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/statfs.h>
  12 #include <linux/buffer_head.h>
  13 #include <linux/backing-dev.h>
  14 #include <linux/kthread.h>
  15 #include <linux/parser.h>
  16 #include <linux/mount.h>
  17 #include <linux/seq_file.h>
  18 #include <linux/proc_fs.h>
  19 #include <linux/random.h>
  20 #include <linux/exportfs.h>
  21 #include <linux/blkdev.h>
  22 #include <linux/quotaops.h>
  23 #include <linux/f2fs_fs.h>
  24 #include <linux/sysfs.h>
  25 #include <linux/quota.h>
  26 #include <linux/unicode.h>
  27 #include <linux/part_stat.h>
  28
  29 #include "f2fs.h"
  30 #include "node.h"
  31 #include "segment.h"
  32 #include "xattr.h"
  33 #include "gc.h"
  34 #include "trace.h"
  35
  36 #define CREATE_TRACE_POINTS
  37 #include <trace/events/f2fs.h>
  38
  39 static struct kmem_cache *f2fs_inode_cachep;
  40
  41 #ifdef CONFIG_F2FS_FAULT_INJECTION
  42
  43 const char *f2fs_fault_name[FAULT_MAX] = {
  44         [FAULT_KMALLOC]         = "kmalloc",
  45         [FAULT_KVMALLOC]        = "kvmalloc",
  46         [FAULT_PAGE_ALLOC]      = "page alloc",
  47         [FAULT_PAGE_GET]        = "page get",
  48         [FAULT_ALLOC_BIO]       = "alloc bio",
  49         [FAULT_ALLOC_NID]       = "alloc nid",
  50         [FAULT_ORPHAN]          = "orphan",
  51         [FAULT_BLOCK]           = "no more block",
  52         [FAULT_DIR_DEPTH]       = "too big dir depth",
  53         [FAULT_EVICT_INODE]     = "evict_inode fail",
  54         [FAULT_TRUNCATE]        = "truncate fail",
  55         [FAULT_READ_IO]         = "read IO error",
  56         [FAULT_CHECKPOINT]      = "checkpoint error",
  57         [FAULT_DISCARD]         = "discard error",
  58         [FAULT_WRITE_IO]        = "write IO error",
  59 };
  60
  61 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  62                                                         unsigned int type)
  63 {
  64         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  65
  66         if (rate) {
  67                 atomic_set(&ffi->inject_ops, 0);
  68                 ffi->inject_rate = rate;
  69         }
  70
  71         if (type)
  72                 ffi->inject_type = type;
  73
  74         if (!rate && !type)
  75                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  76 }
  77 #endif
  78
  79 /* f2fs-wide shrinker description */
  80 static struct shrinker f2fs_shrinker_info = {
  81         .scan_objects = f2fs_shrink_scan,
  82         .count_objects = f2fs_shrink_count,
  83         .seeks = DEFAULT_SEEKS,
  84 };
  85
  86 enum {
  87         Opt_gc_background,
  88         Opt_disable_roll_forward,
  89         Opt_norecovery,
  90         Opt_discard,
  91         Opt_nodiscard,
  92         Opt_noheap,
  93         Opt_heap,
  94         Opt_user_xattr,
  95         Opt_nouser_xattr,
  96         Opt_acl,
  97         Opt_noacl,
  98         Opt_active_logs,
  99         Opt_disable_ext_identify,
 100         Opt_inline_xattr,
 101         Opt_noinline_xattr,
 102         Opt_inline_xattr_size,
 103         Opt_inline_data,
 104         Opt_inline_dentry,
 105         Opt_noinline_dentry,
 106         Opt_flush_merge,
 107         Opt_noflush_merge,
 108         Opt_nobarrier,
 109         Opt_fastboot,
 110         Opt_extent_cache,
 111         Opt_noextent_cache,
 112         Opt_noinline_data,
 113         Opt_data_flush,
 114         Opt_reserve_root,
 115         Opt_resgid,
 116         Opt_resuid,
 117         Opt_mode,
 118         Opt_io_size_bits,
 119         Opt_fault_injection,
 120         Opt_fault_type,
 121         Opt_lazytime,
 122         Opt_nolazytime,
 123         Opt_quota,
 124         Opt_noquota,
 125         Opt_usrquota,
 126         Opt_grpquota,
 127         Opt_prjquota,
 128         Opt_usrjquota,
 129         Opt_grpjquota,
 130         Opt_prjjquota,
 131         Opt_offusrjquota,
 132         Opt_offgrpjquota,
 133         Opt_offprjjquota,
 134         Opt_jqfmt_vfsold,
 135         Opt_jqfmt_vfsv0,
 136         Opt_jqfmt_vfsv1,
 137         Opt_whint,
 138         Opt_alloc,
 139         Opt_fsync,
 140         Opt_test_dummy_encryption,
 141         Opt_inlinecrypt,
 142         Opt_checkpoint_disable,
 143         Opt_checkpoint_disable_cap,
 144         Opt_checkpoint_disable_cap_perc,
 145         Opt_checkpoint_enable,
 146         Opt_compress_algorithm,
 147         Opt_compress_log_size,
 148         Opt_compress_extension,
 149         Opt_atgc,
 150         Opt_err,
 151 };
 152
 153 static match_table_t f2fs_tokens = {
 154         {Opt_gc_background, "background_gc=%s"},
 155         {Opt_disable_roll_forward, "disable_roll_forward"},
 156         {Opt_norecovery, "norecovery"},
 157         {Opt_discard, "discard"},
 158         {Opt_nodiscard, "nodiscard"},
 159         {Opt_noheap, "no_heap"},
 160         {Opt_heap, "heap"},
 161         {Opt_user_xattr, "user_xattr"},
 162         {Opt_nouser_xattr, "nouser_xattr"},
 163         {Opt_acl, "acl"},
 164         {Opt_noacl, "noacl"},
 165         {Opt_active_logs, "active_logs=%u"},
 166         {Opt_disable_ext_identify, "disable_ext_identify"},
 167         {Opt_inline_xattr, "inline_xattr"},
 168         {Opt_noinline_xattr, "noinline_xattr"},
 169         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 170         {Opt_inline_data, "inline_data"},
 171         {Opt_inline_dentry, "inline_dentry"},
 172         {Opt_noinline_dentry, "noinline_dentry"},
 173         {Opt_flush_merge, "flush_merge"},
 174         {Opt_noflush_merge, "noflush_merge"},
 175         {Opt_nobarrier, "nobarrier"},
 176         {Opt_fastboot, "fastboot"},
 177         {Opt_extent_cache, "extent_cache"},
 178         {Opt_noextent_cache, "noextent_cache"},
 179         {Opt_noinline_data, "noinline_data"},
 180         {Opt_data_flush, "data_flush"},
 181         {Opt_reserve_root, "reserve_root=%u"},
 182         {Opt_resgid, "resgid=%u"},
 183         {Opt_resuid, "resuid=%u"},
 184         {Opt_mode, "mode=%s"},
 185         {Opt_io_size_bits, "io_bits=%u"},
 186         {Opt_fault_injection, "fault_injection=%u"},
 187         {Opt_fault_type, "fault_type=%u"},
 188         {Opt_lazytime, "lazytime"},
 189         {Opt_nolazytime, "nolazytime"},
 190         {Opt_quota, "quota"},
 191         {Opt_noquota, "noquota"},
 192         {Opt_usrquota, "usrquota"},
 193         {Opt_grpquota, "grpquota"},
 194         {Opt_prjquota, "prjquota"},
 195         {Opt_usrjquota, "usrjquota=%s"},
 196         {Opt_grpjquota, "grpjquota=%s"},
 197         {Opt_prjjquota, "prjjquota=%s"},
 198         {Opt_offusrjquota, "usrjquota="},
 199         {Opt_offgrpjquota, "grpjquota="},
 200         {Opt_offprjjquota, "prjjquota="},
 201         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 202         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 203         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 204         {Opt_whint, "whint_mode=%s"},
 205         {Opt_alloc, "alloc_mode=%s"},
 206         {Opt_fsync, "fsync_mode=%s"},
 207         {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
 208         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 209         {Opt_inlinecrypt, "inlinecrypt"},
 210         {Opt_checkpoint_disable, "checkpoint=disable"},
 211         {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 212         {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 213         {Opt_checkpoint_enable, "checkpoint=enable"},
 214         {Opt_compress_algorithm, "compress_algorithm=%s"},
 215         {Opt_compress_log_size, "compress_log_size=%u"},
 216         {Opt_compress_extension, "compress_extension=%s"},
 217         {Opt_atgc, "atgc"},
 218         {Opt_err, NULL},
 219 };
 220
 221 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 222 {
 223         struct va_format vaf;
 224         va_list args;
 225         int level;
 226
 227         va_start(args, fmt);
 228
 229         level = printk_get_level(fmt);
 230         vaf.fmt = printk_skip_level(fmt);
 231         vaf.va = &args;
 232         printk("%c%cF2FS-fs (%s): %pV\n",
 233                KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 234
 235         va_end(args);
 236 }
 237
 238 #ifdef CONFIG_UNICODE
 239 static const struct f2fs_sb_encodings {
 240         __u16 magic;
 241         char *name;
 242         char *version;
 243 } f2fs_sb_encoding_map[] = {
 244         {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
 245 };
 246
 247 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
 248                                  const struct f2fs_sb_encodings **encoding,
 249                                  __u16 *flags)
 250 {
 251         __u16 magic = le16_to_cpu(sb->s_encoding);
 252         int i;
 253
 254         for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
 255                 if (magic == f2fs_sb_encoding_map[i].magic)
 256                         break;
 257
 258         if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
 259                 return -EINVAL;
 260
 261         *encoding = &f2fs_sb_encoding_map[i];
 262         *flags = le16_to_cpu(sb->s_encoding_flags);
 263
 264         return 0;
 265 }
 266 #endif
 267
 268 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 269 {
 270         block_t limit = min((sbi->user_block_count << 1) / 1000,
 271                         sbi->user_block_count - sbi->reserved_blocks);
 272
 273         /* limit is 0.2% */
 274         if (test_opt(sbi, RESERVE_ROOT) &&
 275                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 276                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 277                 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 278                           F2FS_OPTION(sbi).root_reserved_blocks);
 279         }
 280         if (!test_opt(sbi, RESERVE_ROOT) &&
 281                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 282                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 283                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 284                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 285                 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 286                           from_kuid_munged(&init_user_ns,
 287                                            F2FS_OPTION(sbi).s_resuid),
 288                           from_kgid_munged(&init_user_ns,
 289                                            F2FS_OPTION(sbi).s_resgid));
 290 }
 291
 292 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
 293 {
 294         if (!F2FS_OPTION(sbi).unusable_cap_perc)
 295                 return;
 296
 297         if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
 298                 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
 299         else
 300                 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
 301                                         F2FS_OPTION(sbi).unusable_cap_perc;
 302
 303         f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
 304                         F2FS_OPTION(sbi).unusable_cap,
 305                         F2FS_OPTION(sbi).unusable_cap_perc);
 306 }
 307
 308 static void init_once(void *foo)
 309 {
 310         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 311
 312         inode_init_once(&fi->vfs_inode);
 313 }
 314
 315 #ifdef CONFIG_QUOTA
 316 static const char * const quotatypes[] = INITQFNAMES;
 317 #define QTYPE2NAME(t) (quotatypes[t])
 318 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 319                                                         substring_t *args)
 320 {
 321         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 322         char *qname;
 323         int ret = -EINVAL;
 324
 325         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 326                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 327                 return -EINVAL;
 328         }
 329         if (f2fs_sb_has_quota_ino(sbi)) {
 330                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 331                 return 0;
 332         }
 333
 334         qname = match_strdup(args);
 335         if (!qname) {
 336                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 337                 return -ENOMEM;
 338         }
 339         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 340                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 341                         ret = 0;
 342                 else
 343                         f2fs_err(sbi, "%s quota file already specified",
 344                                  QTYPE2NAME(qtype));
 345                 goto errout;
 346         }
 347         if (strchr(qname, '/')) {
 348                 f2fs_err(sbi, "quotafile must be on filesystem root");
 349                 goto errout;
 350         }
 351         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 352         set_opt(sbi, QUOTA);
 353         return 0;
 354 errout:
 355         kfree(qname);
 356         return ret;
 357 }
 358
 359 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 360 {
 361         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 362
 363         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 364                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 365                 return -EINVAL;
 366         }
 367         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 368         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 369         return 0;
 370 }
 371
 372 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 373 {
 374         /*
 375          * We do the test below only for project quotas. 'usrquota' and
 376          * 'grpquota' mount options are allowed even without quota feature
 377          * to support legacy quotas in quota files.
 378          */
 379         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 380                 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 381                 return -1;
 382         }
 383         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 384                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 385                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 386                 if (test_opt(sbi, USRQUOTA) &&
 387                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 388                         clear_opt(sbi, USRQUOTA);
 389
 390                 if (test_opt(sbi, GRPQUOTA) &&
 391                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 392                         clear_opt(sbi, GRPQUOTA);
 393
 394                 if (test_opt(sbi, PRJQUOTA) &&
 395                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 396                         clear_opt(sbi, PRJQUOTA);
 397
 398                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 399                                 test_opt(sbi, PRJQUOTA)) {
 400                         f2fs_err(sbi, "old and new quota format mixing");
 401                         return -1;
 402                 }
 403
 404                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 405                         f2fs_err(sbi, "journaled quota format not specified");
 406                         return -1;
 407                 }
 408         }
 409
 410         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 411                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 412                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 413         }
 414         return 0;
 415 }
 416 #endif
 417
 418 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
 419                                           const char *opt,
 420                                           const substring_t *arg,
 421                                           bool is_remount)
 422 {
 423         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 424 #ifdef CONFIG_FS_ENCRYPTION
 425         int err;
 426
 427         if (!f2fs_sb_has_encrypt(sbi)) {
 428                 f2fs_err(sbi, "Encrypt feature is off");
 429                 return -EINVAL;
 430         }
 431
 432         /*
 433          * This mount option is just for testing, and it's not worthwhile to
 434          * implement the extra complexity (e.g. RCU protection) that would be
 435          * needed to allow it to be set or changed during remount.  We do allow
 436          * it to be specified during remount, but only if there is no change.
 437          */
 438         if (is_remount && !F2FS_OPTION(sbi).dummy_enc_ctx.ctx) {
 439                 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
 440                 return -EINVAL;
 441         }
 442         err = fscrypt_set_test_dummy_encryption(
 443                 sb, arg, &F2FS_OPTION(sbi).dummy_enc_ctx);
 444         if (err) {
 445                 if (err == -EEXIST)
 446                         f2fs_warn(sbi,
 447                                   "Can't change test_dummy_encryption on remount");
 448                 else if (err == -EINVAL)
 449                         f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
 450                                   opt);
 451                 else
 452                         f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
 453                                   opt, err);
 454                 return -EINVAL;
 455         }
 456         f2fs_warn(sbi, "Test dummy encryption mode enabled");
 457 #else
 458         f2fs_warn(sbi, "Test dummy encryption mount option ignored");
 459 #endif
 460         return 0;
 461 }
 462
 463 static int parse_options(struct super_block *sb, char *options, bool is_remount)
 464 {
 465         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 466         substring_t args[MAX_OPT_ARGS];
 467 #ifdef CONFIG_F2FS_FS_COMPRESSION
 468         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 469         int ext_cnt;
 470 #endif
 471         char *p, *name;
 472         int arg = 0;
 473         kuid_t uid;
 474         kgid_t gid;
 475         int ret;
 476
 477         if (!options)
 478                 return 0;
 479
 480         while ((p = strsep(&options, ",")) != NULL) {
 481                 int token;
 482                 if (!*p)
 483                         continue;
 484                 /*
 485                  * Initialize args struct so we know whether arg was
 486                  * found; some options take optional arguments.
