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_compress_chksum,
 150         Opt_compress_mode,
 151         Opt_atgc,
 152         Opt_err,
 153 };
 154
 155 static match_table_t f2fs_tokens = {
 156         {Opt_gc_background, "background_gc=%s"},
 157         {Opt_disable_roll_forward, "disable_roll_forward"},
 158         {Opt_norecovery, "norecovery"},
 159         {Opt_discard, "discard"},
 160         {Opt_nodiscard, "nodiscard"},
 161         {Opt_noheap, "no_heap"},
 162         {Opt_heap, "heap"},
 163         {Opt_user_xattr, "user_xattr"},
 164         {Opt_nouser_xattr, "nouser_xattr"},
 165         {Opt_acl, "acl"},
 166         {Opt_noacl, "noacl"},
 167         {Opt_active_logs, "active_logs=%u"},
 168         {Opt_disable_ext_identify, "disable_ext_identify"},
 169         {Opt_inline_xattr, "inline_xattr"},
 170         {Opt_noinline_xattr, "noinline_xattr"},
 171         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 172         {Opt_inline_data, "inline_data"},
 173         {Opt_inline_dentry, "inline_dentry"},
 174         {Opt_noinline_dentry, "noinline_dentry"},
 175         {Opt_flush_merge, "flush_merge"},
 176         {Opt_noflush_merge, "noflush_merge"},
 177         {Opt_nobarrier, "nobarrier"},
 178         {Opt_fastboot, "fastboot"},
 179         {Opt_extent_cache, "extent_cache"},
 180         {Opt_noextent_cache, "noextent_cache"},
 181         {Opt_noinline_data, "noinline_data"},
 182         {Opt_data_flush, "data_flush"},
 183         {Opt_reserve_root, "reserve_root=%u"},
 184         {Opt_resgid, "resgid=%u"},
 185         {Opt_resuid, "resuid=%u"},
 186         {Opt_mode, "mode=%s"},
 187         {Opt_io_size_bits, "io_bits=%u"},
 188         {Opt_fault_injection, "fault_injection=%u"},
 189         {Opt_fault_type, "fault_type=%u"},
 190         {Opt_lazytime, "lazytime"},
 191         {Opt_nolazytime, "nolazytime"},
 192         {Opt_quota, "quota"},
 193         {Opt_noquota, "noquota"},
 194         {Opt_usrquota, "usrquota"},
 195         {Opt_grpquota, "grpquota"},
 196         {Opt_prjquota, "prjquota"},
 197         {Opt_usrjquota, "usrjquota=%s"},
 198         {Opt_grpjquota, "grpjquota=%s"},
 199         {Opt_prjjquota, "prjjquota=%s"},
 200         {Opt_offusrjquota, "usrjquota="},
 201         {Opt_offgrpjquota, "grpjquota="},
 202         {Opt_offprjjquota, "prjjquota="},
 203         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 204         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 205         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 206         {Opt_whint, "whint_mode=%s"},
 207         {Opt_alloc, "alloc_mode=%s"},
 208         {Opt_fsync, "fsync_mode=%s"},
 209         {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
 210         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 211         {Opt_inlinecrypt, "inlinecrypt"},
 212         {Opt_checkpoint_disable, "checkpoint=disable"},
 213         {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 214         {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 215         {Opt_checkpoint_enable, "checkpoint=enable"},
 216         {Opt_compress_algorithm, "compress_algorithm=%s"},
 217         {Opt_compress_log_size, "compress_log_size=%u"},
 218         {Opt_compress_extension, "compress_extension=%s"},
 219         {Opt_compress_chksum, "compress_chksum"},
 220         {Opt_compress_mode, "compress_mode=%s"},
 221         {Opt_atgc, "atgc"},
 222         {Opt_err, NULL},
 223 };
 224
 225 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 226 {
 227         struct va_format vaf;
 228         va_list args;
 229         int level;
 230
 231         va_start(args, fmt);
 232
 233         level = printk_get_level(fmt);
 234         vaf.fmt = printk_skip_level(fmt);
 235         vaf.va = &args;
 236         printk("%c%cF2FS-fs (%s): %pV\n",
 237                KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 238
 239         va_end(args);
 240 }
 241
 242 #ifdef CONFIG_UNICODE
 243 static const struct f2fs_sb_encodings {
 244         __u16 magic;
 245         char *name;
 246         char *version;
 247 } f2fs_sb_encoding_map[] = {
 248         {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
 249 };
 250
 251 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
 252                                  const struct f2fs_sb_encodings **encoding,
 253                                  __u16 *flags)
 254 {
 255         __u16 magic = le16_to_cpu(sb->s_encoding);
 256         int i;
 257
 258         for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
 259                 if (magic == f2fs_sb_encoding_map[i].magic)
 260                         break;
 261
 262         if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
 263                 return -EINVAL;
 264
 265         *encoding = &f2fs_sb_encoding_map[i];
 266         *flags = le16_to_cpu(sb->s_encoding_flags);
 267
 268         return 0;
 269 }
 270 #endif
 271
 272 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 273 {
 274         block_t limit = min((sbi->user_block_count >> 3),
 275                         sbi->user_block_count - sbi->reserved_blocks);
 276
 277         /* limit is 12.5% */
 278         if (test_opt(sbi, RESERVE_ROOT) &&
 279                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 280                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 281                 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 282                           F2FS_OPTION(sbi).root_reserved_blocks);
 283         }
 284         if (!test_opt(sbi, RESERVE_ROOT) &&
 285                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 286                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 287                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 288                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 289                 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 290                           from_kuid_munged(&init_user_ns,
 291                                            F2FS_OPTION(sbi).s_resuid),
 292                           from_kgid_munged(&init_user_ns,
 293                                            F2FS_OPTION(sbi).s_resgid));
 294 }
 295
 296 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
 297 {
 298         unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
 299         unsigned int avg_vblocks;
 300         unsigned int wanted_reserved_segments;
 301         block_t avail_user_block_count;
 302
 303         if (!F2FS_IO_ALIGNED(sbi))
 304                 return 0;
 305
 306         /* average valid block count in section in worst case */
 307         avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
 308
 309         /*
 310          * we need enough free space when migrating one section in worst case
 311          */
 312         wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
 313                                                 reserved_segments(sbi);
 314         wanted_reserved_segments -= reserved_segments(sbi);
 315
 316         avail_user_block_count = sbi->user_block_count -
 317                                 sbi->current_reserved_blocks -
 318                                 F2FS_OPTION(sbi).root_reserved_blocks;
 319
 320         if (wanted_reserved_segments * sbi->blocks_per_seg >
 321                                         avail_user_block_count) {
 322                 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
 323                         wanted_reserved_segments,
 324                         avail_user_block_count >> sbi->log_blocks_per_seg);
 325                 return -ENOSPC;
 326         }
 327
 328         SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
 329
 330         f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
 331                          wanted_reserved_segments);
 332
 333         return 0;
 334 }
 335
 336 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
 337 {
 338         if (!F2FS_OPTION(sbi).unusable_cap_perc)
 339                 return;
 340
 341         if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
 342                 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
 343         else
 344                 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
 345                                         F2FS_OPTION(sbi).unusable_cap_perc;
 346
 347         f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
 348                         F2FS_OPTION(sbi).unusable_cap,
 349                         F2FS_OPTION(sbi).unusable_cap_perc);
 350 }
 351
 352 static void init_once(void *foo)
 353 {
 354         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 355
 356         inode_init_once(&fi->vfs_inode);
 357 }
 358
 359 #ifdef CONFIG_QUOTA
 360 static const char * const quotatypes[] = INITQFNAMES;
 361 #define QTYPE2NAME(t) (quotatypes[t])
 362 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 363                                                         substring_t *args)
 364 {
 365         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 366         char *qname;
 367         int ret = -EINVAL;
 368
 369         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 370                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 371                 return -EINVAL;
 372         }
 373         if (f2fs_sb_has_quota_ino(sbi)) {
 374                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 375                 return 0;
 376         }
 377
 378         qname = match_strdup(args);
 379         if (!qname) {
 380                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 381                 return -ENOMEM;
 382         }
 383         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 384                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 385                         ret = 0;
 386                 else
 387                         f2fs_err(sbi, "%s quota file already specified",
 388                                  QTYPE2NAME(qtype));
 389                 goto errout;
 390         }
 391         if (strchr(qname, '/')) {
 392                 f2fs_err(sbi, "quotafile must be on filesystem root");
 393                 goto errout;
 394         }
 395         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 396         set_opt(sbi, QUOTA);
 397         return 0;
 398 errout:
 399         kfree(qname);
 400         return ret;
 401 }
 402
 403 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 404 {
 405         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 406
 407         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 408                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 409                 return -EINVAL;
 410         }
 411         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 412         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 413         return 0;
 414 }
 415
 416 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 417 {
 418         /*
 419          * We do the test below only for project quotas. 'usrquota' and
 420          * 'grpquota' mount options are allowed even without quota feature
 421          * to support legacy quotas in quota files.
 422          */
 423         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 424                 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 425                 return -1;
 426         }
 427         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 428                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 429                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 430                 if (test_opt(sbi, USRQUOTA) &&
 431                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 432                         clear_opt(sbi, USRQUOTA);
 433
 434                 if (test_opt(sbi, GRPQUOTA) &&
 435                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 436                         clear_opt(sbi, GRPQUOTA);
 437
 438                 if (test_opt(sbi, PRJQUOTA) &&
 439                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 440                         clear_opt(sbi, PRJQUOTA);
 441
 442                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 443                                 test_opt(sbi, PRJQUOTA)) {
 444                         f2fs_err(sbi, "old and new quota format mixing");
 445                         return -1;
 446                 }
 447
 448                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 449                         f2fs_err(sbi, "journaled quota format not specified");
 450                         return -1;
 451                 }
 452         }
 453
 454         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 455                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 456                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 457         }
 458         return 0;
 459 }
 460 #endif
 461
 462 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
 463                                           const char *opt,
 464                                           const substring_t *arg,
 465                                           bool is_remount)
 466 {
 467         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 468 #ifdef CONFIG_FS_ENCRYPTION
 469         int err;
 470
 471         if (!f2fs_sb_has_encrypt(sbi)) {
 472                 f2fs_err(sbi, "Encrypt feature is off");
 473                 return -EINVAL;
 474         }
 475
 476         /*
 477          * This mount option is just for testing, and it's not worthwhile to
 478          * implement the extra complexity (e.g. RCU protection) that would be
 479          * needed to allow it to be set or changed during remount.  We do allow
 480          * it to be specified during remount, but only if there is no change.
 481          */
 482         if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
 483                 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
 484                 return -EINVAL;
 485         }
 486         err = fscrypt_set_test_dummy_encryption(
 487                 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
 488         if (err) {
 489                 if (err == -EEXIST)
 490                         f2fs_warn(sbi,
 491                                   "Can't change test_dummy_encryption on remount");
 492                 else if (err == -EINVAL)
 493                         f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
 494                                   opt);
 495                 else
 496                         f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
 497                                   opt, err);
 498                 return -EINVAL;
 499         }
 500         f2fs_warn(sbi, "Test dummy encryption mode enabled");
 501 #else
 502         f2fs_warn(sbi, "Test dummy encryption mount option ignored");
 503 #endif
 504         return 0;
 505 }
 506
 507 static int parse_options(struct super_block *sb, char *options, bool is_remount)
 508 {
 509         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 510         substring_t args[MAX_OPT_ARGS];
 511 #ifdef CONFIG_F2FS_FS_COMPRESSION
 512         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 513         int ext_cnt;
 514 #endif
 515         char *p, *name;
 516         int arg = 0;
 517         kuid_t uid;
 518         kgid_t gid;
 519         int ret;
 520
 521         if (!options)
 522                 return 0;
 523
 524         while ((p = strsep(&options, ",")) != NULL) {
 525                 int token;
 526                 if (!*p)
 527                         continue;
 528                 /*
 529                  * Initialize args struct so we know whether arg was
 530                  * found; some options take optional arguments.
