIZE);
735 F2FS_OPTION(sbi).inline_xattr_size = arg;
739 f2fs_info(sbi, "user_xattr options not supported");
741 case Opt_nouser_xattr:
742 f2fs_info(sbi, "nouser_xattr options not supported");
744 case Opt_inline_xattr:
745 f2fs_info(sbi, "inline_xattr options not supported");
747 case Opt_noinline_xattr:
748 f2fs_info(sbi, "noinline_xattr options not supported");
751 #ifdef CONFIG_F2FS_FS_POSIX_ACL
753 set_opt(sbi, POSIX_ACL);
756 clear_opt(sbi, POSIX_ACL);
760 f2fs_info(sbi, "acl options not supported");
763 f2fs_info(sbi, "noacl options not supported");
766 case Opt_active_logs:
767 if (args->from && match_int(args, &arg))
769 if (arg != 2 && arg != 4 &&
770 arg != NR_CURSEG_PERSIST_TYPE)
772 F2FS_OPTION(sbi).active_logs = arg;
774 case Opt_disable_ext_identify:
775 set_opt(sbi, DISABLE_EXT_IDENTIFY);
777 case Opt_inline_data:
778 set_opt(sbi, INLINE_DATA);
780 case Opt_inline_dentry:
781 set_opt(sbi, INLINE_DENTRY);
783 case Opt_noinline_dentry:
784 clear_opt(sbi, INLINE_DENTRY);
786 case Opt_flush_merge:
787 set_opt(sbi, FLUSH_MERGE);
789 case Opt_noflush_merge:
790 clear_opt(sbi, FLUSH_MERGE);
793 set_opt(sbi, NOBARRIER);
796 set_opt(sbi, FASTBOOT);
798 case Opt_extent_cache:
799 set_opt(sbi, EXTENT_CACHE);
801 case Opt_noextent_cache:
802 clear_opt(sbi, EXTENT_CACHE);
804 case Opt_noinline_data:
805 clear_opt(sbi, INLINE_DATA);
808 set_opt(sbi, DATA_FLUSH);
810 case Opt_reserve_root:
811 if (args->from && match_int(args, &arg))
813 if (test_opt(sbi, RESERVE_ROOT)) {
814 f2fs_info(sbi, "Preserve previous reserve_root=%u",
815 F2FS_OPTION(sbi).root_reserved_blocks);
817 F2FS_OPTION(sbi).root_reserved_blocks = arg;
818 set_opt(sbi, RESERVE_ROOT);
822 if (args->from && match_int(args, &arg))
824 uid = make_kuid(current_user_ns(), arg);
825 if (!uid_valid(uid)) {
826 f2fs_err(sbi, "Invalid uid value %d", arg);
829 F2FS_OPTION(sbi).s_resuid = uid;
832 if (args->from && match_int(args, &arg))
834 gid = make_kgid(current_user_ns(), arg);
835 if (!gid_valid(gid)) {
836 f2fs_err(sbi, "Invalid gid value %d", arg);
839 F2FS_OPTION(sbi).s_resgid = gid;
842 name = match_strdup(&args[0]);
846 if (!strcmp(name, "adaptive")) {
847 if (f2fs_sb_has_blkzoned(sbi)) {
848 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
852 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
853 } else if (!strcmp(name, "lfs")) {
854 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
855 } else if (!strcmp(name, "fragment:segment")) {
856 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
857 } else if (!strcmp(name, "fragment:block")) {
858 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
865 case Opt_io_size_bits:
866 if (args->from && match_int(args, &arg))
868 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
869 f2fs_warn(sbi, "Not support %d, larger than %d",
870 1 << arg, BIO_MAX_VECS);
873 F2FS_OPTION(sbi).write_io_size_bits = arg;
875 #ifdef CONFIG_F2FS_FAULT_INJECTION
876 case Opt_fault_injection:
877 if (args->from && match_int(args, &arg))
879 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
880 set_opt(sbi, FAULT_INJECTION);
884 if (args->from && match_int(args, &arg))
886 f2fs_build_fault_attr(sbi, 0, arg);
887 set_opt(sbi, FAULT_INJECTION);
890 case Opt_fault_injection:
891 f2fs_info(sbi, "fault_injection options not supported");
895 f2fs_info(sbi, "fault_type options not supported");
899 sb->s_flags |= SB_LAZYTIME;
902 sb->s_flags &= ~SB_LAZYTIME;
907 set_opt(sbi, USRQUOTA);
910 set_opt(sbi, GRPQUOTA);
913 set_opt(sbi, PRJQUOTA);
916 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
921 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
926 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
930 case Opt_offusrjquota:
931 ret = f2fs_clear_qf_name(sb, USRQUOTA);
935 case Opt_offgrpjquota:
936 ret = f2fs_clear_qf_name(sb, GRPQUOTA);
940 case Opt_offprjjquota:
941 ret = f2fs_clear_qf_name(sb, PRJQUOTA);
945 case Opt_jqfmt_vfsold:
946 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
948 case Opt_jqfmt_vfsv0:
949 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
951 case Opt_jqfmt_vfsv1:
952 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
955 clear_opt(sbi, QUOTA);
956 clear_opt(sbi, USRQUOTA);
957 clear_opt(sbi, GRPQUOTA);
958 clear_opt(sbi, PRJQUOTA);
968 case Opt_offusrjquota:
969 case Opt_offgrpjquota:
970 case Opt_offprjjquota:
971 case Opt_jqfmt_vfsold:
972 case Opt_jqfmt_vfsv0:
973 case Opt_jqfmt_vfsv1:
975 f2fs_info(sbi, "quota operations not supported");
979 name = match_strdup(&args[0]);
982 if (!strcmp(name, "user-based")) {
983 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
984 } else if (!strcmp(name, "off")) {
985 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
986 } else if (!strcmp(name, "fs-based")) {
987 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
995 name = match_strdup(&args[0]);
999 if (!strcmp(name, "default")) {
1000 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1001 } else if (!strcmp(name, "reuse")) {
1002 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1010 name = match_strdup(&args[0]);
1013 if (!strcmp(name, "posix")) {
1014 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1015 } else if (!strcmp(name, "strict")) {
1016 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1017 } else if (!strcmp(name, "nobarrier")) {
1018 F2FS_OPTION(sbi).fsync_mode =
1019 FSYNC_MODE_NOBARRIER;
1026 case Opt_test_dummy_encryption:
1027 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1032 case Opt_inlinecrypt:
1033 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1034 sb->s_flags |= SB_INLINECRYPT;
1036 f2fs_info(sbi, "inline encryption not supported");
1039 case Opt_checkpoint_disable_cap_perc:
1040 if (args->from && match_int(args, &arg))
1042 if (arg < 0 || arg > 100)
1044 F2FS_OPTION(sbi).unusable_cap_perc = arg;
1045 set_opt(sbi, DISABLE_CHECKPOINT);
1047 case Opt_checkpoint_disable_cap:
1048 if (args->from && match_int(args, &arg))
1050 F2FS_OPTION(sbi).unusable_cap = arg;
1051 set_opt(sbi, DISABLE_CHECKPOINT);
1053 case Opt_checkpoint_disable:
1054 set_opt(sbi, DISABLE_CHECKPOINT);
1056 case Opt_checkpoint_enable:
1057 clear_opt(sbi, DISABLE_CHECKPOINT);
1059 case Opt_checkpoint_merge:
1060 set_opt(sbi, MERGE_CHECKPOINT);
1062 case Opt_nocheckpoint_merge:
1063 clear_opt(sbi, MERGE_CHECKPOINT);
1065 #ifdef CONFIG_F2FS_FS_COMPRESSION
1066 case Opt_compress_algorithm:
1067 if (!f2fs_sb_has_compression(sbi)) {
1068 f2fs_info(sbi, "Image doesn't support compression");
1071 name = match_strdup(&args[0]);
1074 if (!strcmp(name, "lzo")) {
1075 #ifdef CONFIG_F2FS_FS_LZO
1076 F2FS_OPTION(sbi).compress_level = 0;
1077 F2FS_OPTION(sbi).compress_algorithm =
1080 f2fs_info(sbi, "kernel doesn't support lzo compression");
1082 } else if (!strncmp(name, "lz4", 3)) {
1083 #ifdef CONFIG_F2FS_FS_LZ4
1084 ret = f2fs_set_lz4hc_level(sbi, name);
1089 F2FS_OPTION(sbi).compress_algorithm =
1092 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1094 } else if (!strncmp(name, "zstd", 4)) {
1095 #ifdef CONFIG_F2FS_FS_ZSTD
1096 ret = f2fs_set_zstd_level(sbi, name);
1101 F2FS_OPTION(sbi).compress_algorithm =
1104 f2fs_info(sbi, "kernel doesn't support zstd compression");
1106 } else if (!strcmp(name, "lzo-rle")) {
1107 #ifdef CONFIG_F2FS_FS_LZORLE
1108 F2FS_OPTION(sbi).compress_level = 0;
1109 F2FS_OPTION(sbi).compress_algorithm =
1112 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1120 case Opt_compress_log_size:
1121 if (!f2fs_sb_has_compression(sbi)) {
1122 f2fs_info(sbi, "Image doesn't support compression");
1125 if (args->from && match_int(args, &arg))
1127 if (arg < MIN_COMPRESS_LOG_SIZE ||
1128 arg > MAX_COMPRESS_LOG_SIZE) {
1130 "Compress cluster log size is out of range");
1133 F2FS_OPTION(sbi).compress_log_size = arg;
1135 case Opt_compress_extension:
1136 if (!f2fs_sb_has_compression(sbi)) {
1137 f2fs_info(sbi, "Image doesn't support compression");
1140 name = match_strdup(&args[0]);
1144 ext = F2FS_OPTION(sbi).extensions;
1145 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1147 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1148 ext_cnt >= COMPRESS_EXT_NUM) {
1150 "invalid extension length/number");
1155 strcpy(ext[ext_cnt], name);
1156 F2FS_OPTION(sbi).compress_ext_cnt++;
1159 case Opt_nocompress_extension:
1160 if (!f2fs_sb_has_compression(sbi)) {
1161 f2fs_info(sbi, "Image doesn't support compression");
1164 name = match_strdup(&args[0]);
1168 noext = F2FS_OPTION(sbi).noextensions;
1169 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1171 if (strlen(name) >= F2FS_EXTENSION_LEN ||
1172 noext_cnt >= COMPRESS_EXT_NUM) {
1174 "invalid extension length/number");
1179 strcpy(noext[noext_cnt], name);
1180 F2FS_OPTION(sbi).nocompress_ext_cnt++;
1183 case Opt_compress_chksum:
1184 F2FS_OPTION(sbi).compress_chksum = true;
1186 case Opt_compress_mode:
1187 name = match_strdup(&args[0]);
1190 if (!strcmp(name, "fs")) {
1191 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1192 } else if (!strcmp(name, "user")) {
1193 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1200 case Opt_compress_cache:
1201 set_opt(sbi, COMPRESS_CACHE);
1204 case Opt_compress_algorithm:
1205 case Opt_compress_log_size:
1206 case Opt_compress_extension:
1207 case Opt_nocompress_extension:
1208 case Opt_compress_chksum:
1209 case Opt_compress_mode:
1210 case Opt_compress_cache:
1211 f2fs_info(sbi, "compression options not supported");
1218 set_opt(sbi, GC_MERGE);
1220 case Opt_nogc_merge:
1221 clear_opt(sbi, GC_MERGE);
1223 case Opt_discard_unit:
1224 name = match_strdup(&args[0]);
1227 if (!strcmp(name, "block")) {
1228 F2FS_OPTION(sbi).discard_unit =
1230 } else if (!strcmp(name, "segment")) {
1231 F2FS_OPTION(sbi).discard_unit =
1232 DISCARD_UNIT_SEGMENT;
1233 } else if (!strcmp(name, "section")) {
1234 F2FS_OPTION(sbi).discard_unit =
1235 DISCARD_UNIT_SECTION;
1243 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1250 if (f2fs_check_quota_options(sbi))
1253 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1254 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1257 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1258 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1262 #if !IS_ENABLED(CONFIG_UNICODE)
1263 if (f2fs_sb_has_casefold(sbi)) {
1265 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1270 * The BLKZONED feature indicates that the drive was formatted with
1271 * zone alignment optimization. This is optional for host-aware
1272 * devices, but mandatory for host-managed zoned block devices.