 487                  */
 488                 args[0].to = args[0].from = NULL;
 489                 token = match_token(p, f2fs_tokens, args);
 490
 491                 switch (token) {
 492                 case Opt_gc_background:
 493                         name = match_strdup(&args[0]);
 494
 495                         if (!name)
 496                                 return -ENOMEM;
 497                         if (!strcmp(name, "on")) {
 498                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
 499                         } else if (!strcmp(name, "off")) {
 500                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
 501                         } else if (!strcmp(name, "sync")) {
 502                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
 503                         } else {
 504                                 kfree(name);
 505                                 return -EINVAL;
 506                         }
 507                         kfree(name);
 508                         break;
 509                 case Opt_disable_roll_forward:
 510                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 511                         break;
 512                 case Opt_norecovery:
 513                         /* this option mounts f2fs with ro */
 514                         set_opt(sbi, NORECOVERY);
 515                         if (!f2fs_readonly(sb))
 516                                 return -EINVAL;
 517                         break;
 518                 case Opt_discard:
 519                         set_opt(sbi, DISCARD);
 520                         break;
 521                 case Opt_nodiscard:
 522                         if (f2fs_sb_has_blkzoned(sbi)) {
 523                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 524                                 return -EINVAL;
 525                         }
 526                         clear_opt(sbi, DISCARD);
 527                         break;
 528                 case Opt_noheap:
 529                         set_opt(sbi, NOHEAP);
 530                         break;
 531                 case Opt_heap:
 532                         clear_opt(sbi, NOHEAP);
 533                         break;
 534 #ifdef CONFIG_F2FS_FS_XATTR
 535                 case Opt_user_xattr:
 536                         set_opt(sbi, XATTR_USER);
 537                         break;
 538                 case Opt_nouser_xattr:
 539                         clear_opt(sbi, XATTR_USER);
 540                         break;
 541                 case Opt_inline_xattr:
 542                         set_opt(sbi, INLINE_XATTR);
 543                         break;
 544                 case Opt_noinline_xattr:
 545                         clear_opt(sbi, INLINE_XATTR);
 546                         break;
 547                 case Opt_inline_xattr_size:
 548                         if (args->from && match_int(args, &arg))
 549                                 return -EINVAL;
 550                         set_opt(sbi, INLINE_XATTR_SIZE);
 551                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 552                         break;
 553 #else
 554                 case Opt_user_xattr:
 555                         f2fs_info(sbi, "user_xattr options not supported");
 556                         break;
 557                 case Opt_nouser_xattr:
 558                         f2fs_info(sbi, "nouser_xattr options not supported");
 559                         break;
 560                 case Opt_inline_xattr:
 561                         f2fs_info(sbi, "inline_xattr options not supported");
 562                         break;
 563                 case Opt_noinline_xattr:
 564                         f2fs_info(sbi, "noinline_xattr options not supported");
 565                         break;
 566 #endif
 567 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 568                 case Opt_acl:
 569                         set_opt(sbi, POSIX_ACL);
 570                         break;
 571                 case Opt_noacl:
 572                         clear_opt(sbi, POSIX_ACL);
 573                         break;
 574 #else
 575                 case Opt_acl:
 576                         f2fs_info(sbi, "acl options not supported");
 577                         break;
 578                 case Opt_noacl:
 579                         f2fs_info(sbi, "noacl options not supported");
 580                         break;
 581 #endif
 582                 case Opt_active_logs:
 583                         if (args->from && match_int(args, &arg))
 584                                 return -EINVAL;
 585                         if (arg != 2 && arg != 4 &&
 586                                 arg != NR_CURSEG_PERSIST_TYPE)
 587                                 return -EINVAL;
 588                         F2FS_OPTION(sbi).active_logs = arg;
 589                         break;
 590                 case Opt_disable_ext_identify:
 591                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 592                         break;
 593                 case Opt_inline_data:
 594                         set_opt(sbi, INLINE_DATA);
 595                         break;
 596                 case Opt_inline_dentry:
 597                         set_opt(sbi, INLINE_DENTRY);
 598                         break;
 599                 case Opt_noinline_dentry:
 600                         clear_opt(sbi, INLINE_DENTRY);
 601                         break;
 602                 case Opt_flush_merge:
 603                         set_opt(sbi, FLUSH_MERGE);
 604                         break;
 605                 case Opt_noflush_merge:
 606                         clear_opt(sbi, FLUSH_MERGE);
 607                         break;
 608                 case Opt_nobarrier:
 609                         set_opt(sbi, NOBARRIER);
 610                         break;
 611                 case Opt_fastboot:
 612                         set_opt(sbi, FASTBOOT);
 613                         break;
 614                 case Opt_extent_cache:
 615                         set_opt(sbi, EXTENT_CACHE);
 616                         break;
 617                 case Opt_noextent_cache:
 618                         clear_opt(sbi, EXTENT_CACHE);
 619                         break;
 620                 case Opt_noinline_data:
 621                         clear_opt(sbi, INLINE_DATA);
 622                         break;
 623                 case Opt_data_flush:
 624                         set_opt(sbi, DATA_FLUSH);
 625                         break;
 626                 case Opt_reserve_root:
 627                         if (args->from && match_int(args, &arg))
 628                                 return -EINVAL;
 629                         if (test_opt(sbi, RESERVE_ROOT)) {
 630                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 631                                           F2FS_OPTION(sbi).root_reserved_blocks);
 632                         } else {
 633                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 634                                 set_opt(sbi, RESERVE_ROOT);
 635                         }
 636                         break;
 637                 case Opt_resuid:
 638                         if (args->from && match_int(args, &arg))
 639                                 return -EINVAL;
 640                         uid = make_kuid(current_user_ns(), arg);
 641                         if (!uid_valid(uid)) {
 642                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 643                                 return -EINVAL;
 644                         }
 645                         F2FS_OPTION(sbi).s_resuid = uid;
 646                         break;
 647                 case Opt_resgid:
 648                         if (args->from && match_int(args, &arg))
 649                                 return -EINVAL;
 650                         gid = make_kgid(current_user_ns(), arg);
 651                         if (!gid_valid(gid)) {
 652                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 653                                 return -EINVAL;
 654                         }
 655                         F2FS_OPTION(sbi).s_resgid = gid;
 656                         break;
 657                 case Opt_mode:
 658                         name = match_strdup(&args[0]);
 659
 660                         if (!name)
 661                                 return -ENOMEM;
 662                         if (!strcmp(name, "adaptive")) {
 663                                 if (f2fs_sb_has_blkzoned(sbi)) {
 664                                         f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
 665                                         kfree(name);
 666                                         return -EINVAL;
 667                                 }
 668                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
 669                         } else if (!strcmp(name, "lfs")) {
 670                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
 671                         } else {
 672                                 kfree(name);
 673                                 return -EINVAL;
 674                         }
 675                         kfree(name);
 676                         break;
 677                 case Opt_io_size_bits:
 678                         if (args->from && match_int(args, &arg))
 679                                 return -EINVAL;
 680                         if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
 681                                 f2fs_warn(sbi, "Not support %d, larger than %d",
 682                                           1 << arg, BIO_MAX_PAGES);
 683                                 return -EINVAL;
 684                         }
 685                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 686                         break;
 687 #ifdef CONFIG_F2FS_FAULT_INJECTION
 688                 case Opt_fault_injection:
 689                         if (args->from && match_int(args, &arg))
 690                                 return -EINVAL;
 691                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 692                         set_opt(sbi, FAULT_INJECTION);
 693                         break;
 694
 695                 case Opt_fault_type:
 696                         if (args->from && match_int(args, &arg))
 697                                 return -EINVAL;
 698                         f2fs_build_fault_attr(sbi, 0, arg);
 699                         set_opt(sbi, FAULT_INJECTION);
 700                         break;
 701 #else
 702                 case Opt_fault_injection:
 703                         f2fs_info(sbi, "fault_injection options not supported");
 704                         break;
 705
 706                 case Opt_fault_type:
 707                         f2fs_info(sbi, "fault_type options not supported");
 708                         break;
 709 #endif
 710                 case Opt_lazytime:
 711                         sb->s_flags |= SB_LAZYTIME;
 712                         break;
 713                 case Opt_nolazytime:
 714                         sb->s_flags &= ~SB_LAZYTIME;
 715                         break;
 716 #ifdef CONFIG_QUOTA
 717                 case Opt_quota:
 718                 case Opt_usrquota:
 719                         set_opt(sbi, USRQUOTA);
 720                         break;
 721                 case Opt_grpquota:
 722                         set_opt(sbi, GRPQUOTA);
 723                         break;
 724                 case Opt_prjquota:
 725                         set_opt(sbi, PRJQUOTA);
 726                         break;
 727                 case Opt_usrjquota:
 728                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 729                         if (ret)
 730                                 return ret;
 731                         break;
 732                 case Opt_grpjquota:
 733                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 734                         if (ret)
 735                                 return ret;
 736                         break;
 737                 case Opt_prjjquota:
 738                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 739                         if (ret)
 740                                 return ret;
 741                         break;
 742                 case Opt_offusrjquota:
 743                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 744                         if (ret)
 745                                 return ret;
 746                         break;
 747                 case Opt_offgrpjquota:
 748                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 749                         if (ret)
 750                                 return ret;
 751                         break;
 752                 case Opt_offprjjquota:
 753                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 754                         if (ret)
 755                                 return ret;
 756                         break;
 757                 case Opt_jqfmt_vfsold:
 758                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 759                         break;
 760                 case Opt_jqfmt_vfsv0:
 761                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 762                         break;
 763                 case Opt_jqfmt_vfsv1:
 764                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 765                         break;
 766                 case Opt_noquota:
 767                         clear_opt(sbi, QUOTA);
 768                         clear_opt(sbi, USRQUOTA);
 769                         clear_opt(sbi, GRPQUOTA);
 770                         clear_opt(sbi, PRJQUOTA);
 771                         break;
 772 #else
 773                 case Opt_quota:
 774                 case Opt_usrquota:
 775                 case Opt_grpquota:
 776                 case Opt_prjquota:
 777                 case Opt_usrjquota:
 778                 case Opt_grpjquota:
 779                 case Opt_prjjquota:
 780                 case Opt_offusrjquota:
 781                 case Opt_offgrpjquota:
 782                 case Opt_offprjjquota:
 783                 case Opt_jqfmt_vfsold:
 784                 case Opt_jqfmt_vfsv0:
 785                 case Opt_jqfmt_vfsv1:
 786                 case Opt_noquota:
 787                         f2fs_info(sbi, "quota operations not supported");
 788                         break;
 789 #endif
 790                 case Opt_whint:
 791                         name = match_strdup(&args[0]);
 792                         if (!name)
 793                                 return -ENOMEM;
 794                         if (!strcmp(name, "user-based")) {
 795                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 796                         } else if (!strcmp(name, "off")) {
 797                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 798                         } else if (!strcmp(name, "fs-based")) {
 799                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 800                         } else {
 801                                 kfree(name);
 802                                 return -EINVAL;
 803                         }
 804                         kfree(name);
 805                         break;
 806                 case Opt_alloc:
 807                         name = match_strdup(&args[0]);
 808                         if (!name)
 809                                 return -ENOMEM;
 810
 811                         if (!strcmp(name, "default")) {
 812                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 813                         } else if (!strcmp(name, "reuse")) {
 814                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 815                         } else {
 816                                 kfree(name);
 817                                 return -EINVAL;
 818                         }
 819                         kfree(name);
 820                         break;
 821                 case Opt_fsync:
 822                         name = match_strdup(&args[0]);
 823                         if (!name)
 824                                 return -ENOMEM;
 825                         if (!strcmp(name, "posix")) {
 826                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 827                         } else if (!strcmp(name, "strict")) {
 828                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 829                         } else if (!strcmp(name, "nobarrier")) {
 830                                 F2FS_OPTION(sbi).fsync_mode =
 831                                                         FSYNC_MODE_NOBARRIER;
 832                         } else {
 833                                 kfree(name);
 834                                 return -EINVAL;
 835                         }
 836                         kfree(name);
 837                         break;
 838                 case Opt_test_dummy_encryption:
 839                         ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
 840                                                              is_remount);
 841                         if (ret)
 842                                 return ret;
 843                         break;
 844                 case Opt_inlinecrypt:
 845 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
 846                         sb->s_flags |= SB_INLINECRYPT;
 847 #else
 848                         f2fs_info(sbi, "inline encryption not supported");
 849 #endif
 850                         break;
 851                 case Opt_checkpoint_disable_cap_perc:
 852                         if (args->from && match_int(args, &arg))
 853                                 return -EINVAL;
 854                         if (arg < 0 || arg > 100)
 855                                 return -EINVAL;
 856                         F2FS_OPTION(sbi).unusable_cap_perc = arg;
 857                         set_opt(sbi, DISABLE_CHECKPOINT);
 858                         break;
 859                 case Opt_checkpoint_disable_cap:
 860                         if (args->from && match_int(args, &arg))
 861                                 return -EINVAL;
 862                         F2FS_OPTION(sbi).unusable_cap = arg;
 863                         set_opt(sbi, DISABLE_CHECKPOINT);
 864                         break;
 865                 case Opt_checkpoint_disable:
 866                         set_opt(sbi, DISABLE_CHECKPOINT);
 867                         break;
 868                 case Opt_checkpoint_enable:
 869                         clear_opt(sbi, DISABLE_CHECKPOINT);
 870                         break;
 871 #ifdef CONFIG_F2FS_FS_COMPRESSION
 872                 case Opt_compress_algorithm:
 873                         if (!f2fs_sb_has_compression(sbi)) {
 874                                 f2fs_info(sbi, "Image doesn't support compression");
 875                                 break;
 876                         }
 877                         name = match_strdup(&args[0]);
 878                         if (!name)
 879                                 return -ENOMEM;
 880                         if (!strcmp(name, "lzo")) {
 881                                 F2FS_OPTION(sbi).compress_algorithm =
 882                                                                 COMPRESS_LZO;
 883                         } else if (!strcmp(name, "lz4")) {
 884                                 F2FS_OPTION(sbi).compress_algorithm =
 885                                                                 COMPRESS_LZ4;
 886                         } else if (!strcmp(name, "zstd")) {
 887                                 F2FS_OPTION(sbi).compress_algorithm =
 888                                                                 COMPRESS_ZSTD;
 889                         } else if (!strcmp(name, "lzo-rle")) {
 890                                 F2FS_OPTION(sbi).compress_algorithm =
 891                                                                 COMPRESS_LZORLE;
 892                         } else {
 893                                 kfree(name);
 894                                 return -EINVAL;
 895                         }
 896                         kfree(name);
 897                         break;
 898                 case Opt_compress_log_size:
 899                         if (!f2fs_sb_has_compression(sbi)) {
 900                                 f2fs_info(sbi, "Image doesn't support compression");
 901                                 break;
 902                         }
 903                         if (args->from && match_int(args, &arg))
 904                                 return -EINVAL;
 905                         if (arg < MIN_COMPRESS_LOG_SIZE ||
 906                                 arg > MAX_COMPRESS_LOG_SIZE) {
 907                                 f2fs_err(sbi,
 908                                         "Compress cluster log size is out of range");
 909                                 return -EINVAL;
 910                         }
 911                         F2FS_OPTION(sbi).compress_log_size = arg;
 912                         break;
 913                 case Opt_compress_extension:
 914                         if (!f2fs_sb_has_compression(sbi)) {
 915                                 f2fs_info(sbi, "Image doesn't support compression");
 916                                 break;
 917                         }
 918                         name = match_strdup(&args[0]);
 919                         if (!name)
 920                                 return -ENOMEM;
 921
 922                         ext = F2FS_OPTION(sbi).extensions;
 923                         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
 924
 925                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
 926                                 ext_cnt >= COMPRESS_EXT_NUM) {
 927                                 f2fs_err(sbi,
 928                                         "invalid extension length/number");
 929                                 kfree(name);
 930                                 return -EINVAL;
 931                         }
 932
 933                         strcpy(ext[ext_cnt], name);
 934                         F2FS_OPTION(sbi).compress_ext_cnt++;
 935                         kfree(name);
 936                         break;
 937 #else
 938                 case Opt_compress_algorithm:
 939                 case Opt_compress_log_size:
 940                 case Opt_compress_extension:
 941                         f2fs_info(sbi, "compression options not supported");
 942                         break;
 943 #endif
 944                 case Opt_atgc:
 945                         set_opt(sbi, ATGC);
 946                         break;
 947                 default:
 948                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
 949                                  p);
 950                         return -EINVAL;
 951                 }
 952         }
 953 #ifdef CONFIG_QUOTA
 954         if (f2fs_check_quota_options(sbi))
 955                 return -EINVAL;
 956 #else
 957         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
 958                 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 959                 return -EINVAL;
 960         }
 961         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
 962                 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 963                 return -EINVAL;
 964         }
 965 #endif
 966 #ifndef CONFIG_UNICODE
 967         if (f2fs_sb_has_casefold(sbi)) {
 968                 f2fs_err(sbi,
 969                         "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
 970                 return -EINVAL;
 971         }
 972 #endif
 973
 974         if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
 975                 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
 976                          F2FS_IO_SIZE_KB(sbi));
 977                 return -EINVAL;
 978         }
 979
 980         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 981                 int min_size, max_size;
 982
 983                 if (!f2fs_sb_has_extra_attr(sbi) ||
 984                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 985                         f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
 986                         return -EINVAL;
 987                 }
 988                 if (!test_opt(sbi, INLINE_XATTR)) {
 989                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
 990                         return -EINVAL;
 991                 }
 992
 993                 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
 994                 max_size = MAX_INLINE_XATTR_SIZE;
 995
 996                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
 997                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
 998                         f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
 999                                  min_size, max_size);
1000                         return -EINVAL;
1001                 }
1002         }
1003
1004         if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1005                 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
1006                 return -EINVAL;
1007         }
1008
1009         /* Not pass down write hints if the number of active logs is lesser
1010          * than NR_CURSEG_PERSIST_TYPE.