 531                  */
 532                 args[0].to = args[0].from = NULL;
 533                 token = match_token(p, f2fs_tokens, args);
 534
 535                 switch (token) {
 536                 case Opt_gc_background:
 537                         name = match_strdup(&args[0]);
 538
 539                         if (!name)
 540                                 return -ENOMEM;
 541                         if (!strcmp(name, "on")) {
 542                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
 543                         } else if (!strcmp(name, "off")) {
 544                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
 545                         } else if (!strcmp(name, "sync")) {
 546                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
 547                         } else {
 548                                 kfree(name);
 549                                 return -EINVAL;
 550                         }
 551                         kfree(name);
 552                         break;
 553                 case Opt_disable_roll_forward:
 554                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 555                         break;
 556                 case Opt_norecovery:
 557                         /* this option mounts f2fs with ro */
 558                         set_opt(sbi, NORECOVERY);
 559                         if (!f2fs_readonly(sb))
 560                                 return -EINVAL;
 561                         break;
 562                 case Opt_discard:
 563                         set_opt(sbi, DISCARD);
 564                         break;
 565                 case Opt_nodiscard:
 566                         if (f2fs_sb_has_blkzoned(sbi)) {
 567                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 568                                 return -EINVAL;
 569                         }
 570                         clear_opt(sbi, DISCARD);
 571                         break;
 572                 case Opt_noheap:
 573                         set_opt(sbi, NOHEAP);
 574                         break;
 575                 case Opt_heap:
 576                         clear_opt(sbi, NOHEAP);
 577                         break;
 578 #ifdef CONFIG_F2FS_FS_XATTR
 579                 case Opt_user_xattr:
 580                         set_opt(sbi, XATTR_USER);
 581                         break;
 582                 case Opt_nouser_xattr:
 583                         clear_opt(sbi, XATTR_USER);
 584                         break;
 585                 case Opt_inline_xattr:
 586                         set_opt(sbi, INLINE_XATTR);
 587                         break;
 588                 case Opt_noinline_xattr:
 589                         clear_opt(sbi, INLINE_XATTR);
 590                         break;
 591                 case Opt_inline_xattr_size:
 592                         if (args->from && match_int(args, &arg))
 593                                 return -EINVAL;
 594                         set_opt(sbi, INLINE_XATTR_SIZE);
 595                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 596                         break;
 597 #else
 598                 case Opt_user_xattr:
 599                         f2fs_info(sbi, "user_xattr options not supported");
 600                         break;
 601                 case Opt_nouser_xattr:
 602                         f2fs_info(sbi, "nouser_xattr options not supported");
 603                         break;
 604                 case Opt_inline_xattr:
 605                         f2fs_info(sbi, "inline_xattr options not supported");
 606                         break;
 607                 case Opt_noinline_xattr:
 608                         f2fs_info(sbi, "noinline_xattr options not supported");
 609                         break;
 610 #endif
 611 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 612                 case Opt_acl:
 613                         set_opt(sbi, POSIX_ACL);
 614                         break;
 615                 case Opt_noacl:
 616                         clear_opt(sbi, POSIX_ACL);
 617                         break;
 618 #else
 619                 case Opt_acl:
 620                         f2fs_info(sbi, "acl options not supported");
 621                         break;
 622                 case Opt_noacl:
 623                         f2fs_info(sbi, "noacl options not supported");
 624                         break;
 625 #endif
 626                 case Opt_active_logs:
 627                         if (args->from && match_int(args, &arg))
 628                                 return -EINVAL;
 629                         if (arg != 2 && arg != 4 &&
 630                                 arg != NR_CURSEG_PERSIST_TYPE)
 631                                 return -EINVAL;
 632                         F2FS_OPTION(sbi).active_logs = arg;
 633                         break;
 634                 case Opt_disable_ext_identify:
 635                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 636                         break;
 637                 case Opt_inline_data:
 638                         set_opt(sbi, INLINE_DATA);
 639                         break;
 640                 case Opt_inline_dentry:
 641                         set_opt(sbi, INLINE_DENTRY);
 642                         break;
 643                 case Opt_noinline_dentry:
 644                         clear_opt(sbi, INLINE_DENTRY);
 645                         break;
 646                 case Opt_flush_merge:
 647                         set_opt(sbi, FLUSH_MERGE);
 648                         break;
 649                 case Opt_noflush_merge:
 650                         clear_opt(sbi, FLUSH_MERGE);
 651                         break;
 652                 case Opt_nobarrier:
 653                         set_opt(sbi, NOBARRIER);
 654                         break;
 655                 case Opt_fastboot:
 656                         set_opt(sbi, FASTBOOT);
 657                         break;
 658                 case Opt_extent_cache:
 659                         set_opt(sbi, EXTENT_CACHE);
 660                         break;
 661                 case Opt_noextent_cache:
 662                         clear_opt(sbi, EXTENT_CACHE);
 663                         break;
 664                 case Opt_noinline_data:
 665                         clear_opt(sbi, INLINE_DATA);
 666                         break;
 667                 case Opt_data_flush:
 668                         set_opt(sbi, DATA_FLUSH);
 669                         break;
 670                 case Opt_reserve_root:
 671                         if (args->from && match_int(args, &arg))
 672                                 return -EINVAL;
 673                         if (test_opt(sbi, RESERVE_ROOT)) {
 674                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 675                                           F2FS_OPTION(sbi).root_reserved_blocks);
 676                         } else {
 677                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 678                                 set_opt(sbi, RESERVE_ROOT);
 679                         }
 680                         break;
 681                 case Opt_resuid:
 682                         if (args->from && match_int(args, &arg))
 683                                 return -EINVAL;
 684                         uid = make_kuid(current_user_ns(), arg);
 685                         if (!uid_valid(uid)) {
 686                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 687                                 return -EINVAL;
 688                         }
 689                         F2FS_OPTION(sbi).s_resuid = uid;
 690                         break;
 691                 case Opt_resgid:
 692                         if (args->from && match_int(args, &arg))
 693                                 return -EINVAL;
 694                         gid = make_kgid(current_user_ns(), arg);
 695                         if (!gid_valid(gid)) {
 696                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 697                                 return -EINVAL;
 698                         }
 699                         F2FS_OPTION(sbi).s_resgid = gid;
 700                         break;
 701                 case Opt_mode:
 702                         name = match_strdup(&args[0]);
 703
 704                         if (!name)
 705                                 return -ENOMEM;
 706                         if (!strcmp(name, "adaptive")) {
 707                                 if (f2fs_sb_has_blkzoned(sbi)) {
 708                                         f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
 709                                         kfree(name);
 710                                         return -EINVAL;
 711                                 }
 712                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
 713                         } else if (!strcmp(name, "lfs")) {
 714                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
 715                         } else {
 716                                 kfree(name);
 717                                 return -EINVAL;
 718                         }
 719                         kfree(name);
 720                         break;
 721                 case Opt_io_size_bits:
 722                         if (args->from && match_int(args, &arg))
 723                                 return -EINVAL;
 724                         if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
 725                                 f2fs_warn(sbi, "Not support %d, larger than %d",
 726                                           1 << arg, BIO_MAX_PAGES);
 727                                 return -EINVAL;
 728                         }
 729                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 730                         break;
 731 #ifdef CONFIG_F2FS_FAULT_INJECTION
 732                 case Opt_fault_injection:
 733                         if (args->from && match_int(args, &arg))
 734                                 return -EINVAL;
 735                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 736                         set_opt(sbi, FAULT_INJECTION);
 737                         break;
 738
 739                 case Opt_fault_type:
 740                         if (args->from && match_int(args, &arg))
 741                                 return -EINVAL;
 742                         f2fs_build_fault_attr(sbi, 0, arg);
 743                         set_opt(sbi, FAULT_INJECTION);
 744                         break;
 745 #else
 746                 case Opt_fault_injection:
 747                         f2fs_info(sbi, "fault_injection options not supported");
 748                         break;
 749
 750                 case Opt_fault_type:
 751                         f2fs_info(sbi, "fault_type options not supported");
 752                         break;
 753 #endif
 754                 case Opt_lazytime:
 755                         sb->s_flags |= SB_LAZYTIME;
 756                         break;
 757                 case Opt_nolazytime:
 758                         sb->s_flags &= ~SB_LAZYTIME;
 759                         break;
 760 #ifdef CONFIG_QUOTA
 761                 case Opt_quota:
 762                 case Opt_usrquota:
 763                         set_opt(sbi, USRQUOTA);
 764                         break;
 765                 case Opt_grpquota:
 766                         set_opt(sbi, GRPQUOTA);
 767                         break;
 768                 case Opt_prjquota:
 769                         set_opt(sbi, PRJQUOTA);
 770                         break;
 771                 case Opt_usrjquota:
 772                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 773                         if (ret)
 774                                 return ret;
 775                         break;
 776                 case Opt_grpjquota:
 777                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 778                         if (ret)
 779                                 return ret;
 780                         break;
 781                 case Opt_prjjquota:
 782                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 783                         if (ret)
 784                                 return ret;
 785                         break;
 786                 case Opt_offusrjquota:
 787                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 788                         if (ret)
 789                                 return ret;
 790                         break;
 791                 case Opt_offgrpjquota:
 792                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 793                         if (ret)
 794                                 return ret;
 795                         break;
 796                 case Opt_offprjjquota:
 797                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 798                         if (ret)
 799                                 return ret;
 800                         break;
 801                 case Opt_jqfmt_vfsold:
 802                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 803                         break;
 804                 case Opt_jqfmt_vfsv0:
 805                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 806                         break;
 807                 case Opt_jqfmt_vfsv1:
 808                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 809                         break;
 810                 case Opt_noquota:
 811                         clear_opt(sbi, QUOTA);
 812                         clear_opt(sbi, USRQUOTA);
 813                         clear_opt(sbi, GRPQUOTA);
 814                         clear_opt(sbi, PRJQUOTA);
 815                         break;
 816 #else
 817                 case Opt_quota:
 818                 case Opt_usrquota:
 819                 case Opt_grpquota:
 820                 case Opt_prjquota:
 821                 case Opt_usrjquota:
 822                 case Opt_grpjquota:
 823                 case Opt_prjjquota:
 824                 case Opt_offusrjquota:
 825                 case Opt_offgrpjquota:
 826                 case Opt_offprjjquota:
 827                 case Opt_jqfmt_vfsold:
 828                 case Opt_jqfmt_vfsv0:
 829                 case Opt_jqfmt_vfsv1:
 830                 case Opt_noquota:
 831                         f2fs_info(sbi, "quota operations not supported");
 832                         break;
 833 #endif
 834                 case Opt_whint:
 835                         name = match_strdup(&args[0]);
 836                         if (!name)
 837                                 return -ENOMEM;
 838                         if (!strcmp(name, "user-based")) {
 839                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 840                         } else if (!strcmp(name, "off")) {
 841                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 842                         } else if (!strcmp(name, "fs-based")) {
 843                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 844                         } else {
 845                                 kfree(name);
 846                                 return -EINVAL;
 847                         }
 848                         kfree(name);
 849                         break;
 850                 case Opt_alloc:
 851                         name = match_strdup(&args[0]);
 852                         if (!name)
 853                                 return -ENOMEM;
 854
 855                         if (!strcmp(name, "default")) {
 856                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 857                         } else if (!strcmp(name, "reuse")) {
 858                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 859                         } else {
 860                                 kfree(name);
 861                                 return -EINVAL;
 862                         }
 863                         kfree(name);
 864                         break;
 865                 case Opt_fsync:
 866                         name = match_strdup(&args[0]);
 867                         if (!name)
 868                                 return -ENOMEM;
 869                         if (!strcmp(name, "posix")) {
 870                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 871                         } else if (!strcmp(name, "strict")) {
 872                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 873                         } else if (!strcmp(name, "nobarrier")) {
 874                                 F2FS_OPTION(sbi).fsync_mode =
 875                                                         FSYNC_MODE_NOBARRIER;
 876                         } else {
 877                                 kfree(name);
 878                                 return -EINVAL;
 879                         }
 880                         kfree(name);
 881                         break;
 882                 case Opt_test_dummy_encryption:
 883                         ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
 884                                                              is_remount);
 885                         if (ret)
 886                                 return ret;
 887                         break;
 888                 case Opt_inlinecrypt:
 889 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
 890                         sb->s_flags |= SB_INLINECRYPT;
 891 #else
 892                         f2fs_info(sbi, "inline encryption not supported");
 893 #endif
 894                         break;
 895                 case Opt_checkpoint_disable_cap_perc:
 896                         if (args->from && match_int(args, &arg))
 897                                 return -EINVAL;
 898                         if (arg < 0 || arg > 100)
 899                                 return -EINVAL;
 900                         F2FS_OPTION(sbi).unusable_cap_perc = arg;
 901                         set_opt(sbi, DISABLE_CHECKPOINT);
 902                         break;
 903                 case Opt_checkpoint_disable_cap:
 904                         if (args->from && match_int(args, &arg))
 905                                 return -EINVAL;
 906                         F2FS_OPTION(sbi).unusable_cap = arg;
 907                         set_opt(sbi, DISABLE_CHECKPOINT);
 908                         break;
 909                 case Opt_checkpoint_disable:
 910                         set_opt(sbi, DISABLE_CHECKPOINT);
 911                         break;
 912                 case Opt_checkpoint_enable:
 913                         clear_opt(sbi, DISABLE_CHECKPOINT);
 914                         break;
 915 #ifdef CONFIG_F2FS_FS_COMPRESSION
 916                 case Opt_compress_algorithm:
 917                         if (!f2fs_sb_has_compression(sbi)) {
 918                                 f2fs_info(sbi, "Image doesn't support compression");
 919                                 break;
 920                         }
 921                         name = match_strdup(&args[0]);
 922                         if (!name)
 923                                 return -ENOMEM;
 924                         if (!strcmp(name, "lzo")) {
 925                                 F2FS_OPTION(sbi).compress_algorithm =
 926                                                                 COMPRESS_LZO;
 927                         } else if (!strcmp(name, "lz4")) {
 928                                 F2FS_OPTION(sbi).compress_algorithm =
 929                                                                 COMPRESS_LZ4;
 930                         } else if (!strcmp(name, "zstd")) {
 931                                 F2FS_OPTION(sbi).compress_algorithm =
 932                                                                 COMPRESS_ZSTD;
 933                         } else if (!strcmp(name, "lzo-rle")) {
 934                                 F2FS_OPTION(sbi).compress_algorithm =
 935                                                                 COMPRESS_LZORLE;
 936                         } else {
 937                                 kfree(name);
 938                                 return -EINVAL;
 939                         }
 940                         kfree(name);
 941                         break;
 942                 case Opt_compress_log_size:
 943                         if (!f2fs_sb_has_compression(sbi)) {
 944                                 f2fs_info(sbi, "Image doesn't support compression");
 945                                 break;
 946                         }
 947                         if (args->from && match_int(args, &arg))
 948                                 return -EINVAL;
 949                         if (arg < MIN_COMPRESS_LOG_SIZE ||
 950                                 arg > MAX_COMPRESS_LOG_SIZE) {
 951                                 f2fs_err(sbi,
 952                                         "Compress cluster log size is out of range");
 953                                 return -EINVAL;
 954                         }
 955                         F2FS_OPTION(sbi).compress_log_size = arg;
 956                         break;
 957                 case Opt_compress_extension:
 958                         if (!f2fs_sb_has_compression(sbi)) {
 959                                 f2fs_info(sbi, "Image doesn't support compression");
 960                                 break;
 961                         }
 962                         name = match_strdup(&args[0]);
 963                         if (!name)
 964                                 return -ENOMEM;
 965
 966                         ext = F2FS_OPTION(sbi).extensions;
 967                         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
 968
 969                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
 970                                 ext_cnt >= COMPRESS_EXT_NUM) {
 971                                 f2fs_err(sbi,
 972                                         "invalid extension length/number");
 973                                 kfree(name);
 974                                 return -EINVAL;
 975                         }
 976
 977                         strcpy(ext[ext_cnt], name);
 978                         F2FS_OPTION(sbi).compress_ext_cnt++;
 979                         kfree(name);
 980                         break;
 981                 case Opt_compress_chksum:
 982                         F2FS_OPTION(sbi).compress_chksum = true;
 983                         break;
 984                 case Opt_compress_mode:
 985                         name = match_strdup(&args[0]);
 986                         if (!name)
 987                                 return -ENOMEM;
 988                         if (!strcmp(name, "fs")) {
 989                                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
 990                         } else if (!strcmp(name, "user")) {
 991                                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
 992                         } else {
 993                                 kfree(name);
 994                                 return -EINVAL;
 995                         }
 996                         kfree(name);
 997                         break;
 998 #else
 999                 case Opt_compress_algorithm:
1000                 case Opt_compress_log_size:
1001                 case Opt_compress_extension:
1002                 case Opt_compress_chksum:
1003                 case Opt_compress_mode:
1004                         f2fs_info(sbi, "compression options not supported");
1005                         break;
1006 #endif
1007                 case Opt_atgc:
1008                         set_opt(sbi, ATGC);
1009                         break;
1010                 default:
1011                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1012                                  p);
1013                         return -EINVAL;
1014                 }
1015         }
1016 #ifdef CONFIG_QUOTA
1017         if (f2fs_check_quota_options(sbi))
1018                 return -EINVAL;
1019 #else
1020         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1021                 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1022                 return -EINVAL;
1023         }
1024         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1025                 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1026                 return -EINVAL;
1027         }
1028 #endif
1029 #ifndef CONFIG_UNICODE
1030         if (f2fs_sb_has_casefold(sbi)) {
1031                 f2fs_err(sbi,
1032                         "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1033                 return -EINVAL;
1034         }
1035 #endif
1036         /*
1037          * The BLKZONED feature indicates that the drive was formatted with
1038          * zone alignment optimization. This is optional for host-aware
1039          * devices, but mandatory for host-managed zoned block devices.
1040          */
1041 #ifndef CONFIG_BLK_DEV_ZONED
1042         if (f2fs_sb_has_blkzoned(sbi)) {
1043                 f2fs_err(sbi, "Zoned block device support is not enabled");
1044                 return -EINVAL;
1045         }
1046 #endif
1047
1048         if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1049                 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1050                          F2FS_IO_SIZE_KB(sbi));
1051                 return -EINVAL;
1052         }
1053
1054         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1055                 int min_size, max_size;
1056
1057                 if (!f2fs_sb_has_extra_attr(sbi) ||
1058                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1059                         f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1060                         return -EINVAL;
1061                 }
1062                 if (!test_opt(sbi, INLINE_XATTR)) {
1063                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1064                         return -EINVAL;
1065                 }
1066
1067                 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1068                 max_size = MAX_INLINE_XATTR_SIZE;
1069
1070                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1071                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1072                         f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1073                                  min_size, max_size);
1074                         return -EINVAL;
1075                 }
1076         }
1077
1078         if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1079                 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
1080                 return -EINVAL;
1081         }
1082
1083         /* Not pass down write hints if the number of active logs is lesser
1084          * than NR_CURSEG_PERSIST_TYPE.
