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
  28 #include "f2fs.h"
  29 #include "node.h"
  30 #include "segment.h"
  31 #include "xattr.h"
  32 #include "gc.h"
  33 #include "trace.h"
  34
  35 #define CREATE_TRACE_POINTS
  36 #include <trace/events/f2fs.h>
  37
  38 static struct kmem_cache *f2fs_inode_cachep;
  39
  40 #ifdef CONFIG_F2FS_FAULT_INJECTION
  41
  42 const char *f2fs_fault_name[FAULT_MAX] = {
  43         [FAULT_KMALLOC]         = "kmalloc",
  44         [FAULT_KVMALLOC]        = "kvmalloc",
  45         [FAULT_PAGE_ALLOC]      = "page alloc",
  46         [FAULT_PAGE_GET]        = "page get",
  47         [FAULT_ALLOC_BIO]       = "alloc bio",
  48         [FAULT_ALLOC_NID]       = "alloc nid",
  49         [FAULT_ORPHAN]          = "orphan",
  50         [FAULT_BLOCK]           = "no more block",
  51         [FAULT_DIR_DEPTH]       = "too big dir depth",
  52         [FAULT_EVICT_INODE]     = "evict_inode fail",
  53         [FAULT_TRUNCATE]        = "truncate fail",
  54         [FAULT_READ_IO]         = "read IO error",
  55         [FAULT_CHECKPOINT]      = "checkpoint error",
  56         [FAULT_DISCARD]         = "discard error",
  57         [FAULT_WRITE_IO]        = "write IO error",
  58 };
  59
  60 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  61                                                         unsigned int type)
  62 {
  63         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  64
  65         if (rate) {
  66                 atomic_set(&ffi->inject_ops, 0);
  67                 ffi->inject_rate = rate;
  68         }
  69
  70         if (type)
  71                 ffi->inject_type = type;
  72
  73         if (!rate && !type)
  74                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  75 }
  76 #endif
  77
  78 /* f2fs-wide shrinker description */
  79 static struct shrinker f2fs_shrinker_info = {
  80         .scan_objects = f2fs_shrink_scan,
  81         .count_objects = f2fs_shrink_count,
  82         .seeks = DEFAULT_SEEKS,
  83 };
  84
  85 enum {
  86         Opt_gc_background,
  87         Opt_disable_roll_forward,
  88         Opt_norecovery,
  89         Opt_discard,
  90         Opt_nodiscard,
  91         Opt_noheap,
  92         Opt_heap,
  93         Opt_user_xattr,
  94         Opt_nouser_xattr,
  95         Opt_acl,
  96         Opt_noacl,
  97         Opt_active_logs,
  98         Opt_disable_ext_identify,
  99         Opt_inline_xattr,
 100         Opt_noinline_xattr,
 101         Opt_inline_xattr_size,
 102         Opt_inline_data,
 103         Opt_inline_dentry,
 104         Opt_noinline_dentry,
 105         Opt_flush_merge,
 106         Opt_noflush_merge,
 107         Opt_nobarrier,
 108         Opt_fastboot,
 109         Opt_extent_cache,
 110         Opt_noextent_cache,
 111         Opt_noinline_data,
 112         Opt_data_flush,
 113         Opt_reserve_root,
 114         Opt_resgid,
 115         Opt_resuid,
 116         Opt_mode,
 117         Opt_io_size_bits,
 118         Opt_fault_injection,
 119         Opt_fault_type,
 120         Opt_lazytime,
 121         Opt_nolazytime,
 122         Opt_quota,
 123         Opt_noquota,
 124         Opt_usrquota,
 125         Opt_grpquota,
 126         Opt_prjquota,
 127         Opt_usrjquota,
 128         Opt_grpjquota,
 129         Opt_prjjquota,
 130         Opt_offusrjquota,
 131         Opt_offgrpjquota,
 132         Opt_offprjjquota,
 133         Opt_jqfmt_vfsold,
 134         Opt_jqfmt_vfsv0,
 135         Opt_jqfmt_vfsv1,
 136         Opt_whint,
 137         Opt_alloc,
 138         Opt_fsync,
 139         Opt_test_dummy_encryption,
 140         Opt_checkpoint_disable,
 141         Opt_checkpoint_disable_cap,
 142         Opt_checkpoint_disable_cap_perc,
 143         Opt_checkpoint_enable,
 144         Opt_err,
 145 };
 146
 147 static match_table_t f2fs_tokens = {
 148         {Opt_gc_background, "background_gc=%s"},
 149         {Opt_disable_roll_forward, "disable_roll_forward"},
 150         {Opt_norecovery, "norecovery"},
 151         {Opt_discard, "discard"},
 152         {Opt_nodiscard, "nodiscard"},
 153         {Opt_noheap, "no_heap"},
 154         {Opt_heap, "heap"},
 155         {Opt_user_xattr, "user_xattr"},
 156         {Opt_nouser_xattr, "nouser_xattr"},
 157         {Opt_acl, "acl"},
 158         {Opt_noacl, "noacl"},
 159         {Opt_active_logs, "active_logs=%u"},
 160         {Opt_disable_ext_identify, "disable_ext_identify"},
 161         {Opt_inline_xattr, "inline_xattr"},
 162         {Opt_noinline_xattr, "noinline_xattr"},
 163         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 164         {Opt_inline_data, "inline_data"},
 165         {Opt_inline_dentry, "inline_dentry"},
 166         {Opt_noinline_dentry, "noinline_dentry"},
 167         {Opt_flush_merge, "flush_merge"},
 168         {Opt_noflush_merge, "noflush_merge"},
 169         {Opt_nobarrier, "nobarrier"},
 170         {Opt_fastboot, "fastboot"},
 171         {Opt_extent_cache, "extent_cache"},
 172         {Opt_noextent_cache, "noextent_cache"},
 173         {Opt_noinline_data, "noinline_data"},
 174         {Opt_data_flush, "data_flush"},
 175         {Opt_reserve_root, "reserve_root=%u"},
 176         {Opt_resgid, "resgid=%u"},
 177         {Opt_resuid, "resuid=%u"},
 178         {Opt_mode, "mode=%s"},
 179         {Opt_io_size_bits, "io_bits=%u"},
 180         {Opt_fault_injection, "fault_injection=%u"},
 181         {Opt_fault_type, "fault_type=%u"},
 182         {Opt_lazytime, "lazytime"},
 183         {Opt_nolazytime, "nolazytime"},
 184         {Opt_quota, "quota"},
 185         {Opt_noquota, "noquota"},
 186         {Opt_usrquota, "usrquota"},
 187         {Opt_grpquota, "grpquota"},
 188         {Opt_prjquota, "prjquota"},
 189         {Opt_usrjquota, "usrjquota=%s"},
 190         {Opt_grpjquota, "grpjquota=%s"},
 191         {Opt_prjjquota, "prjjquota=%s"},
 192         {Opt_offusrjquota, "usrjquota="},
 193         {Opt_offgrpjquota, "grpjquota="},
 194         {Opt_offprjjquota, "prjjquota="},
 195         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 196         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 197         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 198         {Opt_whint, "whint_mode=%s"},
 199         {Opt_alloc, "alloc_mode=%s"},
 200         {Opt_fsync, "fsync_mode=%s"},
 201         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 202         {Opt_checkpoint_disable, "checkpoint=disable"},
 203         {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 204         {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 205         {Opt_checkpoint_enable, "checkpoint=enable"},
 206         {Opt_err, NULL},
 207 };
 208
 209 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 210 {
 211         struct va_format vaf;
 212         va_list args;
 213         int level;
 214
 215         va_start(args, fmt);
 216
 217         level = printk_get_level(fmt);
 218         vaf.fmt = printk_skip_level(fmt);
 219         vaf.va = &args;
 220         printk("%c%cF2FS-fs (%s): %pV\n",
 221                KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 222
 223         va_end(args);
 224 }
 225
 226 #ifdef CONFIG_UNICODE
 227 static const struct f2fs_sb_encodings {
 228         __u16 magic;
 229         char *name;
 230         char *version;
 231 } f2fs_sb_encoding_map[] = {
 232         {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
 233 };
 234
 235 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
 236                                  const struct f2fs_sb_encodings **encoding,
 237                                  __u16 *flags)
 238 {
 239         __u16 magic = le16_to_cpu(sb->s_encoding);
 240         int i;
 241
 242         for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
 243                 if (magic == f2fs_sb_encoding_map[i].magic)
 244                         break;
 245
 246         if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
 247                 return -EINVAL;
 248
 249         *encoding = &f2fs_sb_encoding_map[i];
 250         *flags = le16_to_cpu(sb->s_encoding_flags);
 251
 252         return 0;
 253 }
 254 #endif
 255
 256 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 257 {
 258         block_t limit = min((sbi->user_block_count << 1) / 1000,
 259                         sbi->user_block_count - sbi->reserved_blocks);
 260
 261         /* limit is 0.2% */
 262         if (test_opt(sbi, RESERVE_ROOT) &&
 263                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 264                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 265                 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 266                           F2FS_OPTION(sbi).root_reserved_blocks);
 267         }
 268         if (!test_opt(sbi, RESERVE_ROOT) &&
 269                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 270                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 271                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 272                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 273                 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 274                           from_kuid_munged(&init_user_ns,
 275                                            F2FS_OPTION(sbi).s_resuid),
 276                           from_kgid_munged(&init_user_ns,
 277                                            F2FS_OPTION(sbi).s_resgid));
 278 }
 279
 280 static void init_once(void *foo)
 281 {
 282         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 283
 284         inode_init_once(&fi->vfs_inode);
 285 }
 286
 287 #ifdef CONFIG_QUOTA
 288 static const char * const quotatypes[] = INITQFNAMES;
 289 #define QTYPE2NAME(t) (quotatypes[t])
 290 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 291                                                         substring_t *args)
 292 {
 293         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 294         char *qname;
 295         int ret = -EINVAL;
 296
 297         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 298                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 299                 return -EINVAL;
 300         }
 301         if (f2fs_sb_has_quota_ino(sbi)) {
 302                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 303                 return 0;
 304         }
 305
 306         qname = match_strdup(args);
 307         if (!qname) {
 308                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 309                 return -ENOMEM;
 310         }
 311         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 312                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 313                         ret = 0;
 314                 else
 315                         f2fs_err(sbi, "%s quota file already specified",
 316                                  QTYPE2NAME(qtype));
 317                 goto errout;
 318         }
 319         if (strchr(qname, '/')) {
 320                 f2fs_err(sbi, "quotafile must be on filesystem root");
 321                 goto errout;
 322         }
 323         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 324         set_opt(sbi, QUOTA);
 325         return 0;
 326 errout:
 327         kvfree(qname);
 328         return ret;
 329 }
 330
 331 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 332 {
 333         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 334
 335         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 336                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 337                 return -EINVAL;
 338         }
 339         kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 340         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 341         return 0;
 342 }
 343
 344 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 345 {
 346         /*
 347          * We do the test below only for project quotas. 'usrquota' and
 348          * 'grpquota' mount options are allowed even without quota feature
 349          * to support legacy quotas in quota files.
