fs/f2fs/super.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * fs/f2fs/super.c
   4  *
   5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   6  *             http://www.samsung.com/
   7  */
   8 #include <linux/module.h>
   9 #include <linux/init.h>
  10 #include <linux/fs.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>
  30
  31 #include "f2fs.h"
  32 #include "node.h"
  33 #include "segment.h"
  34 #include "xattr.h"
  35 #include "gc.h"
  36 #include "iostat.h"
  37
  38 #define CREATE_TRACE_POINTS
  39 #include <trace/events/f2fs.h>
  40
  41 static struct kmem_cache *f2fs_inode_cachep;
  42
  43 #ifdef CONFIG_F2FS_FAULT_INJECTION
  44
  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",
  63 };
  64
  65 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  66                                                         unsigned int type)
  67 {
  68         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  69
  70         if (rate) {
  71                 atomic_set(&ffi->inject_ops, 0);
  72                 ffi->inject_rate = rate;
  73         }
  74
  75         if (type)
  76                 ffi->inject_type = type;
  77
  78         if (!rate && !type)
  79                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  80 }
  81 #endif
  82
  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,
  88 };
  89
  90 enum {
  91         Opt_gc_background,
  92         Opt_disable_roll_forward,
  93         Opt_norecovery,
  94         Opt_discard,
  95         Opt_nodiscard,
  96         Opt_noheap,
  97         Opt_heap,
  98         Opt_user_xattr,
  99         Opt_nouser_xattr,
 100         Opt_acl,
 101         Opt_noacl,
 102         Opt_active_logs,
 103         Opt_disable_ext_identify,
 104         Opt_inline_xattr,
 105         Opt_noinline_xattr,
 106         Opt_inline_xattr_size,
 107         Opt_inline_data,
 108         Opt_inline_dentry,
 109         Opt_noinline_dentry,
 110         Opt_flush_merge,
 111         Opt_noflush_merge,
 112         Opt_nobarrier,
 113         Opt_fastboot,
 114         Opt_extent_cache,
 115         Opt_noextent_cache,
 116         Opt_noinline_data,
 117         Opt_data_flush,
 118         Opt_reserve_root,
 119         Opt_resgid,
 120         Opt_resuid,
 121         Opt_mode,
 122         Opt_io_size_bits,
 123         Opt_fault_injection,
 124         Opt_fault_type,
 125         Opt_lazytime,
 126         Opt_nolazytime,
 127         Opt_quota,
 128         Opt_noquota,
 129         Opt_usrquota,
 130         Opt_grpquota,
 131         Opt_prjquota,
 132         Opt_usrjquota,
 133         Opt_grpjquota,
 134         Opt_prjjquota,
 135         Opt_offusrjquota,
 136         Opt_offgrpjquota,
 137         Opt_offprjjquota,
 138         Opt_jqfmt_vfsold,
 139         Opt_jqfmt_vfsv0,
 140         Opt_jqfmt_vfsv1,
 141         Opt_whint,
 142         Opt_alloc,
 143         Opt_fsync,
 144         Opt_test_dummy_encryption,
 145         Opt_inlinecrypt,
 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,
 156         Opt_compress_chksum,
 157         Opt_compress_mode,
 158         Opt_compress_cache,
 159         Opt_atgc,
 160         Opt_gc_merge,
 161         Opt_nogc_merge,
 162         Opt_discard_unit,
 163         Opt_err,
 164 };
 165
 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"},
 173         {Opt_heap, "heap"},
 174         {Opt_user_xattr, "user_xattr"},
 175         {Opt_nouser_xattr, "nouser_xattr"},
 176         {Opt_acl, "acl"},
 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"},
 236         {Opt_atgc, "atgc"},
 237         {Opt_gc_merge, "gc_merge"},
 238         {Opt_nogc_merge, "nogc_merge"},
 239         {Opt_discard_unit, "discard_unit=%s"},
 240         {Opt_err, NULL},
 241 };
 242
 243 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 244 {
 245         struct va_format vaf;
 246         va_list args;
 247         int level;
 248
 249         va_start(args, fmt);
 250
 251         level = printk_get_level(fmt);
 252         vaf.fmt = printk_skip_level(fmt);
 253         vaf.va = &args;
 254         printk("%c%cF2FS-fs (%s): %pV\n",
 255                KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 256
 257         va_end(args);
 258 }
 259
 260 #if IS_ENABLED(CONFIG_UNICODE)
 261 static const struct f2fs_sb_encodings {
 262         __u16 magic;
 263         char *name;
 264         unsigned int version;
 265 } f2fs_sb_encoding_map[] = {
 266         {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
 267 };
 268
 269 static const struct f2fs_sb_encodings *
 270 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
 271 {
 272         __u16 magic = le16_to_cpu(sb->s_encoding);
 273         int i;
 274
 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];
 278
 279         return NULL;
 280 }
 281
 282 struct kmem_cache *f2fs_cf_name_slab;
 283 static int __init f2fs_create_casefold_cache(void)
 284 {
 285         f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
 286                                                         F2FS_NAME_LEN);
 287         if (!f2fs_cf_name_slab)
 288                 return -ENOMEM;
 289         return 0;
 290 }
 291
 292 static void f2fs_destroy_casefold_cache(void)
 293 {
 294         kmem_cache_destroy(f2fs_cf_name_slab);
 295 }
 296 #else
 297 static int __init f2fs_create_casefold_cache(void) { return 0; }
 298 static void f2fs_destroy_casefold_cache(void) { }
 299 #endif
 300
 301 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 302 {
 303         block_t limit = min((sbi->user_block_count << 1) / 1000,
 304                         sbi->user_block_count - sbi->reserved_blocks);
 305
 306         /* limit is 0.2% */
 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);
 312         }
 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));
 323 }
 324
 325 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
 326 {
 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;
 331
 332         if (!F2FS_IO_ALIGNED(sbi))
 333                 return 0;
 334
 335         /* average valid block count in section in worst case */
 336         avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
 337
 338         /*
 339          * we need enough free space when migrating one section in worst case
 340          */
 341         wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
 342                                                 reserved_segments(sbi);
 343         wanted_reserved_segments -= reserved_segments(sbi);
 344
 345         avail_user_block_count = sbi->user_block_count -
 346                                 sbi->current_reserved_blocks -
 347                                 F2FS_OPTION(sbi).root_reserved_blocks;
 348
 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);
 354                 return -ENOSPC;
 355         }
 356
 357         SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
 358
 359         f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
 360                          wanted_reserved_segments);
 361
 362         return 0;
 363 }
 364
 365 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
 366 {
 367         if (!F2FS_OPTION(sbi).unusable_cap_perc)
 368                 return;
 369
 370         if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
 371                 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
 372         else
 373                 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
 374                                         F2FS_OPTION(sbi).unusable_cap_perc;
 375
 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);
 379 }
 380
 381 static void init_once(void *foo)
 382 {
 383         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 384
 385         inode_init_once(&fi->vfs_inode);
 386 }
 387
 388 #ifdef CONFIG_QUOTA
 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,
 392                                                         substring_t *args)
 393 {
 394         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 395         char *qname;
 396         int ret = -EINVAL;
 397
 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");
 400                 return -EINVAL;
 401         }
 402         if (f2fs_sb_has_quota_ino(sbi)) {
 403                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 404                 return 0;
 405         }
 406
 407         qname = match_strdup(args);
 408         if (!qname) {
 409                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 410                 return -ENOMEM;
 411         }
 412         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 413                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 414                         ret = 0;
 415                 else
 416                         f2fs_err(sbi, "%s quota file already specified",
 417                                  QTYPE2NAME(qtype));
 418                 goto errout;
 419         }
 420         if (strchr(qname, '/')) {
 421                 f2fs_err(sbi, "quotafile must be on filesystem root");
 422                 goto errout;
 423         }
 424         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 425         set_opt(sbi, QUOTA);
 426         return 0;
 427 errout:
 428         kfree(qname);
 429         return ret;
 430 }
 431
 432 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 433 {
 434         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 435
 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");
 438                 return -EINVAL;
 439         }
 440         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 441         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 442         return 0;
 443 }
 444
 445 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 446 {
 447         /*
 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.
 451          */
 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.");
 454                 return -1;
 455         }
 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);
 462
 463                 if (test_opt(sbi, GRPQUOTA) &&
 464                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 465                         clear_opt(sbi, GRPQUOTA);
 466
 467                 if (test_opt(sbi, PRJQUOTA) &&
 468                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 469                         clear_opt(sbi, PRJQUOTA);
 470
 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");
 474                         return -1;
 475                 }
 476
 477                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 478                         f2fs_err(sbi, "journaled quota format not specified");
 479                         return -1;
 480                 }
 481         }
 482
 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;
 486         }
 487         return 0;
 488 }
 489 #endif
 490
 491 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
 492                                           const char *opt,
 493                                           const substring_t *arg,
 494                                           bool is_remount)
 495 {
 496         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 497 #ifdef CONFIG_FS_ENCRYPTION
 498         int err;
 499
 500         if (!f2fs_sb_has_encrypt(sbi)) {
 501                 f2fs_err(sbi, "Encrypt feature is off");
 502                 return -EINVAL;
 503         }
 504
 505         /*
 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.
 510          */
 511         if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
 512                 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
 513                 return -EINVAL;
 514         }
 515         err = fscrypt_set_test_dummy_encryption(
 516                 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
 517         if (err) {
 518                 if (err == -EEXIST)
 519                         f2fs_warn(sbi,
 520                                   "Can't change test_dummy_encryption on remount");
 521                 else if (err == -EINVAL)
 522                         f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
 523                                   opt);
 524                 else
 525                         f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
 526                                   opt, err);
 527                 return -EINVAL;
 528         }
 529         f2fs_warn(sbi, "Test dummy encryption mode enabled");
 530 #else
 531         f2fs_warn(sbi, "Test dummy encryption mount option ignored");
 532 #endif
 533         return 0;
 534 }
 535
 536 #ifdef CONFIG_F2FS_FS_COMPRESSION
 537 /*
 538  * 1. The same extension name cannot not appear in both compress and non-compress extension
 539  * at the same time.
 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.
 543  */
 544 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
 545 {
 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;
 549
 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;
 554
 555         if (!noext_cnt)
 556                 return 0;
 557
 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");
 561                         return -EINVAL;
 562                 }
 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",
 566                                                 ext[index]);
 567                                 return -EINVAL;
 568                         }
 569                 }
 570         }
 571         return 0;
 572 }
 573
 574 #ifdef CONFIG_F2FS_FS_LZ4
 575 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
 576 {
 577 #ifdef CONFIG_F2FS_FS_LZ4HC
 578         unsigned int level;
 579 #endif
 580
 581         if (strlen(str) == 3) {
 582                 F2FS_OPTION(sbi).compress_level = 0;
 583                 return 0;
 584         }
 585
 586 #ifdef CONFIG_F2FS_FS_LZ4HC
 587         str += 3;
 588
 589         if (str[0] != ':') {
 590                 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
 591                 return -EINVAL;
 592         }
 593         if (kstrtouint(str + 1, 10, &level))
 594                 return -EINVAL;
 595
 596         if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
 597                 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
 598                 return -EINVAL;
 599         }
 600
 601         F2FS_OPTION(sbi).compress_level = level;
 602         return 0;
 603 #else
 604         f2fs_info(sbi, "kernel doesn't support lz4hc compression");
 605         return -EINVAL;
 606 #endif
 607 }
 608 #endif
 609
 610 #ifdef CONFIG_F2FS_FS_ZSTD
 611 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
 612 {
 613         unsigned int level;
 614         int len = 4;
 615
 616         if (strlen(str) == len) {
 617                 F2FS_OPTION(sbi).compress_level = 0;
 618                 return 0;
 619         }
 620
 621         str += len;
 622
 623         if (str[0] != ':') {
 624                 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
 625                 return -EINVAL;
 626         }
 627         if (kstrtouint(str + 1, 10, &level))
 628                 return -EINVAL;
 629
 630         if (!level || level > zstd_max_clevel()) {
 631                 f2fs_info(sbi, "invalid zstd compress level: %d", level);
 632                 return -EINVAL;
 633         }
 634
 635         F2FS_OPTION(sbi).compress_level = level;
 636         return 0;
 637 }
 638 #endif
 639 #endif
 640
 641 static int parse_options(struct super_block *sb, char *options, bool is_remount)
 642 {
 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;
 649 #endif
 650         char *p, *name;
 651         int arg = 0;
 652         kuid_t uid;
 653         kgid_t gid;
 654         int ret;
 655
 656         if (!options)
 657                 goto default_check;
 658
 659         while ((p = strsep(&options, ",")) != NULL) {
 660                 int token;
 661
 662                 if (!*p)
 663                         continue;
 664                 /*
 665                  * Initialize args struct so we know whether arg was
 666                  * found; some options take optional arguments.