1274 #ifndef CONFIG_BLK_DEV_ZONED
1275 if (f2fs_sb_has_blkzoned(sbi)) {
1276 f2fs_err(sbi, "Zoned block device support is not enabled");
1280 if (f2fs_sb_has_blkzoned(sbi)) {
1281 if (F2FS_OPTION(sbi).discard_unit !=
1282 DISCARD_UNIT_SECTION) {
1283 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1284 F2FS_OPTION(sbi).discard_unit =
1285 DISCARD_UNIT_SECTION;
1289 #ifdef CONFIG_F2FS_FS_COMPRESSION
1290 if (f2fs_test_compress_extension(sbi)) {
1291 f2fs_err(sbi, "invalid compress or nocompress extension");
1296 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1297 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1298 F2FS_IO_SIZE_KB(sbi));
1302 if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1303 int min_size, max_size;
1305 if (!f2fs_sb_has_extra_attr(sbi) ||
1306 !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1307 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1310 if (!test_opt(sbi, INLINE_XATTR)) {
1311 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1315 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1316 max_size = MAX_INLINE_XATTR_SIZE;
1318 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1319 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1320 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1321 min_size, max_size);
1326 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1327 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1331 /* Not pass down write hints if the number of active logs is lesser
1332 * than NR_CURSEG_PERSIST_TYPE.
1334 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE)
1335 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1337 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1338 f2fs_err(sbi, "Allow to mount readonly mode only");
1344 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1346 struct f2fs_inode_info *fi;
1348 fi = f2fs_kmem_cache_alloc(f2fs_inode_cachep,
1349 GFP_F2FS_ZERO, false, F2FS_SB(sb));
1353 init_once((void *) fi);
1355 /* Initialize f2fs-specific inode info */
1356 atomic_set(&fi->dirty_pages, 0);
1357 atomic_set(&fi->i_compr_blocks, 0);
1358 init_f2fs_rwsem(&fi->i_sem);
1359 spin_lock_init(&fi->i_size_lock);
1360 INIT_LIST_HEAD(&fi->dirty_list);
1361 INIT_LIST_HEAD(&fi->gdirty_list);
1362 INIT_LIST_HEAD(&fi->inmem_ilist);
1363 INIT_LIST_HEAD(&fi->inmem_pages);
1364 mutex_init(&fi->inmem_lock);
1365 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1366 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1367 init_f2fs_rwsem(&fi->i_xattr_sem);
1369 /* Will be used by directory only */
1370 fi->i_dir_level = F2FS_SB(sb)->dir_level;
1372 return &fi->vfs_inode;
1375 static int f2fs_drop_inode(struct inode *inode)
1377 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1381 * during filesystem shutdown, if checkpoint is disabled,
1382 * drop useless meta/node dirty pages.
1384 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1385 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1386 inode->i_ino == F2FS_META_INO(sbi)) {
1387 trace_f2fs_drop_inode(inode, 1);
1393 * This is to avoid a deadlock condition like below.
1394 * writeback_single_inode(inode)
1395 * - f2fs_write_data_page
1396 * - f2fs_gc -> iput -> evict
1397 * - inode_wait_for_writeback(inode)
1399 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1400 if (!inode->i_nlink && !is_bad_inode(inode)) {
1401 /* to avoid evict_inode call simultaneously */
1402 atomic_inc(&inode->i_count);
1403 spin_unlock(&inode->i_lock);
1405 /* some remained atomic pages should discarded */
1406 if (f2fs_is_atomic_file(inode))
1407 f2fs_drop_inmem_pages(inode);
1409 /* should remain fi->extent_tree for writepage */
1410 f2fs_destroy_extent_node(inode);
1412 sb_start_intwrite(inode->i_sb);
1413 f2fs_i_size_write(inode, 0);
1415 f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1416 inode, NULL, 0, DATA);
1417 truncate_inode_pages_final(inode->i_mapping);
1419 if (F2FS_HAS_BLOCKS(inode))
1420 f2fs_truncate(inode);
1422 sb_end_intwrite(inode->i_sb);
1424 spin_lock(&inode->i_lock);
1425 atomic_dec(&inode->i_count);
1427 trace_f2fs_drop_inode(inode, 0);
1430 ret = generic_drop_inode(inode);
1432 ret = fscrypt_drop_inode(inode);
1433 trace_f2fs_drop_inode(inode, ret);
1437 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1439 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1442 spin_lock(&sbi->inode_lock[DIRTY_META]);
1443 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1446 set_inode_flag(inode, FI_DIRTY_INODE);
1447 stat_inc_dirty_inode(sbi, DIRTY_META);
1449 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1450 list_add_tail(&F2FS_I(inode)->gdirty_list,
1451 &sbi->inode_list[DIRTY_META]);
1452 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1454 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1458 void f2fs_inode_synced(struct inode *inode)
1460 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1462 spin_lock(&sbi->inode_lock[DIRTY_META]);
1463 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1464 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1467 if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1468 list_del_init(&F2FS_I(inode)->gdirty_list);
1469 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1471 clear_inode_flag(inode, FI_DIRTY_INODE);
1472 clear_inode_flag(inode, FI_AUTO_RECOVER);
1473 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1474 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1478 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1480 * We should call set_dirty_inode to write the dirty inode through write_inode.
1482 static void f2fs_dirty_inode(struct inode *inode, int flags)
1484 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1486 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1487 inode->i_ino == F2FS_META_INO(sbi))
1490 if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1491 clear_inode_flag(inode, FI_AUTO_RECOVER);
1493 f2fs_inode_dirtied(inode, false);
1496 static void f2fs_free_inode(struct inode *inode)
1498 fscrypt_free_inode(inode);
1499 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1502 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1504 percpu_counter_destroy(&sbi->total_valid_inode_count);
1505 percpu_counter_destroy(&sbi->rf_node_block_count);
1506 percpu_counter_destroy(&sbi->alloc_valid_block_count);
1509 static void destroy_device_list(struct f2fs_sb_info *sbi)
1513 for (i = 0; i < sbi->s_ndevs; i++) {
1514 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1515 #ifdef CONFIG_BLK_DEV_ZONED
1516 kvfree(FDEV(i).blkz_seq);
1517 kfree(FDEV(i).zone_capacity_blocks);
1523 static void f2fs_put_super(struct super_block *sb)
1525 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1529 /* unregister procfs/sysfs entries in advance to avoid race case */
1530 f2fs_unregister_sysfs(sbi);
1532 f2fs_quota_off_umount(sb);
1534 /* prevent remaining shrinker jobs */
1535 mutex_lock(&sbi->umount_mutex);
1538 * flush all issued checkpoints and stop checkpoint issue thread.
1539 * after then, all checkpoints should be done by each process context.
1541 f2fs_stop_ckpt_thread(sbi);
1544 * We don't need to do checkpoint when superblock is clean.
1545 * But, the previous checkpoint was not done by umount, it needs to do
1546 * clean checkpoint again.
1548 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1549 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1550 struct cp_control cpc = {
1551 .reason = CP_UMOUNT,
1553 f2fs_write_checkpoint(sbi, &cpc);
1556 /* be sure to wait for any on-going discard commands */
1557 dropped = f2fs_issue_discard_timeout(sbi);
1559 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1560 !sbi->discard_blks && !dropped) {
1561 struct cp_control cpc = {
1562 .reason = CP_UMOUNT | CP_TRIMMED,
1564 f2fs_write_checkpoint(sbi, &cpc);
1568 * normally superblock is clean, so we need to release this.
1569 * In addition, EIO will skip do checkpoint, we need this as well.
1571 f2fs_release_ino_entry(sbi, true);
1573 f2fs_leave_shrinker(sbi);
1574 mutex_unlock(&sbi->umount_mutex);
1576 /* our cp_error case, we can wait for any writeback page */
1577 f2fs_flush_merged_writes(sbi);
1579 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1581 f2fs_bug_on(sbi, sbi->fsync_node_num);
1583 f2fs_destroy_compress_inode(sbi);
1585 iput(sbi->node_inode);
1586 sbi->node_inode = NULL;
1588 iput(sbi->meta_inode);
1589 sbi->meta_inode = NULL;
1592 * iput() can update stat information, if f2fs_write_checkpoint()
1593 * above failed with error.
1595 f2fs_destroy_stats(sbi);
1597 /* destroy f2fs internal modules */
1598 f2fs_destroy_node_manager(sbi);
1599 f2fs_destroy_segment_manager(sbi);
1601 f2fs_destroy_post_read_wq(sbi);
1605 sb->s_fs_info = NULL;
1606 if (sbi->s_chksum_driver)
1607 crypto_free_shash(sbi->s_chksum_driver);
1608 kfree(sbi->raw_super);
1610 destroy_device_list(sbi);
1611 f2fs_destroy_page_array_cache(sbi);
1612 f2fs_destroy_xattr_caches(sbi);
1613 mempool_destroy(sbi->write_io_dummy);
1615 for (i = 0; i < MAXQUOTAS; i++)
1616 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1618 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1619 destroy_percpu_info(sbi);
1620 f2fs_destroy_iostat(sbi);
1621 for (i = 0; i < NR_PAGE_TYPE; i++)
1622 kvfree(sbi->write_io[i]);
1623 #if IS_ENABLED(CONFIG_UNICODE)
1624 utf8_unload(sb->s_encoding);
1629 int f2fs_sync_fs(struct super_block *sb, int sync)
1631 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1634 if (unlikely(f2fs_cp_error(sbi)))
1636 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1639 trace_f2fs_sync_fs(sb, sync);
1641 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1645 err = f2fs_issue_checkpoint(sbi);
1650 static int f2fs_freeze(struct super_block *sb)
1652 if (f2fs_readonly(sb))
1655 /* IO error happened before */
1656 if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1659 /* must be clean, since sync_filesystem() was already called */
1660 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1663 /* ensure no checkpoint required */
1664 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list))
1667 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1668 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1672 static int f2fs_unfreeze(struct super_block *sb)
1674 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1679 static int f2fs_statfs_project(struct super_block *sb,
1680 kprojid_t projid, struct kstatfs *buf)
1683 struct dquot *dquot;
1687 qid = make_kqid_projid(projid);
1688 dquot = dqget(sb, qid);
1690 return PTR_ERR(dquot);
1691 spin_lock(&dquot->dq_dqb_lock);
1693 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1694 dquot->dq_dqb.dqb_bhardlimit);
1696 limit >>= sb->s_blocksize_bits;
1698 if (limit && buf->f_blocks > limit) {
1699 curblock = (dquot->dq_dqb.dqb_curspace +
1700 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1701 buf->f_blocks = limit;
1702 buf->f_bfree = buf->f_bavail =
1703 (buf->f_blocks > curblock) ?