1085          */
1086         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE)
1087                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1088         return 0;
1089 }
1090
1091 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1092 {
1093         struct f2fs_inode_info *fi;
1094
1095         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
1096         if (!fi)
1097                 return NULL;
1098
1099         init_once((void *) fi);
1100
1101         /* Initialize f2fs-specific inode info */
1102         atomic_set(&fi->dirty_pages, 0);
1103         atomic_set(&fi->i_compr_blocks, 0);
1104         init_rwsem(&fi->i_sem);
1105         spin_lock_init(&fi->i_size_lock);
1106         INIT_LIST_HEAD(&fi->dirty_list);
1107         INIT_LIST_HEAD(&fi->gdirty_list);
1108         INIT_LIST_HEAD(&fi->inmem_ilist);
1109         INIT_LIST_HEAD(&fi->inmem_pages);
1110         mutex_init(&fi->inmem_lock);
1111         init_rwsem(&fi->i_gc_rwsem[READ]);
1112         init_rwsem(&fi->i_gc_rwsem[WRITE]);
1113         init_rwsem(&fi->i_mmap_sem);
1114         init_rwsem(&fi->i_xattr_sem);
1115
1116         /* Will be used by directory only */
1117         fi->i_dir_level = F2FS_SB(sb)->dir_level;
1118
1119         fi->ra_offset = -1;
1120
1121         return &fi->vfs_inode;
1122 }
1123
1124 static int f2fs_drop_inode(struct inode *inode)
1125 {
1126         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1127         int ret;
1128
1129         /*
1130          * during filesystem shutdown, if checkpoint is disabled,
1131          * drop useless meta/node dirty pages.
1132          */
1133         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1134                 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1135                         inode->i_ino == F2FS_META_INO(sbi)) {
1136                         trace_f2fs_drop_inode(inode, 1);
1137                         return 1;
1138                 }
1139         }
1140
1141         /*
1142          * This is to avoid a deadlock condition like below.
1143          * writeback_single_inode(inode)
1144          *  - f2fs_write_data_page
1145          *    - f2fs_gc -> iput -> evict
1146          *       - inode_wait_for_writeback(inode)
1147          */
1148         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1149                 if (!inode->i_nlink && !is_bad_inode(inode)) {
1150                         /* to avoid evict_inode call simultaneously */
1151                         atomic_inc(&inode->i_count);
1152                         spin_unlock(&inode->i_lock);
1153
1154                         /* some remained atomic pages should discarded */
1155                         if (f2fs_is_atomic_file(inode))
1156                                 f2fs_drop_inmem_pages(inode);
1157
1158                         /* should remain fi->extent_tree for writepage */
1159                         f2fs_destroy_extent_node(inode);
1160
1161                         sb_start_intwrite(inode->i_sb);
1162                         f2fs_i_size_write(inode, 0);
1163
1164                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1165                                         inode, NULL, 0, DATA);
1166                         truncate_inode_pages_final(inode->i_mapping);
1167
1168                         if (F2FS_HAS_BLOCKS(inode))
1169                                 f2fs_truncate(inode);
1170
1171                         sb_end_intwrite(inode->i_sb);
1172
1173                         spin_lock(&inode->i_lock);
1174                         atomic_dec(&inode->i_count);
1175                 }
1176                 trace_f2fs_drop_inode(inode, 0);
1177                 return 0;
1178         }
1179         ret = generic_drop_inode(inode);
1180         if (!ret)
1181                 ret = fscrypt_drop_inode(inode);
1182         trace_f2fs_drop_inode(inode, ret);
1183         return ret;
1184 }
1185
1186 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1187 {
1188         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1189         int ret = 0;
1190
1191         spin_lock(&sbi->inode_lock[DIRTY_META]);
1192         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1193                 ret = 1;
1194         } else {
1195                 set_inode_flag(inode, FI_DIRTY_INODE);
1196                 stat_inc_dirty_inode(sbi, DIRTY_META);
1197         }
1198         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1199                 list_add_tail(&F2FS_I(inode)->gdirty_list,
1200                                 &sbi->inode_list[DIRTY_META]);
1201                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1202         }
1203         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1204         return ret;
1205 }
1206
1207 void f2fs_inode_synced(struct inode *inode)
1208 {
1209         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1210
1211         spin_lock(&sbi->inode_lock[DIRTY_META]);
1212         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1213                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1214                 return;
1215         }
1216         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1217                 list_del_init(&F2FS_I(inode)->gdirty_list);
1218                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1219         }
1220         clear_inode_flag(inode, FI_DIRTY_INODE);
1221         clear_inode_flag(inode, FI_AUTO_RECOVER);
1222         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1223         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1224 }
1225
1226 /*
1227  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1228  *
1229  * We should call set_dirty_inode to write the dirty inode through write_inode.
1230  */
1231 static void f2fs_dirty_inode(struct inode *inode, int flags)
1232 {
1233         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1234
1235         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1236                         inode->i_ino == F2FS_META_INO(sbi))
1237                 return;
1238
1239         if (flags == I_DIRTY_TIME)
1240                 return;
1241
1242         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1243                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1244
1245         f2fs_inode_dirtied(inode, false);
1246 }
1247
1248 static void f2fs_free_inode(struct inode *inode)
1249 {
1250         fscrypt_free_inode(inode);
1251         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1252 }
1253
1254 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1255 {
1256         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1257         percpu_counter_destroy(&sbi->total_valid_inode_count);
1258 }
1259
1260 static void destroy_device_list(struct f2fs_sb_info *sbi)
1261 {
1262         int i;
1263
1264         for (i = 0; i < sbi->s_ndevs; i++) {
1265                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1266 #ifdef CONFIG_BLK_DEV_ZONED
1267                 kvfree(FDEV(i).blkz_seq);
1268 #endif
1269         }
1270         kvfree(sbi->devs);
1271 }
1272
1273 static void f2fs_put_super(struct super_block *sb)
1274 {
1275         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1276         int i;
1277         bool dropped;
1278
1279         /* unregister procfs/sysfs entries in advance to avoid race case */
1280         f2fs_unregister_sysfs(sbi);
1281
1282         f2fs_quota_off_umount(sb);
1283
1284         /* prevent remaining shrinker jobs */
1285         mutex_lock(&sbi->umount_mutex);
1286
1287         /*
1288          * We don't need to do checkpoint when superblock is clean.
1289          * But, the previous checkpoint was not done by umount, it needs to do
1290          * clean checkpoint again.
1291          */
1292         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1293                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1294                 struct cp_control cpc = {
1295                         .reason = CP_UMOUNT,
1296                 };
1297                 f2fs_write_checkpoint(sbi, &cpc);
1298         }
1299
1300         /* be sure to wait for any on-going discard commands */
1301         dropped = f2fs_issue_discard_timeout(sbi);
1302
1303         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1304                                         !sbi->discard_blks && !dropped) {
1305                 struct cp_control cpc = {
1306                         .reason = CP_UMOUNT | CP_TRIMMED,
1307                 };
1308                 f2fs_write_checkpoint(sbi, &cpc);
1309         }
1310
1311         /*
1312          * normally superblock is clean, so we need to release this.
1313          * In addition, EIO will skip do checkpoint, we need this as well.
1314          */
1315         f2fs_release_ino_entry(sbi, true);
1316
1317         f2fs_leave_shrinker(sbi);
1318         mutex_unlock(&sbi->umount_mutex);
1319
1320         /* our cp_error case, we can wait for any writeback page */
1321         f2fs_flush_merged_writes(sbi);
1322
1323         f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1324
1325         f2fs_bug_on(sbi, sbi->fsync_node_num);
1326
1327         iput(sbi->node_inode);
1328         sbi->node_inode = NULL;
1329
1330         iput(sbi->meta_inode);
1331         sbi->meta_inode = NULL;
1332
1333         /*
1334          * iput() can update stat information, if f2fs_write_checkpoint()
1335          * above failed with error.
1336          */
1337         f2fs_destroy_stats(sbi);
1338
1339         /* destroy f2fs internal modules */
1340         f2fs_destroy_node_manager(sbi);
1341         f2fs_destroy_segment_manager(sbi);
1342
1343         f2fs_destroy_post_read_wq(sbi);
1344
1345         kvfree(sbi->ckpt);
1346
1347         sb->s_fs_info = NULL;
1348         if (sbi->s_chksum_driver)
1349                 crypto_free_shash(sbi->s_chksum_driver);
1350         kfree(sbi->raw_super);
1351
1352         destroy_device_list(sbi);
1353         f2fs_destroy_page_array_cache(sbi);
1354         f2fs_destroy_xattr_caches(sbi);
1355         mempool_destroy(sbi->write_io_dummy);
1356 #ifdef CONFIG_QUOTA
1357         for (i = 0; i < MAXQUOTAS; i++)
1358                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1359 #endif
1360         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1361         destroy_percpu_info(sbi);
1362         for (i = 0; i < NR_PAGE_TYPE; i++)
1363                 kvfree(sbi->write_io[i]);
1364 #ifdef CONFIG_UNICODE
1365         utf8_unload(sb->s_encoding);
1366 #endif
1367         kfree(sbi);
1368 }
1369
1370 int f2fs_sync_fs(struct super_block *sb, int sync)
1371 {
1372         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1373         int err = 0;
1374
1375         if (unlikely(f2fs_cp_error(sbi)))
1376                 return 0;
1377         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1378                 return 0;
1379
1380         trace_f2fs_sync_fs(sb, sync);
1381
1382         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1383                 return -EAGAIN;
1384
1385         if (sync) {
1386                 struct cp_control cpc;
1387
1388                 cpc.reason = __get_cp_reason(sbi);
1389
1390                 down_write(&sbi->gc_lock);
1391                 err = f2fs_write_checkpoint(sbi, &cpc);
1392                 up_write(&sbi->gc_lock);
1393         }
1394         f2fs_trace_ios(NULL, 1);
1395
1396         return err;
1397 }
1398
1399 static int f2fs_freeze(struct super_block *sb)
1400 {
1401         if (f2fs_readonly(sb))
1402                 return 0;
1403
1404         /* IO error happened before */
1405         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1406                 return -EIO;
1407
1408         /* must be clean, since sync_filesystem() was already called */
1409         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1410                 return -EINVAL;
1411         return 0;
1412 }
1413
1414 static int f2fs_unfreeze(struct super_block *sb)
1415 {
1416         return 0;
1417 }
1418
1419 #ifdef CONFIG_QUOTA
1420 static int f2fs_statfs_project(struct super_block *sb,
1421                                 kprojid_t projid, struct kstatfs *buf)
1422 {
1423         struct kqid qid;
1424         struct dquot *dquot;
1425         u64 limit;
1426         u64 curblock;
1427
1428         qid = make_kqid_projid(projid);
1429         dquot = dqget(sb, qid);
1430         if (IS_ERR(dquot))
1431                 return PTR_ERR(dquot);
1432         spin_lock(&dquot->dq_dqb_lock);
1433
1434         limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1435                                         dquot->dq_dqb.dqb_bhardlimit);
1436         if (limit)
1437                 limit >>= sb->s_blocksize_bits;
1438
1439         if (limit && buf->f_blocks > limit) {
1440                 curblock = (dquot->dq_dqb.dqb_curspace +
1441                             dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1442                 buf->f_blocks = limit;
1443                 buf->f_bfree = buf->f_bavail =
1444                         (buf->f_blocks > curblock) ?
1445                          (buf->f_blocks - curblock) : 0;
1446         }
1447
1448         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1449                                         dquot->dq_dqb.dqb_ihardlimit);
1450
1451         if (limit && buf->f_files > limit) {
1452                 buf->f_files = limit;
1453                 buf->f_ffree =
1454                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1455                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1456         }
1457
1458         spin_unlock(&dquot->dq_dqb_lock);
1459         dqput(dquot);
1460         return 0;
1461 }
1462 #endif
1463
1464 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1465 {
1466         struct super_block *sb = dentry->d_sb;
1467         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1468         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1469         block_t total_count, user_block_count, start_count;
1470         u64 avail_node_count;
1471
1472         total_count = le64_to_cpu(sbi->raw_super->block_count);
1473         user_block_count = sbi->user_block_count;
1474         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1475         buf->f_type = F2FS_SUPER_MAGIC;
1476         buf->f_bsize = sbi->blocksize;
1477
1478         buf->f_blocks = total_count - start_count;
1479         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1480                                                 sbi->current_reserved_blocks;
1481
1482         spin_lock(&sbi->stat_lock);
1483         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1484                 buf->f_bfree = 0;
1485         else
1486                 buf->f_bfree -= sbi->unusable_block_count;
1487         spin_unlock(&sbi->stat_lock);
1488
1489         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1490                 buf->f_bavail = buf->f_bfree -
1491                                 F2FS_OPTION(sbi).root_reserved_blocks;
1492         else
1493                 buf->f_bavail = 0;
1494
1495         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1496
1497         if (avail_node_count > user_block_count) {
1498                 buf->f_files = user_block_count;
1499                 buf->f_ffree = buf->f_bavail;
1500         } else {
1501                 buf->f_files = avail_node_count;
1502                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1503                                         buf->f_bavail);
1504         }
1505
1506         buf->f_namelen = F2FS_NAME_LEN;
1507         buf->f_fsid    = u64_to_fsid(id);
1508
1509 #ifdef CONFIG_QUOTA
1510         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1511                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1512                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1513         }
1514 #endif
1515         return 0;
1516 }
1517
1518 static inline void f2fs_show_quota_options(struct seq_file *seq,
1519                                            struct super_block *sb)
1520 {
1521 #ifdef CONFIG_QUOTA
1522         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1523
1524         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1525                 char *fmtname = "";
1526
1527                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1528                 case QFMT_VFS_OLD:
1529                         fmtname = "vfsold";
1530                         break;
1531                 case QFMT_VFS_V0:
1532                         fmtname = "vfsv0";
1533                         break;
1534                 case QFMT_VFS_V1:
1535                         fmtname = "vfsv1";
1536                         break;
1537                 }
1538                 seq_printf(seq, ",jqfmt=%s", fmtname);
1539         }
1540
1541         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1542                 seq_show_option(seq, "usrjquota",
1543                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1544
1545         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1546                 seq_show_option(seq, "grpjquota",
1547                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1548
1549         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1550                 seq_show_option(seq, "prjjquota",
1551                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1552 #endif
1553 }
1554
1555 static inline void f2fs_show_compress_options(struct seq_file *seq,
1556                                                         struct super_block *sb)
1557 {
1558         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1559         char *algtype = "";
1560         int i;
1561
1562         if (!f2fs_sb_has_compression(sbi))
1563                 return;
1564
1565         switch (F2FS_OPTION(sbi).compress_algorithm) {
1566         case COMPRESS_LZO:
1567                 algtype = "lzo";
1568                 break;
1569         case COMPRESS_LZ4:
1570                 algtype = "lz4";
1571                 break;
1572         case COMPRESS_ZSTD:
1573                 algtype = "zstd";