 350          */
 351         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 352                 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 353                 return -1;
 354         }
 355         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 356                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 357                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 358                 if (test_opt(sbi, USRQUOTA) &&
 359                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 360                         clear_opt(sbi, USRQUOTA);
 361
 362                 if (test_opt(sbi, GRPQUOTA) &&
 363                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 364                         clear_opt(sbi, GRPQUOTA);
 365
 366                 if (test_opt(sbi, PRJQUOTA) &&
 367                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 368                         clear_opt(sbi, PRJQUOTA);
 369
 370                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 371                                 test_opt(sbi, PRJQUOTA)) {
 372                         f2fs_err(sbi, "old and new quota format mixing");
 373                         return -1;
 374                 }
 375
 376                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 377                         f2fs_err(sbi, "journaled quota format not specified");
 378                         return -1;
 379                 }
 380         }
 381
 382         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 383                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 384                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 385         }
 386         return 0;
 387 }
 388 #endif
 389
 390 static int parse_options(struct super_block *sb, char *options)
 391 {
 392         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 393         substring_t args[MAX_OPT_ARGS];
 394         char *p, *name;
 395         int arg = 0;
 396         kuid_t uid;
 397         kgid_t gid;
 398 #ifdef CONFIG_QUOTA
 399         int ret;
 400 #endif
 401
 402         if (!options)
 403                 return 0;
 404
 405         while ((p = strsep(&options, ",")) != NULL) {
 406                 int token;
 407                 if (!*p)
 408                         continue;
 409                 /*
 410                  * Initialize args struct so we know whether arg was
 411                  * found; some options take optional arguments.
 412                  */
 413                 args[0].to = args[0].from = NULL;
 414                 token = match_token(p, f2fs_tokens, args);
 415
 416                 switch (token) {
 417                 case Opt_gc_background:
 418                         name = match_strdup(&args[0]);
 419
 420                         if (!name)
 421                                 return -ENOMEM;
 422                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 423                                 set_opt(sbi, BG_GC);
 424                                 clear_opt(sbi, FORCE_FG_GC);
 425                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 426                                 clear_opt(sbi, BG_GC);
 427                                 clear_opt(sbi, FORCE_FG_GC);
 428                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 429                                 set_opt(sbi, BG_GC);
 430                                 set_opt(sbi, FORCE_FG_GC);
 431                         } else {
 432                                 kvfree(name);
 433                                 return -EINVAL;
 434                         }
 435                         kvfree(name);
 436                         break;
 437                 case Opt_disable_roll_forward:
 438                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 439                         break;
 440                 case Opt_norecovery:
 441                         /* this option mounts f2fs with ro */
 442                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 443                         if (!f2fs_readonly(sb))
 444                                 return -EINVAL;
 445                         break;
 446                 case Opt_discard:
 447                         set_opt(sbi, DISCARD);
 448                         break;
 449                 case Opt_nodiscard:
 450                         if (f2fs_sb_has_blkzoned(sbi)) {
 451                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 452                                 return -EINVAL;
 453                         }
 454                         clear_opt(sbi, DISCARD);
 455                         break;
 456                 case Opt_noheap:
 457                         set_opt(sbi, NOHEAP);
 458                         break;
 459                 case Opt_heap:
 460                         clear_opt(sbi, NOHEAP);
 461                         break;
 462 #ifdef CONFIG_F2FS_FS_XATTR
 463                 case Opt_user_xattr:
 464                         set_opt(sbi, XATTR_USER);
 465                         break;
 466                 case Opt_nouser_xattr:
 467                         clear_opt(sbi, XATTR_USER);
 468                         break;
 469                 case Opt_inline_xattr:
 470                         set_opt(sbi, INLINE_XATTR);
 471                         break;
 472                 case Opt_noinline_xattr:
 473                         clear_opt(sbi, INLINE_XATTR);
 474                         break;
 475                 case Opt_inline_xattr_size:
 476                         if (args->from && match_int(args, &arg))
 477                                 return -EINVAL;
 478                         set_opt(sbi, INLINE_XATTR_SIZE);
 479                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 480                         break;
 481 #else
 482                 case Opt_user_xattr:
 483                         f2fs_info(sbi, "user_xattr options not supported");
 484                         break;
 485                 case Opt_nouser_xattr:
 486                         f2fs_info(sbi, "nouser_xattr options not supported");
 487                         break;
 488                 case Opt_inline_xattr:
 489                         f2fs_info(sbi, "inline_xattr options not supported");
 490                         break;
 491                 case Opt_noinline_xattr:
 492                         f2fs_info(sbi, "noinline_xattr options not supported");
 493                         break;
 494 #endif
 495 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 496                 case Opt_acl:
 497                         set_opt(sbi, POSIX_ACL);
 498                         break;
 499                 case Opt_noacl:
 500                         clear_opt(sbi, POSIX_ACL);
 501                         break;
 502 #else
 503                 case Opt_acl:
 504                         f2fs_info(sbi, "acl options not supported");
 505                         break;
 506                 case Opt_noacl:
 507                         f2fs_info(sbi, "noacl options not supported");
 508                         break;
 509 #endif
 510                 case Opt_active_logs:
 511                         if (args->from && match_int(args, &arg))
 512                                 return -EINVAL;
 513                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 514                                 return -EINVAL;
 515                         F2FS_OPTION(sbi).active_logs = arg;
 516                         break;
 517                 case Opt_disable_ext_identify:
 518                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 519                         break;
 520                 case Opt_inline_data:
 521                         set_opt(sbi, INLINE_DATA);
 522                         break;
 523                 case Opt_inline_dentry:
 524                         set_opt(sbi, INLINE_DENTRY);
 525                         break;
 526                 case Opt_noinline_dentry:
 527                         clear_opt(sbi, INLINE_DENTRY);
 528                         break;
 529                 case Opt_flush_merge:
 530                         set_opt(sbi, FLUSH_MERGE);
 531                         break;
 532                 case Opt_noflush_merge:
 533                         clear_opt(sbi, FLUSH_MERGE);
 534                         break;
 535                 case Opt_nobarrier:
 536                         set_opt(sbi, NOBARRIER);
 537                         break;
 538                 case Opt_fastboot:
 539                         set_opt(sbi, FASTBOOT);
 540                         break;
 541                 case Opt_extent_cache:
 542                         set_opt(sbi, EXTENT_CACHE);
 543                         break;
 544                 case Opt_noextent_cache:
 545                         clear_opt(sbi, EXTENT_CACHE);
 546                         break;
 547                 case Opt_noinline_data:
 548                         clear_opt(sbi, INLINE_DATA);
 549                         break;
 550                 case Opt_data_flush:
 551                         set_opt(sbi, DATA_FLUSH);
 552                         break;
 553                 case Opt_reserve_root:
 554                         if (args->from && match_int(args, &arg))
 555                                 return -EINVAL;
 556                         if (test_opt(sbi, RESERVE_ROOT)) {
 557                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 558                                           F2FS_OPTION(sbi).root_reserved_blocks);
 559                         } else {
 560                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 561                                 set_opt(sbi, RESERVE_ROOT);
 562                         }
 563                         break;
 564                 case Opt_resuid:
 565                         if (args->from && match_int(args, &arg))
 566                                 return -EINVAL;
 567                         uid = make_kuid(current_user_ns(), arg);
 568                         if (!uid_valid(uid)) {
 569                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 570                                 return -EINVAL;
 571                         }
 572                         F2FS_OPTION(sbi).s_resuid = uid;
 573                         break;
 574                 case Opt_resgid:
 575                         if (args->from && match_int(args, &arg))
 576                                 return -EINVAL;
 577                         gid = make_kgid(current_user_ns(), arg);
 578                         if (!gid_valid(gid)) {
 579                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 580                                 return -EINVAL;
 581                         }
 582                         F2FS_OPTION(sbi).s_resgid = gid;
 583                         break;
 584                 case Opt_mode:
 585                         name = match_strdup(&args[0]);
 586
 587                         if (!name)
 588                                 return -ENOMEM;
 589                         if (strlen(name) == 8 &&
 590                                         !strncmp(name, "adaptive", 8)) {
 591                                 if (f2fs_sb_has_blkzoned(sbi)) {
 592                                         f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
 593                                         kvfree(name);
 594                                         return -EINVAL;
 595                                 }
 596                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 597                         } else if (strlen(name) == 3 &&
 598                                         !strncmp(name, "lfs", 3)) {
 599                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 600                         } else {
 601                                 kvfree(name);
 602                                 return -EINVAL;
 603                         }
 604                         kvfree(name);
 605                         break;
 606                 case Opt_io_size_bits:
 607                         if (args->from && match_int(args, &arg))
 608                                 return -EINVAL;
 609                         if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
 610                                 f2fs_warn(sbi, "Not support %d, larger than %d",
 611                                           1 << arg, BIO_MAX_PAGES);
 612                                 return -EINVAL;
 613                         }
 614                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 615                         break;
 616 #ifdef CONFIG_F2FS_FAULT_INJECTION
 617                 case Opt_fault_injection:
 618                         if (args->from && match_int(args, &arg))
 619                                 return -EINVAL;
 620                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 621                         set_opt(sbi, FAULT_INJECTION);
 622                         break;
 623
 624                 case Opt_fault_type:
 625                         if (args->from && match_int(args, &arg))
 626                                 return -EINVAL;
 627                         f2fs_build_fault_attr(sbi, 0, arg);
 628                         set_opt(sbi, FAULT_INJECTION);
 629                         break;
 630 #else
 631                 case Opt_fault_injection:
 632                         f2fs_info(sbi, "fault_injection options not supported");
 633                         break;
 634
 635                 case Opt_fault_type:
 636                         f2fs_info(sbi, "fault_type options not supported");
 637                         break;
 638 #endif
 639                 case Opt_lazytime:
 640                         sb->s_flags |= SB_LAZYTIME;
 641                         break;
 642                 case Opt_nolazytime:
 643                         sb->s_flags &= ~SB_LAZYTIME;
 644                         break;
 645 #ifdef CONFIG_QUOTA
 646                 case Opt_quota:
 647                 case Opt_usrquota:
 648                         set_opt(sbi, USRQUOTA);
 649                         break;
 650                 case Opt_grpquota:
 651                         set_opt(sbi, GRPQUOTA);
 652                         break;
 653                 case Opt_prjquota:
 654                         set_opt(sbi, PRJQUOTA);
 655                         break;
 656                 case Opt_usrjquota:
 657                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 658                         if (ret)
 659                                 return ret;
 660                         break;
 661                 case Opt_grpjquota:
 662                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 663                         if (ret)
 664                                 return ret;
 665                         break;
 666                 case Opt_prjjquota:
 667                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 668                         if (ret)
 669                                 return ret;
 670                         break;
 671                 case Opt_offusrjquota:
 672                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 673                         if (ret)
 674                                 return ret;