1704 (buf->f_blocks - curblock) : 0;
1707 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1708 dquot->dq_dqb.dqb_ihardlimit);
1710 if (limit && buf->f_files > limit) {
1711 buf->f_files = limit;
1713 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1714 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1717 spin_unlock(&dquot->dq_dqb_lock);
1723 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1725 struct super_block *sb = dentry->d_sb;
1726 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1727 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1728 block_t total_count, user_block_count, start_count;
1729 u64 avail_node_count;
1731 total_count = le64_to_cpu(sbi->raw_super->block_count);
1732 user_block_count = sbi->user_block_count;
1733 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1734 buf->f_type = F2FS_SUPER_MAGIC;
1735 buf->f_bsize = sbi->blocksize;
1737 buf->f_blocks = total_count - start_count;
1738 buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1739 sbi->current_reserved_blocks;
1741 spin_lock(&sbi->stat_lock);
1742 if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1745 buf->f_bfree -= sbi->unusable_block_count;
1746 spin_unlock(&sbi->stat_lock);
1748 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1749 buf->f_bavail = buf->f_bfree -
1750 F2FS_OPTION(sbi).root_reserved_blocks;
1754 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1756 if (avail_node_count > user_block_count) {
1757 buf->f_files = user_block_count;
1758 buf->f_ffree = buf->f_bavail;
1760 buf->f_files = avail_node_count;
1761 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1765 buf->f_namelen = F2FS_NAME_LEN;
1766 buf->f_fsid = u64_to_fsid(id);
1769 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1770 sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1771 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1777 static inline void f2fs_show_quota_options(struct seq_file *seq,
1778 struct super_block *sb)
1781 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1783 if (F2FS_OPTION(sbi).s_jquota_fmt) {
1786 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1797 seq_printf(seq, ",jqfmt=%s", fmtname);
1800 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1801 seq_show_option(seq, "usrjquota",
1802 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1804 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1805 seq_show_option(seq, "grpjquota",
1806 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1808 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1809 seq_show_option(seq, "prjjquota",
1810 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1814 #ifdef CONFIG_F2FS_FS_COMPRESSION
1815 static inline void f2fs_show_compress_options(struct seq_file *seq,
1816 struct super_block *sb)
1818 struct f2fs_sb_info *sbi = F2FS_SB(sb);
1822 if (!f2fs_sb_has_compression(sbi))
1825 switch (F2FS_OPTION(sbi).compress_algorithm) {
1835 case COMPRESS_LZORLE:
1836 algtype = "lzo-rle";
1839 seq_printf(seq, ",compress_algorithm=%s", algtype);
1841 if (F2FS_OPTION(sbi).compress_level)
1842 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1844 seq_printf(seq, ",compress_log_size=%u",
1845 F2FS_OPTION(sbi).compress_log_size);
1847 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1848 seq_printf(seq, ",compress_extension=%s",
1849 F2FS_OPTION(sbi).extensions[i]);
1852 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1853 seq_printf(seq, ",nocompress_extension=%s",
1854 F2FS_OPTION(sbi).noextensions[i]);
1857 if (F2FS_OPTION(sbi).compress_chksum)
1858 seq_puts(seq, ",compress_chksum");
1860 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1861 seq_printf(seq, ",compress_mode=%s", "fs");
1862 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1863 seq_printf(seq, ",compress_mode=%s", "user");
1865 if (test_opt(sbi, COMPRESS_CACHE))
1866 seq_puts(seq, ",compress_cache");
1870 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1872 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1874 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1875 seq_printf(seq, ",background_gc=%s", "sync");
1876 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1877 seq_printf(seq, ",background_gc=%s", "on");
1878 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1879 seq_printf(seq, ",background_gc=%s", "off");
1881 if (test_opt(sbi, GC_MERGE))
1882 seq_puts(seq, ",gc_merge");
1884 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1885 seq_puts(seq, ",disable_roll_forward");
1886 if (test_opt(sbi, NORECOVERY))
1887 seq_puts(seq, ",norecovery");
1888 if (test_opt(sbi, DISCARD))
1889 seq_puts(seq, ",discard");
1891 seq_puts(seq, ",nodiscard");
1892 if (test_opt(sbi, NOHEAP))
1893 seq_puts(seq, ",no_heap");
1895 seq_puts(seq, ",heap");
1896 #ifdef CONFIG_F2FS_FS_XATTR
1897 if (test_opt(sbi, XATTR_USER))
1898 seq_puts(seq, ",user_xattr");
1900 seq_puts(seq, ",nouser_xattr");
1901 if (test_opt(sbi, INLINE_XATTR))
1902 seq_puts(seq, ",inline_xattr");
1904 seq_puts(seq, ",noinline_xattr");
1905 if (test_opt(sbi, INLINE_XATTR_SIZE))
1906 seq_printf(seq, ",inline_xattr_size=%u",
1907 F2FS_OPTION(sbi).inline_xattr_size);
1909 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1910 if (test_opt(sbi, POSIX_ACL))
1911 seq_puts(seq, ",acl");
1913 seq_puts(seq, ",noacl");
1915 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1916 seq_puts(seq, ",disable_ext_identify");
1917 if (test_opt(sbi, INLINE_DATA))
1918 seq_puts(seq, ",inline_data");
1920 seq_puts(seq, ",noinline_data");
1921 if (test_opt(sbi, INLINE_DENTRY))
1922 seq_puts(seq, ",inline_dentry");
1924 seq_puts(seq, ",noinline_dentry");
1925 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1926 seq_puts(seq, ",flush_merge");
1927 if (test_opt(sbi, NOBARRIER))
1928 seq_puts(seq, ",nobarrier");
1929 if (test_opt(sbi, FASTBOOT))
1930 seq_puts(seq, ",fastboot");
1931 if (test_opt(sbi, EXTENT_CACHE))
1932 seq_puts(seq, ",extent_cache");
1934 seq_puts(seq, ",noextent_cache");
1935 if (test_opt(sbi, DATA_FLUSH))
1936 seq_puts(seq, ",data_flush");
1938 seq_puts(seq, ",mode=");
1939 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1940 seq_puts(seq, "adaptive");
1941 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1942 seq_puts(seq, "lfs");
1943 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1944 seq_puts(seq, "fragment:segment");
1945 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1946 seq_puts(seq, "fragment:block");
1947 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1948 if (test_opt(sbi, RESERVE_ROOT))
1949 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1950 F2FS_OPTION(sbi).root_reserved_blocks,
1951 from_kuid_munged(&init_user_ns,
1952 F2FS_OPTION(sbi).s_resuid),
1953 from_kgid_munged(&init_user_ns,
1954 F2FS_OPTION(sbi).s_resgid));
1955 if (F2FS_IO_SIZE_BITS(sbi))
1956 seq_printf(seq, ",io_bits=%u",
1957 F2FS_OPTION(sbi).write_io_size_bits);
1958 #ifdef CONFIG_F2FS_FAULT_INJECTION
1959 if (test_opt(sbi, FAULT_INJECTION)) {
1960 seq_printf(seq, ",fault_injection=%u",
1961 F2FS_OPTION(sbi).fault_info.inject_rate);
1962 seq_printf(seq, ",fault_type=%u",
1963 F2FS_OPTION(sbi).fault_info.inject_type);
1967 if (test_opt(sbi, QUOTA))
1968 seq_puts(seq, ",quota");
1969 if (test_opt(sbi, USRQUOTA))
1970 seq_puts(seq, ",usrquota");
1971 if (test_opt(sbi, GRPQUOTA))
1972 seq_puts(seq, ",grpquota");
1973 if (test_opt(sbi, PRJQUOTA))
1974 seq_puts(seq, ",prjquota");
1976 f2fs_show_quota_options(seq, sbi->sb);
1977 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1978 seq_printf(seq, ",whint_mode=%s", "user-based");
1979 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1980 seq_printf(seq, ",whint_mode=%s", "fs-based");
1982 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1984 if (sbi->sb->s_flags & SB_INLINECRYPT)
1985 seq_puts(seq, ",inlinecrypt");
1987 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1988 seq_printf(seq, ",alloc_mode=%s", "default");
1989 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1990 seq_printf(seq, ",alloc_mode=%s", "reuse");
1992 if (test_opt(sbi, DISABLE_CHECKPOINT))
1993 seq_printf(seq, ",checkpoint=disable:%u",
1994 F2FS_OPTION(sbi).unusable_cap);
1995 if (test_opt(sbi, MERGE_CHECKPOINT))
1996 seq_puts(seq, ",checkpoint_merge");
1998 seq_puts(seq, ",nocheckpoint_merge");
1999 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2000 seq_printf(seq, ",fsync_mode=%s", "posix");
2001 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2002 seq_printf(seq, ",fsync_mode=%s", "strict");
2003 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2004 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2006 #ifdef CONFIG_F2FS_FS_COMPRESSION
2007 f2fs_show_compress_options(seq, sbi->sb);
2010 if (test_opt(sbi, ATGC))
2011 seq_puts(seq, ",atgc");
2013 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2014 seq_printf(seq, ",discard_unit=%s", "block");
2015 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2016 seq_printf(seq, ",discard_unit=%s", "segment");
2017 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2018 seq_printf(seq, ",discard_unit=%s", "section");
2023 static void default_options(struct f2fs_sb_info *sbi)
2025 /* init some FS parameters */
2026 if (f2fs_sb_has_readonly(sbi))
2027 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2029 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2031 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2032 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
2033 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2034 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2035 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2036 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2037 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2038 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2039 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2040 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2041 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2043 sbi->sb->s_flags &= ~SB_INLINECRYPT;
2045 set_opt(sbi, INLINE_XATTR);
2046 set_opt(sbi, INLINE_DATA);
2047 set_opt(sbi, INLINE_DENTRY);
2048 set_opt(sbi, EXTENT_CACHE);
2049 set_opt(sbi, NOHEAP);
2050 clear_opt(sbi, DISABLE_CHECKPOINT);
2051 set_opt(sbi, MERGE_CHECKPOINT);
2052 F2FS_OPTION(sbi).unusable_cap = 0;
2053 sbi->sb->s_flags |= SB_LAZYTIME;
2054 set_opt(sbi, FLUSH_MERGE);
2055 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2056 set_opt(sbi, DISCARD);
2057 if (f2fs_sb_has_blkzoned(sbi)) {
2058 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2059 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2061 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2062 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2065 #ifdef CONFIG_F2FS_FS_XATTR
2066 set_opt(sbi, XATTR_USER);
2068 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2069 set_opt(sbi, POSIX_ACL);
2072 f2fs_build_fault_attr(sbi, 0, 0);
2076 static int f2fs_enable_quotas(struct super_block *sb);
2079 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2081 unsigned int s_flags = sbi->sb->s_flags;
2082 struct cp_control cpc;
2083 unsigned int gc_mode;
2088 if (s_flags & SB_RDONLY) {
2089 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2092 sbi->sb->s_flags |= SB_ACTIVE;
2094 f2fs_update_time(sbi, DISABLE_TIME);
2096 gc_mode = sbi->gc_mode;
2097 sbi->gc_mode = GC_URGENT_HIGH;
2099 while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2100 f2fs_down_write(&sbi->gc_lock);
2101 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
2102 if (err == -ENODATA) {
2106 if (err && err != -EAGAIN)
2110 ret = sync_filesystem(sbi->sb);
2112 err = ret ? ret : err;
2116 unusable = f2fs_get_unusable_blocks(sbi);
2117 if (f2fs_disable_cp_again(sbi, unusable)) {
2122 f2fs_down_write(&sbi->gc_lock);
2123 cpc.reason = CP_PAUSE;
2124 set_sbi_flag(sbi, SBI_CP_DISABLED);
2125 err = f2fs_write_checkpoint(sbi, &cpc);
2129 spin_lock(&sbi->stat_lock);
2130 sbi->unusable_block_count = unusable;
2131 spin_unlock(&sbi->stat_lock);
2134 f2fs_up_write(&sbi->gc_lock);
2136 sbi->gc_mode = gc_mode;
2137 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2141 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2143 int retry = DEFAULT_RETRY_IO_COUNT;
2145 /* we should flush all the data to keep data consistency */
2147 sync_inodes_sb(sbi->sb);
2149 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
2150 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2152 if (unlikely(retry < 0))
2153 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2155 f2fs_down_write(&sbi->gc_lock);
2156 f2fs_dirty_to_prefree(sbi);
2158 clear_sbi_flag(sbi, SBI_CP_DISABLED);
2159 set_sbi_flag(sbi, SBI_IS_DIRTY);
2160 f2fs_up_write(&sbi->gc_lock);
2162 f2fs_sync_fs(sbi->sb, 1);
2165 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2167 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2168 struct f2fs_mount_info org_mount_opt;
2169 unsigned long old_sb_flags;
2171 bool need_restart_gc = false, need_stop_gc = false;
2172 bool need_restart_ckpt = false, need_stop_ckpt = false;
2173 bool need_restart_flush = false, need_stop_flush = false;
2174 bool need_restart_discard = false, need_stop_discard = false;
2175 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
2176 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2177 bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2178 bool no_atgc = !test_opt(sbi, ATGC);
2179 bool no_discard = !test_opt(sbi, DISCARD);
2180 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2181 bool block_unit_discard = f2fs_block_unit_discard(sbi);
2182 struct discard_cmd_control *dcc;
2188 * Save the old mount options in case we
2189 * need to restore them.