1011          */
1012         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
1013                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1014         return 0;
1015 }
1016
1017 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1018 {
1019         struct f2fs_inode_info *fi;
1020
1021         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
1022         if (!fi)
1023                 return NULL;
1024
1025         init_once((void *) fi);
1026
1027         /* Initialize f2fs-specific inode info */
1028         atomic_set(&fi->dirty_pages, 0);
1029         atomic_set(&fi->i_compr_blocks, 0);
1030         init_rwsem(&fi->i_sem);
1031         spin_lock_init(&fi->i_size_lock);
1032         INIT_LIST_HEAD(&fi->dirty_list);
1033         INIT_LIST_HEAD(&fi->gdirty_list);
1034         INIT_LIST_HEAD(&fi->inmem_ilist);
1035         INIT_LIST_HEAD(&fi->inmem_pages);
1036         mutex_init(&fi->inmem_lock);
1037         init_rwsem(&fi->i_gc_rwsem[READ]);
1038         init_rwsem(&fi->i_gc_rwsem[WRITE]);
1039         init_rwsem(&fi->i_mmap_sem);
1040         init_rwsem(&fi->i_xattr_sem);
1041
1042         /* Will be used by directory only */
1043         fi->i_dir_level = F2FS_SB(sb)->dir_level;
1044
1045         fi->ra_offset = -1;
1046
1047         return &fi->vfs_inode;
1048 }
1049
1050 static int f2fs_drop_inode(struct inode *inode)
1051 {
1052         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1053         int ret;
1054
1055         /*
1056          * during filesystem shutdown, if checkpoint is disabled,
1057          * drop useless meta/node dirty pages.
1058          */
1059         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1060                 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1061                         inode->i_ino == F2FS_META_INO(sbi)) {
1062                         trace_f2fs_drop_inode(inode, 1);
1063                         return 1;
1064                 }
1065         }
1066
1067         /*
1068          * This is to avoid a deadlock condition like below.
1069          * writeback_single_inode(inode)
1070          *  - f2fs_write_data_page
1071          *    - f2fs_gc -> iput -> evict
1072          *       - inode_wait_for_writeback(inode)
1073          */
1074         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1075                 if (!inode->i_nlink && !is_bad_inode(inode)) {
1076                         /* to avoid evict_inode call simultaneously */
1077                         atomic_inc(&inode->i_count);
1078                         spin_unlock(&inode->i_lock);
1079
1080                         /* some remained atomic pages should discarded */
1081                         if (f2fs_is_atomic_file(inode))
1082                                 f2fs_drop_inmem_pages(inode);
1083
1084                         /* should remain fi->extent_tree for writepage */
1085                         f2fs_destroy_extent_node(inode);
1086
1087                         sb_start_intwrite(inode->i_sb);
1088                         f2fs_i_size_write(inode, 0);
1089
1090                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1091                                         inode, NULL, 0, DATA);
1092                         truncate_inode_pages_final(inode->i_mapping);
1093
1094                         if (F2FS_HAS_BLOCKS(inode))
1095                                 f2fs_truncate(inode);
1096
1097                         sb_end_intwrite(inode->i_sb);
1098
1099                         spin_lock(&inode->i_lock);
1100                         atomic_dec(&inode->i_count);
1101                 }
1102                 trace_f2fs_drop_inode(inode, 0);
1103                 return 0;
1104         }
1105         ret = generic_drop_inode(inode);
1106         if (!ret)
1107                 ret = fscrypt_drop_inode(inode);
1108         trace_f2fs_drop_inode(inode, ret);
1109         return ret;
1110 }
1111
1112 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1113 {
1114         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1115         int ret = 0;
1116
1117         spin_lock(&sbi->inode_lock[DIRTY_META]);
1118         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1119                 ret = 1;
1120         } else {
1121                 set_inode_flag(inode, FI_DIRTY_INODE);
1122                 stat_inc_dirty_inode(sbi, DIRTY_META);
1123         }
1124         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1125                 list_add_tail(&F2FS_I(inode)->gdirty_list,
1126                                 &sbi->inode_list[DIRTY_META]);
1127                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1128         }
1129         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1130         return ret;
1131 }
1132
1133 void f2fs_inode_synced(struct inode *inode)
1134 {
1135         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1136
1137         spin_lock(&sbi->inode_lock[DIRTY_META]);
1138         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1139                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1140                 return;
1141         }
1142         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1143                 list_del_init(&F2FS_I(inode)->gdirty_list);
1144                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1145         }
1146         clear_inode_flag(inode, FI_DIRTY_INODE);
1147         clear_inode_flag(inode, FI_AUTO_RECOVER);
1148         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1149         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1150 }
1151
1152 /*
1153  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1154  *
1155  * We should call set_dirty_inode to write the dirty inode through write_inode.
1156  */
1157 static void f2fs_dirty_inode(struct inode *inode, int flags)
1158 {
1159         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1160
1161         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1162                         inode->i_ino == F2FS_META_INO(sbi))
1163                 return;
1164
1165         if (flags == I_DIRTY_TIME)
1166                 return;
1167
1168         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1169                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1170
1171         f2fs_inode_dirtied(inode, false);
1172 }
1173
1174 static void f2fs_free_inode(struct inode *inode)
1175 {
1176         fscrypt_free_inode(inode);
1177         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1178 }
1179
1180 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1181 {
1182         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1183         percpu_counter_destroy(&sbi->total_valid_inode_count);
1184 }
1185
1186 static void destroy_device_list(struct f2fs_sb_info *sbi)
1187 {
1188         int i;
1189
1190         for (i = 0; i < sbi->s_ndevs; i++) {
1191                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1192 #ifdef CONFIG_BLK_DEV_ZONED
1193                 kvfree(FDEV(i).blkz_seq);
1194                 kfree(FDEV(i).zone_capacity_blocks);
1195 #endif
1196         }
1197         kvfree(sbi->devs);
1198 }
1199
1200 static void f2fs_put_super(struct super_block *sb)
1201 {
1202         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1203         int i;
1204         bool dropped;
1205
1206         /* unregister procfs/sysfs entries in advance to avoid race case */
1207         f2fs_unregister_sysfs(sbi);
1208
1209         f2fs_quota_off_umount(sb);
1210
1211         /* prevent remaining shrinker jobs */
1212         mutex_lock(&sbi->umount_mutex);
1213
1214         /*
1215          * We don't need to do checkpoint when superblock is clean.
1216          * But, the previous checkpoint was not done by umount, it needs to do
1217          * clean checkpoint again.
1218          */
1219         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1220                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1221                 struct cp_control cpc = {
1222                         .reason = CP_UMOUNT,
1223                 };
1224                 f2fs_write_checkpoint(sbi, &cpc);
1225         }
1226
1227         /* be sure to wait for any on-going discard commands */
1228         dropped = f2fs_issue_discard_timeout(sbi);
1229
1230         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1231                                         !sbi->discard_blks && !dropped) {
1232                 struct cp_control cpc = {
1233                         .reason = CP_UMOUNT | CP_TRIMMED,
1234                 };
1235                 f2fs_write_checkpoint(sbi, &cpc);
1236         }
1237
1238         /*
1239          * normally superblock is clean, so we need to release this.
1240          * In addition, EIO will skip do checkpoint, we need this as well.
1241          */
1242         f2fs_release_ino_entry(sbi, true);
1243
1244         f2fs_leave_shrinker(sbi);
1245         mutex_unlock(&sbi->umount_mutex);
1246
1247         /* our cp_error case, we can wait for any writeback page */
1248         f2fs_flush_merged_writes(sbi);
1249
1250         f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1251
1252         f2fs_bug_on(sbi, sbi->fsync_node_num);
1253
1254         iput(sbi->node_inode);
1255         sbi->node_inode = NULL;
1256
1257         iput(sbi->meta_inode);
1258         sbi->meta_inode = NULL;
1259
1260         /*
1261          * iput() can update stat information, if f2fs_write_checkpoint()
1262          * above failed with error.
1263          */
1264         f2fs_destroy_stats(sbi);
1265
1266         /* destroy f2fs internal modules */
1267         f2fs_destroy_node_manager(sbi);
1268         f2fs_destroy_segment_manager(sbi);
1269
1270         f2fs_destroy_post_read_wq(sbi);
1271
1272         kvfree(sbi->ckpt);
1273
1274         sb->s_fs_info = NULL;
1275         if (sbi->s_chksum_driver)
1276                 crypto_free_shash(sbi->s_chksum_driver);
1277         kfree(sbi->raw_super);
1278
1279         destroy_device_list(sbi);
1280         f2fs_destroy_xattr_caches(sbi);
1281         mempool_destroy(sbi->write_io_dummy);
1282 #ifdef CONFIG_QUOTA
1283         for (i = 0; i < MAXQUOTAS; i++)
1284                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1285 #endif
1286         fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx);
1287         destroy_percpu_info(sbi);
1288         for (i = 0; i < NR_PAGE_TYPE; i++)
1289                 kvfree(sbi->write_io[i]);
1290 #ifdef CONFIG_UNICODE
1291         utf8_unload(sb->s_encoding);
1292 #endif
1293         kfree(sbi);
1294 }
1295
1296 int f2fs_sync_fs(struct super_block *sb, int sync)
1297 {
1298         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1299         int err = 0;
1300
1301         if (unlikely(f2fs_cp_error(sbi)))
1302                 return 0;
1303         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1304                 return 0;
1305
1306         trace_f2fs_sync_fs(sb, sync);
1307
1308         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1309                 return -EAGAIN;
1310
1311         if (sync) {
1312                 struct cp_control cpc;
1313
1314                 cpc.reason = __get_cp_reason(sbi);
1315
1316                 down_write(&sbi->gc_lock);
1317                 err = f2fs_write_checkpoint(sbi, &cpc);
1318                 up_write(&sbi->gc_lock);
1319         }
1320         f2fs_trace_ios(NULL, 1);
1321
1322         return err;
1323 }
1324
1325 static int f2fs_freeze(struct super_block *sb)
1326 {
1327         if (f2fs_readonly(sb))
1328                 return 0;
1329
1330         /* IO error happened before */
1331         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1332                 return -EIO;
1333
1334         /* must be clean, since sync_filesystem() was already called */
1335         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1336                 return -EINVAL;
1337         return 0;
1338 }
1339
1340 static int f2fs_unfreeze(struct super_block *sb)
1341 {
1342         return 0;
1343 }
1344
1345 #ifdef CONFIG_QUOTA
1346 static int f2fs_statfs_project(struct super_block *sb,
1347                                 kprojid_t projid, struct kstatfs *buf)
1348 {
1349         struct kqid qid;
1350         struct dquot *dquot;
1351         u64 limit;
1352         u64 curblock;
1353
1354         qid = make_kqid_projid(projid);
1355         dquot = dqget(sb, qid);
1356         if (IS_ERR(dquot))
1357                 return PTR_ERR(dquot);
1358         spin_lock(&dquot->dq_dqb_lock);
1359
1360         limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1361                                         dquot->dq_dqb.dqb_bhardlimit);
1362         if (limit)
1363                 limit >>= sb->s_blocksize_bits;
1364
1365         if (limit && buf->f_blocks > limit) {
1366                 curblock = (dquot->dq_dqb.dqb_curspace +
1367                             dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1368                 buf->f_blocks = limit;
1369                 buf->f_bfree = buf->f_bavail =
1370                         (buf->f_blocks > curblock) ?