 667                  */
 668                 args[0].to = args[0].from = NULL;
 669                 token = match_token(p, f2fs_tokens, args);
 670
 671                 switch (token) {
 672                 case Opt_gc_background:
 673                         name = match_strdup(&args[0]);
 674
 675                         if (!name)
 676                                 return -ENOMEM;
 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;
 683                         } else {
 684                                 kfree(name);
 685                                 return -EINVAL;
 686                         }
 687                         kfree(name);
 688                         break;
 689                 case Opt_disable_roll_forward:
 690                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 691                         break;
 692                 case Opt_norecovery:
 693                         /* this option mounts f2fs with ro */
 694                         set_opt(sbi, NORECOVERY);
 695                         if (!f2fs_readonly(sb))
 696                                 return -EINVAL;
 697                         break;
 698                 case Opt_discard:
 699                         if (!f2fs_hw_support_discard(sbi)) {
 700                                 f2fs_warn(sbi, "device does not support discard");
 701                                 break;
 702                         }
 703                         set_opt(sbi, DISCARD);
 704                         break;
 705                 case Opt_nodiscard:
 706                         if (f2fs_hw_should_discard(sbi)) {
 707                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 708                                 return -EINVAL;
 709                         }
 710                         clear_opt(sbi, DISCARD);
 711                         break;
 712                 case Opt_noheap:
 713                         set_opt(sbi, NOHEAP);
 714                         break;
 715                 case Opt_heap:
 716                         clear_opt(sbi, NOHEAP);
 717                         break;
 718 #ifdef CONFIG_F2FS_FS_XATTR
 719                 case Opt_user_xattr:
 720                         set_opt(sbi, XATTR_USER);
 721                         break;
 722                 case Opt_nouser_xattr:
 723                         clear_opt(sbi, XATTR_USER);
 724                         break;
 725                 case Opt_inline_xattr:
 726                         set_opt(sbi, INLINE_XATTR);
 727                         break;
 728                 case Opt_noinline_xattr:
 729                         clear_opt(sbi, INLINE_XATTR);
 730                         break;
 731                 case Opt_inline_xattr_size:
 732                         if (args->from && match_int(args, &arg))
 733                                 return -EINVAL;
 734                         set_opt(sbi, INLINE_XATTR_SIZE);
 735                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 736                         break;
 737 #else
 738                 case Opt_user_xattr:
 739                         f2fs_info(sbi, "user_xattr options not supported");
 740                         break;
 741                 case Opt_nouser_xattr:
 742                         f2fs_info(sbi, "nouser_xattr options not supported");
 743                         break;
 744                 case Opt_inline_xattr:
 745                         f2fs_info(sbi, "inline_xattr options not supported");
 746                         break;
 747                 case Opt_noinline_xattr:
 748                         f2fs_info(sbi, "noinline_xattr options not supported");
 749                         break;
 750 #endif
 751 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 752                 case Opt_acl:
 753                         set_opt(sbi, POSIX_ACL);
 754                         break;
 755                 case Opt_noacl:
 756                         clear_opt(sbi, POSIX_ACL);
 757                         break;
 758 #else
 759                 case Opt_acl:
 760                         f2fs_info(sbi, "acl options not supported");
 761                         break;
 762                 case Opt_noacl:
 763                         f2fs_info(sbi, "noacl options not supported");
 764                         break;
 765 #endif
 766                 case Opt_active_logs:
 767                         if (args->from && match_int(args, &arg))
 768                                 return -EINVAL;
 769                         if (arg != 2 && arg != 4 &&
 770                                 arg != NR_CURSEG_PERSIST_TYPE)
 771                                 return -EINVAL;
 772                         F2FS_OPTION(sbi).active_logs = arg;
 773                         break;
 774                 case Opt_disable_ext_identify:
 775                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 776                         break;
 777                 case Opt_inline_data:
 778                         set_opt(sbi, INLINE_DATA);
 779                         break;
 780                 case Opt_inline_dentry:
 781                         set_opt(sbi, INLINE_DENTRY);
 782                         break;
 783                 case Opt_noinline_dentry:
 784                         clear_opt(sbi, INLINE_DENTRY);
 785                         break;
 786                 case Opt_flush_merge:
 787                         set_opt(sbi, FLUSH_MERGE);
 788                         break;
 789                 case Opt_noflush_merge:
 790                         clear_opt(sbi, FLUSH_MERGE);
 791                         break;
 792                 case Opt_nobarrier:
 793                         set_opt(sbi, NOBARRIER);
 794                         break;
 795                 case Opt_fastboot:
 796                         set_opt(sbi, FASTBOOT);
 797                         break;
 798                 case Opt_extent_cache:
 799                         set_opt(sbi, EXTENT_CACHE);
 800                         break;
 801                 case Opt_noextent_cache:
 802                         clear_opt(sbi, EXTENT_CACHE);
 803                         break;
 804                 case Opt_noinline_data:
 805                         clear_opt(sbi, INLINE_DATA);
 806                         break;
 807                 case Opt_data_flush:
 808                         set_opt(sbi, DATA_FLUSH);
 809                         break;
 810                 case Opt_reserve_root:
 811                         if (args->from && match_int(args, &arg))
 812                                 return -EINVAL;
 813                         if (test_opt(sbi, RESERVE_ROOT)) {
 814                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 815                                           F2FS_OPTION(sbi).root_reserved_blocks);
 816                         } else {
 817                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 818                                 set_opt(sbi, RESERVE_ROOT);
 819                         }
 820                         break;
 821                 case Opt_resuid:
 822                         if (args->from && match_int(args, &arg))
 823                                 return -EINVAL;
 824                         uid = make_kuid(current_user_ns(), arg);
 825                         if (!uid_valid(uid)) {
 826                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 827                                 return -EINVAL;
 828                         }
 829                         F2FS_OPTION(sbi).s_resuid = uid;
 830                         break;
 831                 case Opt_resgid:
 832                         if (args->from && match_int(args, &arg))
 833                                 return -EINVAL;
 834                         gid = make_kgid(current_user_ns(), arg);
 835                         if (!gid_valid(gid)) {
 836                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 837                                 return -EINVAL;
 838                         }
 839                         F2FS_OPTION(sbi).s_resgid = gid;
 840                         break;
 841                 case Opt_mode:
 842                         name = match_strdup(&args[0]);
 843
 844                         if (!name)
 845                                 return -ENOMEM;
 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");
 849                                         kfree(name);
 850                                         return -EINVAL;
 851                                 }
 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;
 859                         } else {
 860                                 kfree(name);
 861                                 return -EINVAL;
 862                         }
 863                         kfree(name);
 864                         break;
 865                 case Opt_io_size_bits:
 866                         if (args->from && match_int(args, &arg))
 867                                 return -EINVAL;
 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);
 871                                 return -EINVAL;
 872                         }
 873                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 874                         break;
 875 #ifdef CONFIG_F2FS_FAULT_INJECTION
 876                 case Opt_fault_injection:
 877                         if (args->from && match_int(args, &arg))
 878                                 return -EINVAL;
 879                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 880                         set_opt(sbi, FAULT_INJECTION);
 881                         break;
 882
 883                 case Opt_fault_type:
 884                         if (args->from && match_int(args, &arg))
 885                                 return -EINVAL;
 886                         f2fs_build_fault_attr(sbi, 0, arg);
 887                         set_opt(sbi, FAULT_INJECTION);
 888                         break;
 889 #else
 890                 case Opt_fault_injection:
 891                         f2fs_info(sbi, "fault_injection options not supported");
 892                         break;
 893
 894                 case Opt_fault_type:
 895                         f2fs_info(sbi, "fault_type options not supported");
 896                         break;
 897 #endif
 898                 case Opt_lazytime:
 899                         sb->s_flags |= SB_LAZYTIME;
 900                         break;
 901                 case Opt_nolazytime:
 902                         sb->s_flags &= ~SB_LAZYTIME;
 903                         break;
 904 #ifdef CONFIG_QUOTA
 905                 case Opt_quota:
 906                 case Opt_usrquota:
 907                         set_opt(sbi, USRQUOTA);
 908                         break;
 909                 case Opt_grpquota:
 910                         set_opt(sbi, GRPQUOTA);
 911                         break;
 912                 case Opt_prjquota:
 913                         set_opt(sbi, PRJQUOTA);
 914                         break;
 915                 case Opt_usrjquota:
 916                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 917                         if (ret)
 918                                 return ret;
 919                         break;
 920                 case Opt_grpjquota:
 921                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 922                         if (ret)
 923                                 return ret;
 924                         break;
 925                 case Opt_prjjquota:
 926                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 927                         if (ret)
 928                                 return ret;
 929                         break;
 930                 case Opt_offusrjquota:
 931                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 932                         if (ret)
 933                                 return ret;
 934                         break;
 935                 case Opt_offgrpjquota:
 936                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 937                         if (ret)
 938                                 return ret;
 939                         break;
 940                 case Opt_offprjjquota:
 941                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 942                         if (ret)
 943                                 return ret;
 944                         break;
 945                 case Opt_jqfmt_vfsold:
 946                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 947                         break;
 948                 case Opt_jqfmt_vfsv0:
 949                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 950                         break;
 951                 case Opt_jqfmt_vfsv1:
 952                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 953                         break;
 954                 case Opt_noquota:
 955                         clear_opt(sbi, QUOTA);
 956                         clear_opt(sbi, USRQUOTA);
 957                         clear_opt(sbi, GRPQUOTA);
 958                         clear_opt(sbi, PRJQUOTA);
 959                         break;
 960 #else
 961                 case Opt_quota:
 962                 case Opt_usrquota:
 963                 case Opt_grpquota:
 964                 case Opt_prjquota:
 965                 case Opt_usrjquota:
 966                 case Opt_grpjquota:
 967                 case Opt_prjjquota:
 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:
 974                 case Opt_noquota:
 975                         f2fs_info(sbi, "quota operations not supported");
 976                         break;
 977 #endif
 978                 case Opt_whint:
 979                         name = match_strdup(&args[0]);
 980                         if (!name)
 981                                 return -ENOMEM;
 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;
 988                         } else {
 989                                 kfree(name);
 990                                 return -EINVAL;
 991                         }
 992                         kfree(name);
 993                         break;
 994                 case Opt_alloc:
 995                         name = match_strdup(&args[0]);
 996                         if (!name)
 997                                 return -ENOMEM;
 998
 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;
1003                         } else {
1004                                 kfree(name);
1005                                 return -EINVAL;
1006                         }
1007                         kfree(name);
1008                         break;
1009                 case Opt_fsync:
1010                         name = match_strdup(&args[0]);
1011                         if (!name)
1012                                 return -ENOMEM;
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;
1020                         } else {
1021                                 kfree(name);
1022                                 return -EINVAL;
1023                         }
1024                         kfree(name);
1025                         break;
1026                 case Opt_test_dummy_encryption:
1027                         ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1028                                                              is_remount);
1029                         if (ret)
1030                                 return ret;
1031                         break;
1032                 case Opt_inlinecrypt:
1033 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1034                         sb->s_flags |= SB_INLINECRYPT;
1035 #else
1036                         f2fs_info(sbi, "inline encryption not supported");
1037 #endif
1038                         break;
1039                 case Opt_checkpoint_disable_cap_perc:
1040                         if (args->from && match_int(args, &arg))
1041                                 return -EINVAL;
1042                         if (arg < 0 || arg > 100)
1043                                 return -EINVAL;
1044                         F2FS_OPTION(sbi).unusable_cap_perc = arg;
1045                         set_opt(sbi, DISABLE_CHECKPOINT);
1046                         break;
1047                 case Opt_checkpoint_disable_cap:
1048                         if (args->from && match_int(args, &arg))
1049                                 return -EINVAL;
1050                         F2FS_OPTION(sbi).unusable_cap = arg;
1051                         set_opt(sbi, DISABLE_CHECKPOINT);
1052                         break;
1053                 case Opt_checkpoint_disable:
1054                         set_opt(sbi, DISABLE_CHECKPOINT);
1055                         break;
1056                 case Opt_checkpoint_enable:
1057                         clear_opt(sbi, DISABLE_CHECKPOINT);
1058                         break;
1059                 case Opt_checkpoint_merge:
1060                         set_opt(sbi, MERGE_CHECKPOINT);
1061                         break;
1062                 case Opt_nocheckpoint_merge:
1063                         clear_opt(sbi, MERGE_CHECKPOINT);
1064                         break;
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");
1069                                 break;
1070                         }
1071                         name = match_strdup(&args[0]);
1072                         if (!name)
1073                                 return -ENOMEM;
1074                         if (!strcmp(name, "lzo")) {
1075 #ifdef CONFIG_F2FS_FS_LZO
1076                                 F2FS_OPTION(sbi).compress_level = 0;
1077                                 F2FS_OPTION(sbi).compress_algorithm =
1078                                                                 COMPRESS_LZO;
1079 #else
1080                                 f2fs_info(sbi, "kernel doesn't support lzo compression");
1081 #endif
1082                         } else if (!strncmp(name, "lz4", 3)) {
1083 #ifdef CONFIG_F2FS_FS_LZ4
1084                                 ret = f2fs_set_lz4hc_level(sbi, name);
1085                                 if (ret) {
1086                                         kfree(name);
1087                                         return -EINVAL;
1088                                 }
1089                                 F2FS_OPTION(sbi).compress_algorithm =
1090                                                                 COMPRESS_LZ4;
1091 #else
1092                                 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1093 #endif
1094                         } else if (!strncmp(name, "zstd", 4)) {
1095 #ifdef CONFIG_F2FS_FS_ZSTD
1096                                 ret = f2fs_set_zstd_level(sbi, name);
1097                                 if (ret) {
1098                                         kfree(name);
1099                                         return -EINVAL;
1100                                 }
1101                                 F2FS_OPTION(sbi).compress_algorithm =
1102                                                                 COMPRESS_ZSTD;
1103 #else
1104                                 f2fs_info(sbi, "kernel doesn't support zstd compression");
1105 #endif
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 =
1110                                                                 COMPRESS_LZORLE;
1111 #else
1112                                 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1113 #endif
1114                         } else {
1115                                 kfree(name);
1116                                 return -EINVAL;
1117                         }
1118                         kfree(name);
1119                         break;
1120                 case Opt_compress_log_size:
1121                         if (!f2fs_sb_has_compression(sbi)) {
1122                                 f2fs_info(sbi, "Image doesn't support compression");
1123                                 break;
1124                         }
1125                         if (args->from && match_int(args, &arg))
1126                                 return -EINVAL;
1127                         if (arg < MIN_COMPRESS_LOG_SIZE ||
1128                                 arg > MAX_COMPRESS_LOG_SIZE) {
1129                                 f2fs_err(sbi,
1130                                         "Compress cluster log size is out of range");
1131                                 return -EINVAL;
1132                         }
1133                         F2FS_OPTION(sbi).compress_log_size = arg;
1134                         break;
1135                 case Opt_compress_extension:
1136                         if (!f2fs_sb_has_compression(sbi)) {
1137                                 f2fs_info(sbi, "Image doesn't support compression");
1138                                 break;
1139                         }
1140                         name = match_strdup(&args[0]);
1141                         if (!name)
1142                                 return -ENOMEM;
1143
1144                         ext = F2FS_OPTION(sbi).extensions;
1145                         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1146
1147                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
1148                                 ext_cnt >= COMPRESS_EXT_NUM) {
1149                                 f2fs_err(sbi,
1150                                         "invalid extension length/number");
1151                                 kfree(name);
1152                                 return -EINVAL;
1153                         }
1154
1155                         strcpy(ext[ext_cnt], name);
1156                         F2FS_OPTION(sbi).compress_ext_cnt++;
1157                         kfree(name);
1158                         break;
1159                 case Opt_nocompress_extension:
1160                         if (!f2fs_sb_has_compression(sbi)) {
1161                                 f2fs_info(sbi, "Image doesn't support compression");
1162                                 break;
1163                         }
1164                         name = match_strdup(&args[0]);
1165                         if (!name)
1166                                 return -ENOMEM;
1167
1168                         noext = F2FS_OPTION(sbi).noextensions;
1169                         noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1170
1171                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
1172                                 noext_cnt >= COMPRESS_EXT_NUM) {
1173                                 f2fs_err(sbi,
1174                                         "invalid extension length/number");
1175                                 kfree(name);
1176                                 return -EINVAL;
1177                         }
1178
1179                         strcpy(noext[noext_cnt], name);
1180                         F2FS_OPTION(sbi).nocompress_ext_cnt++;
1181                         kfree(name);
1182                         break;
1183                 case Opt_compress_chksum:
1184                         F2FS_OPTION(sbi).compress_chksum = true;
1185                         break;
1186                 case Opt_compress_mode:
1187                         name = match_strdup(&args[0]);
1188                         if (!name)
1189                                 return -ENOMEM;
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;
1194                         } else {
1195                                 kfree(name);
1196                                 return -EINVAL;
1197                         }
1198                         kfree(name);
1199                         break;
1200                 case Opt_compress_cache:
1201                         set_opt(sbi, COMPRESS_CACHE);
1202                         break;
1203 #else
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");
1212                         break;
1213 #endif
1214                 case Opt_atgc:
1215                         set_opt(sbi, ATGC);
1216                         break;
1217                 case Opt_gc_merge:
1218                         set_opt(sbi, GC_MERGE);
1219                         break;
1220                 case Opt_nogc_merge:
1221                         clear_opt(sbi, GC_MERGE);
1222                         break;
1223                 case Opt_discard_unit:
1224                         name = match_strdup(&args[0]);
1225                         if (!name)
1226                                 return -ENOMEM;
1227                         if (!strcmp(name, "block")) {
1228                                 F2FS_OPTION(sbi).discard_unit =
1229                                                 DISCARD_UNIT_BLOCK;
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;
1236                         } else {
1237                                 kfree(name);
1238                                 return -EINVAL;
1239                         }
1240                         kfree(name);
1241                         break;
1242                 default:
1243                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1244                                  p);
1245                         return -EINVAL;
1246                 }
1247         }
1248 default_check:
1249 #ifdef CONFIG_QUOTA
1250         if (f2fs_check_quota_options(sbi))
1251                 return -EINVAL;
1252 #else
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");
1255                 return -EINVAL;
1256         }
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");
1259                 return -EINVAL;
1260         }
1261 #endif
1262 #if !IS_ENABLED(CONFIG_UNICODE)
1263         if (f2fs_sb_has_casefold(sbi)) {
1264                 f2fs_err(sbi,
1265                         "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1266                 return -EINVAL;
1267         }
1268 #endif
1269         /*
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.
1273          */
1274 #ifndef CONFIG_BLK_DEV_ZONED
1275         if (f2fs_sb_has_blkzoned(sbi)) {
1276                 f2fs_err(sbi, "Zoned block device support is not enabled");
1277                 return -EINVAL;
1278         }
1279 #endif
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;
1286                 }
1287         }
1288
1289 #ifdef CONFIG_F2FS_FS_COMPRESSION
1290         if (f2fs_test_compress_extension(sbi)) {
1291                 f2fs_err(sbi, "invalid compress or nocompress extension");
1292                 return -EINVAL;
1293         }
1294 #endif
1295
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));
1299                 return -EINVAL;
1300         }
1301
1302         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1303                 int min_size, max_size;
1304
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");
1308                         return -EINVAL;
1309                 }
1310                 if (!test_opt(sbi, INLINE_XATTR)) {
1311                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1312                         return -EINVAL;
1313                 }
1314
1315                 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1316                 max_size = MAX_INLINE_XATTR_SIZE;
1317
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);
1322                         return -EINVAL;
1323                 }
1324         }
1325
1326         if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1327                 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1328                 return -EINVAL;
1329         }
1330
1331         /* Not pass down write hints if the number of active logs is lesser
1332          * than NR_CURSEG_PERSIST_TYPE.
1333          */
1334         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE)
1335                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1336
1337         if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1338                 f2fs_err(sbi, "Allow to mount readonly mode only");
1339                 return -EROFS;
1340         }
1341         return 0;
1342 }
1343
1344 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1345 {
1346         struct f2fs_inode_info *fi;
1347
1348         if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) {
1349                 f2fs_show_injection_info(F2FS_SB(sb), FAULT_SLAB_ALLOC);
1350                 return NULL;
1351         }
1352
1353         fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1354         if (!fi)
1355                 return NULL;
1356
1357         init_once((void *) fi);
1358
1359         /* Initialize f2fs-specific inode info */
1360         atomic_set(&fi->dirty_pages, 0);
1361         atomic_set(&fi->i_compr_blocks, 0);
1362         init_f2fs_rwsem(&fi->i_sem);
1363         spin_lock_init(&fi->i_size_lock);
1364         INIT_LIST_HEAD(&fi->dirty_list);
1365         INIT_LIST_HEAD(&fi->gdirty_list);
1366         INIT_LIST_HEAD(&fi->inmem_ilist);
1367         INIT_LIST_HEAD(&fi->inmem_pages);
1368         mutex_init(&fi->inmem_lock);
1369         init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1370         init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1371         init_f2fs_rwsem(&fi->i_xattr_sem);
1372
1373         /* Will be used by directory only */
1374         fi->i_dir_level = F2FS_SB(sb)->dir_level;
1375
1376         return &fi->vfs_inode;
1377 }
1378
1379 static int f2fs_drop_inode(struct inode *inode)
1380 {
1381         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1382         int ret;
1383
1384         /*
1385          * during filesystem shutdown, if checkpoint is disabled,
1386          * drop useless meta/node dirty pages.
1387          */
1388         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1389                 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1390                         inode->i_ino == F2FS_META_INO(sbi)) {
1391                         trace_f2fs_drop_inode(inode, 1);
1392                         return 1;
1393                 }
1394         }
1395
1396         /*
1397          * This is to avoid a deadlock condition like below.