1574                 break;
1575         case COMPRESS_LZORLE:
1576                 algtype = "lzo-rle";
1577                 break;
1578         }
1579         seq_printf(seq, ",compress_algorithm=%s", algtype);
1580
1581         seq_printf(seq, ",compress_log_size=%u",
1582                         F2FS_OPTION(sbi).compress_log_size);
1583
1584         for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1585                 seq_printf(seq, ",compress_extension=%s",
1586                         F2FS_OPTION(sbi).extensions[i]);
1587         }
1588
1589         if (F2FS_OPTION(sbi).compress_chksum)
1590                 seq_puts(seq, ",compress_chksum");
1591
1592         if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1593                 seq_printf(seq, ",compress_mode=%s", "fs");
1594         else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1595                 seq_printf(seq, ",compress_mode=%s", "user");
1596 }
1597
1598 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1599 {
1600         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1601
1602         if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1603                 seq_printf(seq, ",background_gc=%s", "sync");
1604         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1605                 seq_printf(seq, ",background_gc=%s", "on");
1606         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1607                 seq_printf(seq, ",background_gc=%s", "off");
1608
1609         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1610                 seq_puts(seq, ",disable_roll_forward");
1611         if (test_opt(sbi, NORECOVERY))
1612                 seq_puts(seq, ",norecovery");
1613         if (test_opt(sbi, DISCARD))
1614                 seq_puts(seq, ",discard");
1615         else
1616                 seq_puts(seq, ",nodiscard");
1617         if (test_opt(sbi, NOHEAP))
1618                 seq_puts(seq, ",no_heap");
1619         else
1620                 seq_puts(seq, ",heap");
1621 #ifdef CONFIG_F2FS_FS_XATTR
1622         if (test_opt(sbi, XATTR_USER))
1623                 seq_puts(seq, ",user_xattr");
1624         else
1625                 seq_puts(seq, ",nouser_xattr");
1626         if (test_opt(sbi, INLINE_XATTR))
1627                 seq_puts(seq, ",inline_xattr");
1628         else
1629                 seq_puts(seq, ",noinline_xattr");
1630         if (test_opt(sbi, INLINE_XATTR_SIZE))
1631                 seq_printf(seq, ",inline_xattr_size=%u",
1632                                         F2FS_OPTION(sbi).inline_xattr_size);
1633 #endif
1634 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1635         if (test_opt(sbi, POSIX_ACL))
1636                 seq_puts(seq, ",acl");
1637         else
1638                 seq_puts(seq, ",noacl");
1639 #endif
1640         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1641                 seq_puts(seq, ",disable_ext_identify");
1642         if (test_opt(sbi, INLINE_DATA))
1643                 seq_puts(seq, ",inline_data");
1644         else
1645                 seq_puts(seq, ",noinline_data");
1646         if (test_opt(sbi, INLINE_DENTRY))
1647                 seq_puts(seq, ",inline_dentry");
1648         else
1649                 seq_puts(seq, ",noinline_dentry");
1650         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1651                 seq_puts(seq, ",flush_merge");
1652         if (test_opt(sbi, NOBARRIER))
1653                 seq_puts(seq, ",nobarrier");
1654         if (test_opt(sbi, FASTBOOT))
1655                 seq_puts(seq, ",fastboot");
1656         if (test_opt(sbi, EXTENT_CACHE))
1657                 seq_puts(seq, ",extent_cache");
1658         else
1659                 seq_puts(seq, ",noextent_cache");
1660         if (test_opt(sbi, DATA_FLUSH))
1661                 seq_puts(seq, ",data_flush");
1662
1663         seq_puts(seq, ",mode=");
1664         if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1665                 seq_puts(seq, "adaptive");
1666         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1667                 seq_puts(seq, "lfs");
1668         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1669         if (test_opt(sbi, RESERVE_ROOT))
1670                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1671                                 F2FS_OPTION(sbi).root_reserved_blocks,
1672                                 from_kuid_munged(&init_user_ns,
1673                                         F2FS_OPTION(sbi).s_resuid),
1674                                 from_kgid_munged(&init_user_ns,
1675                                         F2FS_OPTION(sbi).s_resgid));
1676         if (F2FS_IO_SIZE_BITS(sbi))
1677                 seq_printf(seq, ",io_bits=%u",
1678                                 F2FS_OPTION(sbi).write_io_size_bits);
1679 #ifdef CONFIG_F2FS_FAULT_INJECTION
1680         if (test_opt(sbi, FAULT_INJECTION)) {
1681                 seq_printf(seq, ",fault_injection=%u",
1682                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1683                 seq_printf(seq, ",fault_type=%u",
1684                                 F2FS_OPTION(sbi).fault_info.inject_type);
1685         }
1686 #endif
1687 #ifdef CONFIG_QUOTA
1688         if (test_opt(sbi, QUOTA))
1689                 seq_puts(seq, ",quota");
1690         if (test_opt(sbi, USRQUOTA))
1691                 seq_puts(seq, ",usrquota");
1692         if (test_opt(sbi, GRPQUOTA))
1693                 seq_puts(seq, ",grpquota");
1694         if (test_opt(sbi, PRJQUOTA))
1695                 seq_puts(seq, ",prjquota");
1696 #endif
1697         f2fs_show_quota_options(seq, sbi->sb);
1698         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1699                 seq_printf(seq, ",whint_mode=%s", "user-based");
1700         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1701                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1702
1703         fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1704
1705         if (sbi->sb->s_flags & SB_INLINECRYPT)
1706                 seq_puts(seq, ",inlinecrypt");
1707
1708         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1709                 seq_printf(seq, ",alloc_mode=%s", "default");
1710         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1711                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1712
1713         if (test_opt(sbi, DISABLE_CHECKPOINT))
1714                 seq_printf(seq, ",checkpoint=disable:%u",
1715                                 F2FS_OPTION(sbi).unusable_cap);
1716         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1717                 seq_printf(seq, ",fsync_mode=%s", "posix");
1718         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1719                 seq_printf(seq, ",fsync_mode=%s", "strict");
1720         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1721                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1722
1723 #ifdef CONFIG_F2FS_FS_COMPRESSION
1724         f2fs_show_compress_options(seq, sbi->sb);
1725 #endif
1726
1727         if (test_opt(sbi, ATGC))
1728                 seq_puts(seq, ",atgc");
1729         return 0;
1730 }
1731
1732 static void default_options(struct f2fs_sb_info *sbi)
1733 {
1734         /* init some FS parameters */
1735         F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
1736         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1737         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1738         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1739         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1740         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1741         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1742         F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1743         F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1744         F2FS_OPTION(sbi).compress_ext_cnt = 0;
1745         F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1746         F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1747
1748         sbi->sb->s_flags &= ~SB_INLINECRYPT;
1749
1750         set_opt(sbi, INLINE_XATTR);
1751         set_opt(sbi, INLINE_DATA);
1752         set_opt(sbi, INLINE_DENTRY);
1753         set_opt(sbi, EXTENT_CACHE);
1754         set_opt(sbi, NOHEAP);
1755         clear_opt(sbi, DISABLE_CHECKPOINT);
1756         F2FS_OPTION(sbi).unusable_cap = 0;
1757         sbi->sb->s_flags |= SB_LAZYTIME;
1758         set_opt(sbi, FLUSH_MERGE);
1759         set_opt(sbi, DISCARD);
1760         if (f2fs_sb_has_blkzoned(sbi))
1761                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1762         else
1763                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1764
1765 #ifdef CONFIG_F2FS_FS_XATTR
1766         set_opt(sbi, XATTR_USER);
1767 #endif
1768 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1769         set_opt(sbi, POSIX_ACL);
1770 #endif
1771
1772         f2fs_build_fault_attr(sbi, 0, 0);
1773 }
1774
1775 #ifdef CONFIG_QUOTA
1776 static int f2fs_enable_quotas(struct super_block *sb);
1777 #endif
1778
1779 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1780 {
1781         unsigned int s_flags = sbi->sb->s_flags;
1782         struct cp_control cpc;
1783         int err = 0;
1784         int ret;
1785         block_t unusable;
1786
1787         if (s_flags & SB_RDONLY) {
1788                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1789                 return -EINVAL;
1790         }
1791         sbi->sb->s_flags |= SB_ACTIVE;
1792
1793         f2fs_update_time(sbi, DISABLE_TIME);
1794
1795         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1796                 down_write(&sbi->gc_lock);
1797                 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
1798                 if (err == -ENODATA) {
1799                         err = 0;
1800                         break;
1801                 }
1802                 if (err && err != -EAGAIN)
1803                         break;
1804         }
1805
1806         ret = sync_filesystem(sbi->sb);
1807         if (ret || err) {
1808                 err = ret ? ret: err;
1809                 goto restore_flag;
1810         }
1811
1812         unusable = f2fs_get_unusable_blocks(sbi);
1813         if (f2fs_disable_cp_again(sbi, unusable)) {
1814                 err = -EAGAIN;
1815                 goto restore_flag;
1816         }
1817
1818         down_write(&sbi->gc_lock);
1819         cpc.reason = CP_PAUSE;
1820         set_sbi_flag(sbi, SBI_CP_DISABLED);
1821         err = f2fs_write_checkpoint(sbi, &cpc);
1822         if (err)
1823                 goto out_unlock;
1824
1825         spin_lock(&sbi->stat_lock);
1826         sbi->unusable_block_count = unusable;
1827         spin_unlock(&sbi->stat_lock);
1828
1829 out_unlock:
1830         up_write(&sbi->gc_lock);
1831 restore_flag:
1832         sbi->sb->s_flags = s_flags;     /* Restore SB_RDONLY status */
1833         return err;
1834 }
1835
1836 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1837 {
1838         int retry = DEFAULT_RETRY_IO_COUNT;
1839
1840         /* we should flush all the data to keep data consistency */
1841         do {
1842                 sync_inodes_sb(sbi->sb);
1843                 cond_resched();
1844                 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
1845         } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
1846
1847         if (unlikely(retry < 0))
1848                 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
1849
1850         down_write(&sbi->gc_lock);
1851         f2fs_dirty_to_prefree(sbi);
1852
1853         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1854         set_sbi_flag(sbi, SBI_IS_DIRTY);
1855         up_write(&sbi->gc_lock);
1856
1857         f2fs_sync_fs(sbi->sb, 1);
1858 }
1859
1860 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1861 {
1862         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1863         struct f2fs_mount_info org_mount_opt;
1864         unsigned long old_sb_flags;
1865         int err;
1866         bool need_restart_gc = false;
1867         bool need_stop_gc = false;
1868         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1869         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1870         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1871         bool no_atgc = !test_opt(sbi, ATGC);
1872         bool checkpoint_changed;
1873 #ifdef CONFIG_QUOTA
1874         int i, j;
1875 #endif
1876
1877         /*
1878          * Save the old mount options in case we
1879          * need to restore them.
1880          */
1881         org_mount_opt = sbi->mount_opt;
1882         old_sb_flags = sb->s_flags;
1883
1884 #ifdef CONFIG_QUOTA
1885         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1886         for (i = 0; i < MAXQUOTAS; i++) {
1887                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1888                         org_mount_opt.s_qf_names[i] =
1889                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1890                                 GFP_KERNEL);
1891                         if (!org_mount_opt.s_qf_names[i]) {
1892                                 for (j = 0; j < i; j++)
1893                                         kfree(org_mount_opt.s_qf_names[j]);
1894                                 return -ENOMEM;
1895                         }
1896                 } else {
1897                         org_mount_opt.s_qf_names[i] = NULL;
1898                 }
1899         }
1900 #endif
1901
1902         /* recover superblocks we couldn't write due to previous RO mount */
1903         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1904                 err = f2fs_commit_super(sbi, false);
1905                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1906                           err);
1907                 if (!err)
1908                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1909         }
1910
1911         default_options(sbi);
1912
1913         /* parse mount options */
1914         err = parse_options(sb, data, true);
1915         if (err)
1916                 goto restore_opts;
1917         checkpoint_changed =
1918                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1919
1920         /*
1921          * Previous and new state of filesystem is RO,
1922          * so skip checking GC and FLUSH_MERGE conditions.
1923          */
1924         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1925                 goto skip;
1926
1927 #ifdef CONFIG_QUOTA
1928         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1929                 err = dquot_suspend(sb, -1);
1930                 if (err < 0)
1931                         goto restore_opts;
1932         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1933                 /* dquot_resume needs RW */
1934                 sb->s_flags &= ~SB_RDONLY;
1935                 if (sb_any_quota_suspended(sb)) {
1936                         dquot_resume(sb, -1);
1937                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1938                         err = f2fs_enable_quotas(sb);
1939                         if (err)
1940                                 goto restore_opts;
1941                 }
1942         }
1943 #endif
1944         /* disallow enable atgc dynamically */
1945         if (no_atgc == !!test_opt(sbi, ATGC)) {
1946                 err = -EINVAL;
1947                 f2fs_warn(sbi, "switch atgc option is not allowed");
1948                 goto restore_opts;
1949         }
1950
1951         /* disallow enable/disable extent_cache dynamically */
1952         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1953                 err = -EINVAL;
1954                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1955                 goto restore_opts;
1956         }
1957
1958         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1959                 err = -EINVAL;
1960                 f2fs_warn(sbi, "switch io_bits option is not allowed");
1961                 goto restore_opts;
1962         }
1963
1964         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1965                 err = -EINVAL;
1966                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1967                 goto restore_opts;
1968         }
1969
1970         /*
1971          * We stop the GC thread if FS is mounted as RO
1972          * or if background_gc = off is passed in mount
1973          * option. Also sync the filesystem.
1974          */
1975         if ((*flags & SB_RDONLY) ||
1976                         F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) {
1977                 if (sbi->gc_thread) {
1978                         f2fs_stop_gc_thread(sbi);
1979                         need_restart_gc = true;
1980                 }
1981         } else if (!sbi->gc_thread) {
1982                 err = f2fs_start_gc_thread(sbi);
1983                 if (err)
1984                         goto restore_opts;
1985                 need_stop_gc = true;
1986         }
1987
1988         if (*flags & SB_RDONLY ||
1989                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1990                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1991                 sync_inodes_sb(sb);
1992
1993                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1994                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1995                 f2fs_sync_fs(sb, 1);
1996                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1997         }
1998
1999         if (checkpoint_changed) {
2000                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2001                         err = f2fs_disable_checkpoint(sbi);
2002                         if (err)
2003                                 goto restore_gc;
2004                 } else {
2005                         f2fs_enable_checkpoint(sbi);
2006                 }
2007         }
2008
2009         /*
2010          * We stop issue flush thread if FS is mounted as RO
2011          * or if flush_merge is not passed in mount option.