1371                          (buf->f_blocks - curblock) : 0;
1372         }
1373
1374         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1375                                         dquot->dq_dqb.dqb_ihardlimit);
1376
1377         if (limit && buf->f_files > limit) {
1378                 buf->f_files = limit;
1379                 buf->f_ffree =
1380                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1381                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1382         }
1383
1384         spin_unlock(&dquot->dq_dqb_lock);
1385         dqput(dquot);
1386         return 0;
1387 }
1388 #endif
1389
1390 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1391 {
1392         struct super_block *sb = dentry->d_sb;
1393         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1394         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1395         block_t total_count, user_block_count, start_count;
1396         u64 avail_node_count;
1397
1398         total_count = le64_to_cpu(sbi->raw_super->block_count);
1399         user_block_count = sbi->user_block_count;
1400         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1401         buf->f_type = F2FS_SUPER_MAGIC;
1402         buf->f_bsize = sbi->blocksize;
1403
1404         buf->f_blocks = total_count - start_count;
1405         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1406                                                 sbi->current_reserved_blocks;
1407
1408         spin_lock(&sbi->stat_lock);
1409         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1410                 buf->f_bfree = 0;
1411         else
1412                 buf->f_bfree -= sbi->unusable_block_count;
1413         spin_unlock(&sbi->stat_lock);
1414
1415         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1416                 buf->f_bavail = buf->f_bfree -
1417                                 F2FS_OPTION(sbi).root_reserved_blocks;
1418         else
1419                 buf->f_bavail = 0;
1420
1421         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1422
1423         if (avail_node_count > user_block_count) {
1424                 buf->f_files = user_block_count;
1425                 buf->f_ffree = buf->f_bavail;
1426         } else {
1427                 buf->f_files = avail_node_count;
1428                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1429                                         buf->f_bavail);
1430         }
1431
1432         buf->f_namelen = F2FS_NAME_LEN;
1433         buf->f_fsid.val[0] = (u32)id;
1434         buf->f_fsid.val[1] = (u32)(id >> 32);
1435
1436 #ifdef CONFIG_QUOTA
1437         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1438                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1439                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1440         }
1441 #endif
1442         return 0;
1443 }
1444
1445 static inline void f2fs_show_quota_options(struct seq_file *seq,
1446                                            struct super_block *sb)
1447 {
1448 #ifdef CONFIG_QUOTA
1449         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1450
1451         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1452                 char *fmtname = "";
1453
1454                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1455                 case QFMT_VFS_OLD:
1456                         fmtname = "vfsold";
1457                         break;
1458                 case QFMT_VFS_V0:
1459                         fmtname = "vfsv0";
1460                         break;
1461                 case QFMT_VFS_V1:
1462                         fmtname = "vfsv1";
1463                         break;
1464                 }
1465                 seq_printf(seq, ",jqfmt=%s", fmtname);
1466         }
1467
1468         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1469                 seq_show_option(seq, "usrjquota",
1470                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1471
1472         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1473                 seq_show_option(seq, "grpjquota",
1474                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1475
1476         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1477                 seq_show_option(seq, "prjjquota",
1478                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1479 #endif
1480 }
1481
1482 static inline void f2fs_show_compress_options(struct seq_file *seq,
1483                                                         struct super_block *sb)
1484 {
1485         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1486         char *algtype = "";
1487         int i;
1488
1489         if (!f2fs_sb_has_compression(sbi))
1490                 return;
1491
1492         switch (F2FS_OPTION(sbi).compress_algorithm) {
1493         case COMPRESS_LZO:
1494                 algtype = "lzo";
1495                 break;
1496         case COMPRESS_LZ4:
1497                 algtype = "lz4";
1498                 break;
1499         case COMPRESS_ZSTD:
1500                 algtype = "zstd";
1501                 break;
1502         case COMPRESS_LZORLE:
1503                 algtype = "lzo-rle";
1504                 break;
1505         }
1506         seq_printf(seq, ",compress_algorithm=%s", algtype);
1507
1508         seq_printf(seq, ",compress_log_size=%u",
1509                         F2FS_OPTION(sbi).compress_log_size);
1510
1511         for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1512                 seq_printf(seq, ",compress_extension=%s",
1513                         F2FS_OPTION(sbi).extensions[i]);
1514         }
1515 }
1516
1517 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1518 {
1519         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1520
1521         if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1522                 seq_printf(seq, ",background_gc=%s", "sync");
1523         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1524                 seq_printf(seq, ",background_gc=%s", "on");
1525         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1526                 seq_printf(seq, ",background_gc=%s", "off");
1527
1528         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1529                 seq_puts(seq, ",disable_roll_forward");
1530         if (test_opt(sbi, NORECOVERY))
1531                 seq_puts(seq, ",norecovery");
1532         if (test_opt(sbi, DISCARD))
1533                 seq_puts(seq, ",discard");
1534         else
1535                 seq_puts(seq, ",nodiscard");
1536         if (test_opt(sbi, NOHEAP))
1537                 seq_puts(seq, ",no_heap");
1538         else
1539                 seq_puts(seq, ",heap");
1540 #ifdef CONFIG_F2FS_FS_XATTR
1541         if (test_opt(sbi, XATTR_USER))
1542                 seq_puts(seq, ",user_xattr");
1543         else
1544                 seq_puts(seq, ",nouser_xattr");
1545         if (test_opt(sbi, INLINE_XATTR))
1546                 seq_puts(seq, ",inline_xattr");
1547         else
1548                 seq_puts(seq, ",noinline_xattr");
1549         if (test_opt(sbi, INLINE_XATTR_SIZE))
1550                 seq_printf(seq, ",inline_xattr_size=%u",
1551                                         F2FS_OPTION(sbi).inline_xattr_size);
1552 #endif
1553 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1554         if (test_opt(sbi, POSIX_ACL))
1555                 seq_puts(seq, ",acl");
1556         else
1557                 seq_puts(seq, ",noacl");
1558 #endif
1559         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1560                 seq_puts(seq, ",disable_ext_identify");
1561         if (test_opt(sbi, INLINE_DATA))
1562                 seq_puts(seq, ",inline_data");
1563         else
1564                 seq_puts(seq, ",noinline_data");
1565         if (test_opt(sbi, INLINE_DENTRY))
1566                 seq_puts(seq, ",inline_dentry");
1567         else
1568                 seq_puts(seq, ",noinline_dentry");
1569         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1570                 seq_puts(seq, ",flush_merge");
1571         if (test_opt(sbi, NOBARRIER))
1572                 seq_puts(seq, ",nobarrier");
1573         if (test_opt(sbi, FASTBOOT))
1574                 seq_puts(seq, ",fastboot");
1575         if (test_opt(sbi, EXTENT_CACHE))
1576                 seq_puts(seq, ",extent_cache");
1577         else
1578                 seq_puts(seq, ",noextent_cache");
1579         if (test_opt(sbi, DATA_FLUSH))
1580                 seq_puts(seq, ",data_flush");
1581
1582         seq_puts(seq, ",mode=");
1583         if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1584                 seq_puts(seq, "adaptive");
1585         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1586                 seq_puts(seq, "lfs");
1587         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1588         if (test_opt(sbi, RESERVE_ROOT))
1589                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1590                                 F2FS_OPTION(sbi).root_reserved_blocks,
1591                                 from_kuid_munged(&init_user_ns,
1592                                         F2FS_OPTION(sbi).s_resuid),
1593                                 from_kgid_munged(&init_user_ns,
1594                                         F2FS_OPTION(sbi).s_resgid));
1595         if (F2FS_IO_SIZE_BITS(sbi))
1596                 seq_printf(seq, ",io_bits=%u",
1597                                 F2FS_OPTION(sbi).write_io_size_bits);
1598 #ifdef CONFIG_F2FS_FAULT_INJECTION
1599         if (test_opt(sbi, FAULT_INJECTION)) {
1600                 seq_printf(seq, ",fault_injection=%u",
1601                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1602                 seq_printf(seq, ",fault_type=%u",
1603                                 F2FS_OPTION(sbi).fault_info.inject_type);
1604         }
1605 #endif
1606 #ifdef CONFIG_QUOTA
1607         if (test_opt(sbi, QUOTA))
1608                 seq_puts(seq, ",quota");
1609         if (test_opt(sbi, USRQUOTA))
1610                 seq_puts(seq, ",usrquota");
1611         if (test_opt(sbi, GRPQUOTA))
1612                 seq_puts(seq, ",grpquota");
1613         if (test_opt(sbi, PRJQUOTA))
1614                 seq_puts(seq, ",prjquota");
1615 #endif
1616         f2fs_show_quota_options(seq, sbi->sb);
1617         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1618                 seq_printf(seq, ",whint_mode=%s", "user-based");
1619         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1620                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1621
1622         fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1623
1624         if (sbi->sb->s_flags & SB_INLINECRYPT)
1625                 seq_puts(seq, ",inlinecrypt");
1626
1627         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1628                 seq_printf(seq, ",alloc_mode=%s", "default");
1629         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1630                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1631
1632         if (test_opt(sbi, DISABLE_CHECKPOINT))
1633                 seq_printf(seq, ",checkpoint=disable:%u",
1634                                 F2FS_OPTION(sbi).unusable_cap);
1635         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1636                 seq_printf(seq, ",fsync_mode=%s", "posix");
1637         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1638                 seq_printf(seq, ",fsync_mode=%s", "strict");
1639         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1640                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1641
1642 #ifdef CONFIG_F2FS_FS_COMPRESSION
1643         f2fs_show_compress_options(seq, sbi->sb);
1644 #endif
1645
1646         if (test_opt(sbi, ATGC))
1647                 seq_puts(seq, ",atgc");
1648         return 0;
1649 }
1650
1651 static void default_options(struct f2fs_sb_info *sbi)
1652 {
1653         /* init some FS parameters */
1654         F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
1655         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1656         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1657         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1658         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1659         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1660         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1661         F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1662         F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1663         F2FS_OPTION(sbi).compress_ext_cnt = 0;
1664         F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1665
1666         sbi->sb->s_flags &= ~SB_INLINECRYPT;
1667
1668         set_opt(sbi, INLINE_XATTR);
1669         set_opt(sbi, INLINE_DATA);
1670         set_opt(sbi, INLINE_DENTRY);
1671         set_opt(sbi, EXTENT_CACHE);
1672         set_opt(sbi, NOHEAP);
1673         clear_opt(sbi, DISABLE_CHECKPOINT);
1674         F2FS_OPTION(sbi).unusable_cap = 0;
1675         sbi->sb->s_flags |= SB_LAZYTIME;
1676         set_opt(sbi, FLUSH_MERGE);
1677         set_opt(sbi, DISCARD);
1678         if (f2fs_sb_has_blkzoned(sbi))
1679                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1680         else
1681                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1682
1683 #ifdef CONFIG_F2FS_FS_XATTR
1684         set_opt(sbi, XATTR_USER);
1685 #endif
1686 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1687         set_opt(sbi, POSIX_ACL);
1688 #endif
1689
1690         f2fs_build_fault_attr(sbi, 0, 0);
1691 }
1692
1693 #ifdef CONFIG_QUOTA
1694 static int f2fs_enable_quotas(struct super_block *sb);
1695 #endif
1696
1697 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1698 {
1699         unsigned int s_flags = sbi->sb->s_flags;
1700         struct cp_control cpc;
1701         int err = 0;
1702         int ret;
1703         block_t unusable;
1704
1705         if (s_flags & SB_RDONLY) {
1706                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1707                 return -EINVAL;
1708         }
1709         sbi->sb->s_flags |= SB_ACTIVE;
1710
1711         f2fs_update_time(sbi, DISABLE_TIME);
1712
1713         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1714                 down_write(&sbi->gc_lock);
1715                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1716                 if (err == -ENODATA) {
1717                         err = 0;
1718                         break;
1719                 }
1720                 if (err && err != -EAGAIN)
1721                         break;
1722         }
1723
1724         ret = sync_filesystem(sbi->sb);
1725         if (ret || err) {
1726                 err = ret ? ret: err;
1727                 goto restore_flag;
1728         }
1729
1730         unusable = f2fs_get_unusable_blocks(sbi);
1731         if (f2fs_disable_cp_again(sbi, unusable)) {
1732                 err = -EAGAIN;
1733                 goto restore_flag;
1734         }
1735
1736         down_write(&sbi->gc_lock);
1737         cpc.reason = CP_PAUSE;
1738         set_sbi_flag(sbi, SBI_CP_DISABLED);
1739         err = f2fs_write_checkpoint(sbi, &cpc);
1740         if (err)
1741                 goto out_unlock;
1742
1743         spin_lock(&sbi->stat_lock);
1744         sbi->unusable_block_count = unusable;
1745         spin_unlock(&sbi->stat_lock);
1746
1747 out_unlock:
1748         up_write(&sbi->gc_lock);
1749 restore_flag:
1750         sbi->sb->s_flags = s_flags;     /* Restore SB_RDONLY status */
1751         return err;
1752 }
1753
1754 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1755 {
1756         down_write(&sbi->gc_lock);
1757         f2fs_dirty_to_prefree(sbi);
1758
1759         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1760         set_sbi_flag(sbi, SBI_IS_DIRTY);
1761         up_write(&sbi->gc_lock);
1762
1763         f2fs_sync_fs(sbi->sb, 1);
1764 }
1765
1766 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1767 {
1768         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1769         struct f2fs_mount_info org_mount_opt;
1770         unsigned long old_sb_flags;
1771         int err;
1772         bool need_restart_gc = false;
1773         bool need_stop_gc = false;
1774         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1775         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1776         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1777         bool no_atgc = !test_opt(sbi, ATGC);
1778         bool checkpoint_changed;
1779 #ifdef CONFIG_QUOTA
1780         int i, j;
1781 #endif
1782
1783         /*
1784          * Save the old mount options in case we
1785          * need to restore them.
1786          */
1787         org_mount_opt = sbi->mount_opt;
1788         old_sb_flags = sb->s_flags;
1789
1790 #ifdef CONFIG_QUOTA
1791         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1792         for (i = 0; i < MAXQUOTAS; i++) {
1793                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1794                         org_mount_opt.s_qf_names[i] =
1795                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1796                                 GFP_KERNEL);
1797                         if (!org_mount_opt.s_qf_names[i]) {
1798                                 for (j = 0; j < i; j++)
1799                                         kfree(org_mount_opt.s_qf_names[j]);
1800                                 return -ENOMEM;
1801                         }
1802                 } else {
1803                         org_mount_opt.s_qf_names[i] = NULL;
1804                 }
1805         }
1806 #endif
1807
1808         /* recover superblocks we couldn't write due to previous RO mount */
1809         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1810                 err = f2fs_commit_super(sbi, false);
1811                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1812                           err);
1813                 if (!err)
1814                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1815         }
1816
1817         default_options(sbi);
1818
1819         /* parse mount options */
1820         err = parse_options(sb, data, true);
1821         if (err)
1822                 goto restore_opts;
1823         checkpoint_changed =
1824                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1825
1826         /*
1827          * Previous and new state of filesystem is RO,
1828          * so skip checking GC and FLUSH_MERGE conditions.
1829          */
1830         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1831                 goto skip;
1832
1833 #ifdef CONFIG_QUOTA
1834         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1835                 err = dquot_suspend(sb, -1);
1836                 if (err < 0)
1837                         goto restore_opts;
1838         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1839                 /* dquot_resume needs RW */
1840                 sb->s_flags &= ~SB_RDONLY;
1841                 if (sb_any_quota_suspended(sb)) {
1842                         dquot_resume(sb, -1);
1843                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1844                         err = f2fs_enable_quotas(sb);
1845                         if (err)
1846                                 goto restore_opts;
1847                 }
1848         }
1849 #endif
1850         /* disallow enable atgc dynamically */
1851         if (no_atgc == !!test_opt(sbi, ATGC)) {
1852                 err = -EINVAL;
1853                 f2fs_warn(sbi, "switch atgc option is not allowed");
1854                 goto restore_opts;
1855         }
1856
1857         /* disallow enable/disable extent_cache dynamically */
1858         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1859                 err = -EINVAL;
1860                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1861                 goto restore_opts;
1862         }
1863
1864         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1865                 err = -EINVAL;
1866                 f2fs_warn(sbi, "switch io_bits option is not allowed");
1867                 goto restore_opts;
1868         }
1869
1870         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1871                 err = -EINVAL;
1872                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1873                 goto restore_opts;
1874         }
1875
1876         /*
1877          * We stop the GC thread if FS is mounted as RO
1878          * or if background_gc = off is passed in mount
1879          * option. Also sync the filesystem.
1880          */
1881         if ((*flags & SB_RDONLY) ||
1882                         F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) {
1883                 if (sbi->gc_thread) {
1884                         f2fs_stop_gc_thread(sbi);
1885                         need_restart_gc = true;
1886                 }
1887         } else if (!sbi->gc_thread) {
1888                 err = f2fs_start_gc_thread(sbi);
1889                 if (err)
1890                         goto restore_opts;
1891                 need_stop_gc = true;
1892         }
1893
1894         if (*flags & SB_RDONLY ||
1895                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1896                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1897                 sync_inodes_sb(sb);
1898
1899                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1900                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1901                 f2fs_sync_fs(sb, 1);
1902                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1903         }
1904
1905         if (checkpoint_changed) {
1906                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1907                         err = f2fs_disable_checkpoint(sbi);
1908                         if (err)
1909                                 goto restore_gc;
1910                 } else {
1911                         f2fs_enable_checkpoint(sbi);
1912                 }
1913         }
1914
1915         /*
1916          * We stop issue flush thread if FS is mounted as RO
1917          * or if flush_merge is not passed in mount option.