1398          * writeback_single_inode(inode)
1399          *  - f2fs_write_data_page
1400          *    - f2fs_gc -> iput -> evict
1401          *       - inode_wait_for_writeback(inode)
1402          */
1403         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1404                 if (!inode->i_nlink && !is_bad_inode(inode)) {
1405                         /* to avoid evict_inode call simultaneously */
1406                         atomic_inc(&inode->i_count);
1407                         spin_unlock(&inode->i_lock);
1408
1409                         /* some remained atomic pages should discarded */
1410                         if (f2fs_is_atomic_file(inode))
1411                                 f2fs_drop_inmem_pages(inode);
1412
1413                         /* should remain fi->extent_tree for writepage */
1414                         f2fs_destroy_extent_node(inode);
1415
1416                         sb_start_intwrite(inode->i_sb);
1417                         f2fs_i_size_write(inode, 0);
1418
1419                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1420                                         inode, NULL, 0, DATA);
1421                         truncate_inode_pages_final(inode->i_mapping);
1422
1423                         if (F2FS_HAS_BLOCKS(inode))
1424                                 f2fs_truncate(inode);
1425
1426                         sb_end_intwrite(inode->i_sb);
1427
1428                         spin_lock(&inode->i_lock);
1429                         atomic_dec(&inode->i_count);
1430                 }
1431                 trace_f2fs_drop_inode(inode, 0);
1432                 return 0;
1433         }
1434         ret = generic_drop_inode(inode);
1435         if (!ret)
1436                 ret = fscrypt_drop_inode(inode);
1437         trace_f2fs_drop_inode(inode, ret);
1438         return ret;
1439 }
1440
1441 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1442 {
1443         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1444         int ret = 0;
1445
1446         spin_lock(&sbi->inode_lock[DIRTY_META]);
1447         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1448                 ret = 1;
1449         } else {
1450                 set_inode_flag(inode, FI_DIRTY_INODE);
1451                 stat_inc_dirty_inode(sbi, DIRTY_META);
1452         }
1453         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1454                 list_add_tail(&F2FS_I(inode)->gdirty_list,
1455                                 &sbi->inode_list[DIRTY_META]);
1456                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1457         }
1458         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1459         return ret;
1460 }
1461
1462 void f2fs_inode_synced(struct inode *inode)
1463 {
1464         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1465
1466         spin_lock(&sbi->inode_lock[DIRTY_META]);
1467         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1468                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1469                 return;
1470         }
1471         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1472                 list_del_init(&F2FS_I(inode)->gdirty_list);
1473                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1474         }
1475         clear_inode_flag(inode, FI_DIRTY_INODE);
1476         clear_inode_flag(inode, FI_AUTO_RECOVER);
1477         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1478         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1479 }
1480
1481 /*
1482  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1483  *
1484  * We should call set_dirty_inode to write the dirty inode through write_inode.
1485  */
1486 static void f2fs_dirty_inode(struct inode *inode, int flags)
1487 {
1488         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1489
1490         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1491                         inode->i_ino == F2FS_META_INO(sbi))
1492                 return;
1493
1494         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1495                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1496
1497         f2fs_inode_dirtied(inode, false);
1498 }
1499
1500 static void f2fs_free_inode(struct inode *inode)
1501 {
1502         fscrypt_free_inode(inode);
1503         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1504 }
1505
1506 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1507 {
1508         percpu_counter_destroy(&sbi->total_valid_inode_count);
1509         percpu_counter_destroy(&sbi->rf_node_block_count);
1510         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1511 }
1512
1513 static void destroy_device_list(struct f2fs_sb_info *sbi)
1514 {
1515         int i;
1516
1517         for (i = 0; i < sbi->s_ndevs; i++) {
1518                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1519 #ifdef CONFIG_BLK_DEV_ZONED
1520                 kvfree(FDEV(i).blkz_seq);
1521                 kfree(FDEV(i).zone_capacity_blocks);
1522 #endif
1523         }
1524         kvfree(sbi->devs);
1525 }
1526
1527 static void f2fs_put_super(struct super_block *sb)
1528 {
1529         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1530         int i;
1531         bool dropped;
1532
1533         /* unregister procfs/sysfs entries in advance to avoid race case */
1534         f2fs_unregister_sysfs(sbi);
1535
1536         f2fs_quota_off_umount(sb);
1537
1538         /* prevent remaining shrinker jobs */
1539         mutex_lock(&sbi->umount_mutex);
1540
1541         /*
1542          * flush all issued checkpoints and stop checkpoint issue thread.
1543          * after then, all checkpoints should be done by each process context.
1544          */
1545         f2fs_stop_ckpt_thread(sbi);
1546
1547         /*
1548          * We don't need to do checkpoint when superblock is clean.
1549          * But, the previous checkpoint was not done by umount, it needs to do
1550          * clean checkpoint again.
1551          */
1552         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1553                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1554                 struct cp_control cpc = {
1555                         .reason = CP_UMOUNT,
1556                 };
1557                 f2fs_write_checkpoint(sbi, &cpc);
1558         }
1559
1560         /* be sure to wait for any on-going discard commands */
1561         dropped = f2fs_issue_discard_timeout(sbi);
1562
1563         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1564                                         !sbi->discard_blks && !dropped) {
1565                 struct cp_control cpc = {
1566                         .reason = CP_UMOUNT | CP_TRIMMED,
1567                 };
1568                 f2fs_write_checkpoint(sbi, &cpc);
1569         }
1570
1571         /*
1572          * normally superblock is clean, so we need to release this.
1573          * In addition, EIO will skip do checkpoint, we need this as well.
1574          */
1575         f2fs_release_ino_entry(sbi, true);
1576
1577         f2fs_leave_shrinker(sbi);
1578         mutex_unlock(&sbi->umount_mutex);
1579
1580         /* our cp_error case, we can wait for any writeback page */
1581         f2fs_flush_merged_writes(sbi);
1582
1583         f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1584
1585         f2fs_bug_on(sbi, sbi->fsync_node_num);
1586
1587         f2fs_destroy_compress_inode(sbi);
1588
1589         iput(sbi->node_inode);
1590         sbi->node_inode = NULL;
1591
1592         iput(sbi->meta_inode);
1593         sbi->meta_inode = NULL;
1594
1595         /*
1596          * iput() can update stat information, if f2fs_write_checkpoint()
1597          * above failed with error.
1598          */
1599         f2fs_destroy_stats(sbi);
1600
1601         /* destroy f2fs internal modules */
1602         f2fs_destroy_node_manager(sbi);
1603         f2fs_destroy_segment_manager(sbi);
1604
1605         f2fs_destroy_post_read_wq(sbi);
1606
1607         kvfree(sbi->ckpt);
1608
1609         sb->s_fs_info = NULL;
1610         if (sbi->s_chksum_driver)
1611                 crypto_free_shash(sbi->s_chksum_driver);
1612         kfree(sbi->raw_super);
1613
1614         destroy_device_list(sbi);
1615         f2fs_destroy_page_array_cache(sbi);
1616         f2fs_destroy_xattr_caches(sbi);
1617         mempool_destroy(sbi->write_io_dummy);
1618 #ifdef CONFIG_QUOTA
1619         for (i = 0; i < MAXQUOTAS; i++)
1620                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1621 #endif
1622         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1623         destroy_percpu_info(sbi);
1624         f2fs_destroy_iostat(sbi);
1625         for (i = 0; i < NR_PAGE_TYPE; i++)
1626                 kvfree(sbi->write_io[i]);
1627 #if IS_ENABLED(CONFIG_UNICODE)
1628         utf8_unload(sb->s_encoding);
1629 #endif
1630         kfree(sbi);
1631 }
1632
1633 int f2fs_sync_fs(struct super_block *sb, int sync)
1634 {
1635         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1636         int err = 0;
1637
1638         if (unlikely(f2fs_cp_error(sbi)))
1639                 return 0;
1640         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1641                 return 0;
1642
1643         trace_f2fs_sync_fs(sb, sync);
1644
1645         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1646                 return -EAGAIN;
1647
1648         if (sync)
1649                 err = f2fs_issue_checkpoint(sbi);
1650
1651         return err;
1652 }
1653
1654 static int f2fs_freeze(struct super_block *sb)
1655 {
1656         if (f2fs_readonly(sb))
1657                 return 0;
1658
1659         /* IO error happened before */
1660         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1661                 return -EIO;
1662
1663         /* must be clean, since sync_filesystem() was already called */
1664         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1665                 return -EINVAL;
1666
1667         /* ensure no checkpoint required */
1668         if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list))
1669                 return -EINVAL;
1670
1671         /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1672         set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1673         return 0;
1674 }
1675
1676 static int f2fs_unfreeze(struct super_block *sb)
1677 {
1678         clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1679         return 0;
1680 }
1681
1682 #ifdef CONFIG_QUOTA
1683 static int f2fs_statfs_project(struct super_block *sb,
1684                                 kprojid_t projid, struct kstatfs *buf)
1685 {
1686         struct kqid qid;
1687         struct dquot *dquot;
1688         u64 limit;
1689         u64 curblock;
1690
1691         qid = make_kqid_projid(projid);
1692         dquot = dqget(sb, qid);
1693         if (IS_ERR(dquot))
1694                 return PTR_ERR(dquot);
1695         spin_lock(&dquot->dq_dqb_lock);
1696
1697         limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1698                                         dquot->dq_dqb.dqb_bhardlimit);
1699         if (limit)
1700                 limit >>= sb->s_blocksize_bits;
1701
1702         if (limit && buf->f_blocks > limit) {
1703                 curblock = (dquot->dq_dqb.dqb_curspace +
1704                             dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1705                 buf->f_blocks = limit;
1706                 buf->f_bfree = buf->f_bavail =
1707                         (buf->f_blocks > curblock) ?
1708                          (buf->f_blocks - curblock) : 0;
1709         }
1710
1711         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1712                                         dquot->dq_dqb.dqb_ihardlimit);
1713
1714         if (limit && buf->f_files > limit) {
1715                 buf->f_files = limit;
1716                 buf->f_ffree =
1717                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1718                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1719         }
1720
1721         spin_unlock(&dquot->dq_dqb_lock);
1722         dqput(dquot);
1723         return 0;
1724 }
1725 #endif
1726
1727 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1728 {
1729         struct super_block *sb = dentry->d_sb;
1730         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1731         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1732         block_t total_count, user_block_count, start_count;
1733         u64 avail_node_count;
1734
1735         total_count = le64_to_cpu(sbi->raw_super->block_count);
1736         user_block_count = sbi->user_block_count;
1737         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1738         buf->f_type = F2FS_SUPER_MAGIC;
1739         buf->f_bsize = sbi->blocksize;
1740
1741         buf->f_blocks = total_count - start_count;
1742         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1743                                                 sbi->current_reserved_blocks;
1744
1745         spin_lock(&sbi->stat_lock);
1746         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1747                 buf->f_bfree = 0;
1748         else
1749                 buf->f_bfree -= sbi->unusable_block_count;
1750         spin_unlock(&sbi->stat_lock);
1751
1752         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1753                 buf->f_bavail = buf->f_bfree -
1754                                 F2FS_OPTION(sbi).root_reserved_blocks;
1755         else
1756                 buf->f_bavail = 0;
1757
1758         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1759
1760         if (avail_node_count > user_block_count) {
1761                 buf->f_files = user_block_count;
1762                 buf->f_ffree = buf->f_bavail;
1763         } else {
1764                 buf->f_files = avail_node_count;
1765                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1766                                         buf->f_bavail);
1767         }
1768
1769         buf->f_namelen = F2FS_NAME_LEN;
1770         buf->f_fsid    = u64_to_fsid(id);
1771
1772 #ifdef CONFIG_QUOTA
1773         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1774                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1775                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1776         }
1777 #endif
1778         return 0;
1779 }
1780
1781 static inline void f2fs_show_quota_options(struct seq_file *seq,
1782                                            struct super_block *sb)
1783 {
1784 #ifdef CONFIG_QUOTA
1785         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1786
1787         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1788                 char *fmtname = "";
1789
1790                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1791                 case QFMT_VFS_OLD:
1792                         fmtname = "vfsold";
1793                         break;
1794                 case QFMT_VFS_V0:
1795                         fmtname = "vfsv0";
1796                         break;
1797                 case QFMT_VFS_V1:
1798                         fmtname = "vfsv1";
1799                         break;
1800                 }
1801                 seq_printf(seq, ",jqfmt=%s", fmtname);
1802         }
1803
1804         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1805                 seq_show_option(seq, "usrjquota",
1806                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1807
1808         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1809                 seq_show_option(seq, "grpjquota",
1810                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1811
1812         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1813                 seq_show_option(seq, "prjjquota",
1814                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1815 #endif
1816 }
1817
1818 #ifdef CONFIG_F2FS_FS_COMPRESSION
1819 static inline void f2fs_show_compress_options(struct seq_file *seq,
1820                                                         struct super_block *sb)
1821 {
1822         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1823         char *algtype = "";
1824         int i;
1825
1826         if (!f2fs_sb_has_compression(sbi))
1827                 return;
1828
1829         switch (F2FS_OPTION(sbi).compress_algorithm) {
1830         case COMPRESS_LZO:
1831                 algtype = "lzo";
1832                 break;
1833         case COMPRESS_LZ4:
1834                 algtype = "lz4";
1835                 break;
1836         case COMPRESS_ZSTD:
1837                 algtype = "zstd";
1838                 break;
1839         case COMPRESS_LZORLE:
1840                 algtype = "lzo-rle";
1841                 break;
1842         }
1843         seq_printf(seq, ",compress_algorithm=%s", algtype);
1844
1845         if (F2FS_OPTION(sbi).compress_level)
1846                 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1847
1848         seq_printf(seq, ",compress_log_size=%u",
1849                         F2FS_OPTION(sbi).compress_log_size);
1850
1851         for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1852                 seq_printf(seq, ",compress_extension=%s",
1853                         F2FS_OPTION(sbi).extensions[i]);
1854         }
1855
1856         for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1857                 seq_printf(seq, ",nocompress_extension=%s",
1858                         F2FS_OPTION(sbi).noextensions[i]);
1859         }
1860
1861         if (F2FS_OPTION(sbi).compress_chksum)
1862                 seq_puts(seq, ",compress_chksum");
1863
1864         if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1865                 seq_printf(seq, ",compress_mode=%s", "fs");
1866         else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1867                 seq_printf(seq, ",compress_mode=%s", "user");
1868
1869         if (test_opt(sbi, COMPRESS_CACHE))
1870                 seq_puts(seq, ",compress_cache");
1871 }
1872 #endif
1873
1874 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1875 {
1876         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1877
1878         if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1879                 seq_printf(seq, ",background_gc=%s", "sync");
1880         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1881                 seq_printf(seq, ",background_gc=%s", "on");
1882         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1883                 seq_printf(seq, ",background_gc=%s", "off");
1884
1885         if (test_opt(sbi, GC_MERGE))
1886                 seq_puts(seq, ",gc_merge");
1887
1888         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1889                 seq_puts(seq, ",disable_roll_forward");
1890         if (test_opt(sbi, NORECOVERY))
1891                 seq_puts(seq, ",norecovery");
1892         if (test_opt(sbi, DISCARD))