2191 org_mount_opt = sbi->mount_opt;
2192 old_sb_flags = sb->s_flags;
2195 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2196 for (i = 0; i < MAXQUOTAS; i++) {
2197 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2198 org_mount_opt.s_qf_names[i] =
2199 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2201 if (!org_mount_opt.s_qf_names[i]) {
2202 for (j = 0; j < i; j++)
2203 kfree(org_mount_opt.s_qf_names[j]);
2207 org_mount_opt.s_qf_names[i] = NULL;
2212 /* recover superblocks we couldn't write due to previous RO mount */
2213 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2214 err = f2fs_commit_super(sbi, false);
2215 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2218 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2221 default_options(sbi);
2223 /* parse mount options */
2224 err = parse_options(sb, data, true);
2229 * Previous and new state of filesystem is RO,
2230 * so skip checking GC and FLUSH_MERGE conditions.
2232 if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2235 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2241 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2242 err = dquot_suspend(sb, -1);
2245 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2246 /* dquot_resume needs RW */
2247 sb->s_flags &= ~SB_RDONLY;
2248 if (sb_any_quota_suspended(sb)) {
2249 dquot_resume(sb, -1);
2250 } else if (f2fs_sb_has_quota_ino(sbi)) {
2251 err = f2fs_enable_quotas(sb);
2257 /* disallow enable atgc dynamically */
2258 if (no_atgc == !!test_opt(sbi, ATGC)) {
2260 f2fs_warn(sbi, "switch atgc option is not allowed");
2264 /* disallow enable/disable extent_cache dynamically */
2265 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
2267 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2271 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2273 f2fs_warn(sbi, "switch io_bits option is not allowed");
2277 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2279 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2283 if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2285 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2289 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2291 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2296 * We stop the GC thread if FS is mounted as RO
2297 * or if background_gc = off is passed in mount
2298 * option. Also sync the filesystem.
2300 if ((*flags & SB_RDONLY) ||
2301 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2302 !test_opt(sbi, GC_MERGE))) {
2303 if (sbi->gc_thread) {
2304 f2fs_stop_gc_thread(sbi);
2305 need_restart_gc = true;
2307 } else if (!sbi->gc_thread) {
2308 err = f2fs_start_gc_thread(sbi);
2311 need_stop_gc = true;
2314 if (*flags & SB_RDONLY ||
2315 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
2318 set_sbi_flag(sbi, SBI_IS_DIRTY);
2319 set_sbi_flag(sbi, SBI_IS_CLOSE);
2320 f2fs_sync_fs(sb, 1);
2321 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2324 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2325 !test_opt(sbi, MERGE_CHECKPOINT)) {
2326 f2fs_stop_ckpt_thread(sbi);
2327 need_restart_ckpt = true;
2329 err = f2fs_start_ckpt_thread(sbi);
2332 "Failed to start F2FS issue_checkpoint_thread (%d)",
2336 need_stop_ckpt = true;
2340 * We stop issue flush thread if FS is mounted as RO
2341 * or if flush_merge is not passed in mount option.
2343 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2344 clear_opt(sbi, FLUSH_MERGE);
2345 f2fs_destroy_flush_cmd_control(sbi, false);
2346 need_restart_flush = true;
2348 err = f2fs_create_flush_cmd_control(sbi);
2351 need_stop_flush = true;
2354 if (no_discard == !!test_opt(sbi, DISCARD)) {
2355 if (test_opt(sbi, DISCARD)) {
2356 err = f2fs_start_discard_thread(sbi);
2359 need_stop_discard = true;
2361 dcc = SM_I(sbi)->dcc_info;
2362 f2fs_stop_discard_thread(sbi);
2363 if (atomic_read(&dcc->discard_cmd_cnt))
2364 f2fs_issue_discard_timeout(sbi);
2365 need_restart_discard = true;
2369 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2370 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2371 err = f2fs_disable_checkpoint(sbi);
2373 goto restore_discard;
2375 f2fs_enable_checkpoint(sbi);
2381 /* Release old quota file names */
2382 for (i = 0; i < MAXQUOTAS; i++)
2383 kfree(org_mount_opt.s_qf_names[i]);
2385 /* Update the POSIXACL Flag */
2386 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2387 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2389 limit_reserve_root(sbi);
2390 adjust_unusable_cap_perc(sbi);
2391 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2394 if (need_restart_discard) {
2395 if (f2fs_start_discard_thread(sbi))
2396 f2fs_warn(sbi, "discard has been stopped");
2397 } else if (need_stop_discard) {
2398 f2fs_stop_discard_thread(sbi);
2401 if (need_restart_flush) {
2402 if (f2fs_create_flush_cmd_control(sbi))
2403 f2fs_warn(sbi, "background flush thread has stopped");
2404 } else if (need_stop_flush) {
2405 clear_opt(sbi, FLUSH_MERGE);
2406 f2fs_destroy_flush_cmd_control(sbi, false);
2409 if (need_restart_ckpt) {
2410 if (f2fs_start_ckpt_thread(sbi))
2411 f2fs_warn(sbi, "background ckpt thread has stopped");
2412 } else if (need_stop_ckpt) {
2413 f2fs_stop_ckpt_thread(sbi);
2416 if (need_restart_gc) {
2417 if (f2fs_start_gc_thread(sbi))
2418 f2fs_warn(sbi, "background gc thread has stopped");
2419 } else if (need_stop_gc) {
2420 f2fs_stop_gc_thread(sbi);
2424 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2425 for (i = 0; i < MAXQUOTAS; i++) {
2426 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2427 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2430 sbi->mount_opt = org_mount_opt;
2431 sb->s_flags = old_sb_flags;
2436 /* Read data from quotafile */
2437 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2438 size_t len, loff_t off)
2440 struct inode *inode = sb_dqopt(sb)->files[type];
2441 struct address_space *mapping = inode->i_mapping;
2442 block_t blkidx = F2FS_BYTES_TO_BLK(off);
2443 int offset = off & (sb->s_blocksize - 1);
2446 loff_t i_size = i_size_read(inode);
2453 if (off + len > i_size)
2456 while (toread > 0) {
2457 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2459 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2461 if (PTR_ERR(page) == -ENOMEM) {
2462 memalloc_retry_wait(GFP_NOFS);
2465 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2466 return PTR_ERR(page);
2471 if (unlikely(page->mapping != mapping)) {
2472 f2fs_put_page(page, 1);
2475 if (unlikely(!PageUptodate(page))) {
2476 f2fs_put_page(page, 1);
2477 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2481 kaddr = kmap_atomic(page);
2482 memcpy(data, kaddr + offset, tocopy);
2483 kunmap_atomic(kaddr);
2484 f2fs_put_page(page, 1);
2494 /* Write to quotafile */
2495 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2496 const char *data, size_t len, loff_t off)
2498 struct inode *inode = sb_dqopt(sb)->files[type];
2499 struct address_space *mapping = inode->i_mapping;
2500 const struct address_space_operations *a_ops = mapping->a_ops;
2501 int offset = off & (sb->s_blocksize - 1);
2502 size_t towrite = len;
2504 void *fsdata = NULL;
2509 while (towrite > 0) {
2510 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2513 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2515 if (unlikely(err)) {
2516 if (err == -ENOMEM) {
2517 congestion_wait(BLK_RW_ASYNC,
2518 DEFAULT_IO_TIMEOUT);
2521 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2525 kaddr = kmap_atomic(page);
2526 memcpy(kaddr + offset, data, tocopy);
2527 kunmap_atomic(kaddr);
2528 flush_dcache_page(page);
2530 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2541 inode->i_mtime = inode->i_ctime = current_time(inode);
2542 f2fs_mark_inode_dirty_sync(inode, false);
2543 return len - towrite;
2546 int f2fs_dquot_initialize(struct inode *inode)
2548 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) {
2549 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT);
2553 return dquot_initialize(inode);
2556 static struct dquot **f2fs_get_dquots(struct inode *inode)
2558 return F2FS_I(inode)->i_dquot;
2561 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2563 return &F2FS_I(inode)->i_reserved_quota;
2566 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2568 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2569 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2573 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2574 F2FS_OPTION(sbi).s_jquota_fmt, type);
2577 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2582 if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2583 err = f2fs_enable_quotas(sbi->sb);
2585 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2591 for (i = 0; i < MAXQUOTAS; i++) {
2592 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2593 err = f2fs_quota_on_mount(sbi, i);
2598 f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2605 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2608 struct inode *qf_inode;
2609 unsigned long qf_inum;
2612 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2614 qf_inum = f2fs_qf_ino(sb, type);
2618 qf_inode = f2fs_iget(sb, qf_inum);
2619 if (IS_ERR(qf_inode)) {
2620 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2621 return PTR_ERR(qf_inode);
2624 /* Don't account quota for quota files to avoid recursion */
2625 qf_inode->i_flags |= S_NOQUOTA;
2626 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2631 static int f2fs_enable_quotas(struct super_block *sb)
2633 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2635 unsigned long qf_inum;
2636 bool quota_mopt[MAXQUOTAS] = {
2637 test_opt(sbi, USRQUOTA),
2638 test_opt(sbi, GRPQUOTA),
2639 test_opt(sbi, PRJQUOTA),
2642 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2643 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2647 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2649 for (type = 0; type < MAXQUOTAS; type++) {
2650 qf_inum = f2fs_qf_ino(sb, type);
2652 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2653 DQUOT_USAGE_ENABLED |
2654 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2656 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2658 for (type--; type >= 0; type--)
2659 dquot_quota_off(sb, type);
2660 set_sbi_flag(F2FS_SB(sb),
2661 SBI_QUOTA_NEED_REPAIR);
2669 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2671 struct quota_info *dqopt = sb_dqopt(sbi->sb);
2672 struct address_space *mapping = dqopt->files[type]->i_mapping;
2675 ret = dquot_writeback_dquots(sbi->sb, type);
2679 ret = filemap_fdatawrite(mapping);
2683 /* if we are using journalled quota */
2684 if (is_journalled_quota(sbi))
2687 ret = filemap_fdatawait(mapping);
2689 truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2692 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2696 int f2fs_quota_sync(struct super_block *sb, int type)
2698 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2699 struct quota_info *dqopt = sb_dqopt(sb);
2704 * Now when everything is written we can discard the pagecache so
2705 * that userspace sees the changes.