 675                         break;
 676                 case Opt_offgrpjquota:
 677                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 678                         if (ret)
 679                                 return ret;
 680                         break;
 681                 case Opt_offprjjquota:
 682                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 683                         if (ret)
 684                                 return ret;
 685                         break;
 686                 case Opt_jqfmt_vfsold:
 687                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 688                         break;
 689                 case Opt_jqfmt_vfsv0:
 690                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 691                         break;
 692                 case Opt_jqfmt_vfsv1:
 693                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 694                         break;
 695                 case Opt_noquota:
 696                         clear_opt(sbi, QUOTA);
 697                         clear_opt(sbi, USRQUOTA);
 698                         clear_opt(sbi, GRPQUOTA);
 699                         clear_opt(sbi, PRJQUOTA);
 700                         break;
 701 #else
 702                 case Opt_quota:
 703                 case Opt_usrquota:
 704                 case Opt_grpquota:
 705                 case Opt_prjquota:
 706                 case Opt_usrjquota:
 707                 case Opt_grpjquota:
 708                 case Opt_prjjquota:
 709                 case Opt_offusrjquota:
 710                 case Opt_offgrpjquota:
 711                 case Opt_offprjjquota:
 712                 case Opt_jqfmt_vfsold:
 713                 case Opt_jqfmt_vfsv0:
 714                 case Opt_jqfmt_vfsv1:
 715                 case Opt_noquota:
 716                         f2fs_info(sbi, "quota operations not supported");
 717                         break;
 718 #endif
 719                 case Opt_whint:
 720                         name = match_strdup(&args[0]);
 721                         if (!name)
 722                                 return -ENOMEM;
 723                         if (strlen(name) == 10 &&
 724                                         !strncmp(name, "user-based", 10)) {
 725                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 726                         } else if (strlen(name) == 3 &&
 727                                         !strncmp(name, "off", 3)) {
 728                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 729                         } else if (strlen(name) == 8 &&
 730                                         !strncmp(name, "fs-based", 8)) {
 731                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 732                         } else {
 733                                 kvfree(name);
 734                                 return -EINVAL;
 735                         }
 736                         kvfree(name);
 737                         break;
 738                 case Opt_alloc:
 739                         name = match_strdup(&args[0]);
 740                         if (!name)
 741                                 return -ENOMEM;
 742
 743                         if (strlen(name) == 7 &&
 744                                         !strncmp(name, "default", 7)) {
 745                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 746                         } else if (strlen(name) == 5 &&
 747                                         !strncmp(name, "reuse", 5)) {
 748                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 749                         } else {
 750                                 kvfree(name);
 751                                 return -EINVAL;
 752                         }
 753                         kvfree(name);
 754                         break;
 755                 case Opt_fsync:
 756                         name = match_strdup(&args[0]);
 757                         if (!name)
 758                                 return -ENOMEM;
 759                         if (strlen(name) == 5 &&
 760                                         !strncmp(name, "posix", 5)) {
 761                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 762                         } else if (strlen(name) == 6 &&
 763                                         !strncmp(name, "strict", 6)) {
 764                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 765                         } else if (strlen(name) == 9 &&
 766                                         !strncmp(name, "nobarrier", 9)) {
 767                                 F2FS_OPTION(sbi).fsync_mode =
 768                                                         FSYNC_MODE_NOBARRIER;
 769                         } else {
 770                                 kvfree(name);
 771                                 return -EINVAL;
 772                         }
 773                         kvfree(name);
 774                         break;
 775                 case Opt_test_dummy_encryption:
 776 #ifdef CONFIG_FS_ENCRYPTION
 777                         if (!f2fs_sb_has_encrypt(sbi)) {
 778                                 f2fs_err(sbi, "Encrypt feature is off");
 779                                 return -EINVAL;
 780                         }
 781
 782                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 783                         f2fs_info(sbi, "Test dummy encryption mode enabled");
 784 #else
 785                         f2fs_info(sbi, "Test dummy encryption mount option ignored");
 786 #endif
 787                         break;
 788                 case Opt_checkpoint_disable_cap_perc:
 789                         if (args->from && match_int(args, &arg))
 790                                 return -EINVAL;
 791                         if (arg < 0 || arg > 100)
 792                                 return -EINVAL;
 793                         if (arg == 100)
 794                                 F2FS_OPTION(sbi).unusable_cap =
 795                                         sbi->user_block_count;
 796                         else
 797                                 F2FS_OPTION(sbi).unusable_cap =
 798                                         (sbi->user_block_count / 100) * arg;
 799                         set_opt(sbi, DISABLE_CHECKPOINT);
 800                         break;
 801                 case Opt_checkpoint_disable_cap:
 802                         if (args->from && match_int(args, &arg))
 803                                 return -EINVAL;
 804                         F2FS_OPTION(sbi).unusable_cap = arg;
 805                         set_opt(sbi, DISABLE_CHECKPOINT);
 806                         break;
 807                 case Opt_checkpoint_disable:
 808                         set_opt(sbi, DISABLE_CHECKPOINT);
 809                         break;
 810                 case Opt_checkpoint_enable:
 811                         clear_opt(sbi, DISABLE_CHECKPOINT);
 812                         break;
 813                 default:
 814                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
 815                                  p);
 816                         return -EINVAL;
 817                 }
 818         }
 819 #ifdef CONFIG_QUOTA
 820         if (f2fs_check_quota_options(sbi))
 821                 return -EINVAL;
 822 #else
 823         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
 824                 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 825                 return -EINVAL;
 826         }
 827         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
 828                 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 829                 return -EINVAL;
 830         }
 831 #endif
 832 #ifndef CONFIG_UNICODE
 833         if (f2fs_sb_has_casefold(sbi)) {
 834                 f2fs_err(sbi,
 835                         "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
 836                 return -EINVAL;
 837         }
 838 #endif
 839
 840         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 841                 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
 842                          F2FS_IO_SIZE_KB(sbi));
 843                 return -EINVAL;
 844         }
 845
 846         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 847                 int min_size, max_size;
 848
 849                 if (!f2fs_sb_has_extra_attr(sbi) ||
 850                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 851                         f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
 852                         return -EINVAL;
 853                 }
 854                 if (!test_opt(sbi, INLINE_XATTR)) {
 855                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
 856                         return -EINVAL;
 857                 }
 858
 859                 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
 860                 max_size = MAX_INLINE_XATTR_SIZE;
 861
 862                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
 863                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
 864                         f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
 865                                  min_size, max_size);
 866                         return -EINVAL;
 867                 }
 868         }
 869
 870         if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
 871                 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
 872                 return -EINVAL;
 873         }
 874
 875         /* Not pass down write hints if the number of active logs is lesser
 876          * than NR_CURSEG_TYPE.
 877          */
 878         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 879                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 880         return 0;
 881 }
 882
 883 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 884 {
 885         struct f2fs_inode_info *fi;
 886
 887         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 888         if (!fi)
 889                 return NULL;
 890
 891         init_once((void *) fi);
 892
 893         /* Initialize f2fs-specific inode info */
 894         atomic_set(&fi->dirty_pages, 0);
 895         init_rwsem(&fi->i_sem);
 896         INIT_LIST_HEAD(&fi->dirty_list);
 897         INIT_LIST_HEAD(&fi->gdirty_list);
 898         INIT_LIST_HEAD(&fi->inmem_ilist);
 899         INIT_LIST_HEAD(&fi->inmem_pages);
 900         mutex_init(&fi->inmem_lock);
 901         init_rwsem(&fi->i_gc_rwsem[READ]);
 902         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 903         init_rwsem(&fi->i_mmap_sem);
 904         init_rwsem(&fi->i_xattr_sem);
 905
 906         /* Will be used by directory only */
 907         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 908
 909         return &fi->vfs_inode;
 910 }
 911
 912 static int f2fs_drop_inode(struct inode *inode)
 913 {
 914         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 915         int ret;
 916
 917         /*
 918          * during filesystem shutdown, if checkpoint is disabled,
 919          * drop useless meta/node dirty pages.
 920          */
 921         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
 922                 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 923                         inode->i_ino == F2FS_META_INO(sbi)) {
 924                         trace_f2fs_drop_inode(inode, 1);
 925                         return 1;
 926                 }
 927         }
 928
 929         /*
 930          * This is to avoid a deadlock condition like below.
 931          * writeback_single_inode(inode)
 932          *  - f2fs_write_data_page
 933          *    - f2fs_gc -> iput -> evict
 934          *       - inode_wait_for_writeback(inode)
 935          */
 936         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 937                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 938                         /* to avoid evict_inode call simultaneously */
 939                         atomic_inc(&inode->i_count);
 940                         spin_unlock(&inode->i_lock);
 941
 942                         /* some remained atomic pages should discarded */
 943                         if (f2fs_is_atomic_file(inode))
 944                                 f2fs_drop_inmem_pages(inode);
 945
 946                         /* should remain fi->extent_tree for writepage */
 947                         f2fs_destroy_extent_node(inode);
 948
 949                         sb_start_intwrite(inode->i_sb);
 950                         f2fs_i_size_write(inode, 0);
 951
 952                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
 953                                         inode, NULL, 0, DATA);
 954                         truncate_inode_pages_final(inode->i_mapping);
 955
 956                         if (F2FS_HAS_BLOCKS(inode))
 957                                 f2fs_truncate(inode);
 958
 959                         sb_end_intwrite(inode->i_sb);
 960
 961                         spin_lock(&inode->i_lock);
 962                         atomic_dec(&inode->i_count);
 963                 }
 964                 trace_f2fs_drop_inode(inode, 0);
 965                 return 0;
 966         }
 967         ret = generic_drop_inode(inode);
 968         if (!ret)
 969                 ret = fscrypt_drop_inode(inode);
 970         trace_f2fs_drop_inode(inode, ret);
 971         return ret;
 972 }
 973
 974 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 975 {
 976         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 977         int ret = 0;
 978
 979         spin_lock(&sbi->inode_lock[DIRTY_META]);
 980         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 981                 ret = 1;
 982         } else {
 983                 set_inode_flag(inode, FI_DIRTY_INODE);
 984                 stat_inc_dirty_inode(sbi, DIRTY_META);
 985         }
 986         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 987                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 988                                 &sbi->inode_list[DIRTY_META]);
 989                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 990         }
 991         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 992         return ret;
 993 }
 994
 995 void f2fs_inode_synced(struct inode *inode)
 996 {
 997         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 998
 999         spin_lock(&sbi->inode_lock[DIRTY_META]);
1000         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1001                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1002                 return;
1003         }
1004         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1005                 list_del_init(&F2FS_I(inode)->gdirty_list);
1006                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1007         }
1008         clear_inode_flag(inode, FI_DIRTY_INODE);
1009         clear_inode_flag(inode, FI_AUTO_RECOVER);
1010         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1011         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1012 }
1013
1014 /*
1015  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1016  *
1017  * We should call set_dirty_inode to write the dirty inode through write_inode.