2012          */
2013         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2014                 clear_opt(sbi, FLUSH_MERGE);
2015                 f2fs_destroy_flush_cmd_control(sbi, false);
2016         } else {
2017                 err = f2fs_create_flush_cmd_control(sbi);
2018                 if (err)
2019                         goto restore_gc;
2020         }
2021 skip:
2022 #ifdef CONFIG_QUOTA
2023         /* Release old quota file names */
2024         for (i = 0; i < MAXQUOTAS; i++)
2025                 kfree(org_mount_opt.s_qf_names[i]);
2026 #endif
2027         /* Update the POSIXACL Flag */
2028         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2029                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2030
2031         limit_reserve_root(sbi);
2032         adjust_unusable_cap_perc(sbi);
2033         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2034         return 0;
2035 restore_gc:
2036         if (need_restart_gc) {
2037                 if (f2fs_start_gc_thread(sbi))
2038                         f2fs_warn(sbi, "background gc thread has stopped");
2039         } else if (need_stop_gc) {
2040                 f2fs_stop_gc_thread(sbi);
2041         }
2042 restore_opts:
2043 #ifdef CONFIG_QUOTA
2044         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2045         for (i = 0; i < MAXQUOTAS; i++) {
2046                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2047                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2048         }
2049 #endif
2050         sbi->mount_opt = org_mount_opt;
2051         sb->s_flags = old_sb_flags;
2052         return err;
2053 }
2054
2055 #ifdef CONFIG_QUOTA
2056 /* Read data from quotafile */
2057 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2058                                size_t len, loff_t off)
2059 {
2060         struct inode *inode = sb_dqopt(sb)->files[type];
2061         struct address_space *mapping = inode->i_mapping;
2062         block_t blkidx = F2FS_BYTES_TO_BLK(off);
2063         int offset = off & (sb->s_blocksize - 1);
2064         int tocopy;
2065         size_t toread;
2066         loff_t i_size = i_size_read(inode);
2067         struct page *page;
2068
2069         if (off > i_size)
2070                 return 0;
2071
2072         if (off + len > i_size)
2073                 len = i_size - off;
2074         toread = len;
2075         while (toread > 0) {
2076                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2077 repeat:
2078                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2079                 if (IS_ERR(page)) {
2080                         if (PTR_ERR(page) == -ENOMEM) {
2081                                 congestion_wait(BLK_RW_ASYNC,
2082                                                 DEFAULT_IO_TIMEOUT);
2083                                 goto repeat;
2084                         }
2085                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2086                         return PTR_ERR(page);
2087                 }
2088
2089                 lock_page(page);
2090
2091                 if (unlikely(page->mapping != mapping)) {
2092                         f2fs_put_page(page, 1);
2093                         goto repeat;
2094                 }
2095                 if (unlikely(!PageUptodate(page))) {
2096                         f2fs_put_page(page, 1);
2097                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2098                         return -EIO;
2099                 }
2100
2101                 memcpy_from_page(data, page, offset, tocopy);
2102                 f2fs_put_page(page, 1);
2103
2104                 offset = 0;
2105                 toread -= tocopy;
2106                 data += tocopy;
2107                 blkidx++;
2108         }
2109         return len;
2110 }
2111
2112 /* Write to quotafile */
2113 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2114                                 const char *data, size_t len, loff_t off)
2115 {
2116         struct inode *inode = sb_dqopt(sb)->files[type];
2117         struct address_space *mapping = inode->i_mapping;
2118         const struct address_space_operations *a_ops = mapping->a_ops;
2119         int offset = off & (sb->s_blocksize - 1);
2120         size_t towrite = len;
2121         struct page *page;
2122         void *fsdata = NULL;
2123         int err = 0;
2124         int tocopy;
2125
2126         while (towrite > 0) {
2127                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2128                                                                 towrite);
2129 retry:
2130                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2131                                                         &page, &fsdata);
2132                 if (unlikely(err)) {
2133                         if (err == -ENOMEM) {
2134                                 congestion_wait(BLK_RW_ASYNC,
2135                                                 DEFAULT_IO_TIMEOUT);
2136                                 goto retry;
2137                         }
2138                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2139                         break;
2140                 }
2141
2142                 memcpy_to_page(page, offset, data, tocopy);
2143
2144                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2145                                                 page, fsdata);
2146                 offset = 0;
2147                 towrite -= tocopy;
2148                 off += tocopy;
2149                 data += tocopy;
2150                 cond_resched();
2151         }
2152
2153         if (len == towrite)
2154                 return err;
2155         inode->i_mtime = inode->i_ctime = current_time(inode);
2156         f2fs_mark_inode_dirty_sync(inode, false);
2157         return len - towrite;
2158 }
2159
2160 static struct dquot **f2fs_get_dquots(struct inode *inode)
2161 {
2162         return F2FS_I(inode)->i_dquot;
2163 }
2164
2165 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2166 {
2167         return &F2FS_I(inode)->i_reserved_quota;
2168 }
2169
2170 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2171 {
2172         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2173                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2174                 return 0;
2175         }
2176
2177         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2178                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
2179 }
2180
2181 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2182 {
2183         int enabled = 0;
2184         int i, err;
2185
2186         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2187                 err = f2fs_enable_quotas(sbi->sb);
2188                 if (err) {
2189                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2190                         return 0;
2191                 }
2192                 return 1;
2193         }
2194
2195         for (i = 0; i < MAXQUOTAS; i++) {
2196                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2197                         err = f2fs_quota_on_mount(sbi, i);
2198                         if (!err) {
2199                                 enabled = 1;
2200                                 continue;
2201                         }
2202                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2203                                  err, i);
2204                 }
2205         }
2206         return enabled;
2207 }
2208
2209 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2210                              unsigned int flags)
2211 {
2212         struct inode *qf_inode;
2213         unsigned long qf_inum;
2214         int err;
2215
2216         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2217
2218         qf_inum = f2fs_qf_ino(sb, type);
2219         if (!qf_inum)
2220                 return -EPERM;
2221
2222         qf_inode = f2fs_iget(sb, qf_inum);
2223         if (IS_ERR(qf_inode)) {
2224                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2225                 return PTR_ERR(qf_inode);
2226         }
2227
2228         /* Don't account quota for quota files to avoid recursion */
2229         qf_inode->i_flags |= S_NOQUOTA;
2230         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2231         iput(qf_inode);
2232         return err;
2233 }
2234
2235 static int f2fs_enable_quotas(struct super_block *sb)
2236 {
2237         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2238         int type, err = 0;
2239         unsigned long qf_inum;
2240         bool quota_mopt[MAXQUOTAS] = {
2241                 test_opt(sbi, USRQUOTA),
2242                 test_opt(sbi, GRPQUOTA),
2243                 test_opt(sbi, PRJQUOTA),
2244         };
2245
2246         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2247                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2248                 return 0;
2249         }
2250
2251         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2252
2253         for (type = 0; type < MAXQUOTAS; type++) {
2254                 qf_inum = f2fs_qf_ino(sb, type);
2255                 if (qf_inum) {
2256                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2257                                 DQUOT_USAGE_ENABLED |
2258                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2259                         if (err) {
2260                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2261                                          type, err);
2262                                 for (type--; type >= 0; type--)
2263                                         dquot_quota_off(sb, type);
2264                                 set_sbi_flag(F2FS_SB(sb),
2265                                                 SBI_QUOTA_NEED_REPAIR);
2266                                 return err;
2267                         }
2268                 }
2269         }
2270         return 0;
2271 }
2272
2273 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2274 {
2275         struct quota_info *dqopt = sb_dqopt(sbi->sb);
2276         struct address_space *mapping = dqopt->files[type]->i_mapping;
2277         int ret = 0;
2278
2279         ret = dquot_writeback_dquots(sbi->sb, type);
2280         if (ret)
2281                 goto out;
2282
2283         ret = filemap_fdatawrite(mapping);
2284         if (ret)
2285                 goto out;
2286
2287         /* if we are using journalled quota */
2288         if (is_journalled_quota(sbi))
2289                 goto out;
2290
2291         ret = filemap_fdatawait(mapping);
2292
2293         truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2294 out:
2295         if (ret)
2296                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2297         return ret;
2298 }
2299
2300 int f2fs_quota_sync(struct super_block *sb, int type)
2301 {
2302         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2303         struct quota_info *dqopt = sb_dqopt(sb);
2304         int cnt;
2305         int ret = 0;
2306
2307         /*
2308          * Now when everything is written we can discard the pagecache so
2309          * that userspace sees the changes.
2310          */
2311         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2312
2313                 if (type != -1 && cnt != type)
2314                         continue;
2315
2316                 if (!sb_has_quota_active(sb, cnt))
2317                         continue;
2318
2319                 if (!f2fs_sb_has_quota_ino(sbi))
2320                         inode_lock(dqopt->files[cnt]);
2321
2322                 /*
2323                  * do_quotactl
2324                  *  f2fs_quota_sync
2325                  *  down_read(quota_sem)
2326                  *  dquot_writeback_dquots()
2327                  *  f2fs_dquot_commit
2328                  *                            block_operation
2329                  *                            down_read(quota_sem)
2330                  */
2331                 f2fs_lock_op(sbi);
2332                 down_read(&sbi->quota_sem);
2333
2334                 ret = f2fs_quota_sync_file(sbi, cnt);
2335
2336                 up_read(&sbi->quota_sem);
2337                 f2fs_unlock_op(sbi);
2338
2339                 if (!f2fs_sb_has_quota_ino(sbi))
2340                         inode_unlock(dqopt->files[cnt]);
2341
2342                 if (ret)
2343                         break;
2344         }
2345         return ret;
2346 }
2347
2348 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2349                                                         const struct path *path)
2350 {
2351         struct inode *inode;
2352         int err;
2353
2354         /* if quota sysfile exists, deny enabling quota with specific file */
2355         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2356                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2357                 return -EBUSY;
2358         }
2359
2360         err = f2fs_quota_sync(sb, type);
2361         if (err)
2362                 return err;
2363
2364         err = dquot_quota_on(sb, type, format_id, path);
2365         if (err)
2366                 return err;
2367
2368         inode = d_inode(path->dentry);
2369
2370         inode_lock(inode);
2371         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2372         f2fs_set_inode_flags(inode);
2373         inode_unlock(inode);
2374         f2fs_mark_inode_dirty_sync(inode, false);
2375
2376         return 0;
2377 }
2378
2379 static int __f2fs_quota_off(struct super_block *sb, int type)
2380 {
2381         struct inode *inode = sb_dqopt(sb)->files[type];
2382         int err;
2383
2384         if (!inode || !igrab(inode))
2385                 return dquot_quota_off(sb, type);
2386
2387         err = f2fs_quota_sync(sb, type);
2388         if (err)
2389                 goto out_put;
2390
2391         err = dquot_quota_off(sb, type);
2392         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2393                 goto out_put;
2394
2395         inode_lock(inode);
2396         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2397         f2fs_set_inode_flags(inode);
2398         inode_unlock(inode);
2399         f2fs_mark_inode_dirty_sync(inode, false);
2400 out_put:
2401         iput(inode);
2402         return err;
2403 }
2404
2405 static int f2fs_quota_off(struct super_block *sb, int type)
2406 {
2407         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2408         int err;
2409
2410         err = __f2fs_quota_off(sb, type);
2411
2412         /*
2413          * quotactl can shutdown journalled quota, result in inconsistence
2414          * between quota record and fs data by following updates, tag the
2415          * flag to let fsck be aware of it.
2416          */
2417         if (is_journalled_quota(sbi))
2418                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2419         return err;
2420 }
2421
2422 void f2fs_quota_off_umount(struct super_block *sb)
2423 {
2424         int type;
2425         int err;
2426
2427         for (type = 0; type < MAXQUOTAS; type++) {
2428                 err = __f2fs_quota_off(sb, type);
2429                 if (err) {
2430                         int ret = dquot_quota_off(sb, type);
2431
2432                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2433                                  type, err, ret);
2434                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2435                 }
2436         }
2437         /*
2438          * In case of checkpoint=disable, we must flush quota blocks.
2439          * This can cause NULL exception for node_inode in end_io, since
2440          * put_super already dropped it.
2441          */
2442         sync_filesystem(sb);
2443 }
2444
2445 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2446 {
2447         struct quota_info *dqopt = sb_dqopt(sb);
2448         int type;
2449
2450         for (type = 0; type < MAXQUOTAS; type++) {
2451                 if (!dqopt->files[type])
2452                         continue;
2453                 f2fs_inode_synced(dqopt->files[type]);
2454         }
2455 }
2456
2457 static int f2fs_dquot_commit(struct dquot *dquot)
2458 {
2459         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2460         int ret;
2461
2462         down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2463         ret = dquot_commit(dquot);
2464         if (ret < 0)
2465                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2466         up_read(&sbi->quota_sem);
2467         return ret;
2468 }
2469
2470 static int f2fs_dquot_acquire(struct dquot *dquot)
2471 {
2472         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2473         int ret;
2474
2475         down_read(&sbi->quota_sem);
2476         ret = dquot_acquire(dquot);
2477         if (ret < 0)
2478                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2479         up_read(&sbi->quota_sem);
2480         return ret;
2481 }
2482
2483 static int f2fs_dquot_release(struct dquot *dquot)
2484 {
2485         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2486         int ret = dquot_release(dquot);
2487
2488         if (ret < 0)
2489                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2490         return ret;
2491 }
2492
2493 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2494 {
2495         struct super_block *sb = dquot->dq_sb;
2496         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2497         int ret = dquot_mark_dquot_dirty(dquot);
2498
2499         /* if we are using journalled quota */
2500         if (is_journalled_quota(sbi))
2501                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2502
2503         return ret;
2504 }
2505
2506 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2507 {
2508         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2509         int ret = dquot_commit_info(sb, type);
2510
2511         if (ret < 0)
2512                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2513         return ret;
2514 }
2515
2516 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2517 {
2518         *projid = F2FS_I(inode)->i_projid;
2519         return 0;
2520 }
2521
2522 static const struct dquot_operations f2fs_quota_operations = {
2523         .get_reserved_space = f2fs_get_reserved_space,
2524         .write_dquot    = f2fs_dquot_commit,
2525         .acquire_dquot  = f2fs_dquot_acquire,
2526         .release_dquot  = f2fs_dquot_release,
2527         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2528         .write_info     = f2fs_dquot_commit_info,
2529         .alloc_dquot    = dquot_alloc,
2530         .destroy_dquot  = dquot_destroy,
2531         .get_projid     = f2fs_get_projid,
2532         .get_next_id    = dquot_get_next_id,
2533 };
2534
2535 static const struct quotactl_ops f2fs_quotactl_ops = {
2536         .quota_on       = f2fs_quota_on,
2537         .quota_off      = f2fs_quota_off,
2538         .quota_sync     = f2fs_quota_sync,
2539         .get_state      = dquot_get_state,
2540         .set_info       = dquot_set_dqinfo,
2541         .get_dqblk      = dquot_get_dqblk,
2542         .set_dqblk      = dquot_set_dqblk,
2543         .get_nextdqblk  = dquot_get_next_dqblk,
2544 };
2545 #else
2546 int f2fs_quota_sync(struct super_block *sb, int type)
2547 {
2548         return 0;
2549 }
2550
2551 void f2fs_quota_off_umount(struct super_block *sb)
2552 {
2553 }
2554 #endif
2555
2556 static const struct super_operations f2fs_sops = {
2557         .alloc_inode    = f2fs_alloc_inode,
2558         .free_inode     = f2fs_free_inode,
2559         .drop_inode     = f2fs_drop_inode,
2560         .write_inode    = f2fs_write_inode,
2561         .dirty_inode    = f2fs_dirty_inode,
2562         .show_options   = f2fs_show_options,
2563 #ifdef CONFIG_QUOTA
2564         .quota_read     = f2fs_quota_read,
2565         .quota_write    = f2fs_quota_write,
2566         .get_dquots     = f2fs_get_dquots,
2567 #endif
2568         .evict_inode    = f2fs_evict_inode,
2569         .put_super      = f2fs_put_super,
2570         .sync_fs        = f2fs_sync_fs,
2571         .freeze_fs      = f2fs_freeze,
2572         .unfreeze_fs    = f2fs_unfreeze,
2573         .statfs         = f2fs_statfs,
2574         .remount_fs     = f2fs_remount,
2575 };
2576
2577 #ifdef CONFIG_FS_ENCRYPTION
2578 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2579 {
2580         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2581                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2582                                 ctx, len, NULL);
2583 }
2584
2585 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2586                                                         void *fs_data)
2587 {
2588         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2589
2590         /*
2591          * Encrypting the root directory is not allowed because fsck
2592          * expects lost+found directory to exist and remain unencrypted
2593          * if LOST_FOUND feature is enabled.
2594          *
2595          */
2596         if (f2fs_sb_has_lost_found(sbi) &&
2597                         inode->i_ino == F2FS_ROOT_INO(sbi))
2598                 return -EPERM;
2599
2600         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2601                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2602                                 ctx, len, fs_data, XATTR_CREATE);
2603 }
2604
2605 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
2606 {
2607         return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
2608 }
2609
2610 static bool f2fs_has_stable_inodes(struct super_block *sb)
2611 {
2612         return true;
2613 }
2614
2615 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2616                                        int *ino_bits_ret, int *lblk_bits_ret)
2617 {
2618         *ino_bits_ret = 8 * sizeof(nid_t);
2619         *lblk_bits_ret = 8 * sizeof(block_t);
2620 }
2621
2622 static int f2fs_get_num_devices(struct super_block *sb)
2623 {
2624         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2625
2626         if (f2fs_is_multi_device(sbi))
2627                 return sbi->s_ndevs;
2628         return 1;
2629 }
2630
2631 static void f2fs_get_devices(struct super_block *sb,
2632                              struct request_queue **devs)
2633 {
2634         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2635         int i;
2636
2637         for (i = 0; i < sbi->s_ndevs; i++)
2638                 devs[i] = bdev_get_queue(FDEV(i).bdev);
2639 }
2640
2641 static const struct fscrypt_operations f2fs_cryptops = {
2642         .key_prefix             = "f2fs:",
2643         .get_context            = f2fs_get_context,
2644         .set_context            = f2fs_set_context,
2645         .get_dummy_policy       = f2fs_get_dummy_policy,
2646         .empty_dir              = f2fs_empty_dir,
2647         .max_namelen            = F2FS_NAME_LEN,
2648         .has_stable_inodes      = f2fs_has_stable_inodes,
2649         .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
2650         .get_num_devices        = f2fs_get_num_devices,
2651         .get_devices            = f2fs_get_devices,
2652 };
2653 #endif
2654
2655 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2656                 u64 ino, u32 generation)
2657 {
2658         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2659         struct inode *inode;
2660
2661         if (f2fs_check_nid_range(sbi, ino))
2662                 return ERR_PTR(-ESTALE);
2663
2664         /*
2665          * f2fs_iget isn't quite right if the inode is currently unallocated!