2707 for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2709 if (type != -1 && cnt != type)
2712 if (!sb_has_quota_active(sb, cnt))
2715 inode_lock(dqopt->files[cnt]);
2720 * f2fs_down_read(quota_sem)
2721 * dquot_writeback_dquots()
2724 * f2fs_down_read(quota_sem)
2727 f2fs_down_read(&sbi->quota_sem);
2729 ret = f2fs_quota_sync_file(sbi, cnt);
2731 f2fs_up_read(&sbi->quota_sem);
2732 f2fs_unlock_op(sbi);
2734 inode_unlock(dqopt->files[cnt]);
2742 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2743 const struct path *path)
2745 struct inode *inode;
2748 /* if quota sysfile exists, deny enabling quota with specific file */
2749 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2750 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2754 err = f2fs_quota_sync(sb, type);
2758 err = dquot_quota_on(sb, type, format_id, path);
2762 inode = d_inode(path->dentry);
2765 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2766 f2fs_set_inode_flags(inode);
2767 inode_unlock(inode);
2768 f2fs_mark_inode_dirty_sync(inode, false);
2773 static int __f2fs_quota_off(struct super_block *sb, int type)
2775 struct inode *inode = sb_dqopt(sb)->files[type];
2778 if (!inode || !igrab(inode))
2779 return dquot_quota_off(sb, type);
2781 err = f2fs_quota_sync(sb, type);
2785 err = dquot_quota_off(sb, type);
2786 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2790 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2791 f2fs_set_inode_flags(inode);
2792 inode_unlock(inode);
2793 f2fs_mark_inode_dirty_sync(inode, false);
2799 static int f2fs_quota_off(struct super_block *sb, int type)
2801 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2804 err = __f2fs_quota_off(sb, type);
2807 * quotactl can shutdown journalled quota, result in inconsistence
2808 * between quota record and fs data by following updates, tag the
2809 * flag to let fsck be aware of it.
2811 if (is_journalled_quota(sbi))
2812 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2816 void f2fs_quota_off_umount(struct super_block *sb)
2821 for (type = 0; type < MAXQUOTAS; type++) {
2822 err = __f2fs_quota_off(sb, type);
2824 int ret = dquot_quota_off(sb, type);
2826 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2828 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2832 * In case of checkpoint=disable, we must flush quota blocks.
2833 * This can cause NULL exception for node_inode in end_io, since
2834 * put_super already dropped it.
2836 sync_filesystem(sb);
2839 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2841 struct quota_info *dqopt = sb_dqopt(sb);
2844 for (type = 0; type < MAXQUOTAS; type++) {
2845 if (!dqopt->files[type])
2847 f2fs_inode_synced(dqopt->files[type]);
2851 static int f2fs_dquot_commit(struct dquot *dquot)
2853 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2856 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2857 ret = dquot_commit(dquot);
2859 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2860 f2fs_up_read(&sbi->quota_sem);
2864 static int f2fs_dquot_acquire(struct dquot *dquot)
2866 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2869 f2fs_down_read(&sbi->quota_sem);
2870 ret = dquot_acquire(dquot);
2872 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2873 f2fs_up_read(&sbi->quota_sem);
2877 static int f2fs_dquot_release(struct dquot *dquot)
2879 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2880 int ret = dquot_release(dquot);
2883 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2887 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2889 struct super_block *sb = dquot->dq_sb;
2890 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2891 int ret = dquot_mark_dquot_dirty(dquot);
2893 /* if we are using journalled quota */
2894 if (is_journalled_quota(sbi))
2895 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2900 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2902 struct f2fs_sb_info *sbi = F2FS_SB(sb);
2903 int ret = dquot_commit_info(sb, type);
2906 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2910 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2912 *projid = F2FS_I(inode)->i_projid;
2916 static const struct dquot_operations f2fs_quota_operations = {
2917 .get_reserved_space = f2fs_get_reserved_space,
2918 .write_dquot = f2fs_dquot_commit,
2919 .acquire_dquot = f2fs_dquot_acquire,
2920 .release_dquot = f2fs_dquot_release,
2921 .mark_dirty = f2fs_dquot_mark_dquot_dirty,
2922 .write_info = f2fs_dquot_commit_info,
2923 .alloc_dquot = dquot_alloc,
2924 .destroy_dquot = dquot_destroy,
2925 .get_projid = f2fs_get_projid,
2926 .get_next_id = dquot_get_next_id,
2929 static const struct quotactl_ops f2fs_quotactl_ops = {
2930 .quota_on = f2fs_quota_on,
2931 .quota_off = f2fs_quota_off,
2932 .quota_sync = f2fs_quota_sync,
2933 .get_state = dquot_get_state,
2934 .set_info = dquot_set_dqinfo,
2935 .get_dqblk = dquot_get_dqblk,
2936 .set_dqblk = dquot_set_dqblk,
2937 .get_nextdqblk = dquot_get_next_dqblk,
2940 int f2fs_dquot_initialize(struct inode *inode)
2945 int f2fs_quota_sync(struct super_block *sb, int type)
2950 void f2fs_quota_off_umount(struct super_block *sb)
2955 static const struct super_operations f2fs_sops = {
2956 .alloc_inode = f2fs_alloc_inode,
2957 .free_inode = f2fs_free_inode,
2958 .drop_inode = f2fs_drop_inode,
2959 .write_inode = f2fs_write_inode,
2960 .dirty_inode = f2fs_dirty_inode,
2961 .show_options = f2fs_show_options,
2963 .quota_read = f2fs_quota_read,
2964 .quota_write = f2fs_quota_write,
2965 .get_dquots = f2fs_get_dquots,
2967 .evict_inode = f2fs_evict_inode,
2968 .put_super = f2fs_put_super,
2969 .sync_fs = f2fs_sync_fs,
2970 .freeze_fs = f2fs_freeze,
2971 .unfreeze_fs = f2fs_unfreeze,
2972 .statfs = f2fs_statfs,
2973 .remount_fs = f2fs_remount,
2976 #ifdef CONFIG_FS_ENCRYPTION
2977 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2979 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2980 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2984 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2987 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2990 * Encrypting the root directory is not allowed because fsck
2991 * expects lost+found directory to exist and remain unencrypted
2992 * if LOST_FOUND feature is enabled.
2995 if (f2fs_sb_has_lost_found(sbi) &&
2996 inode->i_ino == F2FS_ROOT_INO(sbi))
2999 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3000 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3001 ctx, len, fs_data, XATTR_CREATE);
3004 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3006 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3009 static bool f2fs_has_stable_inodes(struct super_block *sb)
3014 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3015 int *ino_bits_ret, int *lblk_bits_ret)
3017 *ino_bits_ret = 8 * sizeof(nid_t);
3018 *lblk_bits_ret = 8 * sizeof(block_t);
3021 static int f2fs_get_num_devices(struct super_block *sb)
3023 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3025 if (f2fs_is_multi_device(sbi))
3026 return sbi->s_ndevs;
3030 static void f2fs_get_devices(struct super_block *sb,
3031 struct request_queue **devs)
3033 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3036 for (i = 0; i < sbi->s_ndevs; i++)
3037 devs[i] = bdev_get_queue(FDEV(i).bdev);
3040 static const struct fscrypt_operations f2fs_cryptops = {
3041 .key_prefix = "f2fs:",
3042 .get_context = f2fs_get_context,
3043 .set_context = f2fs_set_context,
3044 .get_dummy_policy = f2fs_get_dummy_policy,
3045 .empty_dir = f2fs_empty_dir,
3046 .has_stable_inodes = f2fs_has_stable_inodes,
3047 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits,
3048 .get_num_devices = f2fs_get_num_devices,
3049 .get_devices = f2fs_get_devices,
3053 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3054 u64 ino, u32 generation)
3056 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3057 struct inode *inode;
3059 if (f2fs_check_nid_range(sbi, ino))
3060 return ERR_PTR(-ESTALE);
3063 * f2fs_iget isn't quite right if the inode is currently unallocated!
3064 * However f2fs_iget currently does appropriate checks to handle stale
3065 * inodes so everything is OK.
3067 inode = f2fs_iget(sb, ino);
3069 return ERR_CAST(inode);
3070 if (unlikely(generation && inode->i_generation != generation)) {
3071 /* we didn't find the right inode.. */
3073 return ERR_PTR(-ESTALE);
3078 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3079 int fh_len, int fh_type)
3081 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3082 f2fs_nfs_get_inode);
3085 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3086 int fh_len, int fh_type)
3088 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3089 f2fs_nfs_get_inode);
3092 static const struct export_operations f2fs_export_ops = {
3093 .fh_to_dentry = f2fs_fh_to_dentry,
3094 .fh_to_parent = f2fs_fh_to_parent,
3095 .get_parent = f2fs_get_parent,
3098 loff_t max_file_blocks(struct inode *inode)
3104 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3105 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3106 * space in inode.i_addr, it will be more safe to reassign
3110 if (inode && f2fs_compressed_file(inode))
3111 leaf_count = ADDRS_PER_BLOCK(inode);
3113 leaf_count = DEF_ADDRS_PER_BLOCK;
3115 /* two direct node blocks */
3116 result += (leaf_count * 2);
3118 /* two indirect node blocks */
3119 leaf_count *= NIDS_PER_BLOCK;
3120 result += (leaf_count * 2);
3122 /* one double indirect node block */
3123 leaf_count *= NIDS_PER_BLOCK;
3124 result += leaf_count;
3129 static int __f2fs_commit_super(struct buffer_head *bh,
3130 struct f2fs_super_block *super)
3134 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3135 set_buffer_dirty(bh);
3138 /* it's rare case, we can do fua all the time */
3139 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3142 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3143 struct buffer_head *bh)
3145 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3146 (bh->b_data + F2FS_SUPER_OFFSET);
3147 struct super_block *sb = sbi->sb;
3148 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3149 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3150 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3151 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3152 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3153 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3154 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3155 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3156 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3157 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3158 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3159 u32 segment_count = le32_to_cpu(raw_super->segment_count);
3160 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3161 u64 main_end_blkaddr = main_blkaddr +
3162 (segment_count_main << log_blocks_per_seg);
3163 u64 seg_end_blkaddr = segment0_blkaddr +
3164 (segment_count << log_blocks_per_seg);
3166 if (segment0_blkaddr != cp_blkaddr) {
3167 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3168 segment0_blkaddr, cp_blkaddr);
3172 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3174 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3175 cp_blkaddr, sit_blkaddr,
3176 segment_count_ckpt << log_blocks_per_seg);
3180 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3182 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3183 sit_blkaddr, nat_blkaddr,
3184 segment_count_sit << log_blocks_per_seg);
3188 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3190 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3191 nat_blkaddr, ssa_blkaddr,
3192 segment_count_nat << log_blocks_per_seg);
3196 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3198 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3199 ssa_blkaddr, main_blkaddr,
3200 segment_count_ssa << log_blocks_per_seg);
3204 if (main_end_blkaddr > seg_end_blkaddr) {
3205 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3206 main_blkaddr, seg_end_blkaddr,
3207 segment_count_main << log_blocks_per_seg);
3209 } else if (main_end_blkaddr < seg_end_blkaddr) {
3213 /* fix in-memory information all the time */
3214 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3215 segment0_blkaddr) >> log_blocks_per_seg);
3217 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3218 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3221 err = __f2fs_commit_super(bh, NULL);
3222 res = err ? "failed" : "done";
3224 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3225 res, main_blkaddr, seg_end_blkaddr,
3226 segment_count_main << log_blocks_per_seg);
3233 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3234 struct buffer_head *bh)
3236 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3237 block_t total_sections, blocks_per_seg;
3238 struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3239 (bh->b_data + F2FS_SUPER_OFFSET);
3240 size_t crc_offset = 0;
3243 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3244 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3245 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3249 /* Check checksum_offset and crc in superblock */
3250 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3251 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3253 offsetof(struct f2fs_super_block, crc)) {
3254 f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3256 return -EFSCORRUPTED;
3258 crc = le32_to_cpu(raw_super->crc);
3259 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3260 f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3261 return -EFSCORRUPTED;
3265 /* Currently, support only 4KB block size */
3266 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3267 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3268 le32_to_cpu(raw_super->log_blocksize),
3270 return -EFSCORRUPTED;
3273 /* check log blocks per segment */