1893                 seq_puts(seq, ",discard");
1894         else
1895                 seq_puts(seq, ",nodiscard");
1896         if (test_opt(sbi, NOHEAP))
1897                 seq_puts(seq, ",no_heap");
1898         else
1899                 seq_puts(seq, ",heap");
1900 #ifdef CONFIG_F2FS_FS_XATTR
1901         if (test_opt(sbi, XATTR_USER))
1902                 seq_puts(seq, ",user_xattr");
1903         else
1904                 seq_puts(seq, ",nouser_xattr");
1905         if (test_opt(sbi, INLINE_XATTR))
1906                 seq_puts(seq, ",inline_xattr");
1907         else
1908                 seq_puts(seq, ",noinline_xattr");
1909         if (test_opt(sbi, INLINE_XATTR_SIZE))
1910                 seq_printf(seq, ",inline_xattr_size=%u",
1911                                         F2FS_OPTION(sbi).inline_xattr_size);
1912 #endif
1913 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1914         if (test_opt(sbi, POSIX_ACL))
1915                 seq_puts(seq, ",acl");
1916         else
1917                 seq_puts(seq, ",noacl");
1918 #endif
1919         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1920                 seq_puts(seq, ",disable_ext_identify");
1921         if (test_opt(sbi, INLINE_DATA))
1922                 seq_puts(seq, ",inline_data");
1923         else
1924                 seq_puts(seq, ",noinline_data");
1925         if (test_opt(sbi, INLINE_DENTRY))
1926                 seq_puts(seq, ",inline_dentry");
1927         else
1928                 seq_puts(seq, ",noinline_dentry");
1929         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1930                 seq_puts(seq, ",flush_merge");
1931         if (test_opt(sbi, NOBARRIER))
1932                 seq_puts(seq, ",nobarrier");
1933         if (test_opt(sbi, FASTBOOT))
1934                 seq_puts(seq, ",fastboot");
1935         if (test_opt(sbi, EXTENT_CACHE))
1936                 seq_puts(seq, ",extent_cache");
1937         else
1938                 seq_puts(seq, ",noextent_cache");
1939         if (test_opt(sbi, DATA_FLUSH))
1940                 seq_puts(seq, ",data_flush");
1941
1942         seq_puts(seq, ",mode=");
1943         if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1944                 seq_puts(seq, "adaptive");
1945         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1946                 seq_puts(seq, "lfs");
1947         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1948                 seq_puts(seq, "fragment:segment");
1949         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1950                 seq_puts(seq, "fragment:block");
1951         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1952         if (test_opt(sbi, RESERVE_ROOT))
1953                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1954                                 F2FS_OPTION(sbi).root_reserved_blocks,
1955                                 from_kuid_munged(&init_user_ns,
1956                                         F2FS_OPTION(sbi).s_resuid),
1957                                 from_kgid_munged(&init_user_ns,
1958                                         F2FS_OPTION(sbi).s_resgid));
1959         if (F2FS_IO_SIZE_BITS(sbi))
1960                 seq_printf(seq, ",io_bits=%u",
1961                                 F2FS_OPTION(sbi).write_io_size_bits);
1962 #ifdef CONFIG_F2FS_FAULT_INJECTION
1963         if (test_opt(sbi, FAULT_INJECTION)) {
1964                 seq_printf(seq, ",fault_injection=%u",
1965                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1966                 seq_printf(seq, ",fault_type=%u",
1967                                 F2FS_OPTION(sbi).fault_info.inject_type);
1968         }
1969 #endif
1970 #ifdef CONFIG_QUOTA
1971         if (test_opt(sbi, QUOTA))
1972                 seq_puts(seq, ",quota");
1973         if (test_opt(sbi, USRQUOTA))
1974                 seq_puts(seq, ",usrquota");
1975         if (test_opt(sbi, GRPQUOTA))
1976                 seq_puts(seq, ",grpquota");
1977         if (test_opt(sbi, PRJQUOTA))
1978                 seq_puts(seq, ",prjquota");
1979 #endif
1980         f2fs_show_quota_options(seq, sbi->sb);
1981         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1982                 seq_printf(seq, ",whint_mode=%s", "user-based");
1983         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1984                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1985
1986         fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1987
1988         if (sbi->sb->s_flags & SB_INLINECRYPT)
1989                 seq_puts(seq, ",inlinecrypt");
1990
1991         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1992                 seq_printf(seq, ",alloc_mode=%s", "default");
1993         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1994                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1995
1996         if (test_opt(sbi, DISABLE_CHECKPOINT))
1997                 seq_printf(seq, ",checkpoint=disable:%u",
1998                                 F2FS_OPTION(sbi).unusable_cap);
1999         if (test_opt(sbi, MERGE_CHECKPOINT))
2000                 seq_puts(seq, ",checkpoint_merge");
2001         else
2002                 seq_puts(seq, ",nocheckpoint_merge");
2003         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2004                 seq_printf(seq, ",fsync_mode=%s", "posix");
2005         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2006                 seq_printf(seq, ",fsync_mode=%s", "strict");
2007         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2008                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2009
2010 #ifdef CONFIG_F2FS_FS_COMPRESSION
2011         f2fs_show_compress_options(seq, sbi->sb);
2012 #endif
2013
2014         if (test_opt(sbi, ATGC))
2015                 seq_puts(seq, ",atgc");
2016
2017         if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2018                 seq_printf(seq, ",discard_unit=%s", "block");
2019         else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2020                 seq_printf(seq, ",discard_unit=%s", "segment");
2021         else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2022                 seq_printf(seq, ",discard_unit=%s", "section");
2023
2024         return 0;
2025 }
2026
2027 static void default_options(struct f2fs_sb_info *sbi)
2028 {
2029         /* init some FS parameters */
2030         if (f2fs_sb_has_readonly(sbi))
2031                 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2032         else
2033                 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2034
2035         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2036         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
2037         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2038         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2039         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2040         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2041         F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2042         F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2043         F2FS_OPTION(sbi).compress_ext_cnt = 0;
2044         F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2045         F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2046
2047         sbi->sb->s_flags &= ~SB_INLINECRYPT;
2048
2049         set_opt(sbi, INLINE_XATTR);
2050         set_opt(sbi, INLINE_DATA);
2051         set_opt(sbi, INLINE_DENTRY);
2052         set_opt(sbi, EXTENT_CACHE);
2053         set_opt(sbi, NOHEAP);
2054         clear_opt(sbi, DISABLE_CHECKPOINT);
2055         set_opt(sbi, MERGE_CHECKPOINT);
2056         F2FS_OPTION(sbi).unusable_cap = 0;
2057         sbi->sb->s_flags |= SB_LAZYTIME;
2058         set_opt(sbi, FLUSH_MERGE);
2059         if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2060                 set_opt(sbi, DISCARD);
2061         if (f2fs_sb_has_blkzoned(sbi)) {
2062                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2063                 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2064         } else {
2065                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2066                 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2067         }
2068
2069 #ifdef CONFIG_F2FS_FS_XATTR
2070         set_opt(sbi, XATTR_USER);
2071 #endif
2072 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2073         set_opt(sbi, POSIX_ACL);
2074 #endif
2075
2076         f2fs_build_fault_attr(sbi, 0, 0);
2077 }
2078
2079 #ifdef CONFIG_QUOTA
2080 static int f2fs_enable_quotas(struct super_block *sb);
2081 #endif
2082
2083 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2084 {
2085         unsigned int s_flags = sbi->sb->s_flags;
2086         struct cp_control cpc;
2087         unsigned int gc_mode;
2088         int err = 0;
2089         int ret;
2090         block_t unusable;
2091
2092         if (s_flags & SB_RDONLY) {
2093                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2094                 return -EINVAL;
2095         }
2096         sbi->sb->s_flags |= SB_ACTIVE;
2097
2098         f2fs_update_time(sbi, DISABLE_TIME);
2099
2100         gc_mode = sbi->gc_mode;
2101         sbi->gc_mode = GC_URGENT_HIGH;
2102
2103         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2104                 f2fs_down_write(&sbi->gc_lock);
2105                 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
2106                 if (err == -ENODATA) {
2107                         err = 0;
2108                         break;
2109                 }
2110                 if (err && err != -EAGAIN)
2111                         break;
2112         }
2113
2114         ret = sync_filesystem(sbi->sb);
2115         if (ret || err) {
2116                 err = ret ? ret : err;
2117                 goto restore_flag;
2118         }
2119
2120         unusable = f2fs_get_unusable_blocks(sbi);
2121         if (f2fs_disable_cp_again(sbi, unusable)) {
2122                 err = -EAGAIN;
2123                 goto restore_flag;
2124         }
2125
2126         f2fs_down_write(&sbi->gc_lock);
2127         cpc.reason = CP_PAUSE;
2128         set_sbi_flag(sbi, SBI_CP_DISABLED);
2129         err = f2fs_write_checkpoint(sbi, &cpc);
2130         if (err)
2131                 goto out_unlock;
2132
2133         spin_lock(&sbi->stat_lock);
2134         sbi->unusable_block_count = unusable;
2135         spin_unlock(&sbi->stat_lock);
2136
2137 out_unlock:
2138         f2fs_up_write(&sbi->gc_lock);
2139 restore_flag:
2140         sbi->gc_mode = gc_mode;
2141         sbi->sb->s_flags = s_flags;     /* Restore SB_RDONLY status */
2142         return err;
2143 }
2144
2145 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2146 {
2147         int retry = DEFAULT_RETRY_IO_COUNT;
2148
2149         /* we should flush all the data to keep data consistency */
2150         do {
2151                 sync_inodes_sb(sbi->sb);
2152                 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2153         } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2154
2155         if (unlikely(retry < 0))
2156                 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2157
2158         f2fs_down_write(&sbi->gc_lock);
2159         f2fs_dirty_to_prefree(sbi);
2160
2161         clear_sbi_flag(sbi, SBI_CP_DISABLED);
2162         set_sbi_flag(sbi, SBI_IS_DIRTY);
2163         f2fs_up_write(&sbi->gc_lock);
2164
2165         f2fs_sync_fs(sbi->sb, 1);
2166 }
2167
2168 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2169 {
2170         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2171         struct f2fs_mount_info org_mount_opt;
2172         unsigned long old_sb_flags;
2173         int err;
2174         bool need_restart_gc = false, need_stop_gc = false;
2175         bool need_restart_ckpt = false, need_stop_ckpt = false;
2176         bool need_restart_flush = false, need_stop_flush = false;
2177         bool need_restart_discard = false, need_stop_discard = false;
2178         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
2179         bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2180         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2181         bool no_atgc = !test_opt(sbi, ATGC);
2182         bool no_discard = !test_opt(sbi, DISCARD);
2183         bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2184         bool block_unit_discard = f2fs_block_unit_discard(sbi);
2185         struct discard_cmd_control *dcc;
2186 #ifdef CONFIG_QUOTA
2187         int i, j;
2188 #endif
2189
2190         /*
2191          * Save the old mount options in case we
2192          * need to restore them.
2193          */
2194         org_mount_opt = sbi->mount_opt;
2195         old_sb_flags = sb->s_flags;
2196
2197 #ifdef CONFIG_QUOTA
2198         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2199         for (i = 0; i < MAXQUOTAS; i++) {
2200                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2201                         org_mount_opt.s_qf_names[i] =
2202                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2203                                 GFP_KERNEL);
2204                         if (!org_mount_opt.s_qf_names[i]) {
2205                                 for (j = 0; j < i; j++)
2206                                         kfree(org_mount_opt.s_qf_names[j]);
2207                                 return -ENOMEM;
2208                         }
2209                 } else {
2210                         org_mount_opt.s_qf_names[i] = NULL;
2211                 }
2212         }
2213 #endif
2214
2215         /* recover superblocks we couldn't write due to previous RO mount */
2216         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2217                 err = f2fs_commit_super(sbi, false);
2218                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2219                           err);
2220                 if (!err)
2221                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2222         }
2223
2224         default_options(sbi);
2225
2226         /* parse mount options */
2227         err = parse_options(sb, data, true);
2228         if (err)
2229                 goto restore_opts;
2230
2231         /*
2232          * Previous and new state of filesystem is RO,
2233          * so skip checking GC and FLUSH_MERGE conditions.
2234          */
2235         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2236                 goto skip;
2237
2238         if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2239                 err = -EROFS;
2240                 goto restore_opts;
2241         }
2242
2243 #ifdef CONFIG_QUOTA
2244         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2245                 err = dquot_suspend(sb, -1);
2246                 if (err < 0)
2247                         goto restore_opts;
2248         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2249                 /* dquot_resume needs RW */
2250                 sb->s_flags &= ~SB_RDONLY;
2251                 if (sb_any_quota_suspended(sb)) {
2252                         dquot_resume(sb, -1);
2253                 } else if (f2fs_sb_has_quota_ino(sbi)) {
2254                         err = f2fs_enable_quotas(sb);
2255                         if (err)
2256                                 goto restore_opts;
2257                 }
2258         }
2259 #endif
2260         /* disallow enable atgc dynamically */
2261         if (no_atgc == !!test_opt(sbi, ATGC)) {
2262                 err = -EINVAL;
2263                 f2fs_warn(sbi, "switch atgc option is not allowed");
2264                 goto restore_opts;
2265         }
2266
2267         /* disallow enable/disable extent_cache dynamically */
2268         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
2269                 err = -EINVAL;
2270                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2271                 goto restore_opts;
2272         }
2273
2274         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2275                 err = -EINVAL;
2276                 f2fs_warn(sbi, "switch io_bits option is not allowed");
2277                 goto restore_opts;
2278         }
2279
2280         if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2281                 err = -EINVAL;
2282                 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2283                 goto restore_opts;
2284         }
2285
2286         if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2287                 err = -EINVAL;
2288                 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2289                 goto restore_opts;
2290         }
2291
2292         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2293                 err = -EINVAL;
2294                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2295                 goto restore_opts;
2296         }
2297
2298         /*
2299          * We stop the GC thread if FS is mounted as RO
2300          * or if background_gc = off is passed in mount
2301          * option. Also sync the filesystem.
2302          */
2303         if ((*flags & SB_RDONLY) ||
2304                         (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2305                         !test_opt(sbi, GC_MERGE))) {
2306                 if (sbi->gc_thread) {
2307                         f2fs_stop_gc_thread(sbi);
2308                         need_restart_gc = true;
2309                 }
2310         } else if (!sbi->gc_thread) {
2311                 err = f2fs_start_gc_thread(sbi);
2312                 if (err)
2313                         goto restore_opts;
2314                 need_stop_gc = true;
2315         }
2316
2317         if (*flags & SB_RDONLY ||
2318                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
2319                 sync_inodes_sb(sb);
2320
2321                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2322                 set_sbi_flag(sbi, SBI_IS_CLOSE);
2323                 f2fs_sync_fs(sb, 1);
2324                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2325         }
2326
2327         if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2328                         !test_opt(sbi, MERGE_CHECKPOINT)) {
2329                 f2fs_stop_ckpt_thread(sbi);
2330                 need_restart_ckpt = true;
2331         } else {
2332                 err = f2fs_start_ckpt_thread(sbi);
2333                 if (err) {
2334                         f2fs_err(sbi,
2335                             "Failed to start F2FS issue_checkpoint_thread (%d)",
2336                             err);
2337                         goto restore_gc;
2338                 }
2339                 need_stop_ckpt = true;
2340         }
2341
2342         /*
2343          * We stop issue flush thread if FS is mounted as RO
2344          * or if flush_merge is not passed in mount option.