1018  */
1019 static void f2fs_dirty_inode(struct inode *inode, int flags)
1020 {
1021         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1022
1023         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1024                         inode->i_ino == F2FS_META_INO(sbi))
1025                 return;
1026
1027         if (flags == I_DIRTY_TIME)
1028                 return;
1029
1030         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1031                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1032
1033         f2fs_inode_dirtied(inode, false);
1034 }
1035
1036 static void f2fs_free_inode(struct inode *inode)
1037 {
1038         fscrypt_free_inode(inode);
1039         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1040 }
1041
1042 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1043 {
1044         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1045         percpu_counter_destroy(&sbi->total_valid_inode_count);
1046 }
1047
1048 static void destroy_device_list(struct f2fs_sb_info *sbi)
1049 {
1050         int i;
1051
1052         for (i = 0; i < sbi->s_ndevs; i++) {
1053                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1054 #ifdef CONFIG_BLK_DEV_ZONED
1055                 kvfree(FDEV(i).blkz_seq);
1056 #endif
1057         }
1058         kvfree(sbi->devs);
1059 }
1060
1061 static void f2fs_put_super(struct super_block *sb)
1062 {
1063         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1064         int i;
1065         bool dropped;
1066
1067         f2fs_quota_off_umount(sb);
1068
1069         /* prevent remaining shrinker jobs */
1070         mutex_lock(&sbi->umount_mutex);
1071
1072         /*
1073          * We don't need to do checkpoint when superblock is clean.
1074          * But, the previous checkpoint was not done by umount, it needs to do
1075          * clean checkpoint again.
1076          */
1077         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1078                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1079                 struct cp_control cpc = {
1080                         .reason = CP_UMOUNT,
1081                 };
1082                 f2fs_write_checkpoint(sbi, &cpc);
1083         }
1084
1085         /* be sure to wait for any on-going discard commands */
1086         dropped = f2fs_issue_discard_timeout(sbi);
1087
1088         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1089                                         !sbi->discard_blks && !dropped) {
1090                 struct cp_control cpc = {
1091                         .reason = CP_UMOUNT | CP_TRIMMED,
1092                 };
1093                 f2fs_write_checkpoint(sbi, &cpc);
1094         }
1095
1096         /*
1097          * normally superblock is clean, so we need to release this.
1098          * In addition, EIO will skip do checkpoint, we need this as well.
1099          */
1100         f2fs_release_ino_entry(sbi, true);
1101
1102         f2fs_leave_shrinker(sbi);
1103         mutex_unlock(&sbi->umount_mutex);
1104
1105         /* our cp_error case, we can wait for any writeback page */
1106         f2fs_flush_merged_writes(sbi);
1107
1108         f2fs_wait_on_all_pages_writeback(sbi);
1109
1110         f2fs_bug_on(sbi, sbi->fsync_node_num);
1111
1112         iput(sbi->node_inode);
1113         sbi->node_inode = NULL;
1114
1115         iput(sbi->meta_inode);
1116         sbi->meta_inode = NULL;
1117
1118         /*
1119          * iput() can update stat information, if f2fs_write_checkpoint()
1120          * above failed with error.
1121          */
1122         f2fs_destroy_stats(sbi);
1123
1124         /* destroy f2fs internal modules */
1125         f2fs_destroy_node_manager(sbi);
1126         f2fs_destroy_segment_manager(sbi);
1127
1128         kvfree(sbi->ckpt);
1129
1130         f2fs_unregister_sysfs(sbi);
1131
1132         sb->s_fs_info = NULL;
1133         if (sbi->s_chksum_driver)
1134                 crypto_free_shash(sbi->s_chksum_driver);
1135         kvfree(sbi->raw_super);
1136
1137         destroy_device_list(sbi);
1138         mempool_destroy(sbi->write_io_dummy);
1139 #ifdef CONFIG_QUOTA
1140         for (i = 0; i < MAXQUOTAS; i++)
1141                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1142 #endif
1143         destroy_percpu_info(sbi);
1144         for (i = 0; i < NR_PAGE_TYPE; i++)
1145                 kvfree(sbi->write_io[i]);
1146 #ifdef CONFIG_UNICODE
1147         utf8_unload(sbi->s_encoding);
1148 #endif
1149         kvfree(sbi);
1150 }
1151
1152 int f2fs_sync_fs(struct super_block *sb, int sync)
1153 {
1154         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1155         int err = 0;
1156
1157         if (unlikely(f2fs_cp_error(sbi)))
1158                 return 0;
1159         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1160                 return 0;
1161
1162         trace_f2fs_sync_fs(sb, sync);
1163
1164         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1165                 return -EAGAIN;
1166
1167         if (sync) {
1168                 struct cp_control cpc;
1169
1170                 cpc.reason = __get_cp_reason(sbi);
1171
1172                 mutex_lock(&sbi->gc_mutex);
1173                 err = f2fs_write_checkpoint(sbi, &cpc);
1174                 mutex_unlock(&sbi->gc_mutex);
1175         }
1176         f2fs_trace_ios(NULL, 1);
1177
1178         return err;
1179 }
1180
1181 static int f2fs_freeze(struct super_block *sb)
1182 {
1183         if (f2fs_readonly(sb))
1184                 return 0;
1185
1186         /* IO error happened before */
1187         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1188                 return -EIO;
1189
1190         /* must be clean, since sync_filesystem() was already called */
1191         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1192                 return -EINVAL;
1193         return 0;
1194 }
1195
1196 static int f2fs_unfreeze(struct super_block *sb)
1197 {
1198         return 0;
1199 }
1200
1201 #ifdef CONFIG_QUOTA
1202 static int f2fs_statfs_project(struct super_block *sb,
1203                                 kprojid_t projid, struct kstatfs *buf)
1204 {
1205         struct kqid qid;
1206         struct dquot *dquot;
1207         u64 limit;
1208         u64 curblock;
1209
1210         qid = make_kqid_projid(projid);
1211         dquot = dqget(sb, qid);
1212         if (IS_ERR(dquot))
1213                 return PTR_ERR(dquot);
1214         spin_lock(&dquot->dq_dqb_lock);
1215
1216         limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1217                                         dquot->dq_dqb.dqb_bhardlimit);
1218         if (limit)
1219                 limit >>= sb->s_blocksize_bits;
1220
1221         if (limit && buf->f_blocks > limit) {
1222                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1223                 buf->f_blocks = limit;
1224                 buf->f_bfree = buf->f_bavail =
1225                         (buf->f_blocks > curblock) ?
1226                          (buf->f_blocks - curblock) : 0;
1227         }
1228
1229         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1230                                         dquot->dq_dqb.dqb_ihardlimit);
1231
1232         if (limit && buf->f_files > limit) {
1233                 buf->f_files = limit;
1234                 buf->f_ffree =
1235                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1236                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1237         }
1238
1239         spin_unlock(&dquot->dq_dqb_lock);
1240         dqput(dquot);
1241         return 0;
1242 }
1243 #endif
1244
1245 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1246 {
1247         struct super_block *sb = dentry->d_sb;
1248         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1249         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1250         block_t total_count, user_block_count, start_count;
1251         u64 avail_node_count;
1252
1253         total_count = le64_to_cpu(sbi->raw_super->block_count);
1254         user_block_count = sbi->user_block_count;
1255         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1256         buf->f_type = F2FS_SUPER_MAGIC;
1257         buf->f_bsize = sbi->blocksize;
1258
1259         buf->f_blocks = total_count - start_count;
1260         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1261                                                 sbi->current_reserved_blocks;
1262
1263         spin_lock(&sbi->stat_lock);
1264         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1265                 buf->f_bfree = 0;
1266         else
1267                 buf->f_bfree -= sbi->unusable_block_count;
1268         spin_unlock(&sbi->stat_lock);
1269
1270         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1271                 buf->f_bavail = buf->f_bfree -
1272                                 F2FS_OPTION(sbi).root_reserved_blocks;
1273         else
1274                 buf->f_bavail = 0;
1275
1276         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1277
1278         if (avail_node_count > user_block_count) {
1279                 buf->f_files = user_block_count;
1280                 buf->f_ffree = buf->f_bavail;
1281         } else {
1282                 buf->f_files = avail_node_count;
1283                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1284                                         buf->f_bavail);
1285         }
1286
1287         buf->f_namelen = F2FS_NAME_LEN;
1288         buf->f_fsid.val[0] = (u32)id;
1289         buf->f_fsid.val[1] = (u32)(id >> 32);
1290
1291 #ifdef CONFIG_QUOTA
1292         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1293                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1294                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1295         }
1296 #endif
1297         return 0;
1298 }
1299
1300 static inline void f2fs_show_quota_options(struct seq_file *seq,
1301                                            struct super_block *sb)
1302 {
1303 #ifdef CONFIG_QUOTA
1304         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1305
1306         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1307                 char *fmtname = "";
1308
1309                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1310                 case QFMT_VFS_OLD:
1311                         fmtname = "vfsold";
1312                         break;
1313                 case QFMT_VFS_V0:
1314                         fmtname = "vfsv0";
1315                         break;
1316                 case QFMT_VFS_V1:
1317                         fmtname = "vfsv1";
1318                         break;
1319                 }
1320                 seq_printf(seq, ",jqfmt=%s", fmtname);
1321         }
1322
1323         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1324                 seq_show_option(seq, "usrjquota",
1325                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1326
1327         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1328                 seq_show_option(seq, "grpjquota",
1329                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1330
1331         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1332                 seq_show_option(seq, "prjjquota",
1333                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1334 #endif
1335 }
1336
1337 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1338 {
1339         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1340
1341         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1342                 if (test_opt(sbi, FORCE_FG_GC))
1343                         seq_printf(seq, ",background_gc=%s", "sync");
1344                 else
1345                         seq_printf(seq, ",background_gc=%s", "on");
1346         } else {
1347                 seq_printf(seq, ",background_gc=%s", "off");
1348         }
1349         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1350                 seq_puts(seq, ",disable_roll_forward");
1351         if (test_opt(sbi, DISCARD))
1352                 seq_puts(seq, ",discard");
1353         else
1354                 seq_puts(seq, ",nodiscard");
1355         if (test_opt(sbi, NOHEAP))
1356                 seq_puts(seq, ",no_heap");
1357         else
1358                 seq_puts(seq, ",heap");
1359 #ifdef CONFIG_F2FS_FS_XATTR
1360         if (test_opt(sbi, XATTR_USER))
1361                 seq_puts(seq, ",user_xattr");
1362         else
1363                 seq_puts(seq, ",nouser_xattr");
1364         if (test_opt(sbi, INLINE_XATTR))
1365                 seq_puts(seq, ",inline_xattr");
1366         else
1367                 seq_puts(seq, ",noinline_xattr");
1368         if (test_opt(sbi, INLINE_XATTR_SIZE))
1369                 seq_printf(seq, ",inline_xattr_size=%u",
1370                                         F2FS_OPTION(sbi).inline_xattr_size);
1371 #endif
1372 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1373         if (test_opt(sbi, POSIX_ACL))
1374                 seq_puts(seq, ",acl");
1375         else
1376                 seq_puts(seq, ",noacl");
1377 #endif
1378         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1379                 seq_puts(seq, ",disable_ext_identify");
1380         if (test_opt(sbi, INLINE_DATA))
1381                 seq_puts(seq, ",inline_data");
1382         else
1383                 seq_puts(seq, ",noinline_data");
1384         if (test_opt(sbi, INLINE_DENTRY))
1385                 seq_puts(seq, ",inline_dentry");