2666          * However f2fs_iget currently does appropriate checks to handle stale
2667          * inodes so everything is OK.
2668          */
2669         inode = f2fs_iget(sb, ino);
2670         if (IS_ERR(inode))
2671                 return ERR_CAST(inode);
2672         if (unlikely(generation && inode->i_generation != generation)) {
2673                 /* we didn't find the right inode.. */
2674                 iput(inode);
2675                 return ERR_PTR(-ESTALE);
2676         }
2677         return inode;
2678 }
2679
2680 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2681                 int fh_len, int fh_type)
2682 {
2683         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2684                                     f2fs_nfs_get_inode);
2685 }
2686
2687 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2688                 int fh_len, int fh_type)
2689 {
2690         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2691                                     f2fs_nfs_get_inode);
2692 }
2693
2694 static const struct export_operations f2fs_export_ops = {
2695         .fh_to_dentry = f2fs_fh_to_dentry,
2696         .fh_to_parent = f2fs_fh_to_parent,
2697         .get_parent = f2fs_get_parent,
2698 };
2699
2700 static loff_t max_file_blocks(void)
2701 {
2702         loff_t result = 0;
2703         loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2704
2705         /*
2706          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2707          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2708          * space in inode.i_addr, it will be more safe to reassign
2709          * result as zero.
2710          */
2711
2712         /* two direct node blocks */
2713         result += (leaf_count * 2);
2714
2715         /* two indirect node blocks */
2716         leaf_count *= NIDS_PER_BLOCK;
2717         result += (leaf_count * 2);
2718
2719         /* one double indirect node block */
2720         leaf_count *= NIDS_PER_BLOCK;
2721         result += leaf_count;
2722
2723         return result;
2724 }
2725
2726 static int __f2fs_commit_super(struct buffer_head *bh,
2727                         struct f2fs_super_block *super)
2728 {
2729         lock_buffer(bh);
2730         if (super)
2731                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2732         set_buffer_dirty(bh);
2733         unlock_buffer(bh);
2734
2735         /* it's rare case, we can do fua all the time */
2736         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2737 }
2738
2739 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2740                                         struct buffer_head *bh)
2741 {
2742         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2743                                         (bh->b_data + F2FS_SUPER_OFFSET);
2744         struct super_block *sb = sbi->sb;
2745         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2746         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2747         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2748         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2749         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2750         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2751         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2752         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2753         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2754         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2755         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2756         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2757         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2758         u64 main_end_blkaddr = main_blkaddr +
2759                                 (segment_count_main << log_blocks_per_seg);
2760         u64 seg_end_blkaddr = segment0_blkaddr +
2761                                 (segment_count << log_blocks_per_seg);
2762
2763         if (segment0_blkaddr != cp_blkaddr) {
2764                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2765                           segment0_blkaddr, cp_blkaddr);
2766                 return true;
2767         }
2768
2769         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2770                                                         sit_blkaddr) {
2771                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2772                           cp_blkaddr, sit_blkaddr,
2773                           segment_count_ckpt << log_blocks_per_seg);
2774                 return true;
2775         }
2776
2777         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2778                                                         nat_blkaddr) {
2779                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2780                           sit_blkaddr, nat_blkaddr,
2781                           segment_count_sit << log_blocks_per_seg);
2782                 return true;
2783         }
2784
2785         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2786                                                         ssa_blkaddr) {
2787                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2788                           nat_blkaddr, ssa_blkaddr,
2789                           segment_count_nat << log_blocks_per_seg);
2790                 return true;
2791         }
2792
2793         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2794                                                         main_blkaddr) {
2795                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2796                           ssa_blkaddr, main_blkaddr,
2797                           segment_count_ssa << log_blocks_per_seg);
2798                 return true;
2799         }
2800
2801         if (main_end_blkaddr > seg_end_blkaddr) {
2802                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
2803                           main_blkaddr, seg_end_blkaddr,
2804                           segment_count_main << log_blocks_per_seg);
2805                 return true;
2806         } else if (main_end_blkaddr < seg_end_blkaddr) {
2807                 int err = 0;
2808                 char *res;
2809
2810                 /* fix in-memory information all the time */
2811                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2812                                 segment0_blkaddr) >> log_blocks_per_seg);
2813
2814                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2815                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2816                         res = "internally";
2817                 } else {
2818                         err = __f2fs_commit_super(bh, NULL);
2819                         res = err ? "failed" : "done";
2820                 }
2821                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
2822                           res, main_blkaddr, seg_end_blkaddr,
2823                           segment_count_main << log_blocks_per_seg);
2824                 if (err)
2825                         return true;
2826         }
2827         return false;
2828 }
2829
2830 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2831                                 struct buffer_head *bh)
2832 {
2833         block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
2834         block_t total_sections, blocks_per_seg;
2835         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2836                                         (bh->b_data + F2FS_SUPER_OFFSET);
2837         size_t crc_offset = 0;
2838         __u32 crc = 0;
2839
2840         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2841                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2842                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2843                 return -EINVAL;
2844         }
2845
2846         /* Check checksum_offset and crc in superblock */
2847         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2848                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2849                 if (crc_offset !=
2850                         offsetof(struct f2fs_super_block, crc)) {
2851                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2852                                   crc_offset);
2853                         return -EFSCORRUPTED;
2854                 }
2855                 crc = le32_to_cpu(raw_super->crc);
2856                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2857                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2858                         return -EFSCORRUPTED;
2859                 }
2860         }
2861
2862         /* Currently, support only 4KB page cache size */
2863         if (F2FS_BLKSIZE != PAGE_SIZE) {
2864                 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2865                           PAGE_SIZE);
2866                 return -EFSCORRUPTED;
2867         }
2868
2869         /* Currently, support only 4KB block size */
2870         if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
2871                 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
2872                           le32_to_cpu(raw_super->log_blocksize),
2873                           F2FS_BLKSIZE_BITS);
2874                 return -EFSCORRUPTED;
2875         }
2876
2877         /* check log blocks per segment */
2878         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2879                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2880                           le32_to_cpu(raw_super->log_blocks_per_seg));
2881                 return -EFSCORRUPTED;
2882         }
2883
2884         /* Currently, support 512/1024/2048/4096 bytes sector size */
2885         if (le32_to_cpu(raw_super->log_sectorsize) >
2886                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2887                 le32_to_cpu(raw_super->log_sectorsize) <
2888                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2889                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2890                           le32_to_cpu(raw_super->log_sectorsize));
2891                 return -EFSCORRUPTED;
2892         }
2893         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2894                 le32_to_cpu(raw_super->log_sectorsize) !=
2895                         F2FS_MAX_LOG_SECTOR_SIZE) {
2896                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2897                           le32_to_cpu(raw_super->log_sectors_per_block),
2898                           le32_to_cpu(raw_super->log_sectorsize));
2899                 return -EFSCORRUPTED;
2900         }
2901
2902         segment_count = le32_to_cpu(raw_super->segment_count);
2903         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2904         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2905         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2906         total_sections = le32_to_cpu(raw_super->section_count);
2907
2908         /* blocks_per_seg should be 512, given the above check */
2909         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2910
2911         if (segment_count > F2FS_MAX_SEGMENT ||
2912                                 segment_count < F2FS_MIN_SEGMENTS) {
2913                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2914                 return -EFSCORRUPTED;
2915         }
2916
2917         if (total_sections > segment_count_main || total_sections < 1 ||
2918                         segs_per_sec > segment_count || !segs_per_sec) {
2919                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2920                           segment_count, total_sections, segs_per_sec);
2921                 return -EFSCORRUPTED;
2922         }
2923
2924         if (segment_count_main != total_sections * segs_per_sec) {
2925                 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
2926                           segment_count_main, total_sections, segs_per_sec);
2927                 return -EFSCORRUPTED;
2928         }
2929
2930         if ((segment_count / segs_per_sec) < total_sections) {
2931                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2932                           segment_count, segs_per_sec, total_sections);
2933                 return -EFSCORRUPTED;
2934         }
2935
2936         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2937                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2938                           segment_count, le64_to_cpu(raw_super->block_count));
2939                 return -EFSCORRUPTED;
2940         }
2941
2942         if (RDEV(0).path[0]) {
2943                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2944                 int i = 1;
2945
2946                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2947                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2948                         i++;
2949                 }
2950                 if (segment_count != dev_seg_count) {
2951                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2952                                         segment_count, dev_seg_count);
2953                         return -EFSCORRUPTED;
2954                 }
2955         } else {
2956                 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
2957                                         !bdev_is_zoned(sbi->sb->s_bdev)) {
2958                         f2fs_info(sbi, "Zoned block device path is missing");
2959                         return -EFSCORRUPTED;
2960                 }
2961         }
2962
2963         if (secs_per_zone > total_sections || !secs_per_zone) {
2964                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2965                           secs_per_zone, total_sections);
2966                 return -EFSCORRUPTED;
2967         }
2968         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2969                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2970                         (le32_to_cpu(raw_super->extension_count) +
2971                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2972                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2973                           le32_to_cpu(raw_super->extension_count),
2974                           raw_super->hot_ext_count,
2975                           F2FS_MAX_EXTENSION);
2976                 return -EFSCORRUPTED;
2977         }
2978
2979         if (le32_to_cpu(raw_super->cp_payload) >=
2980                                 (blocks_per_seg - F2FS_CP_PACKS -
2981                                 NR_CURSEG_PERSIST_TYPE)) {
2982                 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
2983                           le32_to_cpu(raw_super->cp_payload),
2984                           blocks_per_seg - F2FS_CP_PACKS -
2985                           NR_CURSEG_PERSIST_TYPE);
2986                 return -EFSCORRUPTED;
2987         }
2988
2989         /* check reserved ino info */
2990         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2991                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2992                 le32_to_cpu(raw_super->root_ino) != 3) {
2993                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2994                           le32_to_cpu(raw_super->node_ino),
2995                           le32_to_cpu(raw_super->meta_ino),
2996                           le32_to_cpu(raw_super->root_ino));
2997                 return -EFSCORRUPTED;
2998         }
2999
3000         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3001         if (sanity_check_area_boundary(sbi, bh))
3002                 return -EFSCORRUPTED;
3003
3004         return 0;
3005 }
3006
3007 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3008 {
3009         unsigned int total, fsmeta;
3010         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3011         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3012         unsigned int ovp_segments, reserved_segments;
3013         unsigned int main_segs, blocks_per_seg;
3014         unsigned int sit_segs, nat_segs;
3015         unsigned int sit_bitmap_size, nat_bitmap_size;
3016         unsigned int log_blocks_per_seg;
3017         unsigned int segment_count_main;
3018         unsigned int cp_pack_start_sum, cp_payload;
3019         block_t user_block_count, valid_user_blocks;
3020         block_t avail_node_count, valid_node_count;
3021         unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3022         int i, j;
3023
3024         total = le32_to_cpu(raw_super->segment_count);
3025         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3026         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3027         fsmeta += sit_segs;
3028         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3029         fsmeta += nat_segs;
3030         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3031         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3032
3033         if (unlikely(fsmeta >= total))
3034                 return 1;
3035
3036         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3037         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3038
3039         if (unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3040                         ovp_segments == 0 || reserved_segments == 0)) {
3041                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3042                 return 1;
3043         }
3044
3045         user_block_count = le64_to_cpu(ckpt->user_block_count);
3046         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3047         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3048         if (!user_block_count || user_block_count >=
3049                         segment_count_main << log_blocks_per_seg) {
3050                 f2fs_err(sbi, "Wrong user_block_count: %u",
3051                          user_block_count);
3052                 return 1;
3053         }
3054
3055         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3056         if (valid_user_blocks > user_block_count) {
3057                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3058                          valid_user_blocks, user_block_count);
3059                 return 1;
3060         }
3061
3062         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3063         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3064         if (valid_node_count > avail_node_count) {
3065                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3066                          valid_node_count, avail_node_count);
3067                 return 1;
3068         }
3069
3070         main_segs = le32_to_cpu(raw_super->segment_count_main);
3071         blocks_per_seg = sbi->blocks_per_seg;
3072
3073         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3074                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3075                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3076                         return 1;
3077                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3078                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3079                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
3080                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3081                                          i, j,
3082                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3083                                 return 1;
3084                         }
3085                 }
3086         }
3087         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3088                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3089                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3090                         return 1;
3091                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3092                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3093                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3094                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3095                                          i, j,
3096                                          le32_to_cpu(ckpt->cur_data_segno[i]));
3097                                 return 1;
3098                         }
3099                 }
3100         }
3101         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3102                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3103                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3104                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3105                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3106                                          i, j,
3107                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3108                                 return 1;
3109                         }
3110                 }
3111         }
3112
3113         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3114         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3115
3116         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3117                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3118                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3119                          sit_bitmap_size, nat_bitmap_size);
3120                 return 1;
3121         }
3122
3123         cp_pack_start_sum = __start_sum_addr(sbi);
3124         cp_payload = __cp_payload(sbi);
3125         if (cp_pack_start_sum < cp_payload + 1 ||
3126                 cp_pack_start_sum > blocks_per_seg - 1 -
3127                         NR_CURSEG_PERSIST_TYPE) {
3128                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3129                          cp_pack_start_sum);
3130                 return 1;
3131         }
3132
3133         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3134                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3135                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3136                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3137                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3138                           le32_to_cpu(ckpt->checksum_offset));
3139                 return 1;
3140         }
3141
3142         nat_blocks = nat_segs << log_blocks_per_seg;
3143         nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3144         nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3145         if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3146                 (cp_payload + F2FS_CP_PACKS +
3147                 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3148                 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3149                           cp_payload, nat_bits_blocks);
3150                 return 1;
3151         }
3152
3153         if (unlikely(f2fs_cp_error(sbi))) {
3154                 f2fs_err(sbi, "A bug case: need to run fsck");
3155                 return 1;
3156         }
3157         return 0;
3158 }
3159
3160 static void init_sb_info(struct f2fs_sb_info *sbi)
3161 {
3162         struct f2fs_super_block *raw_super = sbi->raw_super;
3163         int i;
3164
3165         sbi->log_sectors_per_block =
3166                 le32_to_cpu(raw_super->log_sectors_per_block);
3167         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3168         sbi->blocksize = 1 << sbi->log_blocksize;
3169         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3170         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3171         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3172         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3173         sbi->total_sections = le32_to_cpu(raw_super->section_count);
3174         sbi->total_node_count =
3175                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3176                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3177         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
3178         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
3179         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
3180         sbi->cur_victim_sec = NULL_SECNO;
3181         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3182         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3183         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3184         sbi->migration_granularity = sbi->segs_per_sec;
3185
3186         sbi->dir_level = DEF_DIR_LEVEL;
3187         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3188         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3189         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3190         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3191         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3192         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3193                                 DEF_UMOUNT_DISCARD_TIMEOUT;
3194         clear_sbi_flag(sbi, SBI_NEED_FSCK);
3195
3196         for (i = 0; i < NR_COUNT_TYPE; i++)
3197                 atomic_set(&sbi->nr_pages[i], 0);
3198
3199         for (i = 0; i < META; i++)
3200                 atomic_set(&sbi->wb_sync_req[i], 0);
3201
3202         INIT_LIST_HEAD(&sbi->s_list);
3203         mutex_init(&sbi->umount_mutex);
3204         init_rwsem(&sbi->io_order_lock);
3205         spin_lock_init(&sbi->cp_lock);
3206
3207         sbi->dirty_device = 0;
3208         spin_lock_init(&sbi->dev_lock);
3209
3210         init_rwsem(&sbi->sb_lock);
3211         init_rwsem(&sbi->pin_sem);
3212 }
3213
3214 static int init_percpu_info(struct f2fs_sb_info *sbi)
3215 {
3216         int err;
3217
3218         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3219         if (err)
3220                 return err;
3221
3222         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3223                                                                 GFP_KERNEL);
3224         if (err)
3225                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
3226
3227         return err;
3228 }
3229
3230 #ifdef CONFIG_BLK_DEV_ZONED
3231
3232 struct f2fs_report_zones_args {
3233         struct f2fs_sb_info *sbi;
3234         struct f2fs_dev_info *dev;
3235 };
3236
3237 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3238                               void *data)
3239 {
3240         struct f2fs_report_zones_args *rz_args = data;
3241         block_t unusable_blocks = (zone->len - zone->capacity) >>
3242                                         F2FS_LOG_SECTORS_PER_BLOCK;
3243
3244         if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3245                 return 0;
3246
3247         set_bit(idx, rz_args->dev->blkz_seq);
3248         if (!rz_args->sbi->unusable_blocks_per_sec) {
3249                 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3250                 return 0;
3251         }
3252         if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3253                 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3254                 return -EINVAL;
3255         }
3256         return 0;
3257 }
3258
3259 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3260 {
3261         struct block_device *bdev = FDEV(devi).bdev;
3262         sector_t nr_sectors = bdev->bd_part->nr_sects;
3263         struct f2fs_report_zones_args rep_zone_arg;
3264         int ret;
3265
3266         if (!f2fs_sb_has_blkzoned(sbi))
3267                 return 0;
3268
3269         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3270                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3271                 return -EINVAL;
3272         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3273         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3274                                 __ilog2_u32(sbi->blocks_per_blkz))
3275                 return -EINVAL;
3276         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3277         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3278                                         sbi->log_blocks_per_blkz;
3279         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3280                 FDEV(devi).nr_blkz++;
3281
3282         FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3283                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
3284                                         * sizeof(unsigned long),
3285                                         GFP_KERNEL);
3286         if (!FDEV(devi).blkz_seq)
3287                 return -ENOMEM;
3288
3289         rep_zone_arg.sbi = sbi;
3290         rep_zone_arg.dev = &FDEV(devi);
3291
3292         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3293                                   &rep_zone_arg);
3294         if (ret < 0)
3295                 return ret;
3296         return 0;
3297 }
3298 #endif
3299
3300 /*
3301  * Read f2fs raw super block.