2345          */
2346         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2347                 clear_opt(sbi, FLUSH_MERGE);
2348                 f2fs_destroy_flush_cmd_control(sbi, false);
2349                 need_restart_flush = true;
2350         } else {
2351                 err = f2fs_create_flush_cmd_control(sbi);
2352                 if (err)
2353                         goto restore_ckpt;
2354                 need_stop_flush = true;
2355         }
2356
2357         if (no_discard == !!test_opt(sbi, DISCARD)) {
2358                 if (test_opt(sbi, DISCARD)) {
2359                         err = f2fs_start_discard_thread(sbi);
2360                         if (err)
2361                                 goto restore_flush;
2362                         need_stop_discard = true;
2363                 } else {
2364                         dcc = SM_I(sbi)->dcc_info;
2365                         f2fs_stop_discard_thread(sbi);
2366                         if (atomic_read(&dcc->discard_cmd_cnt))
2367                                 f2fs_issue_discard_timeout(sbi);
2368                         need_restart_discard = true;
2369                 }
2370         }
2371
2372         if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2373                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2374                         err = f2fs_disable_checkpoint(sbi);
2375                         if (err)
2376                                 goto restore_discard;
2377                 } else {
2378                         f2fs_enable_checkpoint(sbi);
2379                 }
2380         }
2381
2382 skip:
2383 #ifdef CONFIG_QUOTA
2384         /* Release old quota file names */
2385         for (i = 0; i < MAXQUOTAS; i++)
2386                 kfree(org_mount_opt.s_qf_names[i]);
2387 #endif
2388         /* Update the POSIXACL Flag */
2389         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2390                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2391
2392         limit_reserve_root(sbi);
2393         adjust_unusable_cap_perc(sbi);
2394         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2395         return 0;
2396 restore_discard:
2397         if (need_restart_discard) {
2398                 if (f2fs_start_discard_thread(sbi))
2399                         f2fs_warn(sbi, "discard has been stopped");
2400         } else if (need_stop_discard) {
2401                 f2fs_stop_discard_thread(sbi);
2402         }
2403 restore_flush:
2404         if (need_restart_flush) {
2405                 if (f2fs_create_flush_cmd_control(sbi))
2406                         f2fs_warn(sbi, "background flush thread has stopped");
2407         } else if (need_stop_flush) {
2408                 clear_opt(sbi, FLUSH_MERGE);
2409                 f2fs_destroy_flush_cmd_control(sbi, false);
2410         }
2411 restore_ckpt:
2412         if (need_restart_ckpt) {
2413                 if (f2fs_start_ckpt_thread(sbi))
2414                         f2fs_warn(sbi, "background ckpt thread has stopped");
2415         } else if (need_stop_ckpt) {
2416                 f2fs_stop_ckpt_thread(sbi);
2417         }
2418 restore_gc:
2419         if (need_restart_gc) {
2420                 if (f2fs_start_gc_thread(sbi))
2421                         f2fs_warn(sbi, "background gc thread has stopped");
2422         } else if (need_stop_gc) {
2423                 f2fs_stop_gc_thread(sbi);
2424         }
2425 restore_opts:
2426 #ifdef CONFIG_QUOTA
2427         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2428         for (i = 0; i < MAXQUOTAS; i++) {
2429                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2430                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2431         }
2432 #endif
2433         sbi->mount_opt = org_mount_opt;
2434         sb->s_flags = old_sb_flags;
2435         return err;
2436 }
2437
2438 #ifdef CONFIG_QUOTA
2439 /* Read data from quotafile */
2440 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2441                                size_t len, loff_t off)
2442 {
2443         struct inode *inode = sb_dqopt(sb)->files[type];
2444         struct address_space *mapping = inode->i_mapping;
2445         block_t blkidx = F2FS_BYTES_TO_BLK(off);
2446         int offset = off & (sb->s_blocksize - 1);
2447         int tocopy;
2448         size_t toread;
2449         loff_t i_size = i_size_read(inode);
2450         struct page *page;
2451         char *kaddr;
2452
2453         if (off > i_size)
2454                 return 0;
2455
2456         if (off + len > i_size)
2457                 len = i_size - off;
2458         toread = len;
2459         while (toread > 0) {
2460                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2461 repeat:
2462                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2463                 if (IS_ERR(page)) {
2464                         if (PTR_ERR(page) == -ENOMEM) {
2465                                 memalloc_retry_wait(GFP_NOFS);
2466                                 goto repeat;
2467                         }
2468                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2469                         return PTR_ERR(page);
2470                 }
2471
2472                 lock_page(page);
2473
2474                 if (unlikely(page->mapping != mapping)) {
2475                         f2fs_put_page(page, 1);
2476                         goto repeat;
2477                 }
2478                 if (unlikely(!PageUptodate(page))) {
2479                         f2fs_put_page(page, 1);
2480                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2481                         return -EIO;
2482                 }
2483
2484                 kaddr = kmap_atomic(page);
2485                 memcpy(data, kaddr + offset, tocopy);
2486                 kunmap_atomic(kaddr);
2487                 f2fs_put_page(page, 1);
2488
2489                 offset = 0;
2490                 toread -= tocopy;
2491                 data += tocopy;
2492                 blkidx++;
2493         }
2494         return len;
2495 }
2496
2497 /* Write to quotafile */
2498 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2499                                 const char *data, size_t len, loff_t off)
2500 {
2501         struct inode *inode = sb_dqopt(sb)->files[type];
2502         struct address_space *mapping = inode->i_mapping;
2503         const struct address_space_operations *a_ops = mapping->a_ops;
2504         int offset = off & (sb->s_blocksize - 1);
2505         size_t towrite = len;
2506         struct page *page;
2507         void *fsdata = NULL;
2508         char *kaddr;
2509         int err = 0;
2510         int tocopy;
2511
2512         while (towrite > 0) {
2513                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2514                                                                 towrite);
2515 retry:
2516                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2517                                                         &page, &fsdata);
2518                 if (unlikely(err)) {
2519                         if (err == -ENOMEM) {
2520                                 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2521                                 goto retry;
2522                         }
2523                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2524                         break;
2525                 }
2526
2527                 kaddr = kmap_atomic(page);
2528                 memcpy(kaddr + offset, data, tocopy);
2529                 kunmap_atomic(kaddr);
2530                 flush_dcache_page(page);
2531
2532                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2533                                                 page, fsdata);
2534                 offset = 0;
2535                 towrite -= tocopy;
2536                 off += tocopy;
2537                 data += tocopy;
2538                 cond_resched();
2539         }
2540
2541         if (len == towrite)
2542                 return err;
2543         inode->i_mtime = inode->i_ctime = current_time(inode);
2544         f2fs_mark_inode_dirty_sync(inode, false);
2545         return len - towrite;
2546 }
2547
2548 int f2fs_dquot_initialize(struct inode *inode)
2549 {
2550         if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) {
2551                 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT);
2552                 return -ESRCH;
2553         }
2554
2555         return dquot_initialize(inode);
2556 }
2557
2558 static struct dquot **f2fs_get_dquots(struct inode *inode)
2559 {
2560         return F2FS_I(inode)->i_dquot;
2561 }
2562
2563 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2564 {
2565         return &F2FS_I(inode)->i_reserved_quota;
2566 }
2567
2568 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2569 {
2570         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2571                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2572                 return 0;
2573         }
2574
2575         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2576                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
2577 }
2578
2579 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2580 {
2581         int enabled = 0;
2582         int i, err;
2583
2584         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2585                 err = f2fs_enable_quotas(sbi->sb);
2586                 if (err) {
2587                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2588                         return 0;
2589                 }
2590                 return 1;
2591         }
2592
2593         for (i = 0; i < MAXQUOTAS; i++) {
2594                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2595                         err = f2fs_quota_on_mount(sbi, i);
2596                         if (!err) {
2597                                 enabled = 1;
2598                                 continue;
2599                         }
2600                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2601                                  err, i);
2602                 }
2603         }
2604         return enabled;
2605 }
2606
2607 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2608                              unsigned int flags)
2609 {
2610         struct inode *qf_inode;
2611         unsigned long qf_inum;
2612         int err;
2613
2614         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2615
2616         qf_inum = f2fs_qf_ino(sb, type);
2617         if (!qf_inum)
2618                 return -EPERM;
2619
2620         qf_inode = f2fs_iget(sb, qf_inum);
2621         if (IS_ERR(qf_inode)) {
2622                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2623                 return PTR_ERR(qf_inode);
2624         }
2625
2626         /* Don't account quota for quota files to avoid recursion */
2627         qf_inode->i_flags |= S_NOQUOTA;
2628         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2629         iput(qf_inode);
2630         return err;
2631 }
2632
2633 static int f2fs_enable_quotas(struct super_block *sb)
2634 {
2635         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2636         int type, err = 0;
2637         unsigned long qf_inum;
2638         bool quota_mopt[MAXQUOTAS] = {
2639                 test_opt(sbi, USRQUOTA),
2640                 test_opt(sbi, GRPQUOTA),
2641                 test_opt(sbi, PRJQUOTA),
2642         };
2643
2644         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2645                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2646                 return 0;
2647         }
2648
2649         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2650
2651         for (type = 0; type < MAXQUOTAS; type++) {
2652                 qf_inum = f2fs_qf_ino(sb, type);
2653                 if (qf_inum) {
2654                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2655                                 DQUOT_USAGE_ENABLED |
2656                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2657                         if (err) {
2658                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2659                                          type, err);
2660                                 for (type--; type >= 0; type--)
2661                                         dquot_quota_off(sb, type);
2662                                 set_sbi_flag(F2FS_SB(sb),
2663                                                 SBI_QUOTA_NEED_REPAIR);
2664                                 return err;
2665                         }
2666                 }
2667         }
2668         return 0;
2669 }
2670
2671 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2672 {
2673         struct quota_info *dqopt = sb_dqopt(sbi->sb);
2674         struct address_space *mapping = dqopt->files[type]->i_mapping;
2675         int ret = 0;
2676
2677         ret = dquot_writeback_dquots(sbi->sb, type);
2678         if (ret)
2679                 goto out;
2680
2681         ret = filemap_fdatawrite(mapping);
2682         if (ret)
2683                 goto out;
2684
2685         /* if we are using journalled quota */
2686         if (is_journalled_quota(sbi))
2687                 goto out;
2688
2689         ret = filemap_fdatawait(mapping);
2690
2691         truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2692 out:
2693         if (ret)
2694                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2695         return ret;
2696 }
2697
2698 int f2fs_quota_sync(struct super_block *sb, int type)
2699 {
2700         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2701         struct quota_info *dqopt = sb_dqopt(sb);
2702         int cnt;
2703         int ret = 0;
2704
2705         /*
2706          * Now when everything is written we can discard the pagecache so
2707          * that userspace sees the changes.
2708          */
2709         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2710
2711                 if (type != -1 && cnt != type)
2712                         continue;
2713
2714                 if (!sb_has_quota_active(sb, cnt))
2715                         continue;
2716
2717                 inode_lock(dqopt->files[cnt]);
2718
2719                 /*
2720                  * do_quotactl
2721                  *  f2fs_quota_sync
2722                  *  f2fs_down_read(quota_sem)
2723                  *  dquot_writeback_dquots()
2724                  *  f2fs_dquot_commit
2725                  *                            block_operation
2726                  *                            f2fs_down_read(quota_sem)
2727                  */
2728                 f2fs_lock_op(sbi);
2729                 f2fs_down_read(&sbi->quota_sem);
2730
2731                 ret = f2fs_quota_sync_file(sbi, cnt);
2732
2733                 f2fs_up_read(&sbi->quota_sem);
2734                 f2fs_unlock_op(sbi);
2735
2736                 inode_unlock(dqopt->files[cnt]);
2737
2738                 if (ret)
2739                         break;
2740         }
2741         return ret;
2742 }
2743
2744 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2745                                                         const struct path *path)
2746 {
2747         struct inode *inode;
2748         int err;
2749
2750         /* if quota sysfile exists, deny enabling quota with specific file */
2751         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2752                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2753                 return -EBUSY;
2754         }
2755
2756         err = f2fs_quota_sync(sb, type);
2757         if (err)
2758                 return err;
2759
2760         err = dquot_quota_on(sb, type, format_id, path);
2761         if (err)
2762                 return err;
2763
2764         inode = d_inode(path->dentry);
2765
2766         inode_lock(inode);
2767         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2768         f2fs_set_inode_flags(inode);
2769         inode_unlock(inode);
2770         f2fs_mark_inode_dirty_sync(inode, false);
2771
2772         return 0;
2773 }
2774
2775 static int __f2fs_quota_off(struct super_block *sb, int type)
2776 {
2777         struct inode *inode = sb_dqopt(sb)->files[type];
2778         int err;
2779
2780         if (!inode || !igrab(inode))
2781                 return dquot_quota_off(sb, type);
2782
2783         err = f2fs_quota_sync(sb, type);
2784         if (err)
2785                 goto out_put;
2786
2787         err = dquot_quota_off(sb, type);
2788         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2789                 goto out_put;
2790
2791         inode_lock(inode);
2792         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2793         f2fs_set_inode_flags(inode);
2794         inode_unlock(inode);
2795         f2fs_mark_inode_dirty_sync(inode, false);
2796 out_put:
2797         iput(inode);
2798         return err;
2799 }
2800
2801 static int f2fs_quota_off(struct super_block *sb, int type)
2802 {
2803         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2804         int err;
2805
2806         err = __f2fs_quota_off(sb, type);
2807
2808         /*
2809          * quotactl can shutdown journalled quota, result in inconsistence
2810          * between quota record and fs data by following updates, tag the
2811          * flag to let fsck be aware of it.
2812          */
2813         if (is_journalled_quota(sbi))
2814                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2815         return err;
2816 }
2817
2818 void f2fs_quota_off_umount(struct super_block *sb)
2819 {
2820         int type;
2821         int err;
2822
2823         for (type = 0; type < MAXQUOTAS; type++) {
2824                 err = __f2fs_quota_off(sb, type);
2825                 if (err) {
2826                         int ret = dquot_quota_off(sb, type);
2827
2828                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2829                                  type, err, ret);
2830                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2831                 }
2832         }
2833         /*
2834          * In case of checkpoint=disable, we must flush quota blocks.
2835          * This can cause NULL exception for node_inode in end_io, since
2836          * put_super already dropped it.
2837          */
2838         sync_filesystem(sb);
2839 }
2840
2841 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2842 {
2843         struct quota_info *dqopt = sb_dqopt(sb);
2844         int type;
2845
2846         for (type = 0; type < MAXQUOTAS; type++) {
2847                 if (!dqopt->files[type])
2848                         continue;
2849                 f2fs_inode_synced(dqopt->files[type]);
2850         }
2851 }
2852
2853 static int f2fs_dquot_commit(struct dquot *dquot)
2854 {
2855         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2856         int ret;
2857
2858         f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2859         ret = dquot_commit(dquot);
2860         if (ret < 0)
2861                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2862         f2fs_up_read(&sbi->quota_sem);
2863         return ret;
2864 }
2865
2866 static int f2fs_dquot_acquire(struct dquot *dquot)
2867 {
2868         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2869         int ret;
2870
2871         f2fs_down_read(&sbi->quota_sem);
2872         ret = dquot_acquire(dquot);
2873         if (ret < 0)
2874                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2875         f2fs_up_read(&sbi->quota_sem);
2876         return ret;
2877 }
2878
2879 static int f2fs_dquot_release(struct dquot *dquot)
2880 {
2881         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2882         int ret = dquot_release(dquot);
2883
2884         if (ret < 0)
2885                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2886         return ret;
2887 }
2888
2889 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2890 {
2891         struct super_block *sb = dquot->dq_sb;
2892         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2893         int ret = dquot_mark_dquot_dirty(dquot);
2894
2895         /* if we are using journalled quota */
2896         if (is_journalled_quota(sbi))
2897                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2898
2899         return ret;
2900 }
2901
2902 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2903 {
2904         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2905         int ret = dquot_commit_info(sb, type);
2906
2907         if (ret < 0)
2908                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2909         return ret;
2910 }
2911
2912 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2913 {
2914         *projid = F2FS_I(inode)->i_projid;
2915         return 0;
2916 }
2917
2918 static const struct dquot_operations f2fs_quota_operations = {
2919         .get_reserved_space = f2fs_get_reserved_space,
2920         .write_dquot    = f2fs_dquot_commit,
2921         .acquire_dquot  = f2fs_dquot_acquire,
2922         .release_dquot  = f2fs_dquot_release,
2923         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2924         .write_info     = f2fs_dquot_commit_info,
2925         .alloc_dquot    = dquot_alloc,
2926         .destroy_dquot  = dquot_destroy,
2927         .get_projid     = f2fs_get_projid,
2928         .get_next_id    = dquot_get_next_id,
2929 };
2930
2931 static const struct quotactl_ops f2fs_quotactl_ops = {
2932         .quota_on       = f2fs_quota_on,
2933         .quota_off      = f2fs_quota_off,
2934         .quota_sync     = f2fs_quota_sync,
2935         .get_state      = dquot_get_state,
2936         .set_info       = dquot_set_dqinfo,
2937         .get_dqblk      = dquot_get_dqblk,
2938         .set_dqblk      = dquot_set_dqblk,
2939         .get_nextdqblk  = dquot_get_next_dqblk,
2940 };
2941 #else
2942 int f2fs_dquot_initialize(struct inode *inode)
2943 {
2944         return 0;
2945 }
2946
2947 int f2fs_quota_sync(struct super_block *sb, int type)
2948 {
2949         return 0;
2950 }
2951
2952 void f2fs_quota_off_umount(struct super_block *sb)
2953 {
2954 }
2955 #endif
2956
2957 static const struct super_operations f2fs_sops = {
2958         .alloc_inode    = f2fs_alloc_inode,
2959         .free_inode     = f2fs_free_inode,
2960         .drop_inode     = f2fs_drop_inode,
2961         .write_inode    = f2fs_write_inode,
2962         .dirty_inode    = f2fs_dirty_inode,
2963         .show_options   = f2fs_show_options,
2964 #ifdef CONFIG_QUOTA
2965         .quota_read     = f2fs_quota_read,
2966         .quota_write    = f2fs_quota_write,
2967         .get_dquots     = f2fs_get_dquots,
2968 #endif
2969         .evict_inode    = f2fs_evict_inode,
2970         .put_super      = f2fs_put_super,
2971         .sync_fs        = f2fs_sync_fs,
2972         .freeze_fs      = f2fs_freeze,
2973         .unfreeze_fs    = f2fs_unfreeze,
2974         .statfs         = f2fs_statfs,
2975         .remount_fs     = f2fs_remount,
2976 };
2977
2978 #ifdef CONFIG_FS_ENCRYPTION
2979 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2980 {
2981         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2982                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2983                                 ctx, len, NULL);
2984 }
2985
2986 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2987                                                         void *fs_data)
2988 {
2989         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2990
2991         /*
2992          * Encrypting the root directory is not allowed because fsck
2993          * expects lost+found directory to exist and remain unencrypted
2994          * if LOST_FOUND feature is enabled.
2995          *
2996          */
2997         if (f2fs_sb_has_lost_found(sbi) &&
2998                         inode->i_ino == F2FS_ROOT_INO(sbi))
2999                 return -EPERM;
3000
3001         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3002                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3003                                 ctx, len, fs_data, XATTR_CREATE);
3004 }
3005
3006 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3007 {
3008         return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3009 }
3010
3011 static bool f2fs_has_stable_inodes(struct super_block *sb)
3012 {
3013         return true;
3014 }
3015
3016 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3017                                        int *ino_bits_ret, int *lblk_bits_ret)
3018 {
3019         *ino_bits_ret = 8 * sizeof(nid_t);
3020         *lblk_bits_ret = 8 * sizeof(block_t);
3021 }
3022
3023 static int f2fs_get_num_devices(struct super_block *sb)
3024 {
3025         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3026
3027         if (f2fs_is_multi_device(sbi))
3028                 return sbi->s_ndevs;
3029         return 1;
3030 }
3031
3032 static void f2fs_get_devices(struct super_block *sb,
3033                              struct request_queue **devs)
3034 {
3035         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3036         int i;
3037
3038         for (i = 0; i < sbi->s_ndevs; i++)
3039                 devs[i] = bdev_get_queue(FDEV(i).bdev);
3040 }
3041
3042 static const struct fscrypt_operations f2fs_cryptops = {
3043         .key_prefix             = "f2fs:",
3044         .get_context            = f2fs_get_context,
3045         .set_context            = f2fs_set_context,
3046         .get_dummy_policy       = f2fs_get_dummy_policy,
3047         .empty_dir              = f2fs_empty_dir,
3048         .has_stable_inodes      = f2fs_has_stable_inodes,
3049         .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
3050         .get_num_devices        = f2fs_get_num_devices,
3051         .get_devices            = f2fs_get_devices,
3052 };
3053 #endif
3054
3055 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3056                 u64 ino, u32 generation)
3057 {
3058         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3059         struct inode *inode;
3060
3061         if (f2fs_check_nid_range(sbi, ino))
3062                 return ERR_PTR(-ESTALE);
3063
3064         /*
3065          * f2fs_iget isn't quite right if the inode is currently unallocated!