1918          */
1919         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1920                 clear_opt(sbi, FLUSH_MERGE);
1921                 f2fs_destroy_flush_cmd_control(sbi, false);
1922         } else {
1923                 err = f2fs_create_flush_cmd_control(sbi);
1924                 if (err)
1925                         goto restore_gc;
1926         }
1927 skip:
1928 #ifdef CONFIG_QUOTA
1929         /* Release old quota file names */
1930         for (i = 0; i < MAXQUOTAS; i++)
1931                 kfree(org_mount_opt.s_qf_names[i]);
1932 #endif
1933         /* Update the POSIXACL Flag */
1934         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1935                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1936
1937         limit_reserve_root(sbi);
1938         adjust_unusable_cap_perc(sbi);
1939         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1940         return 0;
1941 restore_gc:
1942         if (need_restart_gc) {
1943                 if (f2fs_start_gc_thread(sbi))
1944                         f2fs_warn(sbi, "background gc thread has stopped");
1945         } else if (need_stop_gc) {
1946                 f2fs_stop_gc_thread(sbi);
1947         }
1948 restore_opts:
1949 #ifdef CONFIG_QUOTA
1950         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1951         for (i = 0; i < MAXQUOTAS; i++) {
1952                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1953                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1954         }
1955 #endif
1956         sbi->mount_opt = org_mount_opt;
1957         sb->s_flags = old_sb_flags;
1958         return err;
1959 }
1960
1961 #ifdef CONFIG_QUOTA
1962 /* Read data from quotafile */
1963 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1964                                size_t len, loff_t off)
1965 {
1966         struct inode *inode = sb_dqopt(sb)->files[type];
1967         struct address_space *mapping = inode->i_mapping;
1968         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1969         int offset = off & (sb->s_blocksize - 1);
1970         int tocopy;
1971         size_t toread;
1972         loff_t i_size = i_size_read(inode);
1973         struct page *page;
1974         char *kaddr;
1975
1976         if (off > i_size)
1977                 return 0;
1978
1979         if (off + len > i_size)
1980                 len = i_size - off;
1981         toread = len;
1982         while (toread > 0) {
1983                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1984 repeat:
1985                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1986                 if (IS_ERR(page)) {
1987                         if (PTR_ERR(page) == -ENOMEM) {
1988                                 congestion_wait(BLK_RW_ASYNC,
1989                                                 DEFAULT_IO_TIMEOUT);
1990                                 goto repeat;
1991                         }
1992                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1993                         return PTR_ERR(page);
1994                 }
1995
1996                 lock_page(page);
1997
1998                 if (unlikely(page->mapping != mapping)) {
1999                         f2fs_put_page(page, 1);
2000                         goto repeat;
2001                 }
2002                 if (unlikely(!PageUptodate(page))) {
2003                         f2fs_put_page(page, 1);
2004                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2005                         return -EIO;
2006                 }
2007
2008                 kaddr = kmap_atomic(page);
2009                 memcpy(data, kaddr + offset, tocopy);
2010                 kunmap_atomic(kaddr);
2011                 f2fs_put_page(page, 1);
2012
2013                 offset = 0;
2014                 toread -= tocopy;
2015                 data += tocopy;
2016                 blkidx++;
2017         }
2018         return len;
2019 }
2020
2021 /* Write to quotafile */
2022 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2023                                 const char *data, size_t len, loff_t off)
2024 {
2025         struct inode *inode = sb_dqopt(sb)->files[type];
2026         struct address_space *mapping = inode->i_mapping;
2027         const struct address_space_operations *a_ops = mapping->a_ops;
2028         int offset = off & (sb->s_blocksize - 1);
2029         size_t towrite = len;
2030         struct page *page;
2031         void *fsdata = NULL;
2032         char *kaddr;
2033         int err = 0;
2034         int tocopy;
2035
2036         while (towrite > 0) {
2037                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2038                                                                 towrite);
2039 retry:
2040                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2041                                                         &page, &fsdata);
2042                 if (unlikely(err)) {
2043                         if (err == -ENOMEM) {
2044                                 congestion_wait(BLK_RW_ASYNC,
2045                                                 DEFAULT_IO_TIMEOUT);
2046                                 goto retry;
2047                         }
2048                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2049                         break;
2050                 }
2051
2052                 kaddr = kmap_atomic(page);
2053                 memcpy(kaddr + offset, data, tocopy);
2054                 kunmap_atomic(kaddr);
2055                 flush_dcache_page(page);
2056
2057                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2058                                                 page, fsdata);
2059                 offset = 0;
2060                 towrite -= tocopy;
2061                 off += tocopy;
2062                 data += tocopy;
2063                 cond_resched();
2064         }
2065
2066         if (len == towrite)
2067                 return err;
2068         inode->i_mtime = inode->i_ctime = current_time(inode);
2069         f2fs_mark_inode_dirty_sync(inode, false);
2070         return len - towrite;
2071 }
2072
2073 static struct dquot **f2fs_get_dquots(struct inode *inode)
2074 {
2075         return F2FS_I(inode)->i_dquot;
2076 }
2077
2078 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2079 {
2080         return &F2FS_I(inode)->i_reserved_quota;
2081 }
2082
2083 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2084 {
2085         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2086                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2087                 return 0;
2088         }
2089
2090         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2091                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
2092 }
2093
2094 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2095 {
2096         int enabled = 0;
2097         int i, err;
2098
2099         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2100                 err = f2fs_enable_quotas(sbi->sb);
2101                 if (err) {
2102                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2103                         return 0;
2104                 }
2105                 return 1;
2106         }
2107
2108         for (i = 0; i < MAXQUOTAS; i++) {
2109                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2110                         err = f2fs_quota_on_mount(sbi, i);
2111                         if (!err) {
2112                                 enabled = 1;
2113                                 continue;
2114                         }
2115                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2116                                  err, i);
2117                 }
2118         }
2119         return enabled;
2120 }
2121
2122 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2123                              unsigned int flags)
2124 {
2125         struct inode *qf_inode;
2126         unsigned long qf_inum;
2127         int err;
2128
2129         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2130
2131         qf_inum = f2fs_qf_ino(sb, type);
2132         if (!qf_inum)
2133                 return -EPERM;
2134
2135         qf_inode = f2fs_iget(sb, qf_inum);
2136         if (IS_ERR(qf_inode)) {
2137                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2138                 return PTR_ERR(qf_inode);
2139         }
2140
2141         /* Don't account quota for quota files to avoid recursion */
2142         qf_inode->i_flags |= S_NOQUOTA;
2143         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2144         iput(qf_inode);
2145         return err;
2146 }
2147
2148 static int f2fs_enable_quotas(struct super_block *sb)
2149 {
2150         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2151         int type, err = 0;
2152         unsigned long qf_inum;
2153         bool quota_mopt[MAXQUOTAS] = {
2154                 test_opt(sbi, USRQUOTA),
2155                 test_opt(sbi, GRPQUOTA),
2156                 test_opt(sbi, PRJQUOTA),
2157         };
2158
2159         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2160                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2161                 return 0;
2162         }
2163
2164         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2165
2166         for (type = 0; type < MAXQUOTAS; type++) {
2167                 qf_inum = f2fs_qf_ino(sb, type);
2168                 if (qf_inum) {
2169                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2170                                 DQUOT_USAGE_ENABLED |
2171                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2172                         if (err) {
2173                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2174                                          type, err);
2175                                 for (type--; type >= 0; type--)
2176                                         dquot_quota_off(sb, type);
2177                                 set_sbi_flag(F2FS_SB(sb),
2178                                                 SBI_QUOTA_NEED_REPAIR);
2179                                 return err;
2180                         }
2181                 }
2182         }
2183         return 0;
2184 }
2185
2186 int f2fs_quota_sync(struct super_block *sb, int type)
2187 {
2188         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2189         struct quota_info *dqopt = sb_dqopt(sb);
2190         int cnt;
2191         int ret;
2192
2193         /*
2194          * do_quotactl
2195          *  f2fs_quota_sync
2196          *  down_read(quota_sem)
2197          *  dquot_writeback_dquots()
2198          *  f2fs_dquot_commit
2199          *                            block_operation
2200          *                            down_read(quota_sem)
2201          */
2202         f2fs_lock_op(sbi);
2203
2204         down_read(&sbi->quota_sem);
2205         ret = dquot_writeback_dquots(sb, type);
2206         if (ret)
2207                 goto out;
2208
2209         /*
2210          * Now when everything is written we can discard the pagecache so
2211          * that userspace sees the changes.
2212          */
2213         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2214                 struct address_space *mapping;
2215
2216                 if (type != -1 && cnt != type)
2217                         continue;
2218                 if (!sb_has_quota_active(sb, cnt))
2219                         continue;
2220
2221                 mapping = dqopt->files[cnt]->i_mapping;
2222
2223                 ret = filemap_fdatawrite(mapping);
2224                 if (ret)
2225                         goto out;
2226
2227                 /* if we are using journalled quota */
2228                 if (is_journalled_quota(sbi))
2229                         continue;
2230
2231                 ret = filemap_fdatawait(mapping);
2232                 if (ret)
2233                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2234
2235                 inode_lock(dqopt->files[cnt]);
2236                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2237                 inode_unlock(dqopt->files[cnt]);
2238         }
2239 out:
2240         if (ret)
2241                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2242         up_read(&sbi->quota_sem);
2243         f2fs_unlock_op(sbi);
2244         return ret;
2245 }
2246
2247 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2248                                                         const struct path *path)
2249 {
2250         struct inode *inode;
2251         int err;
2252
2253         /* if quota sysfile exists, deny enabling quota with specific file */
2254         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2255                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2256                 return -EBUSY;
2257         }
2258
2259         err = f2fs_quota_sync(sb, type);
2260         if (err)
2261                 return err;
2262
2263         err = dquot_quota_on(sb, type, format_id, path);
2264         if (err)
2265                 return err;
2266
2267         inode = d_inode(path->dentry);
2268
2269         inode_lock(inode);
2270         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2271         f2fs_set_inode_flags(inode);
2272         inode_unlock(inode);
2273         f2fs_mark_inode_dirty_sync(inode, false);
2274
2275         return 0;
2276 }
2277
2278 static int __f2fs_quota_off(struct super_block *sb, int type)
2279 {
2280         struct inode *inode = sb_dqopt(sb)->files[type];
2281         int err;
2282
2283         if (!inode || !igrab(inode))
2284                 return dquot_quota_off(sb, type);
2285
2286         err = f2fs_quota_sync(sb, type);
2287         if (err)
2288                 goto out_put;
2289
2290         err = dquot_quota_off(sb, type);
2291         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2292                 goto out_put;
2293
2294         inode_lock(inode);
2295         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2296         f2fs_set_inode_flags(inode);
2297         inode_unlock(inode);
2298         f2fs_mark_inode_dirty_sync(inode, false);
2299 out_put:
2300         iput(inode);
2301         return err;
2302 }
2303
2304 static int f2fs_quota_off(struct super_block *sb, int type)
2305 {
2306         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2307         int err;
2308
2309         err = __f2fs_quota_off(sb, type);
2310
2311         /*
2312          * quotactl can shutdown journalled quota, result in inconsistence
2313          * between quota record and fs data by following updates, tag the
2314          * flag to let fsck be aware of it.
2315          */
2316         if (is_journalled_quota(sbi))
2317                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2318         return err;
2319 }
2320
2321 void f2fs_quota_off_umount(struct super_block *sb)
2322 {
2323         int type;
2324         int err;
2325
2326         for (type = 0; type < MAXQUOTAS; type++) {
2327                 err = __f2fs_quota_off(sb, type);
2328                 if (err) {
2329                         int ret = dquot_quota_off(sb, type);
2330
2331                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2332                                  type, err, ret);
2333                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2334                 }
2335         }
2336         /*
2337          * In case of checkpoint=disable, we must flush quota blocks.
2338          * This can cause NULL exception for node_inode in end_io, since
2339          * put_super already dropped it.
2340          */
2341         sync_filesystem(sb);
2342 }
2343
2344 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2345 {
2346         struct quota_info *dqopt = sb_dqopt(sb);
2347         int type;
2348
2349         for (type = 0; type < MAXQUOTAS; type++) {
2350                 if (!dqopt->files[type])
2351                         continue;
2352                 f2fs_inode_synced(dqopt->files[type]);
2353         }
2354 }
2355
2356 static int f2fs_dquot_commit(struct dquot *dquot)
2357 {
2358         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2359         int ret;
2360
2361         down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2362         ret = dquot_commit(dquot);
2363         if (ret < 0)
2364                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2365         up_read(&sbi->quota_sem);
2366         return ret;
2367 }
2368
2369 static int f2fs_dquot_acquire(struct dquot *dquot)
2370 {
2371         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2372         int ret;
2373
2374         down_read(&sbi->quota_sem);
2375         ret = dquot_acquire(dquot);
2376         if (ret < 0)
2377                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2378         up_read(&sbi->quota_sem);
2379         return ret;
2380 }
2381
2382 static int f2fs_dquot_release(struct dquot *dquot)
2383 {
2384         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2385         int ret = dquot_release(dquot);
2386
2387         if (ret < 0)
2388                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2389         return ret;
2390 }
2391
2392 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2393 {
2394         struct super_block *sb = dquot->dq_sb;
2395         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2396         int ret = dquot_mark_dquot_dirty(dquot);
2397
2398         /* if we are using journalled quota */
2399         if (is_journalled_quota(sbi))
2400                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2401
2402         return ret;
2403 }
2404
2405 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2406 {
2407         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2408         int ret = dquot_commit_info(sb, type);
2409
2410         if (ret < 0)
2411                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2412         return ret;
2413 }
2414
2415 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2416 {
2417         *projid = F2FS_I(inode)->i_projid;
2418         return 0;
2419 }
2420
2421 static const struct dquot_operations f2fs_quota_operations = {
2422         .get_reserved_space = f2fs_get_reserved_space,
2423         .write_dquot    = f2fs_dquot_commit,
2424         .acquire_dquot  = f2fs_dquot_acquire,
2425         .release_dquot  = f2fs_dquot_release,
2426         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2427         .write_info     = f2fs_dquot_commit_info,
2428         .alloc_dquot    = dquot_alloc,
2429         .destroy_dquot  = dquot_destroy,
2430         .get_projid     = f2fs_get_projid,
2431         .get_next_id    = dquot_get_next_id,
2432 };
2433
2434 static const struct quotactl_ops f2fs_quotactl_ops = {
2435         .quota_on       = f2fs_quota_on,
2436         .quota_off      = f2fs_quota_off,
2437         .quota_sync     = f2fs_quota_sync,
2438         .get_state      = dquot_get_state,
2439         .set_info       = dquot_set_dqinfo,
2440         .get_dqblk      = dquot_get_dqblk,
2441         .set_dqblk      = dquot_set_dqblk,
2442         .get_nextdqblk  = dquot_get_next_dqblk,
2443 };
2444 #else
2445 int f2fs_quota_sync(struct super_block *sb, int type)
2446 {
2447         return 0;
2448 }
2449
2450 void f2fs_quota_off_umount(struct super_block *sb)
2451 {
2452 }
2453 #endif
2454
2455 static const struct super_operations f2fs_sops = {
2456         .alloc_inode    = f2fs_alloc_inode,
2457         .free_inode     = f2fs_free_inode,
2458         .drop_inode     = f2fs_drop_inode,
2459         .write_inode    = f2fs_write_inode,
2460         .dirty_inode    = f2fs_dirty_inode,
2461         .show_options   = f2fs_show_options,
2462 #ifdef CONFIG_QUOTA
2463         .quota_read     = f2fs_quota_read,
2464         .quota_write    = f2fs_quota_write,
2465         .get_dquots     = f2fs_get_dquots,
2466 #endif
2467         .evict_inode    = f2fs_evict_inode,
2468         .put_super      = f2fs_put_super,
2469         .sync_fs        = f2fs_sync_fs,
2470         .freeze_fs      = f2fs_freeze,
2471         .unfreeze_fs    = f2fs_unfreeze,
2472         .statfs         = f2fs_statfs,
2473         .remount_fs     = f2fs_remount,
2474 };
2475
2476 #ifdef CONFIG_FS_ENCRYPTION
2477 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2478 {
2479         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2480                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2481                                 ctx, len, NULL);
2482 }
2483
2484 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2485                                                         void *fs_data)
2486 {
2487         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2488
2489         /*
2490          * Encrypting the root directory is not allowed because fsck
2491          * expects lost+found directory to exist and remain unencrypted
2492          * if LOST_FOUND feature is enabled.