3274 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3275 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3276 le32_to_cpu(raw_super->log_blocks_per_seg));
3277 return -EFSCORRUPTED;
3280 /* Currently, support 512/1024/2048/4096 bytes sector size */
3281 if (le32_to_cpu(raw_super->log_sectorsize) >
3282 F2FS_MAX_LOG_SECTOR_SIZE ||
3283 le32_to_cpu(raw_super->log_sectorsize) <
3284 F2FS_MIN_LOG_SECTOR_SIZE) {
3285 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3286 le32_to_cpu(raw_super->log_sectorsize));
3287 return -EFSCORRUPTED;
3289 if (le32_to_cpu(raw_super->log_sectors_per_block) +
3290 le32_to_cpu(raw_super->log_sectorsize) !=
3291 F2FS_MAX_LOG_SECTOR_SIZE) {
3292 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3293 le32_to_cpu(raw_super->log_sectors_per_block),
3294 le32_to_cpu(raw_super->log_sectorsize));
3295 return -EFSCORRUPTED;
3298 segment_count = le32_to_cpu(raw_super->segment_count);
3299 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3300 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3301 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3302 total_sections = le32_to_cpu(raw_super->section_count);
3304 /* blocks_per_seg should be 512, given the above check */
3305 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3307 if (segment_count > F2FS_MAX_SEGMENT ||
3308 segment_count < F2FS_MIN_SEGMENTS) {
3309 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3310 return -EFSCORRUPTED;
3313 if (total_sections > segment_count_main || total_sections < 1 ||
3314 segs_per_sec > segment_count || !segs_per_sec) {
3315 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3316 segment_count, total_sections, segs_per_sec);
3317 return -EFSCORRUPTED;
3320 if (segment_count_main != total_sections * segs_per_sec) {
3321 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3322 segment_count_main, total_sections, segs_per_sec);
3323 return -EFSCORRUPTED;
3326 if ((segment_count / segs_per_sec) < total_sections) {
3327 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3328 segment_count, segs_per_sec, total_sections);
3329 return -EFSCORRUPTED;
3332 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3333 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3334 segment_count, le64_to_cpu(raw_super->block_count));
3335 return -EFSCORRUPTED;
3338 if (RDEV(0).path[0]) {
3339 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3342 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3343 dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3346 if (segment_count != dev_seg_count) {
3347 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3348 segment_count, dev_seg_count);
3349 return -EFSCORRUPTED;
3352 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3353 !bdev_is_zoned(sbi->sb->s_bdev)) {
3354 f2fs_info(sbi, "Zoned block device path is missing");
3355 return -EFSCORRUPTED;
3359 if (secs_per_zone > total_sections || !secs_per_zone) {
3360 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3361 secs_per_zone, total_sections);
3362 return -EFSCORRUPTED;
3364 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3365 raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3366 (le32_to_cpu(raw_super->extension_count) +
3367 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3368 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3369 le32_to_cpu(raw_super->extension_count),
3370 raw_super->hot_ext_count,
3371 F2FS_MAX_EXTENSION);
3372 return -EFSCORRUPTED;
3375 if (le32_to_cpu(raw_super->cp_payload) >=
3376 (blocks_per_seg - F2FS_CP_PACKS -
3377 NR_CURSEG_PERSIST_TYPE)) {
3378 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3379 le32_to_cpu(raw_super->cp_payload),
3380 blocks_per_seg - F2FS_CP_PACKS -
3381 NR_CURSEG_PERSIST_TYPE);
3382 return -EFSCORRUPTED;
3385 /* check reserved ino info */
3386 if (le32_to_cpu(raw_super->node_ino) != 1 ||
3387 le32_to_cpu(raw_super->meta_ino) != 2 ||
3388 le32_to_cpu(raw_super->root_ino) != 3) {
3389 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3390 le32_to_cpu(raw_super->node_ino),
3391 le32_to_cpu(raw_super->meta_ino),
3392 le32_to_cpu(raw_super->root_ino));
3393 return -EFSCORRUPTED;
3396 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3397 if (sanity_check_area_boundary(sbi, bh))
3398 return -EFSCORRUPTED;
3403 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3405 unsigned int total, fsmeta;
3406 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3407 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3408 unsigned int ovp_segments, reserved_segments;
3409 unsigned int main_segs, blocks_per_seg;
3410 unsigned int sit_segs, nat_segs;
3411 unsigned int sit_bitmap_size, nat_bitmap_size;
3412 unsigned int log_blocks_per_seg;
3413 unsigned int segment_count_main;
3414 unsigned int cp_pack_start_sum, cp_payload;
3415 block_t user_block_count, valid_user_blocks;
3416 block_t avail_node_count, valid_node_count;
3417 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3420 total = le32_to_cpu(raw_super->segment_count);
3421 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3422 sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3424 nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3426 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3427 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3429 if (unlikely(fsmeta >= total))
3432 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3433 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3435 if (!f2fs_sb_has_readonly(sbi) &&
3436 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3437 ovp_segments == 0 || reserved_segments == 0)) {
3438 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3441 user_block_count = le64_to_cpu(ckpt->user_block_count);
3442 segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3443 (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3444 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3445 if (!user_block_count || user_block_count >=
3446 segment_count_main << log_blocks_per_seg) {
3447 f2fs_err(sbi, "Wrong user_block_count: %u",
3452 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3453 if (valid_user_blocks > user_block_count) {
3454 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3455 valid_user_blocks, user_block_count);
3459 valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3460 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3461 if (valid_node_count > avail_node_count) {
3462 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3463 valid_node_count, avail_node_count);
3467 main_segs = le32_to_cpu(raw_super->segment_count_main);
3468 blocks_per_seg = sbi->blocks_per_seg;
3470 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3471 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3472 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3475 if (f2fs_sb_has_readonly(sbi))
3478 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3479 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3480 le32_to_cpu(ckpt->cur_node_segno[j])) {
3481 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3483 le32_to_cpu(ckpt->cur_node_segno[i]));
3489 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3490 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3491 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3494 if (f2fs_sb_has_readonly(sbi))
3497 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3498 if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3499 le32_to_cpu(ckpt->cur_data_segno[j])) {
3500 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3502 le32_to_cpu(ckpt->cur_data_segno[i]));
3507 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3508 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3509 if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3510 le32_to_cpu(ckpt->cur_data_segno[j])) {
3511 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3513 le32_to_cpu(ckpt->cur_node_segno[i]));
3519 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3520 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3522 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3523 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3524 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3525 sit_bitmap_size, nat_bitmap_size);
3529 cp_pack_start_sum = __start_sum_addr(sbi);
3530 cp_payload = __cp_payload(sbi);
3531 if (cp_pack_start_sum < cp_payload + 1 ||
3532 cp_pack_start_sum > blocks_per_seg - 1 -
3533 NR_CURSEG_PERSIST_TYPE) {
3534 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3539 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3540 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3541 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3542 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3543 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3544 le32_to_cpu(ckpt->checksum_offset));
3548 nat_blocks = nat_segs << log_blocks_per_seg;
3549 nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3550 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3551 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3552 (cp_payload + F2FS_CP_PACKS +
3553 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3554 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3555 cp_payload, nat_bits_blocks);
3559 if (unlikely(f2fs_cp_error(sbi))) {
3560 f2fs_err(sbi, "A bug case: need to run fsck");
3566 static void init_sb_info(struct f2fs_sb_info *sbi)
3568 struct f2fs_super_block *raw_super = sbi->raw_super;
3571 sbi->log_sectors_per_block =
3572 le32_to_cpu(raw_super->log_sectors_per_block);
3573 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3574 sbi->blocksize = 1 << sbi->log_blocksize;
3575 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3576 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3577 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3578 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3579 sbi->total_sections = le32_to_cpu(raw_super->section_count);
3580 sbi->total_node_count =
3581 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3582 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3583 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3584 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3585 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3586 sbi->cur_victim_sec = NULL_SECNO;
3587 sbi->gc_mode = GC_NORMAL;
3588 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3589 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3590 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3591 sbi->migration_granularity = sbi->segs_per_sec;
3592 sbi->seq_file_ra_mul = MIN_RA_MUL;
3593 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3594 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3595 spin_lock_init(&sbi->gc_urgent_high_lock);
3597 sbi->dir_level = DEF_DIR_LEVEL;
3598 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3599 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3600 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3601 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3602 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3603 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3604 DEF_UMOUNT_DISCARD_TIMEOUT;
3605 clear_sbi_flag(sbi, SBI_NEED_FSCK);
3607 for (i = 0; i < NR_COUNT_TYPE; i++)
3608 atomic_set(&sbi->nr_pages[i], 0);
3610 for (i = 0; i < META; i++)
3611 atomic_set(&sbi->wb_sync_req[i], 0);
3613 INIT_LIST_HEAD(&sbi->s_list);
3614 mutex_init(&sbi->umount_mutex);
3615 init_f2fs_rwsem(&sbi->io_order_lock);
3616 spin_lock_init(&sbi->cp_lock);
3618 sbi->dirty_device = 0;
3619 spin_lock_init(&sbi->dev_lock);
3621 init_f2fs_rwsem(&sbi->sb_lock);
3622 init_f2fs_rwsem(&sbi->pin_sem);
3625 static int init_percpu_info(struct f2fs_sb_info *sbi)
3629 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3633 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3635 goto err_valid_block;
3637 err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3640 goto err_node_block;
3644 percpu_counter_destroy(&sbi->rf_node_block_count);
3646 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3650 #ifdef CONFIG_BLK_DEV_ZONED
3652 struct f2fs_report_zones_args {
3653 struct f2fs_dev_info *dev;
3654 bool zone_cap_mismatch;
3657 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3660 struct f2fs_report_zones_args *rz_args = data;
3662 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3665 set_bit(idx, rz_args->dev->blkz_seq);
3666 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >>
3667 F2FS_LOG_SECTORS_PER_BLOCK;
3668 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch)
3669 rz_args->zone_cap_mismatch = true;
3674 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3676 struct block_device *bdev = FDEV(devi).bdev;
3677 sector_t nr_sectors = bdev_nr_sectors(bdev);
3678 struct f2fs_report_zones_args rep_zone_arg;
3681 if (!f2fs_sb_has_blkzoned(sbi))
3684 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3685 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3687 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3688 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3689 __ilog2_u32(sbi->blocks_per_blkz))
3691 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3692 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3693 sbi->log_blocks_per_blkz;
3694 if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3695 FDEV(devi).nr_blkz++;
3697 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3698 BITS_TO_LONGS(FDEV(devi).nr_blkz)
3699 * sizeof(unsigned long),
3701 if (!FDEV(devi).blkz_seq)
3704 /* Get block zones type and zone-capacity */
3705 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi,
3706 FDEV(devi).nr_blkz * sizeof(block_t),
3708 if (!FDEV(devi).zone_capacity_blocks)
3711 rep_zone_arg.dev = &FDEV(devi);
3712 rep_zone_arg.zone_cap_mismatch = false;
3714 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3719 if (!rep_zone_arg.zone_cap_mismatch) {
3720 kfree(FDEV(devi).zone_capacity_blocks);
3721 FDEV(devi).zone_capacity_blocks = NULL;
3729 * Read f2fs raw super block.