3066          * However f2fs_iget currently does appropriate checks to handle stale
3067          * inodes so everything is OK.
3068          */
3069         inode = f2fs_iget(sb, ino);
3070         if (IS_ERR(inode))
3071                 return ERR_CAST(inode);
3072         if (unlikely(generation && inode->i_generation != generation)) {
3073                 /* we didn't find the right inode.. */
3074                 iput(inode);
3075                 return ERR_PTR(-ESTALE);
3076         }
3077         return inode;
3078 }
3079
3080 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3081                 int fh_len, int fh_type)
3082 {
3083         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3084                                     f2fs_nfs_get_inode);
3085 }
3086
3087 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3088                 int fh_len, int fh_type)
3089 {
3090         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3091                                     f2fs_nfs_get_inode);
3092 }
3093
3094 static const struct export_operations f2fs_export_ops = {
3095         .fh_to_dentry = f2fs_fh_to_dentry,
3096         .fh_to_parent = f2fs_fh_to_parent,
3097         .get_parent = f2fs_get_parent,
3098 };
3099
3100 loff_t max_file_blocks(struct inode *inode)
3101 {
3102         loff_t result = 0;
3103         loff_t leaf_count;
3104
3105         /*
3106          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3107          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3108          * space in inode.i_addr, it will be more safe to reassign
3109          * result as zero.
3110          */
3111
3112         if (inode && f2fs_compressed_file(inode))
3113                 leaf_count = ADDRS_PER_BLOCK(inode);
3114         else
3115                 leaf_count = DEF_ADDRS_PER_BLOCK;
3116
3117         /* two direct node blocks */
3118         result += (leaf_count * 2);
3119
3120         /* two indirect node blocks */
3121         leaf_count *= NIDS_PER_BLOCK;
3122         result += (leaf_count * 2);
3123
3124         /* one double indirect node block */
3125         leaf_count *= NIDS_PER_BLOCK;
3126         result += leaf_count;
3127
3128         return result;
3129 }
3130
3131 static int __f2fs_commit_super(struct buffer_head *bh,
3132                         struct f2fs_super_block *super)
3133 {
3134         lock_buffer(bh);
3135         if (super)
3136                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3137         set_buffer_dirty(bh);
3138         unlock_buffer(bh);
3139
3140         /* it's rare case, we can do fua all the time */
3141         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3142 }
3143
3144 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3145                                         struct buffer_head *bh)
3146 {
3147         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3148                                         (bh->b_data + F2FS_SUPER_OFFSET);
3149         struct super_block *sb = sbi->sb;
3150         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3151         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3152         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3153         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3154         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3155         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3156         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3157         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3158         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3159         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3160         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3161         u32 segment_count = le32_to_cpu(raw_super->segment_count);
3162         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3163         u64 main_end_blkaddr = main_blkaddr +
3164                                 (segment_count_main << log_blocks_per_seg);
3165         u64 seg_end_blkaddr = segment0_blkaddr +
3166                                 (segment_count << log_blocks_per_seg);
3167
3168         if (segment0_blkaddr != cp_blkaddr) {
3169                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3170                           segment0_blkaddr, cp_blkaddr);
3171                 return true;
3172         }
3173
3174         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3175                                                         sit_blkaddr) {
3176                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3177                           cp_blkaddr, sit_blkaddr,
3178                           segment_count_ckpt << log_blocks_per_seg);
3179                 return true;
3180         }
3181
3182         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3183                                                         nat_blkaddr) {
3184                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3185                           sit_blkaddr, nat_blkaddr,
3186                           segment_count_sit << log_blocks_per_seg);
3187                 return true;
3188         }
3189
3190         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3191                                                         ssa_blkaddr) {
3192                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3193                           nat_blkaddr, ssa_blkaddr,
3194                           segment_count_nat << log_blocks_per_seg);
3195                 return true;
3196         }
3197
3198         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3199                                                         main_blkaddr) {
3200                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3201                           ssa_blkaddr, main_blkaddr,
3202                           segment_count_ssa << log_blocks_per_seg);
3203                 return true;
3204         }
3205
3206         if (main_end_blkaddr > seg_end_blkaddr) {
3207                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3208                           main_blkaddr, seg_end_blkaddr,
3209                           segment_count_main << log_blocks_per_seg);
3210                 return true;
3211         } else if (main_end_blkaddr < seg_end_blkaddr) {
3212                 int err = 0;
3213                 char *res;
3214
3215                 /* fix in-memory information all the time */
3216                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3217                                 segment0_blkaddr) >> log_blocks_per_seg);
3218
3219                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3220                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3221                         res = "internally";
3222                 } else {
3223                         err = __f2fs_commit_super(bh, NULL);
3224                         res = err ? "failed" : "done";
3225                 }
3226                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3227                           res, main_blkaddr, seg_end_blkaddr,
3228                           segment_count_main << log_blocks_per_seg);
3229                 if (err)
3230                         return true;
3231         }
3232         return false;
3233 }
3234
3235 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3236                                 struct buffer_head *bh)
3237 {
3238         block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3239         block_t total_sections, blocks_per_seg;
3240         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3241                                         (bh->b_data + F2FS_SUPER_OFFSET);
3242         size_t crc_offset = 0;
3243         __u32 crc = 0;
3244
3245         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3246                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3247                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3248                 return -EINVAL;
3249         }
3250
3251         /* Check checksum_offset and crc in superblock */
3252         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3253                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3254                 if (crc_offset !=
3255                         offsetof(struct f2fs_super_block, crc)) {
3256                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3257                                   crc_offset);
3258                         return -EFSCORRUPTED;
3259                 }
3260                 crc = le32_to_cpu(raw_super->crc);
3261                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3262                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3263                         return -EFSCORRUPTED;
3264                 }
3265         }
3266
3267         /* Currently, support only 4KB block size */
3268         if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3269                 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3270                           le32_to_cpu(raw_super->log_blocksize),
3271                           F2FS_BLKSIZE_BITS);
3272                 return -EFSCORRUPTED;
3273         }
3274
3275         /* check log blocks per segment */
3276         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3277                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3278                           le32_to_cpu(raw_super->log_blocks_per_seg));
3279                 return -EFSCORRUPTED;
3280         }
3281
3282         /* Currently, support 512/1024/2048/4096 bytes sector size */
3283         if (le32_to_cpu(raw_super->log_sectorsize) >
3284                                 F2FS_MAX_LOG_SECTOR_SIZE ||
3285                 le32_to_cpu(raw_super->log_sectorsize) <
3286                                 F2FS_MIN_LOG_SECTOR_SIZE) {
3287                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3288                           le32_to_cpu(raw_super->log_sectorsize));
3289                 return -EFSCORRUPTED;
3290         }
3291         if (le32_to_cpu(raw_super->log_sectors_per_block) +
3292                 le32_to_cpu(raw_super->log_sectorsize) !=
3293                         F2FS_MAX_LOG_SECTOR_SIZE) {
3294                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3295                           le32_to_cpu(raw_super->log_sectors_per_block),
3296                           le32_to_cpu(raw_super->log_sectorsize));
3297                 return -EFSCORRUPTED;
3298         }
3299
3300         segment_count = le32_to_cpu(raw_super->segment_count);
3301         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3302         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3303         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3304         total_sections = le32_to_cpu(raw_super->section_count);
3305
3306         /* blocks_per_seg should be 512, given the above check */
3307         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3308
3309         if (segment_count > F2FS_MAX_SEGMENT ||
3310                                 segment_count < F2FS_MIN_SEGMENTS) {
3311                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3312                 return -EFSCORRUPTED;
3313         }
3314
3315         if (total_sections > segment_count_main || total_sections < 1 ||
3316                         segs_per_sec > segment_count || !segs_per_sec) {
3317                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3318                           segment_count, total_sections, segs_per_sec);
3319                 return -EFSCORRUPTED;
3320         }
3321
3322         if (segment_count_main != total_sections * segs_per_sec) {
3323                 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3324                           segment_count_main, total_sections, segs_per_sec);
3325                 return -EFSCORRUPTED;
3326         }
3327
3328         if ((segment_count / segs_per_sec) < total_sections) {
3329                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3330                           segment_count, segs_per_sec, total_sections);
3331                 return -EFSCORRUPTED;
3332         }
3333
3334         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3335                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3336                           segment_count, le64_to_cpu(raw_super->block_count));
3337                 return -EFSCORRUPTED;
3338         }
3339
3340         if (RDEV(0).path[0]) {
3341                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3342                 int i = 1;
3343
3344                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3345                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3346                         i++;
3347                 }
3348                 if (segment_count != dev_seg_count) {
3349                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3350                                         segment_count, dev_seg_count);
3351                         return -EFSCORRUPTED;
3352                 }
3353         } else {
3354                 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3355                                         !bdev_is_zoned(sbi->sb->s_bdev)) {
3356                         f2fs_info(sbi, "Zoned block device path is missing");
3357                         return -EFSCORRUPTED;
3358                 }
3359         }
3360
3361         if (secs_per_zone > total_sections || !secs_per_zone) {
3362                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3363                           secs_per_zone, total_sections);
3364                 return -EFSCORRUPTED;
3365         }
3366         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3367                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3368                         (le32_to_cpu(raw_super->extension_count) +
3369                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3370                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3371                           le32_to_cpu(raw_super->extension_count),
3372                           raw_super->hot_ext_count,
3373                           F2FS_MAX_EXTENSION);
3374                 return -EFSCORRUPTED;
3375         }
3376
3377         if (le32_to_cpu(raw_super->cp_payload) >=
3378                                 (blocks_per_seg - F2FS_CP_PACKS -
3379                                 NR_CURSEG_PERSIST_TYPE)) {
3380                 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3381                           le32_to_cpu(raw_super->cp_payload),
3382                           blocks_per_seg - F2FS_CP_PACKS -
3383                           NR_CURSEG_PERSIST_TYPE);
3384                 return -EFSCORRUPTED;
3385         }
3386
3387         /* check reserved ino info */
3388         if (le32_to_cpu(raw_super->node_ino) != 1 ||
3389                 le32_to_cpu(raw_super->meta_ino) != 2 ||
3390                 le32_to_cpu(raw_super->root_ino) != 3) {
3391                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3392                           le32_to_cpu(raw_super->node_ino),
3393                           le32_to_cpu(raw_super->meta_ino),
3394                           le32_to_cpu(raw_super->root_ino));
3395                 return -EFSCORRUPTED;
3396         }
3397
3398         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3399         if (sanity_check_area_boundary(sbi, bh))
3400                 return -EFSCORRUPTED;
3401
3402         return 0;
3403 }
3404
3405 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3406 {
3407         unsigned int total, fsmeta;
3408         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3409         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3410         unsigned int ovp_segments, reserved_segments;
3411         unsigned int main_segs, blocks_per_seg;
3412         unsigned int sit_segs, nat_segs;
3413         unsigned int sit_bitmap_size, nat_bitmap_size;
3414         unsigned int log_blocks_per_seg;
3415         unsigned int segment_count_main;
3416         unsigned int cp_pack_start_sum, cp_payload;
3417         block_t user_block_count, valid_user_blocks;
3418         block_t avail_node_count, valid_node_count;
3419         unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3420         int i, j;
3421
3422         total = le32_to_cpu(raw_super->segment_count);
3423         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3424         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3425         fsmeta += sit_segs;
3426         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3427         fsmeta += nat_segs;
3428         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3429         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3430
3431         if (unlikely(fsmeta >= total))
3432                 return 1;
3433
3434         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3435         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3436
3437         if (!f2fs_sb_has_readonly(sbi) &&
3438                         unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3439                         ovp_segments == 0 || reserved_segments == 0)) {
3440                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3441                 return 1;
3442         }
3443         user_block_count = le64_to_cpu(ckpt->user_block_count);
3444         segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3445                         (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3446         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3447         if (!user_block_count || user_block_count >=
3448                         segment_count_main << log_blocks_per_seg) {
3449                 f2fs_err(sbi, "Wrong user_block_count: %u",
3450                          user_block_count);
3451                 return 1;
3452         }
3453
3454         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3455         if (valid_user_blocks > user_block_count) {
3456                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3457                          valid_user_blocks, user_block_count);
3458                 return 1;
3459         }
3460
3461         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3462         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3463         if (valid_node_count > avail_node_count) {
3464                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3465                          valid_node_count, avail_node_count);
3466                 return 1;
3467         }
3468
3469         main_segs = le32_to_cpu(raw_super->segment_count_main);
3470         blocks_per_seg = sbi->blocks_per_seg;
3471
3472         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3473                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3474                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3475                         return 1;
3476
3477                 if (f2fs_sb_has_readonly(sbi))
3478                         goto check_data;
3479
3480                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3481                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3482                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
3483                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3484                                          i, j,
3485                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3486                                 return 1;
3487                         }
3488                 }
3489         }
3490 check_data:
3491         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3492                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3493                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3494                         return 1;
3495
3496                 if (f2fs_sb_has_readonly(sbi))
3497                         goto skip_cross;
3498
3499                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3500                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3501                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3502                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3503                                          i, j,
3504                                          le32_to_cpu(ckpt->cur_data_segno[i]));
3505                                 return 1;
3506                         }
3507                 }
3508         }
3509         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3510                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3511                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3512                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3513                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3514                                          i, j,
3515                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3516                                 return 1;
3517                         }
3518                 }
3519         }
3520 skip_cross:
3521         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3522         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3523
3524         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3525                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3526                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3527                          sit_bitmap_size, nat_bitmap_size);
3528                 return 1;
3529         }
3530
3531         cp_pack_start_sum = __start_sum_addr(sbi);
3532         cp_payload = __cp_payload(sbi);
3533         if (cp_pack_start_sum < cp_payload + 1 ||
3534                 cp_pack_start_sum > blocks_per_seg - 1 -
3535                         NR_CURSEG_PERSIST_TYPE) {
3536                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3537                          cp_pack_start_sum);
3538                 return 1;
3539         }
3540
3541         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3542                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3543                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3544                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3545                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3546                           le32_to_cpu(ckpt->checksum_offset));
3547                 return 1;
3548         }
3549
3550         nat_blocks = nat_segs << log_blocks_per_seg;
3551         nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3552         nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3553         if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3554                 (cp_payload + F2FS_CP_PACKS +
3555                 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3556                 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3557                           cp_payload, nat_bits_blocks);
3558                 return 1;
3559         }
3560
3561         if (unlikely(f2fs_cp_error(sbi))) {
3562                 f2fs_err(sbi, "A bug case: need to run fsck");
3563                 return 1;
3564         }
3565         return 0;
3566 }
3567
3568 static void init_sb_info(struct f2fs_sb_info *sbi)
3569 {
3570         struct f2fs_super_block *raw_super = sbi->raw_super;
3571         int i;
3572
3573         sbi->log_sectors_per_block =
3574                 le32_to_cpu(raw_super->log_sectors_per_block);
3575         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3576         sbi->blocksize = 1 << sbi->log_blocksize;
3577         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3578         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3579         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3580         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3581         sbi->total_sections = le32_to_cpu(raw_super->section_count);
3582         sbi->total_node_count =
3583                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3584                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3585         F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3586         F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3587         F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3588         sbi->cur_victim_sec = NULL_SECNO;
3589         sbi->gc_mode = GC_NORMAL;
3590         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3591         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3592         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3593         sbi->migration_granularity = sbi->segs_per_sec;
3594         sbi->seq_file_ra_mul = MIN_RA_MUL;
3595         sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3596         sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3597         spin_lock_init(&sbi->gc_urgent_high_lock);
3598
3599         sbi->dir_level = DEF_DIR_LEVEL;
3600         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3601         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3602         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3603         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3604         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3605         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3606                                 DEF_UMOUNT_DISCARD_TIMEOUT;
3607         clear_sbi_flag(sbi, SBI_NEED_FSCK);
3608
3609         for (i = 0; i < NR_COUNT_TYPE; i++)
3610                 atomic_set(&sbi->nr_pages[i], 0);
3611
3612         for (i = 0; i < META; i++)
3613                 atomic_set(&sbi->wb_sync_req[i], 0);
3614
3615         INIT_LIST_HEAD(&sbi->s_list);
3616         mutex_init(&sbi->umount_mutex);
3617         init_f2fs_rwsem(&sbi->io_order_lock);
3618         spin_lock_init(&sbi->cp_lock);
3619
3620         sbi->dirty_device = 0;
3621         spin_lock_init(&sbi->dev_lock);
3622
3623         init_f2fs_rwsem(&sbi->sb_lock);
3624         init_f2fs_rwsem(&sbi->pin_sem);
3625 }
3626
3627 static int init_percpu_info(struct f2fs_sb_info *sbi)
3628 {
3629         int err;
3630
3631         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3632         if (err)
3633                 return err;
3634
3635         err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3636         if (err)
3637                 goto err_valid_block;
3638
3639         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3640                                                                 GFP_KERNEL);
3641         if (err)
3642                 goto err_node_block;
3643         return 0;
3644
3645 err_node_block:
3646         percpu_counter_destroy(&sbi->rf_node_block_count);
3647 err_valid_block:
3648         percpu_counter_destroy(&sbi->alloc_valid_block_count);
3649         return err;
3650 }
3651
3652 #ifdef CONFIG_BLK_DEV_ZONED
3653
3654 struct f2fs_report_zones_args {
3655         struct f2fs_dev_info *dev;
3656         bool zone_cap_mismatch;
3657 };
3658
3659 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3660                               void *data)
3661 {
3662         struct f2fs_report_zones_args *rz_args = data;
3663
3664         if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3665                 return 0;
3666
3667         set_bit(idx, rz_args->dev->blkz_seq);
3668         rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >>
3669                                                 F2FS_LOG_SECTORS_PER_BLOCK;
3670         if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch)
3671                 rz_args->zone_cap_mismatch = true;
3672
3673         return 0;
3674 }
3675
3676 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3677 {
3678         struct block_device *bdev = FDEV(devi).bdev;
3679         sector_t nr_sectors = bdev_nr_sectors(bdev);
3680         struct f2fs_report_zones_args rep_zone_arg;
3681         int ret;
3682
3683         if (!f2fs_sb_has_blkzoned(sbi))
3684                 return 0;
3685
3686         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3687                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3688                 return -EINVAL;
3689         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3690         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3691                                 __ilog2_u32(sbi->blocks_per_blkz))
3692                 return -EINVAL;
3693         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3694         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3695                                         sbi->log_blocks_per_blkz;
3696         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3697                 FDEV(devi).nr_blkz++;
3698
3699         FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3700                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
3701                                         * sizeof(unsigned long),
3702                                         GFP_KERNEL);
3703         if (!FDEV(devi).blkz_seq)
3704                 return -ENOMEM;
3705
3706         /* Get block zones type and zone-capacity */
3707         FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi,
3708                                         FDEV(devi).nr_blkz * sizeof(block_t),
3709                                         GFP_KERNEL);
3710         if (!FDEV(devi).zone_capacity_blocks)
3711                 return -ENOMEM;
3712
3713         rep_zone_arg.dev = &FDEV(devi);
3714         rep_zone_arg.zone_cap_mismatch = false;
3715
3716         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3717                                   &rep_zone_arg);
3718         if (ret < 0)
3719                 return ret;
3720
3721         if (!rep_zone_arg.zone_cap_mismatch) {
3722                 kfree(FDEV(devi).zone_capacity_blocks);
3723                 FDEV(devi).zone_capacity_blocks = NULL;
3724         }
3725
3726         return 0;
3727 }
3728 #endif
3729
3730 /*
3731  * Read f2fs raw super block.