2493          *
2494          */
2495         if (f2fs_sb_has_lost_found(sbi) &&
2496                         inode->i_ino == F2FS_ROOT_INO(sbi))
2497                 return -EPERM;
2498
2499         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2500                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2501                                 ctx, len, fs_data, XATTR_CREATE);
2502 }
2503
2504 static const union fscrypt_context *
2505 f2fs_get_dummy_context(struct super_block *sb)
2506 {
2507         return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_ctx.ctx;
2508 }
2509
2510 static bool f2fs_has_stable_inodes(struct super_block *sb)
2511 {
2512         return true;
2513 }
2514
2515 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2516                                        int *ino_bits_ret, int *lblk_bits_ret)
2517 {
2518         *ino_bits_ret = 8 * sizeof(nid_t);
2519         *lblk_bits_ret = 8 * sizeof(block_t);
2520 }
2521
2522 static int f2fs_get_num_devices(struct super_block *sb)
2523 {
2524         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2525
2526         if (f2fs_is_multi_device(sbi))
2527                 return sbi->s_ndevs;
2528         return 1;
2529 }
2530
2531 static void f2fs_get_devices(struct super_block *sb,
2532                              struct request_queue **devs)
2533 {
2534         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2535         int i;
2536
2537         for (i = 0; i < sbi->s_ndevs; i++)
2538                 devs[i] = bdev_get_queue(FDEV(i).bdev);
2539 }
2540
2541 static const struct fscrypt_operations f2fs_cryptops = {
2542         .key_prefix             = "f2fs:",
2543         .get_context            = f2fs_get_context,
2544         .set_context            = f2fs_set_context,
2545         .get_dummy_context      = f2fs_get_dummy_context,
2546         .empty_dir              = f2fs_empty_dir,
2547         .max_namelen            = F2FS_NAME_LEN,
2548         .has_stable_inodes      = f2fs_has_stable_inodes,
2549         .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
2550         .get_num_devices        = f2fs_get_num_devices,
2551         .get_devices            = f2fs_get_devices,
2552 };
2553 #endif
2554
2555 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2556                 u64 ino, u32 generation)
2557 {
2558         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2559         struct inode *inode;
2560
2561         if (f2fs_check_nid_range(sbi, ino))
2562                 return ERR_PTR(-ESTALE);
2563
2564         /*
2565          * f2fs_iget isn't quite right if the inode is currently unallocated!
2566          * However f2fs_iget currently does appropriate checks to handle stale
2567          * inodes so everything is OK.
2568          */
2569         inode = f2fs_iget(sb, ino);
2570         if (IS_ERR(inode))
2571                 return ERR_CAST(inode);
2572         if (unlikely(generation && inode->i_generation != generation)) {
2573                 /* we didn't find the right inode.. */
2574                 iput(inode);
2575                 return ERR_PTR(-ESTALE);
2576         }
2577         return inode;
2578 }
2579
2580 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2581                 int fh_len, int fh_type)
2582 {
2583         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2584                                     f2fs_nfs_get_inode);
2585 }
2586
2587 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2588                 int fh_len, int fh_type)
2589 {
2590         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2591                                     f2fs_nfs_get_inode);
2592 }
2593
2594 static const struct export_operations f2fs_export_ops = {
2595         .fh_to_dentry = f2fs_fh_to_dentry,
2596         .fh_to_parent = f2fs_fh_to_parent,
2597         .get_parent = f2fs_get_parent,
2598 };
2599
2600 static loff_t max_file_blocks(void)
2601 {
2602         loff_t result = 0;
2603         loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2604
2605         /*
2606          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2607          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2608          * space in inode.i_addr, it will be more safe to reassign
2609          * result as zero.
2610          */
2611
2612         /* two direct node blocks */
2613         result += (leaf_count * 2);
2614
2615         /* two indirect node blocks */
2616         leaf_count *= NIDS_PER_BLOCK;
2617         result += (leaf_count * 2);
2618
2619         /* one double indirect node block */
2620         leaf_count *= NIDS_PER_BLOCK;
2621         result += leaf_count;
2622
2623         return result;
2624 }
2625
2626 static int __f2fs_commit_super(struct buffer_head *bh,
2627                         struct f2fs_super_block *super)
2628 {
2629         lock_buffer(bh);
2630         if (super)
2631                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2632         set_buffer_dirty(bh);
2633         unlock_buffer(bh);
2634
2635         /* it's rare case, we can do fua all the time */
2636         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2637 }
2638
2639 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2640                                         struct buffer_head *bh)
2641 {
2642         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2643                                         (bh->b_data + F2FS_SUPER_OFFSET);
2644         struct super_block *sb = sbi->sb;
2645         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2646         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2647         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2648         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2649         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2650         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2651         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2652         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2653         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2654         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2655         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2656         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2657         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2658         u64 main_end_blkaddr = main_blkaddr +
2659                                 (segment_count_main << log_blocks_per_seg);
2660         u64 seg_end_blkaddr = segment0_blkaddr +
2661                                 (segment_count << log_blocks_per_seg);
2662
2663         if (segment0_blkaddr != cp_blkaddr) {
2664                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2665                           segment0_blkaddr, cp_blkaddr);
2666                 return true;
2667         }
2668
2669         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2670                                                         sit_blkaddr) {
2671                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2672                           cp_blkaddr, sit_blkaddr,
2673                           segment_count_ckpt << log_blocks_per_seg);
2674                 return true;
2675         }
2676
2677         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2678                                                         nat_blkaddr) {
2679                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2680                           sit_blkaddr, nat_blkaddr,
2681                           segment_count_sit << log_blocks_per_seg);
2682                 return true;
2683         }
2684
2685         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2686                                                         ssa_blkaddr) {
2687                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2688                           nat_blkaddr, ssa_blkaddr,
2689                           segment_count_nat << log_blocks_per_seg);
2690                 return true;
2691         }
2692
2693         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2694                                                         main_blkaddr) {
2695                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2696                           ssa_blkaddr, main_blkaddr,
2697                           segment_count_ssa << log_blocks_per_seg);
2698                 return true;
2699         }
2700
2701         if (main_end_blkaddr > seg_end_blkaddr) {
2702                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2703                           main_blkaddr,
2704                           segment0_blkaddr +
2705                           (segment_count << log_blocks_per_seg),
2706                           segment_count_main << log_blocks_per_seg);
2707                 return true;
2708         } else if (main_end_blkaddr < seg_end_blkaddr) {
2709                 int err = 0;
2710                 char *res;
2711
2712                 /* fix in-memory information all the time */
2713                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2714                                 segment0_blkaddr) >> log_blocks_per_seg);
2715
2716                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2717                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2718                         res = "internally";
2719                 } else {
2720                         err = __f2fs_commit_super(bh, NULL);
2721                         res = err ? "failed" : "done";
2722                 }
2723                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2724                           res, main_blkaddr,
2725                           segment0_blkaddr +
2726                           (segment_count << log_blocks_per_seg),
2727                           segment_count_main << log_blocks_per_seg);
2728                 if (err)
2729                         return true;
2730         }
2731         return false;
2732 }
2733
2734 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2735                                 struct buffer_head *bh)
2736 {
2737         block_t segment_count, segs_per_sec, secs_per_zone;
2738         block_t total_sections, blocks_per_seg;
2739         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2740                                         (bh->b_data + F2FS_SUPER_OFFSET);
2741         unsigned int blocksize;
2742         size_t crc_offset = 0;
2743         __u32 crc = 0;
2744
2745         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2746                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2747                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2748                 return -EINVAL;
2749         }
2750
2751         /* Check checksum_offset and crc in superblock */
2752         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2753                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2754                 if (crc_offset !=
2755                         offsetof(struct f2fs_super_block, crc)) {
2756                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2757                                   crc_offset);
2758                         return -EFSCORRUPTED;
2759                 }
2760                 crc = le32_to_cpu(raw_super->crc);
2761                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2762                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2763                         return -EFSCORRUPTED;
2764                 }
2765         }
2766
2767         /* Currently, support only 4KB page cache size */
2768         if (F2FS_BLKSIZE != PAGE_SIZE) {
2769                 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2770                           PAGE_SIZE);
2771                 return -EFSCORRUPTED;
2772         }
2773
2774         /* Currently, support only 4KB block size */
2775         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2776         if (blocksize != F2FS_BLKSIZE) {
2777                 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2778                           blocksize);
2779                 return -EFSCORRUPTED;
2780         }
2781
2782         /* check log blocks per segment */
2783         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2784                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2785                           le32_to_cpu(raw_super->log_blocks_per_seg));
2786                 return -EFSCORRUPTED;
2787         }
2788
2789         /* Currently, support 512/1024/2048/4096 bytes sector size */
2790         if (le32_to_cpu(raw_super->log_sectorsize) >
2791                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2792                 le32_to_cpu(raw_super->log_sectorsize) <
2793                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2794                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2795                           le32_to_cpu(raw_super->log_sectorsize));
2796                 return -EFSCORRUPTED;
2797         }
2798         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2799                 le32_to_cpu(raw_super->log_sectorsize) !=
2800                         F2FS_MAX_LOG_SECTOR_SIZE) {
2801                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2802                           le32_to_cpu(raw_super->log_sectors_per_block),
2803                           le32_to_cpu(raw_super->log_sectorsize));
2804                 return -EFSCORRUPTED;
2805         }
2806
2807         segment_count = le32_to_cpu(raw_super->segment_count);
2808         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2809         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2810         total_sections = le32_to_cpu(raw_super->section_count);
2811
2812         /* blocks_per_seg should be 512, given the above check */
2813         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2814
2815         if (segment_count > F2FS_MAX_SEGMENT ||
2816                                 segment_count < F2FS_MIN_SEGMENTS) {
2817                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2818                 return -EFSCORRUPTED;
2819         }
2820
2821         if (total_sections > segment_count ||
2822                         total_sections < F2FS_MIN_SEGMENTS ||
2823                         segs_per_sec > segment_count || !segs_per_sec) {
2824                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2825                           segment_count, total_sections, segs_per_sec);
2826                 return -EFSCORRUPTED;
2827         }
2828
2829         if ((segment_count / segs_per_sec) < total_sections) {
2830                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2831                           segment_count, segs_per_sec, total_sections);
2832                 return -EFSCORRUPTED;
2833         }
2834
2835         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2836                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2837                           segment_count, le64_to_cpu(raw_super->block_count));
2838                 return -EFSCORRUPTED;
2839         }
2840
2841         if (RDEV(0).path[0]) {
2842                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2843                 int i = 1;
2844
2845                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2846                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2847                         i++;
2848                 }
2849                 if (segment_count != dev_seg_count) {
2850                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2851                                         segment_count, dev_seg_count);
2852                         return -EFSCORRUPTED;
2853                 }
2854         } else {
2855                 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
2856                                         !bdev_is_zoned(sbi->sb->s_bdev)) {
2857                         f2fs_info(sbi, "Zoned block device path is missing");
2858                         return -EFSCORRUPTED;
2859                 }
2860         }
2861
2862         if (secs_per_zone > total_sections || !secs_per_zone) {
2863                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2864                           secs_per_zone, total_sections);
2865                 return -EFSCORRUPTED;
2866         }
2867         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2868                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2869                         (le32_to_cpu(raw_super->extension_count) +
2870                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2871                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2872                           le32_to_cpu(raw_super->extension_count),
2873                           raw_super->hot_ext_count,
2874                           F2FS_MAX_EXTENSION);
2875                 return -EFSCORRUPTED;
2876         }
2877
2878         if (le32_to_cpu(raw_super->cp_payload) >
2879                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2880                 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2881                           le32_to_cpu(raw_super->cp_payload),
2882                           blocks_per_seg - F2FS_CP_PACKS);
2883                 return -EFSCORRUPTED;
2884         }
2885
2886         /* check reserved ino info */
2887         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2888                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2889                 le32_to_cpu(raw_super->root_ino) != 3) {
2890                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2891                           le32_to_cpu(raw_super->node_ino),
2892                           le32_to_cpu(raw_super->meta_ino),
2893                           le32_to_cpu(raw_super->root_ino));
2894                 return -EFSCORRUPTED;
2895         }
2896
2897         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2898         if (sanity_check_area_boundary(sbi, bh))
2899                 return -EFSCORRUPTED;
2900
2901         return 0;
2902 }
2903
2904 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2905 {
2906         unsigned int total, fsmeta;
2907         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2908         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2909         unsigned int ovp_segments, reserved_segments;
2910         unsigned int main_segs, blocks_per_seg;
2911         unsigned int sit_segs, nat_segs;
2912         unsigned int sit_bitmap_size, nat_bitmap_size;
2913         unsigned int log_blocks_per_seg;
2914         unsigned int segment_count_main;
2915         unsigned int cp_pack_start_sum, cp_payload;
2916         block_t user_block_count, valid_user_blocks;
2917         block_t avail_node_count, valid_node_count;
2918         int i, j;
2919
2920         total = le32_to_cpu(raw_super->segment_count);
2921         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2922         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2923         fsmeta += sit_segs;
2924         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2925         fsmeta += nat_segs;
2926         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2927         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2928
2929         if (unlikely(fsmeta >= total))
2930                 return 1;
2931
2932         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2933         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2934
2935         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2936                         ovp_segments == 0 || reserved_segments == 0)) {
2937                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2938                 return 1;
2939         }
2940
2941         user_block_count = le64_to_cpu(ckpt->user_block_count);
2942         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2943         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2944         if (!user_block_count || user_block_count >=
2945                         segment_count_main << log_blocks_per_seg) {
2946                 f2fs_err(sbi, "Wrong user_block_count: %u",
2947                          user_block_count);
2948                 return 1;
2949         }
2950
2951         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2952         if (valid_user_blocks > user_block_count) {
2953                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2954                          valid_user_blocks, user_block_count);
2955                 return 1;
2956         }
2957
2958         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2959         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2960         if (valid_node_count > avail_node_count) {
2961                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2962                          valid_node_count, avail_node_count);
2963                 return 1;
2964         }
2965
2966         main_segs = le32_to_cpu(raw_super->segment_count_main);
2967         blocks_per_seg = sbi->blocks_per_seg;
2968
2969         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2970                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2971                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2972                         return 1;
2973                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2974                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2975                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2976                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2977                                          i, j,
2978                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2979                                 return 1;
2980                         }
2981                 }
2982         }
2983         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2984                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2985                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2986                         return 1;
2987                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2988                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2989                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2990                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2991                                          i, j,
2992                                          le32_to_cpu(ckpt->cur_data_segno[i]));
2993                                 return 1;
2994                         }
2995                 }
2996         }
2997         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2998                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2999                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3000                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3001                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3002                                          i, j,
3003                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3004                                 return 1;
3005                         }
3006                 }
3007         }
3008
3009         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3010         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3011
3012         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3013                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3014                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3015                          sit_bitmap_size, nat_bitmap_size);
3016                 return 1;
3017         }
3018
3019         cp_pack_start_sum = __start_sum_addr(sbi);
3020         cp_payload = __cp_payload(sbi);
3021         if (cp_pack_start_sum < cp_payload + 1 ||
3022                 cp_pack_start_sum > blocks_per_seg - 1 -
3023                         NR_CURSEG_PERSIST_TYPE) {
3024                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3025                          cp_pack_start_sum);
3026                 return 1;
3027         }
3028
3029         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3030                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3031                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3032                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3033                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3034                           le32_to_cpu(ckpt->checksum_offset));
3035                 return 1;
3036         }
3037
3038         if (unlikely(f2fs_cp_error(sbi))) {
3039                 f2fs_err(sbi, "A bug case: need to run fsck");
3040                 return 1;
3041         }
3042         return 0;
3043 }
3044
3045 static void init_sb_info(struct f2fs_sb_info *sbi)
3046 {
3047         struct f2fs_super_block *raw_super = sbi->raw_super;
3048         int i;
3049
3050         sbi->log_sectors_per_block =
3051                 le32_to_cpu(raw_super->log_sectors_per_block);
3052         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3053         sbi->blocksize = 1 << sbi->log_blocksize;
3054         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3055         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3056         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3057         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3058         sbi->total_sections = le32_to_cpu(raw_super->section_count);
3059         sbi->total_node_count =
3060                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3061                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3062         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
3063         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
3064         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
3065         sbi->cur_victim_sec = NULL_SECNO;
3066         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3067         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3068         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3069         sbi->migration_granularity = sbi->segs_per_sec;
3070
3071         sbi->dir_level = DEF_DIR_LEVEL;
3072         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3073         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3074         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3075         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3076         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3077         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3078                                 DEF_UMOUNT_DISCARD_TIMEOUT;
3079         clear_sbi_flag(sbi, SBI_NEED_FSCK);
3080
3081         for (i = 0; i < NR_COUNT_TYPE; i++)
3082                 atomic_set(&sbi->nr_pages[i], 0);
3083
3084         for (i = 0; i < META; i++)
3085                 atomic_set(&sbi->wb_sync_req[i], 0);
3086
3087         INIT_LIST_HEAD(&sbi->s_list);
3088         mutex_init(&sbi->umount_mutex);
3089         init_rwsem(&sbi->io_order_lock);
3090         spin_lock_init(&sbi->cp_lock);
3091
3092         sbi->dirty_device = 0;
3093         spin_lock_init(&sbi->dev_lock);
3094
3095         init_rwsem(&sbi->sb_lock);
3096         init_rwsem(&sbi->pin_sem);
3097 }
3098
3099 static int init_percpu_info(struct f2fs_sb_info *sbi)
3100 {
3101         int err;
3102
3103         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3104         if (err)
3105                 return err;
3106
3107         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3108                                                                 GFP_KERNEL);
3109         if (err)
3110                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3111
3112         return err;
3113 }
3114
3115 #ifdef CONFIG_BLK_DEV_ZONED
3116
3117 struct f2fs_report_zones_args {
3118         struct f2fs_dev_info *dev;
3119         bool zone_cap_mismatch;
3120 };
3121
3122 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3123                               void *data)
3124 {
3125         struct f2fs_report_zones_args *rz_args = data;
3126
3127         if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3128                 return 0;
3129
3130         set_bit(idx, rz_args->dev->blkz_seq);
3131         rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >>
3132                                                 F2FS_LOG_SECTORS_PER_BLOCK;
3133         if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch)
3134                 rz_args->zone_cap_mismatch = true;
3135
3136         return 0;
3137 }
3138
3139 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3140 {
3141         struct block_device *bdev = FDEV(devi).bdev;
3142         sector_t nr_sectors = bdev->bd_part->nr_sects;
3143         struct f2fs_report_zones_args rep_zone_arg;
3144         int ret;
3145
3146         if (!f2fs_sb_has_blkzoned(sbi))
3147                 return 0;
3148
3149         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3150                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3151                 return -EINVAL;
3152         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3153         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3154                                 __ilog2_u32(sbi->blocks_per_blkz))
3155                 return -EINVAL;
3156         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3157         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3158                                         sbi->log_blocks_per_blkz;
3159         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3160                 FDEV(devi).nr_blkz++;
3161
3162         FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3163                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
3164                                         * sizeof(unsigned long),
3165                                         GFP_KERNEL);
3166         if (!FDEV(devi).blkz_seq)
3167                 return -ENOMEM;
3168
3169         /* Get block zones type and zone-capacity */
3170         FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi,
3171                                         FDEV(devi).nr_blkz * sizeof(block_t),
3172                                         GFP_KERNEL);
3173         if (!FDEV(devi).zone_capacity_blocks)
3174                 return -ENOMEM;
3175
3176         rep_zone_arg.dev = &FDEV(devi);
3177         rep_zone_arg.zone_cap_mismatch = false;
3178
3179         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3180                                   &rep_zone_arg);
3181         if (ret < 0)
3182                 return ret;
3183
3184         if (!rep_zone_arg.zone_cap_mismatch) {
3185                 kfree(FDEV(devi).zone_capacity_blocks);
3186                 FDEV(devi).zone_capacity_blocks = NULL;
3187         }
3188
3189         return 0;
3190 }
3191 #endif
3192
3193 /*
3194  * Read f2fs raw super block.