1386         else
1387                 seq_puts(seq, ",noinline_dentry");
1388         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1389                 seq_puts(seq, ",flush_merge");
1390         if (test_opt(sbi, NOBARRIER))
1391                 seq_puts(seq, ",nobarrier");
1392         if (test_opt(sbi, FASTBOOT))
1393                 seq_puts(seq, ",fastboot");
1394         if (test_opt(sbi, EXTENT_CACHE))
1395                 seq_puts(seq, ",extent_cache");
1396         else
1397                 seq_puts(seq, ",noextent_cache");
1398         if (test_opt(sbi, DATA_FLUSH))
1399                 seq_puts(seq, ",data_flush");
1400
1401         seq_puts(seq, ",mode=");
1402         if (test_opt(sbi, ADAPTIVE))
1403                 seq_puts(seq, "adaptive");
1404         else if (test_opt(sbi, LFS))
1405                 seq_puts(seq, "lfs");
1406         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1407         if (test_opt(sbi, RESERVE_ROOT))
1408                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1409                                 F2FS_OPTION(sbi).root_reserved_blocks,
1410                                 from_kuid_munged(&init_user_ns,
1411                                         F2FS_OPTION(sbi).s_resuid),
1412                                 from_kgid_munged(&init_user_ns,
1413                                         F2FS_OPTION(sbi).s_resgid));
1414         if (F2FS_IO_SIZE_BITS(sbi))
1415                 seq_printf(seq, ",io_bits=%u",
1416                                 F2FS_OPTION(sbi).write_io_size_bits);
1417 #ifdef CONFIG_F2FS_FAULT_INJECTION
1418         if (test_opt(sbi, FAULT_INJECTION)) {
1419                 seq_printf(seq, ",fault_injection=%u",
1420                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1421                 seq_printf(seq, ",fault_type=%u",
1422                                 F2FS_OPTION(sbi).fault_info.inject_type);
1423         }
1424 #endif
1425 #ifdef CONFIG_QUOTA
1426         if (test_opt(sbi, QUOTA))
1427                 seq_puts(seq, ",quota");
1428         if (test_opt(sbi, USRQUOTA))
1429                 seq_puts(seq, ",usrquota");
1430         if (test_opt(sbi, GRPQUOTA))
1431                 seq_puts(seq, ",grpquota");
1432         if (test_opt(sbi, PRJQUOTA))
1433                 seq_puts(seq, ",prjquota");
1434 #endif
1435         f2fs_show_quota_options(seq, sbi->sb);
1436         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1437                 seq_printf(seq, ",whint_mode=%s", "user-based");
1438         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1439                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1440 #ifdef CONFIG_FS_ENCRYPTION
1441         if (F2FS_OPTION(sbi).test_dummy_encryption)
1442                 seq_puts(seq, ",test_dummy_encryption");
1443 #endif
1444
1445         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1446                 seq_printf(seq, ",alloc_mode=%s", "default");
1447         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1448                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1449
1450         if (test_opt(sbi, DISABLE_CHECKPOINT))
1451                 seq_printf(seq, ",checkpoint=disable:%u",
1452                                 F2FS_OPTION(sbi).unusable_cap);
1453         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1454                 seq_printf(seq, ",fsync_mode=%s", "posix");
1455         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1456                 seq_printf(seq, ",fsync_mode=%s", "strict");
1457         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1458                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1459         return 0;
1460 }
1461
1462 static void default_options(struct f2fs_sb_info *sbi)
1463 {
1464         /* init some FS parameters */
1465         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1466         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1467         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1468         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1469         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1470         F2FS_OPTION(sbi).test_dummy_encryption = false;
1471         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1472         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1473
1474         set_opt(sbi, BG_GC);
1475         set_opt(sbi, INLINE_XATTR);
1476         set_opt(sbi, INLINE_DATA);
1477         set_opt(sbi, INLINE_DENTRY);
1478         set_opt(sbi, EXTENT_CACHE);
1479         set_opt(sbi, NOHEAP);
1480         clear_opt(sbi, DISABLE_CHECKPOINT);
1481         F2FS_OPTION(sbi).unusable_cap = 0;
1482         sbi->sb->s_flags |= SB_LAZYTIME;
1483         set_opt(sbi, FLUSH_MERGE);
1484         set_opt(sbi, DISCARD);
1485         if (f2fs_sb_has_blkzoned(sbi))
1486                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1487         else
1488                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1489
1490 #ifdef CONFIG_F2FS_FS_XATTR
1491         set_opt(sbi, XATTR_USER);
1492 #endif
1493 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1494         set_opt(sbi, POSIX_ACL);
1495 #endif
1496
1497         f2fs_build_fault_attr(sbi, 0, 0);
1498 }
1499
1500 #ifdef CONFIG_QUOTA
1501 static int f2fs_enable_quotas(struct super_block *sb);
1502 #endif
1503
1504 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1505 {
1506         unsigned int s_flags = sbi->sb->s_flags;
1507         struct cp_control cpc;
1508         int err = 0;
1509         int ret;
1510         block_t unusable;
1511
1512         if (s_flags & SB_RDONLY) {
1513                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1514                 return -EINVAL;
1515         }
1516         sbi->sb->s_flags |= SB_ACTIVE;
1517
1518         f2fs_update_time(sbi, DISABLE_TIME);
1519
1520         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1521                 mutex_lock(&sbi->gc_mutex);
1522                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1523                 if (err == -ENODATA) {
1524                         err = 0;
1525                         break;
1526                 }
1527                 if (err && err != -EAGAIN)
1528                         break;
1529         }
1530
1531         ret = sync_filesystem(sbi->sb);
1532         if (ret || err) {
1533                 err = ret ? ret: err;
1534                 goto restore_flag;
1535         }
1536
1537         unusable = f2fs_get_unusable_blocks(sbi);
1538         if (f2fs_disable_cp_again(sbi, unusable)) {
1539                 err = -EAGAIN;
1540                 goto restore_flag;
1541         }
1542
1543         mutex_lock(&sbi->gc_mutex);
1544         cpc.reason = CP_PAUSE;
1545         set_sbi_flag(sbi, SBI_CP_DISABLED);
1546         err = f2fs_write_checkpoint(sbi, &cpc);
1547         if (err)
1548                 goto out_unlock;
1549
1550         spin_lock(&sbi->stat_lock);
1551         sbi->unusable_block_count = unusable;
1552         spin_unlock(&sbi->stat_lock);
1553
1554 out_unlock:
1555         mutex_unlock(&sbi->gc_mutex);
1556 restore_flag:
1557         sbi->sb->s_flags = s_flags;     /* Restore MS_RDONLY status */
1558         return err;
1559 }
1560
1561 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1562 {
1563         mutex_lock(&sbi->gc_mutex);
1564         f2fs_dirty_to_prefree(sbi);
1565
1566         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1567         set_sbi_flag(sbi, SBI_IS_DIRTY);
1568         mutex_unlock(&sbi->gc_mutex);
1569
1570         f2fs_sync_fs(sbi->sb, 1);
1571 }
1572
1573 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1574 {
1575         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1576         struct f2fs_mount_info org_mount_opt;
1577         unsigned long old_sb_flags;
1578         int err;
1579         bool need_restart_gc = false;
1580         bool need_stop_gc = false;
1581         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1582         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1583         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1584         bool checkpoint_changed;
1585 #ifdef CONFIG_QUOTA
1586         int i, j;
1587 #endif
1588
1589         /*
1590          * Save the old mount options in case we
1591          * need to restore them.
1592          */
1593         org_mount_opt = sbi->mount_opt;
1594         old_sb_flags = sb->s_flags;
1595
1596 #ifdef CONFIG_QUOTA
1597         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1598         for (i = 0; i < MAXQUOTAS; i++) {
1599                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1600                         org_mount_opt.s_qf_names[i] =
1601                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1602                                 GFP_KERNEL);
1603                         if (!org_mount_opt.s_qf_names[i]) {
1604                                 for (j = 0; j < i; j++)
1605                                         kvfree(org_mount_opt.s_qf_names[j]);
1606                                 return -ENOMEM;
1607                         }
1608                 } else {
1609                         org_mount_opt.s_qf_names[i] = NULL;
1610                 }
1611         }
1612 #endif
1613
1614         /* recover superblocks we couldn't write due to previous RO mount */
1615         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1616                 err = f2fs_commit_super(sbi, false);
1617                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1618                           err);
1619                 if (!err)
1620                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1621         }
1622
1623         default_options(sbi);
1624
1625         /* parse mount options */
1626         err = parse_options(sb, data);
1627         if (err)
1628                 goto restore_opts;
1629         checkpoint_changed =
1630                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1631
1632         /*
1633          * Previous and new state of filesystem is RO,
1634          * so skip checking GC and FLUSH_MERGE conditions.
1635          */
1636         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1637                 goto skip;
1638
1639 #ifdef CONFIG_QUOTA
1640         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1641                 err = dquot_suspend(sb, -1);
1642                 if (err < 0)
1643                         goto restore_opts;
1644         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1645                 /* dquot_resume needs RW */
1646                 sb->s_flags &= ~SB_RDONLY;
1647                 if (sb_any_quota_suspended(sb)) {
1648                         dquot_resume(sb, -1);
1649                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1650                         err = f2fs_enable_quotas(sb);
1651                         if (err)
1652                                 goto restore_opts;
1653                 }
1654         }
1655 #endif
1656         /* disallow enable/disable extent_cache dynamically */
1657         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1658                 err = -EINVAL;
1659                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1660                 goto restore_opts;
1661         }
1662
1663         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1664                 err = -EINVAL;
1665                 f2fs_warn(sbi, "switch io_bits option is not allowed");
1666                 goto restore_opts;
1667         }
1668
1669         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1670                 err = -EINVAL;
1671                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1672                 goto restore_opts;
1673         }
1674
1675         /*
1676          * We stop the GC thread if FS is mounted as RO
1677          * or if background_gc = off is passed in mount
1678          * option. Also sync the filesystem.
1679          */
1680         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1681                 if (sbi->gc_thread) {
1682                         f2fs_stop_gc_thread(sbi);
1683                         need_restart_gc = true;
1684                 }
1685         } else if (!sbi->gc_thread) {
1686                 err = f2fs_start_gc_thread(sbi);
1687                 if (err)
1688                         goto restore_opts;
1689                 need_stop_gc = true;
1690         }
1691
1692         if (*flags & SB_RDONLY ||
1693                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1694                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1695                 sync_inodes_sb(sb);
1696
1697                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1698                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1699                 f2fs_sync_fs(sb, 1);
1700                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1701         }
1702
1703         if (checkpoint_changed) {
1704                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1705                         err = f2fs_disable_checkpoint(sbi);
1706                         if (err)
1707                                 goto restore_gc;
1708                 } else {
1709                         f2fs_enable_checkpoint(sbi);
1710                 }
1711         }
1712
1713         /*
1714          * We stop issue flush thread if FS is mounted as RO
1715          * or if flush_merge is not passed in mount option.