3730 * Because we have two copies of super block, so read both of them
3731 * to get the first valid one. If any one of them is broken, we pass
3732 * them recovery flag back to the caller.
3734 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3735 struct f2fs_super_block **raw_super,
3736 int *valid_super_block, int *recovery)
3738 struct super_block *sb = sbi->sb;
3740 struct buffer_head *bh;
3741 struct f2fs_super_block *super;
3744 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3748 for (block = 0; block < 2; block++) {
3749 bh = sb_bread(sb, block);
3751 f2fs_err(sbi, "Unable to read %dth superblock",
3758 /* sanity checking of raw super */
3759 err = sanity_check_raw_super(sbi, bh);
3761 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3769 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3771 *valid_super_block = block;
3777 /* No valid superblock */
3786 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3788 struct buffer_head *bh;
3792 if ((recover && f2fs_readonly(sbi->sb)) ||
3793 bdev_read_only(sbi->sb->s_bdev)) {
3794 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3798 /* we should update superblock crc here */
3799 if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3800 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3801 offsetof(struct f2fs_super_block, crc));
3802 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3805 /* write back-up superblock first */
3806 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3809 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3812 /* if we are in recovery path, skip writing valid superblock */
3816 /* write current valid superblock */
3817 bh = sb_bread(sbi->sb, sbi->valid_super_block);
3820 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3825 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3827 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3828 unsigned int max_devices = MAX_DEVICES;
3829 unsigned int logical_blksize;
3832 /* Initialize single device information */
3833 if (!RDEV(0).path[0]) {
3834 if (!bdev_is_zoned(sbi->sb->s_bdev))
3840 * Initialize multiple devices information, or single
3841 * zoned block device information.
3843 sbi->devs = f2fs_kzalloc(sbi,
3844 array_size(max_devices,
3845 sizeof(struct f2fs_dev_info)),
3850 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3851 sbi->aligned_blksize = true;
3853 for (i = 0; i < max_devices; i++) {
3855 if (i > 0 && !RDEV(i).path[0])
3858 if (max_devices == 1) {
3859 /* Single zoned block device mount */
3861 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3862 sbi->sb->s_mode, sbi->sb->s_type);
3864 /* Multi-device mount */
3865 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3866 FDEV(i).total_segments =
3867 le32_to_cpu(RDEV(i).total_segments);
3869 FDEV(i).start_blk = 0;
3870 FDEV(i).end_blk = FDEV(i).start_blk +
3871 (FDEV(i).total_segments <<
3872 sbi->log_blocks_per_seg) - 1 +
3873 le32_to_cpu(raw_super->segment0_blkaddr);
3875 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3876 FDEV(i).end_blk = FDEV(i).start_blk +
3877 (FDEV(i).total_segments <<
3878 sbi->log_blocks_per_seg) - 1;
3880 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3881 sbi->sb->s_mode, sbi->sb->s_type);
3883 if (IS_ERR(FDEV(i).bdev))
3884 return PTR_ERR(FDEV(i).bdev);
3886 /* to release errored devices */
3887 sbi->s_ndevs = i + 1;
3889 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
3890 sbi->aligned_blksize = false;
3892 #ifdef CONFIG_BLK_DEV_ZONED
3893 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3894 !f2fs_sb_has_blkzoned(sbi)) {
3895 f2fs_err(sbi, "Zoned block device feature not enabled");
3898 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3899 if (init_blkz_info(sbi, i)) {
3900 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3903 if (max_devices == 1)
3905 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3907 FDEV(i).total_segments,
3908 FDEV(i).start_blk, FDEV(i).end_blk,
3909 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3910 "Host-aware" : "Host-managed");
3914 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3916 FDEV(i).total_segments,
3917 FDEV(i).start_blk, FDEV(i).end_blk);
3920 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3924 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3926 #if IS_ENABLED(CONFIG_UNICODE)
3927 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3928 const struct f2fs_sb_encodings *encoding_info;
3929 struct unicode_map *encoding;
3930 __u16 encoding_flags;
3932 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
3933 if (!encoding_info) {
3935 "Encoding requested by superblock is unknown");
3939 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
3940 encoding = utf8_load(encoding_info->version);
3941 if (IS_ERR(encoding)) {
3943 "can't mount with superblock charset: %s-%u.%u.%u "
3944 "not supported by the kernel. flags: 0x%x.",
3945 encoding_info->name,
3946 unicode_major(encoding_info->version),
3947 unicode_minor(encoding_info->version),
3948 unicode_rev(encoding_info->version),
3950 return PTR_ERR(encoding);
3952 f2fs_info(sbi, "Using encoding defined by superblock: "
3953 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
3954 unicode_major(encoding_info->version),
3955 unicode_minor(encoding_info->version),
3956 unicode_rev(encoding_info->version),
3959 sbi->sb->s_encoding = encoding;
3960 sbi->sb->s_encoding_flags = encoding_flags;
3963 if (f2fs_sb_has_casefold(sbi)) {
3964 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3971 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3973 struct f2fs_sm_info *sm_i = SM_I(sbi);
3975 /* adjust parameters according to the volume size */
3976 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3977 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3978 if (f2fs_block_unit_discard(sbi))
3979 sm_i->dcc_info->discard_granularity = 1;
3980 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE |
3981 1 << F2FS_IPU_HONOR_OPU_WRITE;
3984 sbi->readdir_ra = 1;
3987 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3989 struct f2fs_sb_info *sbi;
3990 struct f2fs_super_block *raw_super;
3993 bool skip_recovery = false, need_fsck = false;
3994 char *options = NULL;
3995 int recovery, i, valid_super_block;
3996 struct curseg_info *seg_i;
4002 valid_super_block = -1;
4005 /* allocate memory for f2fs-specific super block info */
4006 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4012 /* Load the checksum driver */
4013 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4014 if (IS_ERR(sbi->s_chksum_driver)) {
4015 f2fs_err(sbi, "Cannot load crc32 driver.");
4016 err = PTR_ERR(sbi->s_chksum_driver);
4017 sbi->s_chksum_driver = NULL;
4021 /* set a block size */
4022 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4023 f2fs_err(sbi, "unable to set blocksize");
4027 err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4032 sb->s_fs_info = sbi;
4033 sbi->raw_super = raw_super;
4035 /* precompute checksum seed for metadata */
4036 if (f2fs_sb_has_inode_chksum(sbi))
4037 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4038 sizeof(raw_super->uuid));
4040 default_options(sbi);
4041 /* parse mount options */
4042 options = kstrdup((const char *)data, GFP_KERNEL);
4043 if (data && !options) {
4048 err = parse_options(sb, options, false);
4052 sb->s_maxbytes = max_file_blocks(NULL) <<
4053 le32_to_cpu(raw_super->log_blocksize);
4054 sb->s_max_links = F2FS_LINK_MAX;
4056 err = f2fs_setup_casefold(sbi);
4061 sb->dq_op = &f2fs_quota_operations;
4062 sb->s_qcop = &f2fs_quotactl_ops;
4063 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4065 if (f2fs_sb_has_quota_ino(sbi)) {
4066 for (i = 0; i < MAXQUOTAS; i++) {
4067 if (f2fs_qf_ino(sbi->sb, i))
4068 sbi->nquota_files++;
4073 sb->s_op = &f2fs_sops;
4074 #ifdef CONFIG_FS_ENCRYPTION
4075 sb->s_cop = &f2fs_cryptops;
4077 #ifdef CONFIG_FS_VERITY
4078 sb->s_vop = &f2fs_verityops;
4080 sb->s_xattr = f2fs_xattr_handlers;
4081 sb->s_export_op = &f2fs_export_ops;
4082 sb->s_magic = F2FS_SUPER_MAGIC;
4083 sb->s_time_gran = 1;
4084 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4085 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4086 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4087 sb->s_iflags |= SB_I_CGROUPWB;
4089 /* init f2fs-specific super block info */
4090 sbi->valid_super_block = valid_super_block;
4091 init_f2fs_rwsem(&sbi->gc_lock);
4092 mutex_init(&sbi->writepages);
4093 init_f2fs_rwsem(&sbi->cp_global_sem);
4094 init_f2fs_rwsem(&sbi->node_write);
4095 init_f2fs_rwsem(&sbi->node_change);
4097 /* disallow all the data/node/meta page writes */
4098 set_sbi_flag(sbi, SBI_POR_DOING);
4099 spin_lock_init(&sbi->stat_lock);
4101 for (i = 0; i < NR_PAGE_TYPE; i++) {
4102 int n = (i == META) ? 1 : NR_TEMP_TYPE;
4108 sizeof(struct f2fs_bio_info)),
4110 if (!sbi->write_io[i]) {
4115 for (j = HOT; j < n; j++) {
4116 init_f2fs_rwsem(&sbi->write_io[i][j].io_rwsem);
4117 sbi->write_io[i][j].sbi = sbi;
4118 sbi->write_io[i][j].bio = NULL;
4119 spin_lock_init(&sbi->write_io[i][j].io_lock);
4120 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
4121 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
4122 init_f2fs_rwsem(&sbi->write_io[i][j].bio_list_lock);
4126 init_f2fs_rwsem(&sbi->cp_rwsem);
4127 init_f2fs_rwsem(&sbi->quota_sem);
4128 init_waitqueue_head(&sbi->cp_wait);
4131 err = f2fs_init_iostat(sbi);
4135 err = init_percpu_info(sbi);
4139 if (F2FS_IO_ALIGNED(sbi)) {
4140 sbi->write_io_dummy =
4141 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4142 if (!sbi->write_io_dummy) {
4148 /* init per sbi slab cache */
4149 err = f2fs_init_xattr_caches(sbi);
4152 err = f2fs_init_page_array_cache(sbi);
4154 goto free_xattr_cache;
4156 /* get an inode for meta space */
4157 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4158 if (IS_ERR(sbi->meta_inode)) {
4159 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4160 err = PTR_ERR(sbi->meta_inode);
4161 goto free_page_array_cache;
4164 err = f2fs_get_valid_checkpoint(sbi);
4166 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4167 goto free_meta_inode;
4170 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4171 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4172 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4173 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4174 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4177 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4178 set_sbi_flag(sbi, SBI_NEED_FSCK);
4180 /* Initialize device list */
4181 err = f2fs_scan_devices(sbi);
4183 f2fs_err(sbi, "Failed to find devices");
4187 err = f2fs_init_post_read_wq(sbi);
4189 f2fs_err(sbi, "Failed to initialize post read workqueue");
4193 sbi->total_valid_node_count =
4194 le32_to_cpu(sbi->ckpt->valid_node_count);
4195 percpu_counter_set(&sbi->total_valid_inode_count,
4196 le32_to_cpu(sbi->ckpt->valid_inode_count));
4197 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4198 sbi->total_valid_block_count =
4199 le64_to_cpu(sbi->ckpt->valid_block_count);
4200 sbi->last_valid_block_count = sbi->total_valid_block_count;
4201 sbi->reserved_blocks = 0;
4202 sbi->current_reserved_blocks = 0;
4203 limit_reserve_root(sbi);
4204 adjust_unusable_cap_perc(sbi);
4206 for (i = 0; i < NR_INODE_TYPE; i++) {
4207 INIT_LIST_HEAD(&sbi->inode_list[i]);
4208 spin_lock_init(&sbi->inode_lock[i]);
4210 mutex_init(&sbi->flush_lock);
4212 f2fs_init_extent_cache_info(sbi);
4214 f2fs_init_ino_entry_info(sbi);
4216 f2fs_init_fsync_node_info(sbi);
4218 /* setup checkpoint request control and start checkpoint issue thread */
4219 f2fs_init_ckpt_req_control(sbi);
4220 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4221 test_opt(sbi, MERGE_CHECKPOINT)) {
4222 err = f2fs_start_ckpt_thread(sbi);
4225 "Failed to start F2FS issue_checkpoint_thread (%d)",
4227 goto stop_ckpt_thread;
4231 /* setup f2fs internal modules */
4232 err = f2fs_build_segment_manager(sbi);
4234 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4238 err = f2fs_build_node_manager(sbi);
4240 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4245 err = adjust_reserved_segment(sbi);
4249 /* For write statistics */
4250 sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4252 /* Read accumulated write IO statistics if exists */
4253 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4254 if (__exist_node_summaries(sbi))
4255 sbi->kbytes_written =
4256 le64_to_cpu(seg_i->journal->info.kbytes_written);
4258 f2fs_build_gc_manager(sbi);
4260 err = f2fs_build_stats(sbi);
4264 /* get an inode for node space */
4265 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4266 if (IS_ERR(sbi->node_inode)) {
4267 f2fs_err(sbi, "Failed to read node inode");
4268 err = PTR_ERR(sbi->node_inode);
4272 /* read root inode and dentry */
4273 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4275 f2fs_err(sbi, "Failed to read root inode");
4276 err = PTR_ERR(root);
4277 goto free_node_inode;
4279 if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4280 !root->i_size || !root->i_nlink) {
4283 goto free_node_inode;
4286 sb->s_root = d_make_root(root); /* allocate root dentry */
4289 goto free_node_inode;
4292 err = f2fs_init_compress_inode(sbi);
4294 goto free_root_inode;
4296 err = f2fs_register_sysfs(sbi);
4298 goto free_compress_inode;
4301 /* Enable quota usage during mount */
4302 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4303 err = f2fs_enable_quotas(sb);
4305 f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4308 /* if there are any orphan inodes, free them */
4309 err = f2fs_recover_orphan_inodes(sbi);
4313 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4314 goto reset_checkpoint;
4316 /* recover fsynced data */
4317 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4318 !test_opt(sbi, NORECOVERY)) {
4320 * mount should be failed, when device has readonly mode, and
4321 * previous checkpoint was not done by clean system shutdown.