3195  * Because we have two copies of super block, so read both of them
3196  * to get the first valid one. If any one of them is broken, we pass
3197  * them recovery flag back to the caller.
3198  */
3199 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3200                         struct f2fs_super_block **raw_super,
3201                         int *valid_super_block, int *recovery)
3202 {
3203         struct super_block *sb = sbi->sb;
3204         int block;
3205         struct buffer_head *bh;
3206         struct f2fs_super_block *super;
3207         int err = 0;
3208
3209         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3210         if (!super)
3211                 return -ENOMEM;
3212
3213         for (block = 0; block < 2; block++) {
3214                 bh = sb_bread(sb, block);
3215                 if (!bh) {
3216                         f2fs_err(sbi, "Unable to read %dth superblock",
3217                                  block + 1);
3218                         err = -EIO;
3219                         *recovery = 1;
3220                         continue;
3221                 }
3222
3223                 /* sanity checking of raw super */
3224                 err = sanity_check_raw_super(sbi, bh);
3225                 if (err) {
3226                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3227                                  block + 1);
3228                         brelse(bh);
3229                         *recovery = 1;
3230                         continue;
3231                 }
3232
3233                 if (!*raw_super) {
3234                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3235                                                         sizeof(*super));
3236                         *valid_super_block = block;
3237                         *raw_super = super;
3238                 }
3239                 brelse(bh);
3240         }
3241
3242         /* No valid superblock */
3243         if (!*raw_super)
3244                 kfree(super);
3245         else
3246                 err = 0;
3247
3248         return err;
3249 }
3250
3251 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3252 {
3253         struct buffer_head *bh;
3254         __u32 crc = 0;
3255         int err;
3256
3257         if ((recover && f2fs_readonly(sbi->sb)) ||
3258                                 bdev_read_only(sbi->sb->s_bdev)) {
3259                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3260                 return -EROFS;
3261         }
3262
3263         /* we should update superblock crc here */
3264         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3265                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3266                                 offsetof(struct f2fs_super_block, crc));
3267                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3268         }
3269
3270         /* write back-up superblock first */
3271         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3272         if (!bh)
3273                 return -EIO;
3274         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3275         brelse(bh);
3276
3277         /* if we are in recovery path, skip writing valid superblock */
3278         if (recover || err)
3279                 return err;
3280
3281         /* write current valid superblock */
3282         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3283         if (!bh)
3284                 return -EIO;
3285         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3286         brelse(bh);
3287         return err;
3288 }
3289
3290 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3291 {
3292         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3293         unsigned int max_devices = MAX_DEVICES;
3294         int i;
3295
3296         /* Initialize single device information */
3297         if (!RDEV(0).path[0]) {
3298                 if (!bdev_is_zoned(sbi->sb->s_bdev))
3299                         return 0;
3300                 max_devices = 1;
3301         }
3302
3303         /*
3304          * Initialize multiple devices information, or single
3305          * zoned block device information.
3306          */
3307         sbi->devs = f2fs_kzalloc(sbi,
3308                                  array_size(max_devices,
3309                                             sizeof(struct f2fs_dev_info)),
3310                                  GFP_KERNEL);
3311         if (!sbi->devs)
3312                 return -ENOMEM;
3313
3314         for (i = 0; i < max_devices; i++) {
3315
3316                 if (i > 0 && !RDEV(i).path[0])
3317                         break;
3318
3319                 if (max_devices == 1) {
3320                         /* Single zoned block device mount */
3321                         FDEV(0).bdev =
3322                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3323                                         sbi->sb->s_mode, sbi->sb->s_type);
3324                 } else {
3325                         /* Multi-device mount */
3326                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3327                         FDEV(i).total_segments =
3328                                 le32_to_cpu(RDEV(i).total_segments);
3329                         if (i == 0) {
3330                                 FDEV(i).start_blk = 0;
3331                                 FDEV(i).end_blk = FDEV(i).start_blk +
3332                                     (FDEV(i).total_segments <<
3333                                     sbi->log_blocks_per_seg) - 1 +
3334                                     le32_to_cpu(raw_super->segment0_blkaddr);
3335                         } else {
3336                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3337                                 FDEV(i).end_blk = FDEV(i).start_blk +
3338                                         (FDEV(i).total_segments <<
3339                                         sbi->log_blocks_per_seg) - 1;
3340                         }
3341                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3342                                         sbi->sb->s_mode, sbi->sb->s_type);
3343                 }
3344                 if (IS_ERR(FDEV(i).bdev))
3345                         return PTR_ERR(FDEV(i).bdev);
3346
3347                 /* to release errored devices */
3348                 sbi->s_ndevs = i + 1;
3349
3350 #ifdef CONFIG_BLK_DEV_ZONED
3351                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3352                                 !f2fs_sb_has_blkzoned(sbi)) {
3353                         f2fs_err(sbi, "Zoned block device feature not enabled\n");
3354                         return -EINVAL;
3355                 }
3356                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3357                         if (init_blkz_info(sbi, i)) {
3358                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3359                                 return -EINVAL;
3360                         }
3361                         if (max_devices == 1)
3362                                 break;
3363                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3364                                   i, FDEV(i).path,
3365                                   FDEV(i).total_segments,
3366                                   FDEV(i).start_blk, FDEV(i).end_blk,
3367                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3368                                   "Host-aware" : "Host-managed");
3369                         continue;
3370                 }
3371 #endif
3372                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3373                           i, FDEV(i).path,
3374                           FDEV(i).total_segments,
3375                           FDEV(i).start_blk, FDEV(i).end_blk);
3376         }
3377         f2fs_info(sbi,
3378                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3379         return 0;
3380 }
3381
3382 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3383 {
3384 #ifdef CONFIG_UNICODE
3385         if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3386                 const struct f2fs_sb_encodings *encoding_info;
3387                 struct unicode_map *encoding;
3388                 __u16 encoding_flags;
3389
3390                 if (f2fs_sb_has_encrypt(sbi)) {
3391                         f2fs_err(sbi,
3392                                 "Can't mount with encoding and encryption");
3393                         return -EINVAL;
3394                 }
3395
3396                 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3397                                           &encoding_flags)) {
3398                         f2fs_err(sbi,
3399                                  "Encoding requested by superblock is unknown");
3400                         return -EINVAL;
3401                 }
3402
3403                 encoding = utf8_load(encoding_info->version);
3404                 if (IS_ERR(encoding)) {
3405                         f2fs_err(sbi,
3406                                  "can't mount with superblock charset: %s-%s "
3407                                  "not supported by the kernel. flags: 0x%x.",
3408                                  encoding_info->name, encoding_info->version,
3409                                  encoding_flags);
3410                         return PTR_ERR(encoding);
3411                 }
3412                 f2fs_info(sbi, "Using encoding defined by superblock: "
3413                          "%s-%s with flags 0x%hx", encoding_info->name,
3414                          encoding_info->version?:"\b", encoding_flags);
3415
3416                 sbi->sb->s_encoding = encoding;
3417                 sbi->sb->s_encoding_flags = encoding_flags;
3418                 sbi->sb->s_d_op = &f2fs_dentry_ops;
3419         }
3420 #else
3421         if (f2fs_sb_has_casefold(sbi)) {
3422                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3423                 return -EINVAL;
3424         }
3425 #endif
3426         return 0;
3427 }
3428
3429 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3430 {
3431         struct f2fs_sm_info *sm_i = SM_I(sbi);
3432
3433         /* adjust parameters according to the volume size */
3434         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3435                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3436                 sm_i->dcc_info->discard_granularity = 1;
3437                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3438         }
3439
3440         sbi->readdir_ra = 1;
3441 }
3442
3443 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3444 {
3445         struct f2fs_sb_info *sbi;
3446         struct f2fs_super_block *raw_super;
3447         struct inode *root;
3448         int err;
3449         bool skip_recovery = false, need_fsck = false;
3450         char *options = NULL;
3451         int recovery, i, valid_super_block;
3452         struct curseg_info *seg_i;
3453         int retry_cnt = 1;
3454
3455 try_onemore:
3456         err = -EINVAL;
3457         raw_super = NULL;
3458         valid_super_block = -1;
3459         recovery = 0;
3460
3461         /* allocate memory for f2fs-specific super block info */
3462         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3463         if (!sbi)
3464                 return -ENOMEM;
3465
3466         sbi->sb = sb;
3467
3468         /* Load the checksum driver */
3469         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3470         if (IS_ERR(sbi->s_chksum_driver)) {
3471                 f2fs_err(sbi, "Cannot load crc32 driver.");
3472                 err = PTR_ERR(sbi->s_chksum_driver);
3473                 sbi->s_chksum_driver = NULL;
3474                 goto free_sbi;
3475         }
3476
3477         /* set a block size */
3478         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3479                 f2fs_err(sbi, "unable to set blocksize");
3480                 goto free_sbi;
3481         }
3482
3483         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3484                                                                 &recovery);
3485         if (err)
3486                 goto free_sbi;
3487
3488         sb->s_fs_info = sbi;
3489         sbi->raw_super = raw_super;
3490
3491         /* precompute checksum seed for metadata */
3492         if (f2fs_sb_has_inode_chksum(sbi))
3493                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3494                                                 sizeof(raw_super->uuid));
3495
3496         /*
3497          * The BLKZONED feature indicates that the drive was formatted with
3498          * zone alignment optimization. This is optional for host-aware
3499          * devices, but mandatory for host-managed zoned block devices.