3732  * Because we have two copies of super block, so read both of them
3733  * to get the first valid one. If any one of them is broken, we pass
3734  * them recovery flag back to the caller.
3735  */
3736 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3737                         struct f2fs_super_block **raw_super,
3738                         int *valid_super_block, int *recovery)
3739 {
3740         struct super_block *sb = sbi->sb;
3741         int block;
3742         struct buffer_head *bh;
3743         struct f2fs_super_block *super;
3744         int err = 0;
3745
3746         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3747         if (!super)
3748                 return -ENOMEM;
3749
3750         for (block = 0; block < 2; block++) {
3751                 bh = sb_bread(sb, block);
3752                 if (!bh) {
3753                         f2fs_err(sbi, "Unable to read %dth superblock",
3754                                  block + 1);
3755                         err = -EIO;
3756                         *recovery = 1;
3757                         continue;
3758                 }
3759
3760                 /* sanity checking of raw super */
3761                 err = sanity_check_raw_super(sbi, bh);
3762                 if (err) {
3763                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3764                                  block + 1);
3765                         brelse(bh);
3766                         *recovery = 1;
3767                         continue;
3768                 }
3769
3770                 if (!*raw_super) {
3771                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3772                                                         sizeof(*super));
3773                         *valid_super_block = block;
3774                         *raw_super = super;
3775                 }
3776                 brelse(bh);
3777         }
3778
3779         /* No valid superblock */
3780         if (!*raw_super)
3781                 kfree(super);
3782         else
3783                 err = 0;
3784
3785         return err;
3786 }
3787
3788 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3789 {
3790         struct buffer_head *bh;
3791         __u32 crc = 0;
3792         int err;
3793
3794         if ((recover && f2fs_readonly(sbi->sb)) ||
3795                                 bdev_read_only(sbi->sb->s_bdev)) {
3796                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3797                 return -EROFS;
3798         }
3799
3800         /* we should update superblock crc here */
3801         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3802                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3803                                 offsetof(struct f2fs_super_block, crc));
3804                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3805         }
3806
3807         /* write back-up superblock first */
3808         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3809         if (!bh)
3810                 return -EIO;
3811         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3812         brelse(bh);
3813
3814         /* if we are in recovery path, skip writing valid superblock */
3815         if (recover || err)
3816                 return err;
3817
3818         /* write current valid superblock */
3819         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3820         if (!bh)
3821                 return -EIO;
3822         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3823         brelse(bh);
3824         return err;
3825 }
3826
3827 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3828 {
3829         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3830         unsigned int max_devices = MAX_DEVICES;
3831         unsigned int logical_blksize;
3832         int i;
3833
3834         /* Initialize single device information */
3835         if (!RDEV(0).path[0]) {
3836                 if (!bdev_is_zoned(sbi->sb->s_bdev))
3837                         return 0;
3838                 max_devices = 1;
3839         }
3840
3841         /*
3842          * Initialize multiple devices information, or single
3843          * zoned block device information.
3844          */
3845         sbi->devs = f2fs_kzalloc(sbi,
3846                                  array_size(max_devices,
3847                                             sizeof(struct f2fs_dev_info)),
3848                                  GFP_KERNEL);
3849         if (!sbi->devs)
3850                 return -ENOMEM;
3851
3852         logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3853         sbi->aligned_blksize = true;
3854
3855         for (i = 0; i < max_devices; i++) {
3856
3857                 if (i > 0 && !RDEV(i).path[0])
3858                         break;
3859
3860                 if (max_devices == 1) {
3861                         /* Single zoned block device mount */
3862                         FDEV(0).bdev =
3863                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3864                                         sbi->sb->s_mode, sbi->sb->s_type);
3865                 } else {
3866                         /* Multi-device mount */
3867                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3868                         FDEV(i).total_segments =
3869                                 le32_to_cpu(RDEV(i).total_segments);
3870                         if (i == 0) {
3871                                 FDEV(i).start_blk = 0;
3872                                 FDEV(i).end_blk = FDEV(i).start_blk +
3873                                     (FDEV(i).total_segments <<
3874                                     sbi->log_blocks_per_seg) - 1 +
3875                                     le32_to_cpu(raw_super->segment0_blkaddr);
3876                         } else {
3877                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3878                                 FDEV(i).end_blk = FDEV(i).start_blk +
3879                                         (FDEV(i).total_segments <<
3880                                         sbi->log_blocks_per_seg) - 1;
3881                         }
3882                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3883                                         sbi->sb->s_mode, sbi->sb->s_type);
3884                 }
3885                 if (IS_ERR(FDEV(i).bdev))
3886                         return PTR_ERR(FDEV(i).bdev);
3887
3888                 /* to release errored devices */
3889                 sbi->s_ndevs = i + 1;
3890
3891                 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
3892                         sbi->aligned_blksize = false;
3893
3894 #ifdef CONFIG_BLK_DEV_ZONED
3895                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3896                                 !f2fs_sb_has_blkzoned(sbi)) {
3897                         f2fs_err(sbi, "Zoned block device feature not enabled");
3898                         return -EINVAL;
3899                 }
3900                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3901                         if (init_blkz_info(sbi, i)) {
3902                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3903                                 return -EINVAL;
3904                         }
3905                         if (max_devices == 1)
3906                                 break;
3907                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3908                                   i, FDEV(i).path,
3909                                   FDEV(i).total_segments,
3910                                   FDEV(i).start_blk, FDEV(i).end_blk,
3911                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3912                                   "Host-aware" : "Host-managed");
3913                         continue;
3914                 }
3915 #endif
3916                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3917                           i, FDEV(i).path,
3918                           FDEV(i).total_segments,
3919                           FDEV(i).start_blk, FDEV(i).end_blk);
3920         }
3921         f2fs_info(sbi,
3922                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3923         return 0;
3924 }
3925
3926 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3927 {
3928 #if IS_ENABLED(CONFIG_UNICODE)
3929         if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3930                 const struct f2fs_sb_encodings *encoding_info;
3931                 struct unicode_map *encoding;
3932                 __u16 encoding_flags;
3933
3934                 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
3935                 if (!encoding_info) {
3936                         f2fs_err(sbi,
3937                                  "Encoding requested by superblock is unknown");
3938                         return -EINVAL;
3939                 }
3940
3941                 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
3942                 encoding = utf8_load(encoding_info->version);
3943                 if (IS_ERR(encoding)) {
3944                         f2fs_err(sbi,
3945                                  "can't mount with superblock charset: %s-%u.%u.%u "
3946                                  "not supported by the kernel. flags: 0x%x.",
3947                                  encoding_info->name,
3948                                  unicode_major(encoding_info->version),
3949                                  unicode_minor(encoding_info->version),
3950                                  unicode_rev(encoding_info->version),
3951                                  encoding_flags);
3952                         return PTR_ERR(encoding);
3953                 }
3954                 f2fs_info(sbi, "Using encoding defined by superblock: "
3955                          "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
3956                          unicode_major(encoding_info->version),
3957                          unicode_minor(encoding_info->version),
3958                          unicode_rev(encoding_info->version),
3959                          encoding_flags);
3960
3961                 sbi->sb->s_encoding = encoding;
3962                 sbi->sb->s_encoding_flags = encoding_flags;
3963         }
3964 #else
3965         if (f2fs_sb_has_casefold(sbi)) {
3966                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3967                 return -EINVAL;
3968         }
3969 #endif
3970         return 0;
3971 }
3972
3973 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3974 {
3975         struct f2fs_sm_info *sm_i = SM_I(sbi);
3976
3977         /* adjust parameters according to the volume size */
3978         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3979                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3980                 if (f2fs_block_unit_discard(sbi))
3981                         sm_i->dcc_info->discard_granularity = 1;
3982                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE |
3983                                         1 << F2FS_IPU_HONOR_OPU_WRITE;
3984         }
3985
3986         sbi->readdir_ra = 1;
3987 }
3988
3989 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3990 {
3991         struct f2fs_sb_info *sbi;
3992         struct f2fs_super_block *raw_super;
3993         struct inode *root;
3994         int err;
3995         bool skip_recovery = false, need_fsck = false;
3996         char *options = NULL;
3997         int recovery, i, valid_super_block;
3998         struct curseg_info *seg_i;
3999         int retry_cnt = 1;
4000
4001 try_onemore:
4002         err = -EINVAL;
4003         raw_super = NULL;
4004         valid_super_block = -1;
4005         recovery = 0;
4006
4007         /* allocate memory for f2fs-specific super block info */
4008         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4009         if (!sbi)
4010                 return -ENOMEM;
4011
4012         sbi->sb = sb;
4013
4014         /* Load the checksum driver */
4015         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4016         if (IS_ERR(sbi->s_chksum_driver)) {
4017                 f2fs_err(sbi, "Cannot load crc32 driver.");
4018                 err = PTR_ERR(sbi->s_chksum_driver);
4019                 sbi->s_chksum_driver = NULL;
4020                 goto free_sbi;
4021         }
4022
4023         /* set a block size */
4024         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4025                 f2fs_err(sbi, "unable to set blocksize");
4026                 goto free_sbi;
4027         }
4028
4029         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4030                                                                 &recovery);
4031         if (err)
4032                 goto free_sbi;
4033
4034         sb->s_fs_info = sbi;
4035         sbi->raw_super = raw_super;
4036
4037         /* precompute checksum seed for metadata */
4038         if (f2fs_sb_has_inode_chksum(sbi))
4039                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4040                                                 sizeof(raw_super->uuid));
4041
4042         default_options(sbi);
4043         /* parse mount options */
4044         options = kstrdup((const char *)data, GFP_KERNEL);
4045         if (data && !options) {
4046                 err = -ENOMEM;
4047                 goto free_sb_buf;
4048         }
4049
4050         err = parse_options(sb, options, false);
4051         if (err)
4052                 goto free_options;
4053
4054         sb->s_maxbytes = max_file_blocks(NULL) <<
4055                                 le32_to_cpu(raw_super->log_blocksize);
4056         sb->s_max_links = F2FS_LINK_MAX;
4057
4058         err = f2fs_setup_casefold(sbi);
4059         if (err)
4060                 goto free_options;
4061
4062 #ifdef CONFIG_QUOTA
4063         sb->dq_op = &f2fs_quota_operations;
4064         sb->s_qcop = &f2fs_quotactl_ops;
4065         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4066
4067         if (f2fs_sb_has_quota_ino(sbi)) {
4068                 for (i = 0; i < MAXQUOTAS; i++) {
4069                         if (f2fs_qf_ino(sbi->sb, i))
4070                                 sbi->nquota_files++;
4071                 }
4072         }
4073 #endif
4074
4075         sb->s_op = &f2fs_sops;
4076 #ifdef CONFIG_FS_ENCRYPTION
4077         sb->s_cop = &f2fs_cryptops;
4078 #endif
4079 #ifdef CONFIG_FS_VERITY
4080         sb->s_vop = &f2fs_verityops;
4081 #endif
4082         sb->s_xattr = f2fs_xattr_handlers;
4083         sb->s_export_op = &f2fs_export_ops;
4084         sb->s_magic = F2FS_SUPER_MAGIC;
4085         sb->s_time_gran = 1;
4086         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4087                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4088         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4089         sb->s_iflags |= SB_I_CGROUPWB;
4090
4091         /* init f2fs-specific super block info */
4092         sbi->valid_super_block = valid_super_block;
4093         init_f2fs_rwsem(&sbi->gc_lock);
4094         mutex_init(&sbi->writepages);
4095         init_f2fs_rwsem(&sbi->cp_global_sem);
4096         init_f2fs_rwsem(&sbi->node_write);
4097         init_f2fs_rwsem(&sbi->node_change);
4098
4099         /* disallow all the data/node/meta page writes */
4100         set_sbi_flag(sbi, SBI_POR_DOING);
4101         spin_lock_init(&sbi->stat_lock);
4102
4103         for (i = 0; i < NR_PAGE_TYPE; i++) {
4104                 int n = (i == META) ? 1 : NR_TEMP_TYPE;
4105                 int j;
4106
4107                 sbi->write_io[i] =
4108                         f2fs_kmalloc(sbi,
4109                                      array_size(n,
4110                                                 sizeof(struct f2fs_bio_info)),
4111                                      GFP_KERNEL);
4112                 if (!sbi->write_io[i]) {
4113                         err = -ENOMEM;
4114                         goto free_bio_info;
4115                 }
4116
4117                 for (j = HOT; j < n; j++) {
4118                         init_f2fs_rwsem(&sbi->write_io[i][j].io_rwsem);
4119                         sbi->write_io[i][j].sbi = sbi;
4120                         sbi->write_io[i][j].bio = NULL;
4121                         spin_lock_init(&sbi->write_io[i][j].io_lock);
4122                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
4123                         INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
4124                         init_f2fs_rwsem(&sbi->write_io[i][j].bio_list_lock);
4125                 }
4126         }
4127
4128         init_f2fs_rwsem(&sbi->cp_rwsem);
4129         init_f2fs_rwsem(&sbi->quota_sem);
4130         init_waitqueue_head(&sbi->cp_wait);
4131         init_sb_info(sbi);
4132
4133         err = f2fs_init_iostat(sbi);
4134         if (err)
4135                 goto free_bio_info;
4136
4137         err = init_percpu_info(sbi);
4138         if (err)
4139                 goto free_iostat;
4140
4141         if (F2FS_IO_ALIGNED(sbi)) {
4142                 sbi->write_io_dummy =
4143                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4144                 if (!sbi->write_io_dummy) {
4145                         err = -ENOMEM;
4146                         goto free_percpu;
4147                 }
4148         }
4149
4150         /* init per sbi slab cache */
4151         err = f2fs_init_xattr_caches(sbi);
4152         if (err)
4153                 goto free_io_dummy;
4154         err = f2fs_init_page_array_cache(sbi);
4155         if (err)
4156                 goto free_xattr_cache;
4157
4158         /* get an inode for meta space */
4159         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4160         if (IS_ERR(sbi->meta_inode)) {
4161                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4162                 err = PTR_ERR(sbi->meta_inode);
4163                 goto free_page_array_cache;
4164         }
4165
4166         err = f2fs_get_valid_checkpoint(sbi);
4167         if (err) {
4168                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4169                 goto free_meta_inode;
4170         }
4171
4172         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4173                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4174         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4175                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4176                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4177         }
4178
4179         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4180                 set_sbi_flag(sbi, SBI_NEED_FSCK);
4181
4182         /* Initialize device list */
4183         err = f2fs_scan_devices(sbi);
4184         if (err) {
4185                 f2fs_err(sbi, "Failed to find devices");
4186                 goto free_devices;
4187         }
4188
4189         err = f2fs_init_post_read_wq(sbi);
4190         if (err) {
4191                 f2fs_err(sbi, "Failed to initialize post read workqueue");
4192                 goto free_devices;
4193         }
4194
4195         sbi->total_valid_node_count =
4196                                 le32_to_cpu(sbi->ckpt->valid_node_count);
4197         percpu_counter_set(&sbi->total_valid_inode_count,
4198                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
4199         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4200         sbi->total_valid_block_count =
4201                                 le64_to_cpu(sbi->ckpt->valid_block_count);
4202         sbi->last_valid_block_count = sbi->total_valid_block_count;
4203         sbi->reserved_blocks = 0;
4204         sbi->current_reserved_blocks = 0;
4205         limit_reserve_root(sbi);
4206         adjust_unusable_cap_perc(sbi);
4207
4208         for (i = 0; i < NR_INODE_TYPE; i++) {
4209                 INIT_LIST_HEAD(&sbi->inode_list[i]);
4210                 spin_lock_init(&sbi->inode_lock[i]);
4211         }
4212         mutex_init(&sbi->flush_lock);
4213
4214         f2fs_init_extent_cache_info(sbi);
4215
4216         f2fs_init_ino_entry_info(sbi);
4217
4218         f2fs_init_fsync_node_info(sbi);
4219
4220         /* setup checkpoint request control and start checkpoint issue thread */
4221         f2fs_init_ckpt_req_control(sbi);
4222         if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4223                         test_opt(sbi, MERGE_CHECKPOINT)) {
4224                 err = f2fs_start_ckpt_thread(sbi);
4225                 if (err) {
4226                         f2fs_err(sbi,
4227                             "Failed to start F2FS issue_checkpoint_thread (%d)",
4228                             err);
4229                         goto stop_ckpt_thread;
4230                 }
4231         }
4232
4233         /* setup f2fs internal modules */
4234         err = f2fs_build_segment_manager(sbi);
4235         if (err) {
4236                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4237                          err);
4238                 goto free_sm;
4239         }
4240         err = f2fs_build_node_manager(sbi);
4241         if (err) {
4242                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4243                          err);
4244                 goto free_nm;
4245         }
4246
4247         err = adjust_reserved_segment(sbi);
4248         if (err)
4249                 goto free_nm;
4250
4251         /* For write statistics */
4252         sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4253
4254         /* Read accumulated write IO statistics if exists */
4255         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4256         if (__exist_node_summaries(sbi))
4257                 sbi->kbytes_written =
4258                         le64_to_cpu(seg_i->journal->info.kbytes_written);
4259
4260         f2fs_build_gc_manager(sbi);
4261
4262         err = f2fs_build_stats(sbi);
4263         if (err)
4264                 goto free_nm;
4265
4266         /* get an inode for node space */
4267         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4268         if (IS_ERR(sbi->node_inode)) {
4269                 f2fs_err(sbi, "Failed to read node inode");
4270                 err = PTR_ERR(sbi->node_inode);
4271                 goto free_stats;
4272         }
4273
4274         /* read root inode and dentry */
4275         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4276         if (IS_ERR(root)) {
4277                 f2fs_err(sbi, "Failed to read root inode");
4278                 err = PTR_ERR(root);
4279                 goto free_node_inode;
4280         }
4281         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4282                         !root->i_size || !root->i_nlink) {
4283                 iput(root);
4284                 err = -EINVAL;
4285                 goto free_node_inode;
4286         }
4287
4288         sb->s_root = d_make_root(root); /* allocate root dentry */
4289         if (!sb->s_root) {
4290                 err = -ENOMEM;
4291                 goto free_node_inode;
4292         }
4293
4294         err = f2fs_init_compress_inode(sbi);
4295         if (err)
4296                 goto free_root_inode;
4297
4298         err = f2fs_register_sysfs(sbi);
4299         if (err)
4300                 goto free_compress_inode;
4301
4302 #ifdef CONFIG_QUOTA
4303         /* Enable quota usage during mount */
4304         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4305                 err = f2fs_enable_quotas(sb);
4306                 if (err)
4307                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4308         }
4309 #endif
4310         /* if there are any orphan inodes, free them */
4311         err = f2fs_recover_orphan_inodes(sbi);
4312         if (err)
4313                 goto free_meta;
4314
4315         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4316                 goto reset_checkpoint;
4317
4318         /* recover fsynced data */
4319         if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4320                         !test_opt(sbi, NORECOVERY)) {
4321                 /*
4322                  * mount should be failed, when device has readonly mode, and
4323                  * previous checkpoint was not done by clean system shutdown.