3302  * Because we have two copies of super block, so read both of them
3303  * to get the first valid one. If any one of them is broken, we pass
3304  * them recovery flag back to the caller.
3305  */
3306 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3307                         struct f2fs_super_block **raw_super,
3308                         int *valid_super_block, int *recovery)
3309 {
3310         struct super_block *sb = sbi->sb;
3311         int block;
3312         struct buffer_head *bh;
3313         struct f2fs_super_block *super;
3314         int err = 0;
3315
3316         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3317         if (!super)
3318                 return -ENOMEM;
3319
3320         for (block = 0; block < 2; block++) {
3321                 bh = sb_bread(sb, block);
3322                 if (!bh) {
3323                         f2fs_err(sbi, "Unable to read %dth superblock",
3324                                  block + 1);
3325                         err = -EIO;
3326                         *recovery = 1;
3327                         continue;
3328                 }
3329
3330                 /* sanity checking of raw super */
3331                 err = sanity_check_raw_super(sbi, bh);
3332                 if (err) {
3333                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3334                                  block + 1);
3335                         brelse(bh);
3336                         *recovery = 1;
3337                         continue;
3338                 }
3339
3340                 if (!*raw_super) {
3341                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3342                                                         sizeof(*super));
3343                         *valid_super_block = block;
3344                         *raw_super = super;
3345                 }
3346                 brelse(bh);
3347         }
3348
3349         /* No valid superblock */
3350         if (!*raw_super)
3351                 kfree(super);
3352         else
3353                 err = 0;
3354
3355         return err;
3356 }
3357
3358 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3359 {
3360         struct buffer_head *bh;
3361         __u32 crc = 0;
3362         int err;
3363
3364         if ((recover && f2fs_readonly(sbi->sb)) ||
3365                                 bdev_read_only(sbi->sb->s_bdev)) {
3366                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3367                 return -EROFS;
3368         }
3369
3370         /* we should update superblock crc here */
3371         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3372                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3373                                 offsetof(struct f2fs_super_block, crc));
3374                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3375         }
3376
3377         /* write back-up superblock first */
3378         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3379         if (!bh)
3380                 return -EIO;
3381         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3382         brelse(bh);
3383
3384         /* if we are in recovery path, skip writing valid superblock */
3385         if (recover || err)
3386                 return err;
3387
3388         /* write current valid superblock */
3389         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3390         if (!bh)
3391                 return -EIO;
3392         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3393         brelse(bh);
3394         return err;
3395 }
3396
3397 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3398 {
3399         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3400         unsigned int max_devices = MAX_DEVICES;
3401         int i;
3402
3403         /* Initialize single device information */
3404         if (!RDEV(0).path[0]) {
3405                 if (!bdev_is_zoned(sbi->sb->s_bdev))
3406                         return 0;
3407                 max_devices = 1;
3408         }
3409
3410         /*
3411          * Initialize multiple devices information, or single
3412          * zoned block device information.
3413          */
3414         sbi->devs = f2fs_kzalloc(sbi,
3415                                  array_size(max_devices,
3416                                             sizeof(struct f2fs_dev_info)),
3417                                  GFP_KERNEL);
3418         if (!sbi->devs)
3419                 return -ENOMEM;
3420
3421         for (i = 0; i < max_devices; i++) {
3422
3423                 if (i > 0 && !RDEV(i).path[0])
3424                         break;
3425
3426                 if (max_devices == 1) {
3427                         /* Single zoned block device mount */
3428                         FDEV(0).bdev =
3429                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3430                                         sbi->sb->s_mode, sbi->sb->s_type);
3431                 } else {
3432                         /* Multi-device mount */
3433                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3434                         FDEV(i).total_segments =
3435                                 le32_to_cpu(RDEV(i).total_segments);
3436                         if (i == 0) {
3437                                 FDEV(i).start_blk = 0;
3438                                 FDEV(i).end_blk = FDEV(i).start_blk +
3439                                     (FDEV(i).total_segments <<
3440                                     sbi->log_blocks_per_seg) - 1 +
3441                                     le32_to_cpu(raw_super->segment0_blkaddr);
3442                         } else {
3443                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3444                                 FDEV(i).end_blk = FDEV(i).start_blk +
3445                                         (FDEV(i).total_segments <<
3446                                         sbi->log_blocks_per_seg) - 1;
3447                         }
3448                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3449                                         sbi->sb->s_mode, sbi->sb->s_type);
3450                 }
3451                 if (IS_ERR(FDEV(i).bdev))
3452                         return PTR_ERR(FDEV(i).bdev);
3453
3454                 /* to release errored devices */
3455                 sbi->s_ndevs = i + 1;
3456
3457 #ifdef CONFIG_BLK_DEV_ZONED
3458                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3459                                 !f2fs_sb_has_blkzoned(sbi)) {
3460                         f2fs_err(sbi, "Zoned block device feature not enabled\n");
3461                         return -EINVAL;
3462                 }
3463                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3464                         if (init_blkz_info(sbi, i)) {
3465                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3466                                 return -EINVAL;
3467                         }
3468                         if (max_devices == 1)
3469                                 break;
3470                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3471                                   i, FDEV(i).path,
3472                                   FDEV(i).total_segments,
3473                                   FDEV(i).start_blk, FDEV(i).end_blk,
3474                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3475                                   "Host-aware" : "Host-managed");
3476                         continue;
3477                 }
3478 #endif
3479                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3480                           i, FDEV(i).path,
3481                           FDEV(i).total_segments,
3482                           FDEV(i).start_blk, FDEV(i).end_blk);
3483         }
3484         f2fs_info(sbi,
3485                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3486         return 0;
3487 }
3488
3489 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3490 {
3491 #ifdef CONFIG_UNICODE
3492         if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3493                 const struct f2fs_sb_encodings *encoding_info;
3494                 struct unicode_map *encoding;
3495                 __u16 encoding_flags;
3496
3497                 if (f2fs_sb_has_encrypt(sbi)) {
3498                         f2fs_err(sbi,
3499                                 "Can't mount with encoding and encryption");
3500                         return -EINVAL;
3501                 }
3502
3503                 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3504                                           &encoding_flags)) {
3505                         f2fs_err(sbi,
3506                                  "Encoding requested by superblock is unknown");
3507                         return -EINVAL;
3508                 }
3509
3510                 encoding = utf8_load(encoding_info->version);
3511                 if (IS_ERR(encoding)) {
3512                         f2fs_err(sbi,
3513                                  "can't mount with superblock charset: %s-%s "
3514                                  "not supported by the kernel. flags: 0x%x.",
3515                                  encoding_info->name, encoding_info->version,
3516                                  encoding_flags);
3517                         return PTR_ERR(encoding);
3518                 }
3519                 f2fs_info(sbi, "Using encoding defined by superblock: "
3520                          "%s-%s with flags 0x%hx", encoding_info->name,
3521                          encoding_info->version?:"\b", encoding_flags);
3522
3523                 sbi->sb->s_encoding = encoding;
3524                 sbi->sb->s_encoding_flags = encoding_flags;
3525                 sbi->sb->s_d_op = &f2fs_dentry_ops;
3526         }
3527 #else
3528         if (f2fs_sb_has_casefold(sbi)) {
3529                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3530                 return -EINVAL;
3531         }
3532 #endif
3533         return 0;
3534 }
3535
3536 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3537 {
3538         struct f2fs_sm_info *sm_i = SM_I(sbi);
3539
3540         /* adjust parameters according to the volume size */
3541         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3542                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3543                 sm_i->dcc_info->discard_granularity = 1;
3544                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3545         }
3546
3547         sbi->readdir_ra = 1;
3548 }
3549
3550 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3551 {
3552         struct f2fs_sb_info *sbi;
3553         struct f2fs_super_block *raw_super;
3554         struct inode *root;
3555         int err;
3556         bool skip_recovery = false, need_fsck = false;
3557         char *options = NULL;
3558         int recovery, i, valid_super_block;
3559         struct curseg_info *seg_i;
3560         int retry_cnt = 1;
3561
3562 try_onemore:
3563         err = -EINVAL;
3564         raw_super = NULL;
3565         valid_super_block = -1;
3566         recovery = 0;
3567
3568         /* allocate memory for f2fs-specific super block info */
3569         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3570         if (!sbi)
3571                 return -ENOMEM;
3572
3573         sbi->sb = sb;
3574
3575         /* Load the checksum driver */
3576         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3577         if (IS_ERR(sbi->s_chksum_driver)) {
3578                 f2fs_err(sbi, "Cannot load crc32 driver.");
3579                 err = PTR_ERR(sbi->s_chksum_driver);
3580                 sbi->s_chksum_driver = NULL;
3581                 goto free_sbi;
3582         }
3583
3584         /* set a block size */
3585         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3586                 f2fs_err(sbi, "unable to set blocksize");
3587                 goto free_sbi;
3588         }
3589
3590         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3591                                                                 &recovery);
3592         if (err)
3593                 goto free_sbi;
3594
3595         sb->s_fs_info = sbi;
3596         sbi->raw_super = raw_super;
3597
3598         /* precompute checksum seed for metadata */
3599         if (f2fs_sb_has_inode_chksum(sbi))
3600                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3601                                                 sizeof(raw_super->uuid));
3602
3603         default_options(sbi);
3604         /* parse mount options */
3605         options = kstrdup((const char *)data, GFP_KERNEL);
3606         if (data && !options) {
3607                 err = -ENOMEM;
3608                 goto free_sb_buf;
3609         }
3610
3611         err = parse_options(sb, options, false);
3612         if (err)
3613                 goto free_options;
3614
3615         sbi->max_file_blocks = max_file_blocks();
3616         sb->s_maxbytes = sbi->max_file_blocks <<
3617                                 le32_to_cpu(raw_super->log_blocksize);
3618         sb->s_max_links = F2FS_LINK_MAX;
3619
3620         err = f2fs_setup_casefold(sbi);
3621         if (err)
3622                 goto free_options;
3623
3624 #ifdef CONFIG_QUOTA
3625         sb->dq_op = &f2fs_quota_operations;
3626         sb->s_qcop = &f2fs_quotactl_ops;
3627         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3628
3629         if (f2fs_sb_has_quota_ino(sbi)) {
3630                 for (i = 0; i < MAXQUOTAS; i++) {
3631                         if (f2fs_qf_ino(sbi->sb, i))
3632                                 sbi->nquota_files++;
3633                 }
3634         }
3635 #endif
3636
3637         sb->s_op = &f2fs_sops;
3638 #ifdef CONFIG_FS_ENCRYPTION
3639         sb->s_cop = &f2fs_cryptops;
3640 #endif
3641 #ifdef CONFIG_FS_VERITY
3642         sb->s_vop = &f2fs_verityops;
3643 #endif
3644         sb->s_xattr = f2fs_xattr_handlers;
3645         sb->s_export_op = &f2fs_export_ops;
3646         sb->s_magic = F2FS_SUPER_MAGIC;
3647         sb->s_time_gran = 1;
3648         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3649                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3650         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3651         sb->s_iflags |= SB_I_CGROUPWB;
3652
3653         /* init f2fs-specific super block info */
3654         sbi->valid_super_block = valid_super_block;
3655         init_rwsem(&sbi->gc_lock);
3656         mutex_init(&sbi->writepages);
3657         mutex_init(&sbi->cp_mutex);
3658         init_rwsem(&sbi->node_write);
3659         init_rwsem(&sbi->node_change);
3660
3661         /* disallow all the data/node/meta page writes */
3662         set_sbi_flag(sbi, SBI_POR_DOING);
3663         spin_lock_init(&sbi->stat_lock);
3664
3665         /* init iostat info */
3666         spin_lock_init(&sbi->iostat_lock);
3667         sbi->iostat_enable = false;
3668         sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3669
3670         for (i = 0; i < NR_PAGE_TYPE; i++) {
3671                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3672                 int j;
3673
3674                 sbi->write_io[i] =
3675                         f2fs_kmalloc(sbi,
3676                                      array_size(n,
3677                                                 sizeof(struct f2fs_bio_info)),
3678                                      GFP_KERNEL);
3679                 if (!sbi->write_io[i]) {
3680                         err = -ENOMEM;
3681                         goto free_bio_info;
3682                 }
3683
3684                 for (j = HOT; j < n; j++) {
3685                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3686                         sbi->write_io[i][j].sbi = sbi;
3687                         sbi->write_io[i][j].bio = NULL;
3688                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3689                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3690                         INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3691                         init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3692                 }
3693         }
3694
3695         init_rwsem(&sbi->cp_rwsem);
3696         init_rwsem(&sbi->quota_sem);
3697         init_waitqueue_head(&sbi->cp_wait);
3698         init_sb_info(sbi);
3699
3700         err = init_percpu_info(sbi);
3701         if (err)
3702                 goto free_bio_info;
3703
3704         if (F2FS_IO_ALIGNED(sbi)) {
3705                 sbi->write_io_dummy =
3706                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3707                 if (!sbi->write_io_dummy) {
3708                         err = -ENOMEM;
3709                         goto free_percpu;
3710                 }
3711         }
3712
3713         /* init per sbi slab cache */
3714         err = f2fs_init_xattr_caches(sbi);
3715         if (err)
3716                 goto free_io_dummy;
3717         err = f2fs_init_page_array_cache(sbi);
3718         if (err)
3719                 goto free_xattr_cache;
3720
3721         /* get an inode for meta space */
3722         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3723         if (IS_ERR(sbi->meta_inode)) {
3724                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3725                 err = PTR_ERR(sbi->meta_inode);
3726                 goto free_page_array_cache;
3727         }
3728
3729         err = f2fs_get_valid_checkpoint(sbi);
3730         if (err) {
3731                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3732                 goto free_meta_inode;
3733         }
3734
3735         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3736                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3737         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3738                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3739                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3740         }
3741
3742         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3743                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3744
3745         /* Initialize device list */
3746         err = f2fs_scan_devices(sbi);
3747         if (err) {
3748                 f2fs_err(sbi, "Failed to find devices");
3749                 goto free_devices;
3750         }
3751
3752         err = f2fs_init_post_read_wq(sbi);
3753         if (err) {
3754                 f2fs_err(sbi, "Failed to initialize post read workqueue");
3755                 goto free_devices;
3756         }
3757
3758         sbi->total_valid_node_count =
3759                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3760         percpu_counter_set(&sbi->total_valid_inode_count,
3761                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3762         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3763         sbi->total_valid_block_count =
3764                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3765         sbi->last_valid_block_count = sbi->total_valid_block_count;
3766         sbi->reserved_blocks = 0;
3767         sbi->current_reserved_blocks = 0;
3768         limit_reserve_root(sbi);
3769         adjust_unusable_cap_perc(sbi);
3770
3771         for (i = 0; i < NR_INODE_TYPE; i++) {
3772                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3773                 spin_lock_init(&sbi->inode_lock[i]);
3774         }
3775         mutex_init(&sbi->flush_lock);
3776
3777         f2fs_init_extent_cache_info(sbi);
3778
3779         f2fs_init_ino_entry_info(sbi);
3780
3781         f2fs_init_fsync_node_info(sbi);
3782
3783         /* setup f2fs internal modules */
3784         err = f2fs_build_segment_manager(sbi);
3785         if (err) {
3786                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3787                          err);
3788                 goto free_sm;
3789         }
3790         err = f2fs_build_node_manager(sbi);
3791         if (err) {
3792                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3793                          err);
3794                 goto free_nm;
3795         }
3796
3797         err = adjust_reserved_segment(sbi);
3798         if (err)
3799                 goto free_nm;
3800
3801         /* For write statistics */
3802         if (sb->s_bdev->bd_part)
3803                 sbi->sectors_written_start =
3804                         (u64)part_stat_read(sb->s_bdev->bd_part,
3805                                             sectors[STAT_WRITE]);
3806
3807         /* Read accumulated write IO statistics if exists */
3808         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3809         if (__exist_node_summaries(sbi))
3810                 sbi->kbytes_written =
3811                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3812
3813         f2fs_build_gc_manager(sbi);
3814
3815         err = f2fs_build_stats(sbi);
3816         if (err)
3817                 goto free_nm;
3818
3819         /* get an inode for node space */
3820         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3821         if (IS_ERR(sbi->node_inode)) {
3822                 f2fs_err(sbi, "Failed to read node inode");
3823                 err = PTR_ERR(sbi->node_inode);
3824                 goto free_stats;
3825         }
3826
3827         /* read root inode and dentry */
3828         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3829         if (IS_ERR(root)) {
3830                 f2fs_err(sbi, "Failed to read root inode");
3831                 err = PTR_ERR(root);
3832                 goto free_node_inode;
3833         }
3834         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3835                         !root->i_size || !root->i_nlink) {
3836                 iput(root);
3837                 err = -EINVAL;
3838                 goto free_node_inode;
3839         }
3840
3841         sb->s_root = d_make_root(root); /* allocate root dentry */
3842         if (!sb->s_root) {
3843                 err = -ENOMEM;
3844                 goto free_node_inode;
3845         }
3846
3847         err = f2fs_register_sysfs(sbi);
3848         if (err)
3849                 goto free_root_inode;
3850
3851 #ifdef CONFIG_QUOTA
3852         /* Enable quota usage during mount */
3853         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3854                 err = f2fs_enable_quotas(sb);
3855                 if (err)
3856                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3857         }
3858 #endif
3859         /* if there are any orphan inodes, free them */
3860         err = f2fs_recover_orphan_inodes(sbi);
3861         if (err)
3862                 goto free_meta;
3863
3864         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3865                 goto reset_checkpoint;
3866
3867         /* recover fsynced data */
3868         if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3869                         !test_opt(sbi, NORECOVERY)) {
3870                 /*
3871                  * mount should be failed, when device has readonly mode, and
3872                  * previous checkpoint was not done by clean system shutdown.