1716          */
1717         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1718                 clear_opt(sbi, FLUSH_MERGE);
1719                 f2fs_destroy_flush_cmd_control(sbi, false);
1720         } else {
1721                 err = f2fs_create_flush_cmd_control(sbi);
1722                 if (err)
1723                         goto restore_gc;
1724         }
1725 skip:
1726 #ifdef CONFIG_QUOTA
1727         /* Release old quota file names */
1728         for (i = 0; i < MAXQUOTAS; i++)
1729                 kvfree(org_mount_opt.s_qf_names[i]);
1730 #endif
1731         /* Update the POSIXACL Flag */
1732         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1733                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1734
1735         limit_reserve_root(sbi);
1736         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1737         return 0;
1738 restore_gc:
1739         if (need_restart_gc) {
1740                 if (f2fs_start_gc_thread(sbi))
1741                         f2fs_warn(sbi, "background gc thread has stopped");
1742         } else if (need_stop_gc) {
1743                 f2fs_stop_gc_thread(sbi);
1744         }
1745 restore_opts:
1746 #ifdef CONFIG_QUOTA
1747         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1748         for (i = 0; i < MAXQUOTAS; i++) {
1749                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1750                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1751         }
1752 #endif
1753         sbi->mount_opt = org_mount_opt;
1754         sb->s_flags = old_sb_flags;
1755         return err;
1756 }
1757
1758 #ifdef CONFIG_QUOTA
1759 /* Read data from quotafile */
1760 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1761                                size_t len, loff_t off)
1762 {
1763         struct inode *inode = sb_dqopt(sb)->files[type];
1764         struct address_space *mapping = inode->i_mapping;
1765         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1766         int offset = off & (sb->s_blocksize - 1);
1767         int tocopy;
1768         size_t toread;
1769         loff_t i_size = i_size_read(inode);
1770         struct page *page;
1771         char *kaddr;
1772
1773         if (off > i_size)
1774                 return 0;
1775
1776         if (off + len > i_size)
1777                 len = i_size - off;
1778         toread = len;
1779         while (toread > 0) {
1780                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1781 repeat:
1782                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1783                 if (IS_ERR(page)) {
1784                         if (PTR_ERR(page) == -ENOMEM) {
1785                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1786                                 goto repeat;
1787                         }
1788                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1789                         return PTR_ERR(page);
1790                 }
1791
1792                 lock_page(page);
1793
1794                 if (unlikely(page->mapping != mapping)) {
1795                         f2fs_put_page(page, 1);
1796                         goto repeat;
1797                 }
1798                 if (unlikely(!PageUptodate(page))) {
1799                         f2fs_put_page(page, 1);
1800                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1801                         return -EIO;
1802                 }
1803
1804                 kaddr = kmap_atomic(page);
1805                 memcpy(data, kaddr + offset, tocopy);
1806                 kunmap_atomic(kaddr);
1807                 f2fs_put_page(page, 1);
1808
1809                 offset = 0;
1810                 toread -= tocopy;
1811                 data += tocopy;
1812                 blkidx++;
1813         }
1814         return len;
1815 }
1816
1817 /* Write to quotafile */
1818 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1819                                 const char *data, size_t len, loff_t off)
1820 {
1821         struct inode *inode = sb_dqopt(sb)->files[type];
1822         struct address_space *mapping = inode->i_mapping;
1823         const struct address_space_operations *a_ops = mapping->a_ops;
1824         int offset = off & (sb->s_blocksize - 1);
1825         size_t towrite = len;
1826         struct page *page;
1827         char *kaddr;
1828         int err = 0;
1829         int tocopy;
1830
1831         while (towrite > 0) {
1832                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1833                                                                 towrite);
1834 retry:
1835                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1836                                                         &page, NULL);
1837                 if (unlikely(err)) {
1838                         if (err == -ENOMEM) {
1839                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1840                                 goto retry;
1841                         }
1842                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1843                         break;
1844                 }
1845
1846                 kaddr = kmap_atomic(page);
1847                 memcpy(kaddr + offset, data, tocopy);
1848                 kunmap_atomic(kaddr);
1849                 flush_dcache_page(page);
1850
1851                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1852                                                 page, NULL);
1853                 offset = 0;
1854                 towrite -= tocopy;
1855                 off += tocopy;
1856                 data += tocopy;
1857                 cond_resched();
1858         }
1859
1860         if (len == towrite)
1861                 return err;
1862         inode->i_mtime = inode->i_ctime = current_time(inode);
1863         f2fs_mark_inode_dirty_sync(inode, false);
1864         return len - towrite;
1865 }
1866
1867 static struct dquot **f2fs_get_dquots(struct inode *inode)
1868 {
1869         return F2FS_I(inode)->i_dquot;
1870 }
1871
1872 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1873 {
1874         return &F2FS_I(inode)->i_reserved_quota;
1875 }
1876
1877 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1878 {
1879         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1880                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1881                 return 0;
1882         }
1883
1884         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1885                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1886 }
1887
1888 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1889 {
1890         int enabled = 0;
1891         int i, err;
1892
1893         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1894                 err = f2fs_enable_quotas(sbi->sb);
1895                 if (err) {
1896                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
1897                         return 0;
1898                 }
1899                 return 1;
1900         }
1901
1902         for (i = 0; i < MAXQUOTAS; i++) {
1903                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1904                         err = f2fs_quota_on_mount(sbi, i);
1905                         if (!err) {
1906                                 enabled = 1;
1907                                 continue;
1908                         }
1909                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
1910                                  err, i);
1911                 }
1912         }
1913         return enabled;
1914 }
1915
1916 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1917                              unsigned int flags)
1918 {
1919         struct inode *qf_inode;
1920         unsigned long qf_inum;
1921         int err;
1922
1923         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1924
1925         qf_inum = f2fs_qf_ino(sb, type);
1926         if (!qf_inum)
1927                 return -EPERM;
1928
1929         qf_inode = f2fs_iget(sb, qf_inum);
1930         if (IS_ERR(qf_inode)) {
1931                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
1932                 return PTR_ERR(qf_inode);
1933         }
1934
1935         /* Don't account quota for quota files to avoid recursion */
1936         qf_inode->i_flags |= S_NOQUOTA;
1937         err = dquot_enable(qf_inode, type, format_id, flags);
1938         iput(qf_inode);
1939         return err;
1940 }
1941
1942 static int f2fs_enable_quotas(struct super_block *sb)
1943 {
1944         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1945         int type, err = 0;
1946         unsigned long qf_inum;
1947         bool quota_mopt[MAXQUOTAS] = {
1948                 test_opt(sbi, USRQUOTA),
1949                 test_opt(sbi, GRPQUOTA),
1950                 test_opt(sbi, PRJQUOTA),
1951         };
1952
1953         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1954                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
1955                 return 0;
1956         }
1957
1958         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1959
1960         for (type = 0; type < MAXQUOTAS; type++) {
1961                 qf_inum = f2fs_qf_ino(sb, type);
1962                 if (qf_inum) {
1963                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1964                                 DQUOT_USAGE_ENABLED |
1965                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1966                         if (err) {
1967                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
1968                                          type, err);
1969                                 for (type--; type >= 0; type--)
1970                                         dquot_quota_off(sb, type);
1971                                 set_sbi_flag(F2FS_SB(sb),
1972                                                 SBI_QUOTA_NEED_REPAIR);
1973                                 return err;
1974                         }
1975                 }
1976         }
1977         return 0;
1978 }
1979
1980 int f2fs_quota_sync(struct super_block *sb, int type)
1981 {
1982         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1983         struct quota_info *dqopt = sb_dqopt(sb);
1984         int cnt;
1985         int ret;
1986
1987         /*
1988          * do_quotactl
1989          *  f2fs_quota_sync
1990          *  down_read(quota_sem)
1991          *  dquot_writeback_dquots()
1992          *  f2fs_dquot_commit
1993          *                            block_operation
1994          *                            down_read(quota_sem)
1995          */
1996         f2fs_lock_op(sbi);
1997
1998         down_read(&sbi->quota_sem);
1999         ret = dquot_writeback_dquots(sb, type);
2000         if (ret)
2001                 goto out;
2002
2003         /*
2004          * Now when everything is written we can discard the pagecache so
2005          * that userspace sees the changes.
2006          */
2007         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2008                 struct address_space *mapping;
2009
2010                 if (type != -1 && cnt != type)
2011                         continue;
2012                 if (!sb_has_quota_active(sb, cnt))
2013                         continue;
2014
2015                 mapping = dqopt->files[cnt]->i_mapping;
2016
2017                 ret = filemap_fdatawrite(mapping);
2018                 if (ret)
2019                         goto out;
2020
2021                 /* if we are using journalled quota */
2022                 if (is_journalled_quota(sbi))
2023                         continue;
2024
2025                 ret = filemap_fdatawait(mapping);
2026                 if (ret)
2027                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2028
2029                 inode_lock(dqopt->files[cnt]);
2030                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2031                 inode_unlock(dqopt->files[cnt]);
2032         }
2033 out:
2034         if (ret)
2035                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2036         up_read(&sbi->quota_sem);
2037         f2fs_unlock_op(sbi);
2038         return ret;
2039 }
2040
2041 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2042                                                         const struct path *path)
2043 {
2044         struct inode *inode;
2045         int err;
2046
2047         /* if quota sysfile exists, deny enabling quota with specific file */
2048         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2049                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2050                 return -EBUSY;
2051         }
2052
2053         err = f2fs_quota_sync(sb, type);
2054         if (err)
2055                 return err;
2056
2057         err = dquot_quota_on(sb, type, format_id, path);
2058         if (err)
2059                 return err;
2060
2061         inode = d_inode(path->dentry);
2062
2063         inode_lock(inode);
2064         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2065         f2fs_set_inode_flags(inode);
2066         inode_unlock(inode);
2067         f2fs_mark_inode_dirty_sync(inode, false);
2068
2069         return 0;
2070 }
2071
2072 static int __f2fs_quota_off(struct super_block *sb, int type)
2073 {
2074         struct inode *inode = sb_dqopt(sb)->files[type];
2075         int err;
2076
2077         if (!inode || !igrab(inode))
2078                 return dquot_quota_off(sb, type);
2079
2080         err = f2fs_quota_sync(sb, type);
2081         if (err)
2082                 goto out_put;
2083
2084         err = dquot_quota_off(sb, type);
2085         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2086                 goto out_put;
2087
2088         inode_lock(inode);
2089         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2090         f2fs_set_inode_flags(inode);
2091         inode_unlock(inode);
2092         f2fs_mark_inode_dirty_sync(inode, false);
2093 out_put:
2094         iput(inode);
2095         return err;
2096 }
2097
2098 static int f2fs_quota_off(struct super_block *sb, int type)
2099 {
2100         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2101         int err;
2102
2103         err = __f2fs_quota_off(sb, type);
2104
2105         /*
2106          * quotactl can shutdown journalled quota, result in inconsistence
2107          * between quota record and fs data by following updates, tag the
2108          * flag to let fsck be aware of it.
2109          */
2110         if (is_journalled_quota(sbi))
2111                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2112         return err;
2113 }
2114
2115 void f2fs_quota_off_umount(struct super_block *sb)
2116 {
2117         int type;
2118         int err;
2119
2120         for (type = 0; type < MAXQUOTAS; type++) {
2121                 err = __f2fs_quota_off(sb, type);
2122                 if (err) {
2123                         int ret = dquot_quota_off(sb, type);
2124
2125                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2126                                  type, err, ret);
2127                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2128                 }
2129         }
2130         /*
2131          * In case of checkpoint=disable, we must flush quota blocks.