3500          */
3501 #ifndef CONFIG_BLK_DEV_ZONED
3502         if (f2fs_sb_has_blkzoned(sbi)) {
3503                 f2fs_err(sbi, "Zoned block device support is not enabled");
3504                 err = -EOPNOTSUPP;
3505                 goto free_sb_buf;
3506         }
3507 #endif
3508         default_options(sbi);
3509         /* parse mount options */
3510         options = kstrdup((const char *)data, GFP_KERNEL);
3511         if (data && !options) {
3512                 err = -ENOMEM;
3513                 goto free_sb_buf;
3514         }
3515
3516         err = parse_options(sb, options, false);
3517         if (err)
3518                 goto free_options;
3519
3520         sbi->max_file_blocks = max_file_blocks();
3521         sb->s_maxbytes = sbi->max_file_blocks <<
3522                                 le32_to_cpu(raw_super->log_blocksize);
3523         sb->s_max_links = F2FS_LINK_MAX;
3524
3525         err = f2fs_setup_casefold(sbi);
3526         if (err)
3527                 goto free_options;
3528
3529 #ifdef CONFIG_QUOTA
3530         sb->dq_op = &f2fs_quota_operations;
3531         sb->s_qcop = &f2fs_quotactl_ops;
3532         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3533
3534         if (f2fs_sb_has_quota_ino(sbi)) {
3535                 for (i = 0; i < MAXQUOTAS; i++) {
3536                         if (f2fs_qf_ino(sbi->sb, i))
3537                                 sbi->nquota_files++;
3538                 }
3539         }
3540 #endif
3541
3542         sb->s_op = &f2fs_sops;
3543 #ifdef CONFIG_FS_ENCRYPTION
3544         sb->s_cop = &f2fs_cryptops;
3545 #endif
3546 #ifdef CONFIG_FS_VERITY
3547         sb->s_vop = &f2fs_verityops;
3548 #endif
3549         sb->s_xattr = f2fs_xattr_handlers;
3550         sb->s_export_op = &f2fs_export_ops;
3551         sb->s_magic = F2FS_SUPER_MAGIC;
3552         sb->s_time_gran = 1;
3553         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3554                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3555         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3556         sb->s_iflags |= SB_I_CGROUPWB;
3557
3558         /* init f2fs-specific super block info */
3559         sbi->valid_super_block = valid_super_block;
3560         init_rwsem(&sbi->gc_lock);
3561         mutex_init(&sbi->writepages);
3562         mutex_init(&sbi->cp_mutex);
3563         init_rwsem(&sbi->node_write);
3564         init_rwsem(&sbi->node_change);
3565
3566         /* disallow all the data/node/meta page writes */
3567         set_sbi_flag(sbi, SBI_POR_DOING);
3568         spin_lock_init(&sbi->stat_lock);
3569
3570         /* init iostat info */
3571         spin_lock_init(&sbi->iostat_lock);
3572         sbi->iostat_enable = false;
3573         sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3574
3575         for (i = 0; i < NR_PAGE_TYPE; i++) {
3576                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3577                 int j;
3578
3579                 sbi->write_io[i] =
3580                         f2fs_kmalloc(sbi,
3581                                      array_size(n,
3582                                                 sizeof(struct f2fs_bio_info)),
3583                                      GFP_KERNEL);
3584                 if (!sbi->write_io[i]) {
3585                         err = -ENOMEM;
3586                         goto free_bio_info;
3587                 }
3588
3589                 for (j = HOT; j < n; j++) {
3590                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3591                         sbi->write_io[i][j].sbi = sbi;
3592                         sbi->write_io[i][j].bio = NULL;
3593                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3594                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3595                         INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3596                         init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3597                 }
3598         }
3599
3600         init_rwsem(&sbi->cp_rwsem);
3601         init_rwsem(&sbi->quota_sem);
3602         init_waitqueue_head(&sbi->cp_wait);
3603         init_sb_info(sbi);
3604
3605         err = init_percpu_info(sbi);
3606         if (err)
3607                 goto free_bio_info;
3608
3609         if (F2FS_IO_ALIGNED(sbi)) {
3610                 sbi->write_io_dummy =
3611                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3612                 if (!sbi->write_io_dummy) {
3613                         err = -ENOMEM;
3614                         goto free_percpu;
3615                 }
3616         }
3617
3618         /* init per sbi slab cache */
3619         err = f2fs_init_xattr_caches(sbi);
3620         if (err)
3621                 goto free_io_dummy;
3622
3623         /* get an inode for meta space */
3624         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3625         if (IS_ERR(sbi->meta_inode)) {
3626                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3627                 err = PTR_ERR(sbi->meta_inode);
3628                 goto free_xattr_cache;
3629         }
3630
3631         err = f2fs_get_valid_checkpoint(sbi);
3632         if (err) {
3633                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3634                 goto free_meta_inode;
3635         }
3636
3637         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3638                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3639         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3640                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3641                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3642         }
3643
3644         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3645                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3646
3647         /* Initialize device list */
3648         err = f2fs_scan_devices(sbi);
3649         if (err) {
3650                 f2fs_err(sbi, "Failed to find devices");
3651                 goto free_devices;
3652         }
3653
3654         err = f2fs_init_post_read_wq(sbi);
3655         if (err) {
3656                 f2fs_err(sbi, "Failed to initialize post read workqueue");
3657                 goto free_devices;
3658         }
3659
3660         sbi->total_valid_node_count =
3661                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3662         percpu_counter_set(&sbi->total_valid_inode_count,
3663                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3664         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3665         sbi->total_valid_block_count =
3666                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3667         sbi->last_valid_block_count = sbi->total_valid_block_count;
3668         sbi->reserved_blocks = 0;
3669         sbi->current_reserved_blocks = 0;
3670         limit_reserve_root(sbi);
3671         adjust_unusable_cap_perc(sbi);
3672
3673         for (i = 0; i < NR_INODE_TYPE; i++) {
3674                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3675                 spin_lock_init(&sbi->inode_lock[i]);
3676         }
3677         mutex_init(&sbi->flush_lock);
3678
3679         f2fs_init_extent_cache_info(sbi);
3680
3681         f2fs_init_ino_entry_info(sbi);
3682
3683         f2fs_init_fsync_node_info(sbi);
3684
3685         /* setup f2fs internal modules */
3686         err = f2fs_build_segment_manager(sbi);
3687         if (err) {
3688                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3689                          err);
3690                 goto free_sm;
3691         }
3692         err = f2fs_build_node_manager(sbi);
3693         if (err) {
3694                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3695                          err);
3696                 goto free_nm;
3697         }
3698
3699         /* For write statistics */
3700         if (sb->s_bdev->bd_part)
3701                 sbi->sectors_written_start =
3702                         (u64)part_stat_read(sb->s_bdev->bd_part,
3703                                             sectors[STAT_WRITE]);
3704
3705         /* Read accumulated write IO statistics if exists */
3706         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3707         if (__exist_node_summaries(sbi))
3708                 sbi->kbytes_written =
3709                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3710
3711         f2fs_build_gc_manager(sbi);
3712
3713         err = f2fs_build_stats(sbi);
3714         if (err)
3715                 goto free_nm;
3716
3717         /* get an inode for node space */
3718         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3719         if (IS_ERR(sbi->node_inode)) {
3720                 f2fs_err(sbi, "Failed to read node inode");
3721                 err = PTR_ERR(sbi->node_inode);
3722                 goto free_stats;
3723         }
3724
3725         /* read root inode and dentry */
3726         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3727         if (IS_ERR(root)) {
3728                 f2fs_err(sbi, "Failed to read root inode");
3729                 err = PTR_ERR(root);
3730                 goto free_node_inode;
3731         }
3732         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3733                         !root->i_size || !root->i_nlink) {
3734                 iput(root);
3735                 err = -EINVAL;
3736                 goto free_node_inode;
3737         }
3738
3739         sb->s_root = d_make_root(root); /* allocate root dentry */
3740         if (!sb->s_root) {
3741                 err = -ENOMEM;
3742                 goto free_node_inode;
3743         }
3744
3745         err = f2fs_register_sysfs(sbi);
3746         if (err)
3747                 goto free_root_inode;
3748
3749 #ifdef CONFIG_QUOTA
3750         /* Enable quota usage during mount */
3751         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3752                 err = f2fs_enable_quotas(sb);
3753                 if (err)
3754                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3755         }
3756 #endif
3757         /* if there are any orphan inodes, free them */
3758         err = f2fs_recover_orphan_inodes(sbi);
3759         if (err)
3760                 goto free_meta;
3761
3762         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3763                 goto reset_checkpoint;
3764
3765         /* recover fsynced data */
3766         if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3767                         !test_opt(sbi, NORECOVERY)) {
3768                 /*
3769                  * mount should be failed, when device has readonly mode, and
3770                  * previous checkpoint was not done by clean system shutdown.
3771                  */
3772                 if (f2fs_hw_is_readonly(sbi)) {
3773                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3774                                 err = -EROFS;
3775                                 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3776                                 goto free_meta;
3777                         }
3778                         f2fs_info(sbi, "write access unavailable, skipping recovery");
3779                         goto reset_checkpoint;
3780                 }
3781
3782                 if (need_fsck)
3783                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3784
3785                 if (skip_recovery)
3786                         goto reset_checkpoint;
3787
3788                 err = f2fs_recover_fsync_data(sbi, false);
3789                 if (err < 0) {
3790                         if (err != -ENOMEM)
3791                                 skip_recovery = true;
3792                         need_fsck = true;
3793                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3794                                  err);
3795                         goto free_meta;
3796                 }
3797         } else {
3798                 err = f2fs_recover_fsync_data(sbi, true);
3799
3800                 if (!f2fs_readonly(sb) && err > 0) {
3801                         err = -EINVAL;
3802                         f2fs_err(sbi, "Need to recover fsync data");
3803                         goto free_meta;
3804                 }
3805         }
3806
3807         /*
3808          * If the f2fs is not readonly and fsync data recovery succeeds,
3809          * check zoned block devices' write pointer consistency.
3810          */
3811         if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3812                 err = f2fs_check_write_pointer(sbi);
3813                 if (err)
3814                         goto free_meta;
3815         }
3816
3817 reset_checkpoint:
3818         f2fs_init_inmem_curseg(sbi);
3819
3820         /* f2fs_recover_fsync_data() cleared this already */
3821         clear_sbi_flag(sbi, SBI_POR_DOING);
3822
3823         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3824                 err = f2fs_disable_checkpoint(sbi);
3825                 if (err)
3826                         goto sync_free_meta;
3827         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3828                 f2fs_enable_checkpoint(sbi);
3829         }
3830
3831         /*
3832          * If filesystem is not mounted as read-only then
3833          * do start the gc_thread.
3834          */
3835         if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) {
3836                 /* After POR, we can run background GC thread.*/
3837                 err = f2fs_start_gc_thread(sbi);
3838                 if (err)
3839                         goto sync_free_meta;
3840         }
3841         kvfree(options);
3842
3843         /* recover broken superblock */
3844         if (recovery) {
3845                 err = f2fs_commit_super(sbi, true);
3846                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3847                           sbi->valid_super_block ? 1 : 2, err);
3848         }
3849
3850         f2fs_join_shrinker(sbi);
3851
3852         f2fs_tuning_parameters(sbi);
3853
3854         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3855                     cur_cp_version(F2FS_CKPT(sbi)));
3856         f2fs_update_time(sbi, CP_TIME);
3857         f2fs_update_time(sbi, REQ_TIME);
3858         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3859         return 0;
3860
3861 sync_free_meta:
3862         /* safe to flush all the data */
3863         sync_filesystem(sbi->sb);
3864         retry_cnt = 0;
3865
3866 free_meta:
3867 #ifdef CONFIG_QUOTA
3868         f2fs_truncate_quota_inode_pages(sb);
3869         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3870                 f2fs_quota_off_umount(sbi->sb);
3871 #endif
3872         /*
3873          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3874          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3875          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3876          * falls into an infinite loop in f2fs_sync_meta_pages().
3877          */
3878         truncate_inode_pages_final(META_MAPPING(sbi));
3879         /* evict some inodes being cached by GC */
3880         evict_inodes(sb);
3881         f2fs_unregister_sysfs(sbi);
3882 free_root_inode:
3883         dput(sb->s_root);
3884         sb->s_root = NULL;
3885 free_node_inode:
3886         f2fs_release_ino_entry(sbi, true);
3887         truncate_inode_pages_final(NODE_MAPPING(sbi));
3888         iput(sbi->node_inode);
3889         sbi->node_inode = NULL;
3890 free_stats:
3891         f2fs_destroy_stats(sbi);
3892 free_nm:
3893         f2fs_destroy_node_manager(sbi);
3894 free_sm:
3895         f2fs_destroy_segment_manager(sbi);
3896         f2fs_destroy_post_read_wq(sbi);
3897 free_devices:
3898         destroy_device_list(sbi);
3899         kvfree(sbi->ckpt);
3900 free_meta_inode:
3901         make_bad_inode(sbi->meta_inode);
3902         iput(sbi->meta_inode);
3903         sbi->meta_inode = NULL;
3904 free_xattr_cache:
3905         f2fs_destroy_xattr_caches(sbi);
3906 free_io_dummy:
3907         mempool_destroy(sbi->write_io_dummy);
3908 free_percpu:
3909         destroy_percpu_info(sbi);
3910 free_bio_info:
3911         for (i = 0; i < NR_PAGE_TYPE; i++)
3912                 kvfree(sbi->write_io[i]);
3913
3914 #ifdef CONFIG_UNICODE
3915         utf8_unload(sb->s_encoding);
3916 #endif
3917 free_options:
3918 #ifdef CONFIG_QUOTA
3919         for (i = 0; i < MAXQUOTAS; i++)
3920                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
3921 #endif
3922         fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx);
3923         kvfree(options);
3924 free_sb_buf:
3925         kfree(raw_super);
3926 free_sbi:
3927         if (sbi->s_chksum_driver)
3928                 crypto_free_shash(sbi->s_chksum_driver);
3929         kfree(sbi);
3930
3931         /* give only one another chance */
3932         if (retry_cnt > 0 && skip_recovery) {
3933                 retry_cnt--;
3934                 shrink_dcache_sb(sb);
3935                 goto try_onemore;
3936         }
3937         return err;
3938 }
3939
3940 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3941                         const char *dev_name, void *data)
3942 {
3943         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3944 }
3945
3946 static void kill_f2fs_super(struct super_block *sb)
3947 {
3948         if (sb->s_root) {
3949                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3950
3951                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3952                 f2fs_stop_gc_thread(sbi);
3953                 f2fs_stop_discard_thread(sbi);
3954
3955                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3956                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3957                         struct cp_control cpc = {
3958                                 .reason = CP_UMOUNT,
3959                         };
3960                         f2fs_write_checkpoint(sbi, &cpc);
3961                 }
3962
3963                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3964                         sb->s_flags &= ~SB_RDONLY;
3965         }
3966         kill_block_super(sb);
3967 }
3968
3969 static struct file_system_type f2fs_fs_type = {
3970         .owner          = THIS_MODULE,
3971         .name           = "f2fs",
3972         .mount          = f2fs_mount,
3973         .kill_sb        = kill_f2fs_super,
3974         .fs_flags       = FS_REQUIRES_DEV,
3975 };
3976 MODULE_ALIAS_FS("f2fs");
3977
3978 static int __init init_inodecache(void)
3979 {
3980         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3981                         sizeof(struct f2fs_inode_info), 0,
3982                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3983         if (!f2fs_inode_cachep)
3984                 return -ENOMEM;
3985         return 0;
3986 }
3987
3988 static void destroy_inodecache(void)
3989 {
3990         /*
3991          * Make sure all delayed rcu free inodes are flushed before we
3992          * destroy cache.
3993          */
3994         rcu_barrier();
3995         kmem_cache_destroy(f2fs_inode_cachep);
3996 }
3997
3998 static int __init init_f2fs_fs(void)
3999 {
4000         int err;
4001
4002         if (PAGE_SIZE != F2FS_BLKSIZE) {
4003                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4004                                 PAGE_SIZE, F2FS_BLKSIZE);
4005                 return -EINVAL;
4006         }
4007
4008         f2fs_build_trace_ios();
4009
4010         err = init_inodecache();
4011         if (err)
4012                 goto fail;
4013         err = f2fs_create_node_manager_caches();
4014         if (err)
4015                 goto free_inodecache;
4016         err = f2fs_create_segment_manager_caches();
4017         if (err)
4018                 goto free_node_manager_caches;
4019         err = f2fs_create_checkpoint_caches();
4020         if (err)
4021                 goto free_segment_manager_caches;
4022         err = f2fs_create_extent_cache();
4023         if (err)
4024                 goto free_checkpoint_caches;
4025         err = f2fs_create_garbage_collection_cache();
4026         if (err)
4027                 goto free_extent_cache;
4028         err = f2fs_init_sysfs();
4029         if (err)
4030                 goto free_garbage_collection_cache;
4031         err = register_shrinker(&f2fs_shrinker_info);
4032         if (err)
4033                 goto free_sysfs;
4034         err = register_filesystem(&f2fs_fs_type);
4035         if (err)
4036                 goto free_shrinker;
4037         f2fs_create_root_stats();
4038         err = f2fs_init_post_read_processing();
4039         if (err)
4040                 goto free_root_stats;
4041         err = f2fs_init_bio_entry_cache();
4042         if (err)
4043                 goto free_post_read;
4044         err = f2fs_init_bioset();
4045         if (err)
4046                 goto free_bio_enrty_cache;
4047         err = f2fs_init_compress_mempool();
4048         if (err)
4049                 goto free_bioset;
4050         return 0;
4051 free_bioset:
4052         f2fs_destroy_bioset();
4053 free_bio_enrty_cache:
4054         f2fs_destroy_bio_entry_cache();
4055 free_post_read:
4056         f2fs_destroy_post_read_processing();
4057 free_root_stats:
4058         f2fs_destroy_root_stats();
4059         unregister_filesystem(&f2fs_fs_type);
4060 free_shrinker:
4061         unregister_shrinker(&f2fs_shrinker_info);
4062 free_sysfs:
4063         f2fs_exit_sysfs();
4064 free_garbage_collection_cache:
4065         f2fs_destroy_garbage_collection_cache();
4066 free_extent_cache:
4067         f2fs_destroy_extent_cache();
4068 free_checkpoint_caches:
4069         f2fs_destroy_checkpoint_caches();
4070 free_segment_manager_caches:
4071         f2fs_destroy_segment_manager_caches();
4072 free_node_manager_caches:
4073         f2fs_destroy_node_manager_caches();
4074 free_inodecache:
4075         destroy_inodecache();
4076 fail:
4077         return err;
4078 }
4079
4080 static void __exit exit_f2fs_fs(void)
4081 {
4082         f2fs_destroy_compress_mempool();
4083         f2fs_destroy_bioset();
4084         f2fs_destroy_bio_entry_cache();
4085         f2fs_destroy_post_read_processing();
4086         f2fs_destroy_root_stats();
4087         unregister_filesystem(&f2fs_fs_type);
4088         unregister_shrinker(&f2fs_shrinker_info);
4089         f2fs_exit_sysfs();
4090         f2fs_destroy_garbage_collection_cache();
4091         f2fs_destroy_extent_cache();
4092         f2fs_destroy_checkpoint_caches();
4093         f2fs_destroy_segment_manager_caches();
4094         f2fs_destroy_node_manager_caches();
4095         destroy_inodecache();
4096         f2fs_destroy_trace_ios();
4097 }
4098
4099 module_init(init_f2fs_fs)
4100 module_exit(exit_f2fs_fs)
4101
4102 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4103 MODULE_DESCRIPTION("Flash Friendly File System");
4104 MODULE_LICENSE("GPL");
4105