3873                  */
3874                 if (f2fs_hw_is_readonly(sbi)) {
3875                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3876                                 err = -EROFS;
3877                                 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3878                                 goto free_meta;
3879                         }
3880                         f2fs_info(sbi, "write access unavailable, skipping recovery");
3881                         goto reset_checkpoint;
3882                 }
3883
3884                 if (need_fsck)
3885                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3886
3887                 if (skip_recovery)
3888                         goto reset_checkpoint;
3889
3890                 err = f2fs_recover_fsync_data(sbi, false);
3891                 if (err < 0) {
3892                         if (err != -ENOMEM)
3893                                 skip_recovery = true;
3894                         need_fsck = true;
3895                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3896                                  err);
3897                         goto free_meta;
3898                 }
3899         } else {
3900                 err = f2fs_recover_fsync_data(sbi, true);
3901
3902                 if (!f2fs_readonly(sb) && err > 0) {
3903                         err = -EINVAL;
3904                         f2fs_err(sbi, "Need to recover fsync data");
3905                         goto free_meta;
3906                 }
3907         }
3908
3909         /*
3910          * If the f2fs is not readonly and fsync data recovery succeeds,
3911          * check zoned block devices' write pointer consistency.
3912          */
3913         if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3914                 err = f2fs_check_write_pointer(sbi);
3915                 if (err)
3916                         goto free_meta;
3917         }
3918
3919 reset_checkpoint:
3920         f2fs_init_inmem_curseg(sbi);
3921
3922         /* f2fs_recover_fsync_data() cleared this already */
3923         clear_sbi_flag(sbi, SBI_POR_DOING);
3924
3925         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3926                 err = f2fs_disable_checkpoint(sbi);
3927                 if (err)
3928                         goto sync_free_meta;
3929         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3930                 f2fs_enable_checkpoint(sbi);
3931         }
3932
3933         /*
3934          * If filesystem is not mounted as read-only then
3935          * do start the gc_thread.
3936          */
3937         if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) {
3938                 /* After POR, we can run background GC thread.*/
3939                 err = f2fs_start_gc_thread(sbi);
3940                 if (err)
3941                         goto sync_free_meta;
3942         }
3943         kvfree(options);
3944
3945         /* recover broken superblock */
3946         if (recovery) {
3947                 err = f2fs_commit_super(sbi, true);
3948                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3949                           sbi->valid_super_block ? 1 : 2, err);
3950         }
3951
3952         f2fs_join_shrinker(sbi);
3953
3954         f2fs_tuning_parameters(sbi);
3955
3956         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3957                     cur_cp_version(F2FS_CKPT(sbi)));
3958         f2fs_update_time(sbi, CP_TIME);
3959         f2fs_update_time(sbi, REQ_TIME);
3960         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3961         return 0;
3962
3963 sync_free_meta:
3964         /* safe to flush all the data */
3965         sync_filesystem(sbi->sb);
3966         retry_cnt = 0;
3967
3968 free_meta:
3969 #ifdef CONFIG_QUOTA
3970         f2fs_truncate_quota_inode_pages(sb);
3971         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3972                 f2fs_quota_off_umount(sbi->sb);
3973 #endif
3974         /*
3975          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3976          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3977          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3978          * falls into an infinite loop in f2fs_sync_meta_pages().
3979          */
3980         truncate_inode_pages_final(META_MAPPING(sbi));
3981         /* evict some inodes being cached by GC */
3982         evict_inodes(sb);
3983         f2fs_unregister_sysfs(sbi);
3984 free_root_inode:
3985         dput(sb->s_root);
3986         sb->s_root = NULL;
3987 free_node_inode:
3988         f2fs_release_ino_entry(sbi, true);
3989         truncate_inode_pages_final(NODE_MAPPING(sbi));
3990         iput(sbi->node_inode);
3991         sbi->node_inode = NULL;
3992 free_stats:
3993         f2fs_destroy_stats(sbi);
3994 free_nm:
3995         f2fs_destroy_node_manager(sbi);
3996 free_sm:
3997         f2fs_destroy_segment_manager(sbi);
3998         f2fs_destroy_post_read_wq(sbi);
3999 free_devices:
4000         destroy_device_list(sbi);
4001         kvfree(sbi->ckpt);
4002 free_meta_inode:
4003         make_bad_inode(sbi->meta_inode);
4004         iput(sbi->meta_inode);
4005         sbi->meta_inode = NULL;
4006 free_page_array_cache:
4007         f2fs_destroy_page_array_cache(sbi);
4008 free_xattr_cache:
4009         f2fs_destroy_xattr_caches(sbi);
4010 free_io_dummy:
4011         mempool_destroy(sbi->write_io_dummy);
4012 free_percpu:
4013         destroy_percpu_info(sbi);
4014 free_bio_info:
4015         for (i = 0; i < NR_PAGE_TYPE; i++)
4016                 kvfree(sbi->write_io[i]);
4017
4018 #ifdef CONFIG_UNICODE
4019         utf8_unload(sb->s_encoding);
4020         sb->s_encoding = NULL;
4021 #endif
4022 free_options:
4023 #ifdef CONFIG_QUOTA
4024         for (i = 0; i < MAXQUOTAS; i++)
4025                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4026 #endif
4027         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4028         kvfree(options);
4029 free_sb_buf:
4030         kfree(raw_super);
4031 free_sbi:
4032         if (sbi->s_chksum_driver)
4033                 crypto_free_shash(sbi->s_chksum_driver);
4034         kfree(sbi);
4035
4036         /* give only one another chance */
4037         if (retry_cnt > 0 && skip_recovery) {
4038                 retry_cnt--;
4039                 shrink_dcache_sb(sb);
4040                 goto try_onemore;
4041         }
4042         return err;
4043 }
4044
4045 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4046                         const char *dev_name, void *data)
4047 {
4048         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4049 }
4050
4051 static void kill_f2fs_super(struct super_block *sb)
4052 {
4053         if (sb->s_root) {
4054                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4055
4056                 set_sbi_flag(sbi, SBI_IS_CLOSE);
4057                 f2fs_stop_gc_thread(sbi);
4058                 f2fs_stop_discard_thread(sbi);
4059
4060                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4061                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4062                         struct cp_control cpc = {
4063                                 .reason = CP_UMOUNT,
4064                         };
4065                         f2fs_write_checkpoint(sbi, &cpc);
4066                 }
4067
4068                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4069                         sb->s_flags &= ~SB_RDONLY;
4070         }
4071         kill_block_super(sb);
4072 }
4073
4074 static struct file_system_type f2fs_fs_type = {
4075         .owner          = THIS_MODULE,
4076         .name           = "f2fs",
4077         .mount          = f2fs_mount,
4078         .kill_sb        = kill_f2fs_super,
4079         .fs_flags       = FS_REQUIRES_DEV,
4080 };
4081 MODULE_ALIAS_FS("f2fs");
4082
4083 static int __init init_inodecache(void)
4084 {
4085         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4086                         sizeof(struct f2fs_inode_info), 0,
4087                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4088         if (!f2fs_inode_cachep)
4089                 return -ENOMEM;
4090         return 0;
4091 }
4092
4093 static void destroy_inodecache(void)
4094 {
4095         /*
4096          * Make sure all delayed rcu free inodes are flushed before we
4097          * destroy cache.
4098          */
4099         rcu_barrier();
4100         kmem_cache_destroy(f2fs_inode_cachep);
4101 }
4102
4103 static int __init init_f2fs_fs(void)
4104 {
4105         int err;
4106
4107         if (PAGE_SIZE != F2FS_BLKSIZE) {
4108                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4109                                 PAGE_SIZE, F2FS_BLKSIZE);
4110                 return -EINVAL;
4111         }
4112
4113         f2fs_build_trace_ios();
4114
4115         err = init_inodecache();
4116         if (err)
4117                 goto fail;
4118         err = f2fs_create_node_manager_caches();
4119         if (err)
4120                 goto free_inodecache;
4121         err = f2fs_create_segment_manager_caches();
4122         if (err)
4123                 goto free_node_manager_caches;
4124         err = f2fs_create_checkpoint_caches();
4125         if (err)
4126                 goto free_segment_manager_caches;
4127         err = f2fs_create_recovery_cache();
4128         if (err)
4129                 goto free_checkpoint_caches;
4130         err = f2fs_create_extent_cache();
4131         if (err)
4132                 goto free_recovery_cache;
4133         err = f2fs_create_garbage_collection_cache();
4134         if (err)
4135                 goto free_extent_cache;
4136         err = f2fs_init_sysfs();
4137         if (err)
4138                 goto free_garbage_collection_cache;
4139         err = register_shrinker(&f2fs_shrinker_info);
4140         if (err)
4141                 goto free_sysfs;
4142         err = register_filesystem(&f2fs_fs_type);
4143         if (err)
4144                 goto free_shrinker;
4145         f2fs_create_root_stats();
4146         err = f2fs_init_post_read_processing();
4147         if (err)
4148                 goto free_root_stats;
4149         err = f2fs_init_bio_entry_cache();
4150         if (err)
4151                 goto free_post_read;
4152         err = f2fs_init_bioset();
4153         if (err)
4154                 goto free_bio_enrty_cache;
4155         err = f2fs_init_compress_mempool();
4156         if (err)
4157                 goto free_bioset;
4158         err = f2fs_init_compress_cache();
4159         if (err)
4160                 goto free_compress_mempool;
4161         return 0;
4162 free_compress_mempool:
4163         f2fs_destroy_compress_mempool();
4164 free_bioset:
4165         f2fs_destroy_bioset();
4166 free_bio_enrty_cache:
4167         f2fs_destroy_bio_entry_cache();
4168 free_post_read:
4169         f2fs_destroy_post_read_processing();
4170 free_root_stats:
4171         f2fs_destroy_root_stats();
4172         unregister_filesystem(&f2fs_fs_type);
4173 free_shrinker:
4174         unregister_shrinker(&f2fs_shrinker_info);
4175 free_sysfs:
4176         f2fs_exit_sysfs();
4177 free_garbage_collection_cache:
4178         f2fs_destroy_garbage_collection_cache();
4179 free_extent_cache:
4180         f2fs_destroy_extent_cache();
4181 free_recovery_cache:
4182         f2fs_destroy_recovery_cache();
4183 free_checkpoint_caches:
4184         f2fs_destroy_checkpoint_caches();
4185 free_segment_manager_caches:
4186         f2fs_destroy_segment_manager_caches();
4187 free_node_manager_caches:
4188         f2fs_destroy_node_manager_caches();
4189 free_inodecache:
4190         destroy_inodecache();
4191 fail:
4192         return err;
4193 }
4194
4195 static void __exit exit_f2fs_fs(void)
4196 {
4197         f2fs_destroy_compress_cache();
4198         f2fs_destroy_compress_mempool();
4199         f2fs_destroy_bioset();
4200         f2fs_destroy_bio_entry_cache();
4201         f2fs_destroy_post_read_processing();
4202         f2fs_destroy_root_stats();
4203         unregister_filesystem(&f2fs_fs_type);
4204         unregister_shrinker(&f2fs_shrinker_info);
4205         f2fs_exit_sysfs();
4206         f2fs_destroy_garbage_collection_cache();
4207         f2fs_destroy_extent_cache();
4208         f2fs_destroy_recovery_cache();
4209         f2fs_destroy_checkpoint_caches();
4210         f2fs_destroy_segment_manager_caches();
4211         f2fs_destroy_node_manager_caches();
4212         destroy_inodecache();
4213         f2fs_destroy_trace_ios();
4214 }
4215
4216 module_init(init_f2fs_fs)
4217 module_exit(exit_f2fs_fs)
4218
4219 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4220 MODULE_DESCRIPTION("Flash Friendly File System");
4221 MODULE_LICENSE("GPL");
4222 MODULE_SOFTDEP("pre: crc32");
4223