4323 if (f2fs_hw_is_readonly(sbi)) {
4324 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4325 err = f2fs_recover_fsync_data(sbi, true);
4328 f2fs_err(sbi, "Need to recover fsync data, but "
4329 "write access unavailable, please try "
4330 "mount w/ disable_roll_forward or norecovery");
4335 f2fs_info(sbi, "write access unavailable, skipping recovery");
4336 goto reset_checkpoint;
4340 set_sbi_flag(sbi, SBI_NEED_FSCK);
4343 goto reset_checkpoint;
4345 err = f2fs_recover_fsync_data(sbi, false);
4348 skip_recovery = true;
4350 f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4355 err = f2fs_recover_fsync_data(sbi, true);
4357 if (!f2fs_readonly(sb) && err > 0) {
4359 f2fs_err(sbi, "Need to recover fsync data");
4365 * If the f2fs is not readonly and fsync data recovery succeeds,
4366 * check zoned block devices' write pointer consistency.
4368 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4369 err = f2fs_check_write_pointer(sbi);
4375 f2fs_init_inmem_curseg(sbi);
4377 /* f2fs_recover_fsync_data() cleared this already */
4378 clear_sbi_flag(sbi, SBI_POR_DOING);
4380 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4381 err = f2fs_disable_checkpoint(sbi);
4383 goto sync_free_meta;
4384 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4385 f2fs_enable_checkpoint(sbi);
4389 * If filesystem is not mounted as read-only then
4390 * do start the gc_thread.
4392 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4393 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4394 /* After POR, we can run background GC thread.*/
4395 err = f2fs_start_gc_thread(sbi);
4397 goto sync_free_meta;
4401 /* recover broken superblock */
4403 err = f2fs_commit_super(sbi, true);
4404 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4405 sbi->valid_super_block ? 1 : 2, err);
4408 f2fs_join_shrinker(sbi);
4410 f2fs_tuning_parameters(sbi);
4412 f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4413 cur_cp_version(F2FS_CKPT(sbi)));
4414 f2fs_update_time(sbi, CP_TIME);
4415 f2fs_update_time(sbi, REQ_TIME);
4416 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4420 /* safe to flush all the data */
4421 sync_filesystem(sbi->sb);
4426 f2fs_truncate_quota_inode_pages(sb);
4427 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4428 f2fs_quota_off_umount(sbi->sb);
4431 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4432 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4433 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4434 * falls into an infinite loop in f2fs_sync_meta_pages().
4436 truncate_inode_pages_final(META_MAPPING(sbi));
4437 /* evict some inodes being cached by GC */
4439 f2fs_unregister_sysfs(sbi);
4440 free_compress_inode:
4441 f2fs_destroy_compress_inode(sbi);
4446 f2fs_release_ino_entry(sbi, true);
4447 truncate_inode_pages_final(NODE_MAPPING(sbi));
4448 iput(sbi->node_inode);
4449 sbi->node_inode = NULL;
4451 f2fs_destroy_stats(sbi);
4453 /* stop discard thread before destroying node manager */
4454 f2fs_stop_discard_thread(sbi);
4455 f2fs_destroy_node_manager(sbi);
4457 f2fs_destroy_segment_manager(sbi);
4458 f2fs_destroy_post_read_wq(sbi);
4460 f2fs_stop_ckpt_thread(sbi);
4462 destroy_device_list(sbi);
4465 make_bad_inode(sbi->meta_inode);
4466 iput(sbi->meta_inode);
4467 sbi->meta_inode = NULL;
4468 free_page_array_cache:
4469 f2fs_destroy_page_array_cache(sbi);
4471 f2fs_destroy_xattr_caches(sbi);
4473 mempool_destroy(sbi->write_io_dummy);
4475 destroy_percpu_info(sbi);
4477 f2fs_destroy_iostat(sbi);
4479 for (i = 0; i < NR_PAGE_TYPE; i++)
4480 kvfree(sbi->write_io[i]);
4482 #if IS_ENABLED(CONFIG_UNICODE)
4483 utf8_unload(sb->s_encoding);
4484 sb->s_encoding = NULL;
4488 for (i = 0; i < MAXQUOTAS; i++)
4489 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4491 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4496 if (sbi->s_chksum_driver)
4497 crypto_free_shash(sbi->s_chksum_driver);
4500 /* give only one another chance */
4501 if (retry_cnt > 0 && skip_recovery) {
4503 shrink_dcache_sb(sb);
4509 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4510 const char *dev_name, void *data)
4512 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4515 static void kill_f2fs_super(struct super_block *sb)
4518 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4520 set_sbi_flag(sbi, SBI_IS_CLOSE);
4521 f2fs_stop_gc_thread(sbi);
4522 f2fs_stop_discard_thread(sbi);
4524 #ifdef CONFIG_F2FS_FS_COMPRESSION
4526 * latter evict_inode() can bypass checking and invalidating
4527 * compress inode cache.
4529 if (test_opt(sbi, COMPRESS_CACHE))
4530 truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4533 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4534 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4535 struct cp_control cpc = {
4536 .reason = CP_UMOUNT,
4538 f2fs_write_checkpoint(sbi, &cpc);
4541 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4542 sb->s_flags &= ~SB_RDONLY;
4544 kill_block_super(sb);
4547 static struct file_system_type f2fs_fs_type = {
4548 .owner = THIS_MODULE,
4550 .mount = f2fs_mount,
4551 .kill_sb = kill_f2fs_super,
4552 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4554 MODULE_ALIAS_FS("f2fs");
4556 static int __init init_inodecache(void)
4558 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4559 sizeof(struct f2fs_inode_info), 0,
4560 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4561 if (!f2fs_inode_cachep)
4566 static void destroy_inodecache(void)
4569 * Make sure all delayed rcu free inodes are flushed before we
4573 kmem_cache_destroy(f2fs_inode_cachep);
4576 static int __init init_f2fs_fs(void)
4580 if (PAGE_SIZE != F2FS_BLKSIZE) {
4581 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4582 PAGE_SIZE, F2FS_BLKSIZE);
4586 err = init_inodecache();
4589 err = f2fs_create_node_manager_caches();
4591 goto free_inodecache;
4592 err = f2fs_create_segment_manager_caches();
4594 goto free_node_manager_caches;
4595 err = f2fs_create_checkpoint_caches();
4597 goto free_segment_manager_caches;
4598 err = f2fs_create_recovery_cache();
4600 goto free_checkpoint_caches;
4601 err = f2fs_create_extent_cache();
4603 goto free_recovery_cache;
4604 err = f2fs_create_garbage_collection_cache();
4606 goto free_extent_cache;
4607 err = f2fs_init_sysfs();
4609 goto free_garbage_collection_cache;
4610 err = register_shrinker(&f2fs_shrinker_info);
4613 err = register_filesystem(&f2fs_fs_type);
4616 f2fs_create_root_stats();
4617 err = f2fs_init_post_read_processing();
4619 goto free_root_stats;
4620 err = f2fs_init_iostat_processing();
4622 goto free_post_read;
4623 err = f2fs_init_bio_entry_cache();
4626 err = f2fs_init_bioset();
4628 goto free_bio_enrty_cache;
4629 err = f2fs_init_compress_mempool();
4632 err = f2fs_init_compress_cache();
4634 goto free_compress_mempool;
4635 err = f2fs_create_casefold_cache();
4637 goto free_compress_cache;
4639 free_compress_cache:
4640 f2fs_destroy_compress_cache();
4641 free_compress_mempool:
4642 f2fs_destroy_compress_mempool();
4644 f2fs_destroy_bioset();
4645 free_bio_enrty_cache:
4646 f2fs_destroy_bio_entry_cache();
4648 f2fs_destroy_iostat_processing();
4650 f2fs_destroy_post_read_processing();
4652 f2fs_destroy_root_stats();
4653 unregister_filesystem(&f2fs_fs_type);
4655 unregister_shrinker(&f2fs_shrinker_info);
4658 free_garbage_collection_cache:
4659 f2fs_destroy_garbage_collection_cache();
4661 f2fs_destroy_extent_cache();
4662 free_recovery_cache:
4663 f2fs_destroy_recovery_cache();
4664 free_checkpoint_caches:
4665 f2fs_destroy_checkpoint_caches();
4666 free_segment_manager_caches:
4667 f2fs_destroy_segment_manager_caches();
4668 free_node_manager_caches:
4669 f2fs_destroy_node_manager_caches();
4671 destroy_inodecache();
4676 static void __exit exit_f2fs_fs(void)
4678 f2fs_destroy_casefold_cache();
4679 f2fs_destroy_compress_cache();
4680 f2fs_destroy_compress_mempool();
4681 f2fs_destroy_bioset();
4682 f2fs_destroy_bio_entry_cache();
4683 f2fs_destroy_iostat_processing();
4684 f2fs_destroy_post_read_processing();
4685 f2fs_destroy_root_stats();
4686 unregister_filesystem(&f2fs_fs_type);
4687 unregister_shrinker(&f2fs_shrinker_info);
4689 f2fs_destroy_garbage_collection_cache();
4690 f2fs_destroy_extent_cache();
4691 f2fs_destroy_recovery_cache();
4692 f2fs_destroy_checkpoint_caches();
4693 f2fs_destroy_segment_manager_caches();
4694 f2fs_destroy_node_manager_caches();
4695 destroy_inodecache();
4698 module_init(init_f2fs_fs)
4699 module_exit(exit_f2fs_fs)
4701 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4702 MODULE_DESCRIPTION("Flash Friendly File System");
4703 MODULE_LICENSE("GPL");
4704 MODULE_SOFTDEP("pre: crc32");