2132          * This can cause NULL exception for node_inode in end_io, since
2133          * put_super already dropped it.
2134          */
2135         sync_filesystem(sb);
2136 }
2137
2138 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2139 {
2140         struct quota_info *dqopt = sb_dqopt(sb);
2141         int type;
2142
2143         for (type = 0; type < MAXQUOTAS; type++) {
2144                 if (!dqopt->files[type])
2145                         continue;
2146                 f2fs_inode_synced(dqopt->files[type]);
2147         }
2148 }
2149
2150 static int f2fs_dquot_commit(struct dquot *dquot)
2151 {
2152         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2153         int ret;
2154
2155         down_read(&sbi->quota_sem);
2156         ret = dquot_commit(dquot);
2157         if (ret < 0)
2158                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2159         up_read(&sbi->quota_sem);
2160         return ret;
2161 }
2162
2163 static int f2fs_dquot_acquire(struct dquot *dquot)
2164 {
2165         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2166         int ret;
2167
2168         down_read(&sbi->quota_sem);
2169         ret = dquot_acquire(dquot);
2170         if (ret < 0)
2171                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2172         up_read(&sbi->quota_sem);
2173         return ret;
2174 }
2175
2176 static int f2fs_dquot_release(struct dquot *dquot)
2177 {
2178         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2179         int ret;
2180
2181         down_read(&sbi->quota_sem);
2182         ret = dquot_release(dquot);
2183         if (ret < 0)
2184                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2185         up_read(&sbi->quota_sem);
2186         return ret;
2187 }
2188
2189 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2190 {
2191         struct super_block *sb = dquot->dq_sb;
2192         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2193         int ret;
2194
2195         down_read(&sbi->quota_sem);
2196         ret = dquot_mark_dquot_dirty(dquot);
2197
2198         /* if we are using journalled quota */
2199         if (is_journalled_quota(sbi))
2200                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2201
2202         up_read(&sbi->quota_sem);
2203         return ret;
2204 }
2205
2206 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2207 {
2208         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2209         int ret;
2210
2211         down_read(&sbi->quota_sem);
2212         ret = dquot_commit_info(sb, type);
2213         if (ret < 0)
2214                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2215         up_read(&sbi->quota_sem);
2216         return ret;
2217 }
2218
2219 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2220 {
2221         *projid = F2FS_I(inode)->i_projid;
2222         return 0;
2223 }
2224
2225 static const struct dquot_operations f2fs_quota_operations = {
2226         .get_reserved_space = f2fs_get_reserved_space,
2227         .write_dquot    = f2fs_dquot_commit,
2228         .acquire_dquot  = f2fs_dquot_acquire,
2229         .release_dquot  = f2fs_dquot_release,
2230         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2231         .write_info     = f2fs_dquot_commit_info,
2232         .alloc_dquot    = dquot_alloc,
2233         .destroy_dquot  = dquot_destroy,
2234         .get_projid     = f2fs_get_projid,
2235         .get_next_id    = dquot_get_next_id,
2236 };
2237
2238 static const struct quotactl_ops f2fs_quotactl_ops = {
2239         .quota_on       = f2fs_quota_on,
2240         .quota_off      = f2fs_quota_off,
2241         .quota_sync     = f2fs_quota_sync,
2242         .get_state      = dquot_get_state,
2243         .set_info       = dquot_set_dqinfo,
2244         .get_dqblk      = dquot_get_dqblk,
2245         .set_dqblk      = dquot_set_dqblk,
2246         .get_nextdqblk  = dquot_get_next_dqblk,
2247 };
2248 #else
2249 int f2fs_quota_sync(struct super_block *sb, int type)
2250 {
2251         return 0;
2252 }
2253
2254 void f2fs_quota_off_umount(struct super_block *sb)
2255 {
2256 }
2257 #endif
2258
2259 static const struct super_operations f2fs_sops = {
2260         .alloc_inode    = f2fs_alloc_inode,
2261         .free_inode     = f2fs_free_inode,
2262         .drop_inode     = f2fs_drop_inode,
2263         .write_inode    = f2fs_write_inode,
2264         .dirty_inode    = f2fs_dirty_inode,
2265         .show_options   = f2fs_show_options,
2266 #ifdef CONFIG_QUOTA
2267         .quota_read     = f2fs_quota_read,
2268         .quota_write    = f2fs_quota_write,
2269         .get_dquots     = f2fs_get_dquots,
2270 #endif
2271         .evict_inode    = f2fs_evict_inode,
2272         .put_super      = f2fs_put_super,
2273         .sync_fs        = f2fs_sync_fs,
2274         .freeze_fs      = f2fs_freeze,
2275         .unfreeze_fs    = f2fs_unfreeze,
2276         .statfs         = f2fs_statfs,
2277         .remount_fs     = f2fs_remount,
2278 };
2279
2280 #ifdef CONFIG_FS_ENCRYPTION
2281 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2282 {
2283         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2284                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2285                                 ctx, len, NULL);
2286 }
2287
2288 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2289                                                         void *fs_data)
2290 {
2291         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2292
2293         /*
2294          * Encrypting the root directory is not allowed because fsck
2295          * expects lost+found directory to exist and remain unencrypted
2296          * if LOST_FOUND feature is enabled.
2297          *
2298          */
2299         if (f2fs_sb_has_lost_found(sbi) &&
2300                         inode->i_ino == F2FS_ROOT_INO(sbi))
2301                 return -EPERM;
2302
2303         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2304                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2305                                 ctx, len, fs_data, XATTR_CREATE);
2306 }
2307
2308 static bool f2fs_dummy_context(struct inode *inode)
2309 {
2310         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2311 }
2312
2313 static const struct fscrypt_operations f2fs_cryptops = {
2314         .key_prefix     = "f2fs:",
2315         .get_context    = f2fs_get_context,
2316         .set_context    = f2fs_set_context,
2317         .dummy_context  = f2fs_dummy_context,
2318         .empty_dir      = f2fs_empty_dir,
2319         .max_namelen    = F2FS_NAME_LEN,
2320 };
2321 #endif
2322
2323 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2324                 u64 ino, u32 generation)
2325 {
2326         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2327         struct inode *inode;
2328
2329         if (f2fs_check_nid_range(sbi, ino))
2330                 return ERR_PTR(-ESTALE);
2331
2332         /*
2333          * f2fs_iget isn't quite right if the inode is currently unallocated!
2334          * However f2fs_iget currently does appropriate checks to handle stale
2335          * inodes so everything is OK.
2336          */
2337         inode = f2fs_iget(sb, ino);
2338         if (IS_ERR(inode))
2339                 return ERR_CAST(inode);
2340         if (unlikely(generation && inode->i_generation != generation)) {
2341                 /* we didn't find the right inode.. */
2342                 iput(inode);
2343                 return ERR_PTR(-ESTALE);
2344         }
2345         return inode;
2346 }
2347
2348 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2349                 int fh_len, int fh_type)
2350 {
2351         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2352                                     f2fs_nfs_get_inode);
2353 }
2354
2355 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2356                 int fh_len, int fh_type)
2357 {
2358         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2359                                     f2fs_nfs_get_inode);
2360 }
2361
2362 static const struct export_operations f2fs_export_ops = {
2363         .fh_to_dentry = f2fs_fh_to_dentry,
2364         .fh_to_parent = f2fs_fh_to_parent,
2365         .get_parent = f2fs_get_parent,
2366 };
2367
2368 static loff_t max_file_blocks(void)
2369 {
2370         loff_t result = 0;
2371         loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2372
2373         /*
2374          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2375          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2376          * space in inode.i_addr, it will be more safe to reassign
2377          * result as zero.
2378          */
2379
2380         /* two direct node blocks */
2381         result += (leaf_count * 2);
2382
2383         /* two indirect node blocks */
2384         leaf_count *= NIDS_PER_BLOCK;
2385         result += (leaf_count * 2);
2386
2387         /* one double indirect node block */
2388         leaf_count *= NIDS_PER_BLOCK;
2389         result += leaf_count;
2390
2391         return result;
2392 }
2393
2394 static int __f2fs_commit_super(struct buffer_head *bh,
2395                         struct f2fs_super_block *super)
2396 {
2397         lock_buffer(bh);
2398         if (super)
2399                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2400         set_buffer_dirty(bh);
2401         unlock_buffer(bh);
2402
2403         /* it's rare case, we can do fua all the time */
2404         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2405 }
2406
2407 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2408                                         struct buffer_head *bh)
2409 {
2410         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2411                                         (bh->b_data + F2FS_SUPER_OFFSET);
2412         struct super_block *sb = sbi->sb;
2413         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2414         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2415         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2416         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2417         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2418         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2419         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2420         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2421         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2422         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2423         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2424         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2425         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2426         u64 main_end_blkaddr = main_blkaddr +
2427                                 (segment_count_main << log_blocks_per_seg);
2428         u64 seg_end_blkaddr = segment0_blkaddr +
2429                                 (segment_count << log_blocks_per_seg);
2430
2431         if (segment0_blkaddr != cp_blkaddr) {
2432                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2433                           segment0_blkaddr, cp_blkaddr);
2434                 return true;
2435         }
2436
2437         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2438                                                         sit_blkaddr) {
2439                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2440                           cp_blkaddr, sit_blkaddr,
2441                           segment_count_ckpt << log_blocks_per_seg);
2442                 return true;
2443         }
2444
2445         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2446                                                         nat_blkaddr) {
2447                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2448                           sit_blkaddr, nat_blkaddr,
2449                           segment_count_sit << log_blocks_per_seg);
2450                 return true;
2451         }
2452
2453         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2454                                                         ssa_blkaddr) {
2455                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2456                           nat_blkaddr, ssa_blkaddr,
2457                           segment_count_nat << log_blocks_per_seg);
2458                 return true;
2459         }
2460
2461         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2462                                                         main_blkaddr) {
2463                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2464                           ssa_blkaddr, main_blkaddr,
2465                           segment_count_ssa << log_blocks_per_seg);
2466                 return true;
2467         }
2468
2469         if (main_end_blkaddr > seg_end_blkaddr) {
2470                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2471                           main_blkaddr,
2472                           segment0_blkaddr +
2473                           (segment_count << log_blocks_per_seg),
2474                           segment_count_main << log_blocks_per_seg);
2475                 return true;
2476         } else if (main_end_blkaddr < seg_end_blkaddr) {
2477                 int err = 0;
2478                 char *res;
2479
2480                 /* fix in-memory information all the time */
2481                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2482                                 segment0_blkaddr) >> log_blocks_per_seg);
2483
2484                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2485                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2486                         res = "internally";
2487                 } else {
2488                         err = __f2fs_commit_super(bh, NULL);
2489                         res = err ? "failed" : "done";
2490                 }
2491                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2492                           res, main_blkaddr,
2493                           segment0_blkaddr +
2494                           (segment_count << log_blocks_per_seg),
2495                           segment_count_main << log_blocks_per_seg);
2496                 if (err)
2497                         return true;
2498         }
2499         return false;
2500 }
2501
2502 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2503                                 struct buffer_head *bh)
2504 {
2505         block_t segment_count, segs_per_sec, secs_per_zone;
2506         block_t total_sections, blocks_per_seg;
2507         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2508                                         (bh->b_data + F2FS_SUPER_OFFSET);
2509         unsigned int blocksize;
2510         size_t crc_offset = 0;
2511         __u32 crc = 0;
2512
2513         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2514                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2515                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2516                 return -EINVAL;
2517         }
2518
2519         /* Check checksum_offset and crc in superblock */
2520         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2521                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2522                 if (crc_offset !=
2523                         offsetof(struct f2fs_super_block, crc)) {
2524                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2525                                   crc_offset);
2526                         return -EFSCORRUPTED;
2527                 }
2528                 crc = le32_to_cpu(raw_super->crc);
2529                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2530                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2531                         return -EFSCORRUPTED;
2532                 }
2533         }
2534
2535         /* Currently, support only 4KB page cache size */
2536         if (F2FS_BLKSIZE != PAGE_SIZE) {
2537                 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2538                           PAGE_SIZE);
2539                 return -EFSCORRUPTED;
2540         }
2541
2542         /* Currently, support only 4KB block size */
2543         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2544         if (blocksize != F2FS_BLKSIZE) {
2545                 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2546                           blocksize);
2547                 return -EFSCORRUPTED;
2548         }
2549
2550         /* check log blocks per segment */