4324                  */
4325                 if (f2fs_hw_is_readonly(sbi)) {
4326                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4327                                 err = f2fs_recover_fsync_data(sbi, true);
4328                                 if (err > 0) {
4329                                         err = -EROFS;
4330                                         f2fs_err(sbi, "Need to recover fsync data, but "
4331                                                 "write access unavailable, please try "
4332                                                 "mount w/ disable_roll_forward or norecovery");
4333                                 }
4334                                 if (err < 0)
4335                                         goto free_meta;
4336                         }
4337                         f2fs_info(sbi, "write access unavailable, skipping recovery");
4338                         goto reset_checkpoint;
4339                 }
4340
4341                 if (need_fsck)
4342                         set_sbi_flag(sbi, SBI_NEED_FSCK);
4343
4344                 if (skip_recovery)
4345                         goto reset_checkpoint;
4346
4347                 err = f2fs_recover_fsync_data(sbi, false);
4348                 if (err < 0) {
4349                         if (err != -ENOMEM)
4350                                 skip_recovery = true;
4351                         need_fsck = true;
4352                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4353                                  err);
4354                         goto free_meta;
4355                 }
4356         } else {
4357                 err = f2fs_recover_fsync_data(sbi, true);
4358
4359                 if (!f2fs_readonly(sb) && err > 0) {
4360                         err = -EINVAL;
4361                         f2fs_err(sbi, "Need to recover fsync data");
4362                         goto free_meta;
4363                 }
4364         }
4365
4366         /*
4367          * If the f2fs is not readonly and fsync data recovery succeeds,
4368          * check zoned block devices' write pointer consistency.
4369          */
4370         if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4371                 err = f2fs_check_write_pointer(sbi);
4372                 if (err)
4373                         goto free_meta;
4374         }
4375
4376 reset_checkpoint:
4377         f2fs_init_inmem_curseg(sbi);
4378
4379         /* f2fs_recover_fsync_data() cleared this already */
4380         clear_sbi_flag(sbi, SBI_POR_DOING);
4381
4382         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4383                 err = f2fs_disable_checkpoint(sbi);
4384                 if (err)
4385                         goto sync_free_meta;
4386         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4387                 f2fs_enable_checkpoint(sbi);
4388         }
4389
4390         /*
4391          * If filesystem is not mounted as read-only then
4392          * do start the gc_thread.
4393          */
4394         if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4395                 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4396                 /* After POR, we can run background GC thread.*/
4397                 err = f2fs_start_gc_thread(sbi);
4398                 if (err)
4399                         goto sync_free_meta;
4400         }
4401         kvfree(options);
4402
4403         /* recover broken superblock */
4404         if (recovery) {
4405                 err = f2fs_commit_super(sbi, true);
4406                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4407                           sbi->valid_super_block ? 1 : 2, err);
4408         }
4409
4410         f2fs_join_shrinker(sbi);
4411
4412         f2fs_tuning_parameters(sbi);
4413
4414         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4415                     cur_cp_version(F2FS_CKPT(sbi)));
4416         f2fs_update_time(sbi, CP_TIME);
4417         f2fs_update_time(sbi, REQ_TIME);
4418         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4419         return 0;
4420
4421 sync_free_meta:
4422         /* safe to flush all the data */
4423         sync_filesystem(sbi->sb);
4424         retry_cnt = 0;
4425
4426 free_meta:
4427 #ifdef CONFIG_QUOTA
4428         f2fs_truncate_quota_inode_pages(sb);
4429         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4430                 f2fs_quota_off_umount(sbi->sb);
4431 #endif
4432         /*
4433          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4434          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4435          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4436          * falls into an infinite loop in f2fs_sync_meta_pages().
4437          */
4438         truncate_inode_pages_final(META_MAPPING(sbi));
4439         /* evict some inodes being cached by GC */
4440         evict_inodes(sb);
4441         f2fs_unregister_sysfs(sbi);
4442 free_compress_inode:
4443         f2fs_destroy_compress_inode(sbi);
4444 free_root_inode:
4445         dput(sb->s_root);
4446         sb->s_root = NULL;
4447 free_node_inode:
4448         f2fs_release_ino_entry(sbi, true);
4449         truncate_inode_pages_final(NODE_MAPPING(sbi));
4450         iput(sbi->node_inode);
4451         sbi->node_inode = NULL;
4452 free_stats:
4453         f2fs_destroy_stats(sbi);
4454 free_nm:
4455         /* stop discard thread before destroying node manager */
4456         f2fs_stop_discard_thread(sbi);
4457         f2fs_destroy_node_manager(sbi);
4458 free_sm:
4459         f2fs_destroy_segment_manager(sbi);
4460         f2fs_destroy_post_read_wq(sbi);
4461 stop_ckpt_thread:
4462         f2fs_stop_ckpt_thread(sbi);
4463 free_devices:
4464         destroy_device_list(sbi);
4465         kvfree(sbi->ckpt);
4466 free_meta_inode:
4467         make_bad_inode(sbi->meta_inode);
4468         iput(sbi->meta_inode);
4469         sbi->meta_inode = NULL;
4470 free_page_array_cache:
4471         f2fs_destroy_page_array_cache(sbi);
4472 free_xattr_cache:
4473         f2fs_destroy_xattr_caches(sbi);
4474 free_io_dummy:
4475         mempool_destroy(sbi->write_io_dummy);
4476 free_percpu:
4477         destroy_percpu_info(sbi);
4478 free_iostat:
4479         f2fs_destroy_iostat(sbi);
4480 free_bio_info:
4481         for (i = 0; i < NR_PAGE_TYPE; i++)
4482                 kvfree(sbi->write_io[i]);
4483
4484 #if IS_ENABLED(CONFIG_UNICODE)
4485         utf8_unload(sb->s_encoding);
4486         sb->s_encoding = NULL;
4487 #endif
4488 free_options:
4489 #ifdef CONFIG_QUOTA
4490         for (i = 0; i < MAXQUOTAS; i++)
4491                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4492 #endif
4493         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4494         kvfree(options);
4495 free_sb_buf:
4496         kfree(raw_super);
4497 free_sbi:
4498         if (sbi->s_chksum_driver)
4499                 crypto_free_shash(sbi->s_chksum_driver);
4500         kfree(sbi);
4501
4502         /* give only one another chance */
4503         if (retry_cnt > 0 && skip_recovery) {
4504                 retry_cnt--;
4505                 shrink_dcache_sb(sb);
4506                 goto try_onemore;
4507         }
4508         return err;
4509 }
4510
4511 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4512                         const char *dev_name, void *data)
4513 {
4514         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4515 }
4516
4517 static void kill_f2fs_super(struct super_block *sb)
4518 {
4519         if (sb->s_root) {
4520                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4521
4522                 set_sbi_flag(sbi, SBI_IS_CLOSE);
4523                 f2fs_stop_gc_thread(sbi);
4524                 f2fs_stop_discard_thread(sbi);
4525
4526 #ifdef CONFIG_F2FS_FS_COMPRESSION
4527                 /*
4528                  * latter evict_inode() can bypass checking and invalidating
4529                  * compress inode cache.
4530                  */
4531                 if (test_opt(sbi, COMPRESS_CACHE))
4532                         truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4533 #endif
4534
4535                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4536                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4537                         struct cp_control cpc = {
4538                                 .reason = CP_UMOUNT,
4539                         };
4540                         f2fs_write_checkpoint(sbi, &cpc);
4541                 }
4542
4543                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4544                         sb->s_flags &= ~SB_RDONLY;
4545         }
4546         kill_block_super(sb);
4547 }
4548
4549 static struct file_system_type f2fs_fs_type = {
4550         .owner          = THIS_MODULE,
4551         .name           = "f2fs",
4552         .mount          = f2fs_mount,
4553         .kill_sb        = kill_f2fs_super,
4554         .fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4555 };
4556 MODULE_ALIAS_FS("f2fs");
4557
4558 static int __init init_inodecache(void)
4559 {
4560         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4561                         sizeof(struct f2fs_inode_info), 0,
4562                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4563         if (!f2fs_inode_cachep)
4564                 return -ENOMEM;
4565         return 0;
4566 }
4567
4568 static void destroy_inodecache(void)
4569 {
4570         /*
4571          * Make sure all delayed rcu free inodes are flushed before we
4572          * destroy cache.
4573          */
4574         rcu_barrier();
4575         kmem_cache_destroy(f2fs_inode_cachep);
4576 }
4577
4578 static int __init init_f2fs_fs(void)
4579 {
4580         int err;
4581
4582         if (PAGE_SIZE != F2FS_BLKSIZE) {
4583                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4584                                 PAGE_SIZE, F2FS_BLKSIZE);
4585                 return -EINVAL;
4586         }
4587
4588         err = init_inodecache();
4589         if (err)
4590                 goto fail;
4591         err = f2fs_create_node_manager_caches();
4592         if (err)
4593                 goto free_inodecache;
4594         err = f2fs_create_segment_manager_caches();
4595         if (err)
4596                 goto free_node_manager_caches;
4597         err = f2fs_create_checkpoint_caches();
4598         if (err)
4599                 goto free_segment_manager_caches;
4600         err = f2fs_create_recovery_cache();
4601         if (err)
4602                 goto free_checkpoint_caches;
4603         err = f2fs_create_extent_cache();
4604         if (err)
4605                 goto free_recovery_cache;
4606         err = f2fs_create_garbage_collection_cache();
4607         if (err)
4608                 goto free_extent_cache;
4609         err = f2fs_init_sysfs();
4610         if (err)
4611                 goto free_garbage_collection_cache;
4612         err = register_shrinker(&f2fs_shrinker_info);
4613         if (err)
4614                 goto free_sysfs;
4615         err = register_filesystem(&f2fs_fs_type);
4616         if (err)
4617                 goto free_shrinker;
4618         f2fs_create_root_stats();
4619         err = f2fs_init_post_read_processing();
4620         if (err)
4621                 goto free_root_stats;
4622         err = f2fs_init_iostat_processing();
4623         if (err)
4624                 goto free_post_read;
4625         err = f2fs_init_bio_entry_cache();
4626         if (err)
4627                 goto free_iostat;
4628         err = f2fs_init_bioset();
4629         if (err)
4630                 goto free_bio_enrty_cache;
4631         err = f2fs_init_compress_mempool();
4632         if (err)
4633                 goto free_bioset;
4634         err = f2fs_init_compress_cache();
4635         if (err)
4636                 goto free_compress_mempool;
4637         err = f2fs_create_casefold_cache();
4638         if (err)
4639                 goto free_compress_cache;
4640         return 0;
4641 free_compress_cache:
4642         f2fs_destroy_compress_cache();
4643 free_compress_mempool:
4644         f2fs_destroy_compress_mempool();
4645 free_bioset:
4646         f2fs_destroy_bioset();
4647 free_bio_enrty_cache:
4648         f2fs_destroy_bio_entry_cache();
4649 free_iostat:
4650         f2fs_destroy_iostat_processing();
4651 free_post_read:
4652         f2fs_destroy_post_read_processing();
4653 free_root_stats:
4654         f2fs_destroy_root_stats();
4655         unregister_filesystem(&f2fs_fs_type);
4656 free_shrinker:
4657         unregister_shrinker(&f2fs_shrinker_info);
4658 free_sysfs:
4659         f2fs_exit_sysfs();
4660 free_garbage_collection_cache:
4661         f2fs_destroy_garbage_collection_cache();
4662 free_extent_cache:
4663         f2fs_destroy_extent_cache();
4664 free_recovery_cache:
4665         f2fs_destroy_recovery_cache();
4666 free_checkpoint_caches:
4667         f2fs_destroy_checkpoint_caches();
4668 free_segment_manager_caches:
4669         f2fs_destroy_segment_manager_caches();
4670 free_node_manager_caches:
4671         f2fs_destroy_node_manager_caches();
4672 free_inodecache:
4673         destroy_inodecache();
4674 fail:
4675         return err;
4676 }
4677
4678 static void __exit exit_f2fs_fs(void)
4679 {
4680         f2fs_destroy_casefold_cache();
4681         f2fs_destroy_compress_cache();
4682         f2fs_destroy_compress_mempool();
4683         f2fs_destroy_bioset();
4684         f2fs_destroy_bio_entry_cache();
4685         f2fs_destroy_iostat_processing();
4686         f2fs_destroy_post_read_processing();
4687         f2fs_destroy_root_stats();
4688         unregister_filesystem(&f2fs_fs_type);
4689         unregister_shrinker(&f2fs_shrinker_info);
4690         f2fs_exit_sysfs();
4691         f2fs_destroy_garbage_collection_cache();
4692         f2fs_destroy_extent_cache();
4693         f2fs_destroy_recovery_cache();
4694         f2fs_destroy_checkpoint_caches();
4695         f2fs_destroy_segment_manager_caches();
4696         f2fs_destroy_node_manager_caches();
4697         destroy_inodecache();
4698 }
4699
4700 module_init(init_f2fs_fs)
4701 module_exit(exit_f2fs_fs)
4702
4703 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4704 MODULE_DESCRIPTION("Flash Friendly File System");
4705 MODULE_LICENSE("GPL");
4706 MODULE_SOFTDEP("pre: crc32");
4707