2551         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2552                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2553                           le32_to_cpu(raw_super->log_blocks_per_seg));
2554                 return -EFSCORRUPTED;
2555         }
2556
2557         /* Currently, support 512/1024/2048/4096 bytes sector size */
2558         if (le32_to_cpu(raw_super->log_sectorsize) >
2559                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2560                 le32_to_cpu(raw_super->log_sectorsize) <
2561                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2562                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2563                           le32_to_cpu(raw_super->log_sectorsize));
2564                 return -EFSCORRUPTED;
2565         }
2566         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2567                 le32_to_cpu(raw_super->log_sectorsize) !=
2568                         F2FS_MAX_LOG_SECTOR_SIZE) {
2569                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2570                           le32_to_cpu(raw_super->log_sectors_per_block),
2571                           le32_to_cpu(raw_super->log_sectorsize));
2572                 return -EFSCORRUPTED;
2573         }
2574
2575         segment_count = le32_to_cpu(raw_super->segment_count);
2576         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2577         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2578         total_sections = le32_to_cpu(raw_super->section_count);
2579
2580         /* blocks_per_seg should be 512, given the above check */
2581         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2582
2583         if (segment_count > F2FS_MAX_SEGMENT ||
2584                                 segment_count < F2FS_MIN_SEGMENTS) {
2585                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2586                 return -EFSCORRUPTED;
2587         }
2588
2589         if (total_sections > segment_count ||
2590                         total_sections < F2FS_MIN_SEGMENTS ||
2591                         segs_per_sec > segment_count || !segs_per_sec) {
2592                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2593                           segment_count, total_sections, segs_per_sec);
2594                 return -EFSCORRUPTED;
2595         }
2596
2597         if ((segment_count / segs_per_sec) < total_sections) {
2598                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2599                           segment_count, segs_per_sec, total_sections);
2600                 return -EFSCORRUPTED;
2601         }
2602
2603         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2604                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2605                           segment_count, le64_to_cpu(raw_super->block_count));
2606                 return -EFSCORRUPTED;
2607         }
2608
2609         if (secs_per_zone > total_sections || !secs_per_zone) {
2610                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2611                           secs_per_zone, total_sections);
2612                 return -EFSCORRUPTED;
2613         }
2614         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2615                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2616                         (le32_to_cpu(raw_super->extension_count) +
2617                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2618                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2619                           le32_to_cpu(raw_super->extension_count),
2620                           raw_super->hot_ext_count,
2621                           F2FS_MAX_EXTENSION);
2622                 return -EFSCORRUPTED;
2623         }
2624
2625         if (le32_to_cpu(raw_super->cp_payload) >
2626                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2627                 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2628                           le32_to_cpu(raw_super->cp_payload),
2629                           blocks_per_seg - F2FS_CP_PACKS);
2630                 return -EFSCORRUPTED;
2631         }
2632
2633         /* check reserved ino info */
2634         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2635                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2636                 le32_to_cpu(raw_super->root_ino) != 3) {
2637                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2638                           le32_to_cpu(raw_super->node_ino),
2639                           le32_to_cpu(raw_super->meta_ino),
2640                           le32_to_cpu(raw_super->root_ino));
2641                 return -EFSCORRUPTED;
2642         }
2643
2644         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2645         if (sanity_check_area_boundary(sbi, bh))
2646                 return -EFSCORRUPTED;
2647
2648         return 0;
2649 }
2650
2651 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2652 {
2653         unsigned int total, fsmeta;
2654         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2655         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2656         unsigned int ovp_segments, reserved_segments;
2657         unsigned int main_segs, blocks_per_seg;
2658         unsigned int sit_segs, nat_segs;
2659         unsigned int sit_bitmap_size, nat_bitmap_size;
2660         unsigned int log_blocks_per_seg;
2661         unsigned int segment_count_main;
2662         unsigned int cp_pack_start_sum, cp_payload;
2663         block_t user_block_count, valid_user_blocks;
2664         block_t avail_node_count, valid_node_count;
2665         int i, j;
2666
2667         total = le32_to_cpu(raw_super->segment_count);
2668         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2669         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2670         fsmeta += sit_segs;
2671         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2672         fsmeta += nat_segs;
2673         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2674         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2675
2676         if (unlikely(fsmeta >= total))
2677                 return 1;
2678
2679         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2680         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2681
2682         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2683                         ovp_segments == 0 || reserved_segments == 0)) {
2684                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2685                 return 1;
2686         }
2687
2688         user_block_count = le64_to_cpu(ckpt->user_block_count);
2689         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2690         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2691         if (!user_block_count || user_block_count >=
2692                         segment_count_main << log_blocks_per_seg) {
2693                 f2fs_err(sbi, "Wrong user_block_count: %u",
2694                          user_block_count);
2695                 return 1;
2696         }
2697
2698         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2699         if (valid_user_blocks > user_block_count) {
2700                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2701                          valid_user_blocks, user_block_count);
2702                 return 1;
2703         }
2704
2705         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2706         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2707         if (valid_node_count > avail_node_count) {
2708                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2709                          valid_node_count, avail_node_count);
2710                 return 1;
2711         }
2712
2713         main_segs = le32_to_cpu(raw_super->segment_count_main);
2714         blocks_per_seg = sbi->blocks_per_seg;
2715
2716         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2717                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2718                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2719                         return 1;
2720                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2721                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2722                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2723                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2724                                          i, j,
2725                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2726                                 return 1;
2727                         }
2728                 }
2729         }
2730         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2731                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2732                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2733                         return 1;
2734                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2735                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2736                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2737                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2738                                          i, j,
2739                                          le32_to_cpu(ckpt->cur_data_segno[i]));
2740                                 return 1;
2741                         }
2742                 }
2743         }
2744         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2745                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2746                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2747                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2748                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2749                                          i, j,
2750                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2751                                 return 1;
2752                         }
2753                 }
2754         }
2755
2756         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2757         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2758
2759         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2760                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2761                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2762                          sit_bitmap_size, nat_bitmap_size);
2763                 return 1;
2764         }
2765
2766         cp_pack_start_sum = __start_sum_addr(sbi);
2767         cp_payload = __cp_payload(sbi);
2768         if (cp_pack_start_sum < cp_payload + 1 ||
2769                 cp_pack_start_sum > blocks_per_seg - 1 -
2770                         NR_CURSEG_TYPE) {
2771                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2772                          cp_pack_start_sum);
2773                 return 1;
2774         }
2775
2776         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2777                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2778                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2779                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2780                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2781                           le32_to_cpu(ckpt->checksum_offset));
2782                 return 1;
2783         }
2784
2785         if (unlikely(f2fs_cp_error(sbi))) {
2786                 f2fs_err(sbi, "A bug case: need to run fsck");
2787                 return 1;
2788         }
2789         return 0;
2790 }
2791
2792 static void init_sb_info(struct f2fs_sb_info *sbi)
2793 {
2794         struct f2fs_super_block *raw_super = sbi->raw_super;
2795         int i;
2796
2797         sbi->log_sectors_per_block =
2798                 le32_to_cpu(raw_super->log_sectors_per_block);
2799         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2800         sbi->blocksize = 1 << sbi->log_blocksize;
2801         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2802         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2803         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2804         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2805         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2806         sbi->total_node_count =
2807                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2808                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2809         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2810         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2811         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2812         sbi->cur_victim_sec = NULL_SECNO;
2813         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2814         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2815         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2816         sbi->migration_granularity = sbi->segs_per_sec;
2817
2818         sbi->dir_level = DEF_DIR_LEVEL;
2819         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2820         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2821         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2822         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2823         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2824         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2825                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2826         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2827
2828         for (i = 0; i < NR_COUNT_TYPE; i++)
2829                 atomic_set(&sbi->nr_pages[i], 0);
2830
2831         for (i = 0; i < META; i++)
2832                 atomic_set(&sbi->wb_sync_req[i], 0);
2833
2834         INIT_LIST_HEAD(&sbi->s_list);
2835         mutex_init(&sbi->umount_mutex);
2836         init_rwsem(&sbi->io_order_lock);
2837         spin_lock_init(&sbi->cp_lock);
2838
2839         sbi->dirty_device = 0;
2840         spin_lock_init(&sbi->dev_lock);
2841
2842         init_rwsem(&sbi->sb_lock);
2843 }
2844
2845 static int init_percpu_info(struct f2fs_sb_info *sbi)
2846 {
2847         int err;
2848
2849         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2850         if (err)
2851                 return err;
2852
2853         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2854                                                                 GFP_KERNEL);
2855         if (err)
2856                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2857
2858         return err;
2859 }
2860
2861 #ifdef CONFIG_BLK_DEV_ZONED
2862 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2863 {
2864         struct block_device *bdev = FDEV(devi).bdev;
2865         sector_t nr_sectors = bdev->bd_part->nr_sects;
2866         sector_t sector = 0;
2867         struct blk_zone *zones;
2868         unsigned int i, nr_zones;
2869         unsigned int n = 0;
2870         int err = -EIO;
2871
2872         if (!f2fs_sb_has_blkzoned(sbi))
2873                 return 0;
2874
2875         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2876                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2877                 return -EINVAL;
2878         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2879         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2880                                 __ilog2_u32(sbi->blocks_per_blkz))
2881                 return -EINVAL;
2882         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2883         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2884                                         sbi->log_blocks_per_blkz;
2885         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2886                 FDEV(devi).nr_blkz++;
2887
2888         FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
2889                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
2890                                         * sizeof(unsigned long),
2891                                         GFP_KERNEL);
2892         if (!FDEV(devi).blkz_seq)
2893                 return -ENOMEM;
2894
2895 #define F2FS_REPORT_NR_ZONES   4096
2896
2897         zones = f2fs_kzalloc(sbi,
2898                              array_size(F2FS_REPORT_NR_ZONES,
2899                                         sizeof(struct blk_zone)),
2900                              GFP_KERNEL);
2901         if (!zones)
2902                 return -ENOMEM;
2903
2904         /* Get block zones type */
2905         while (zones && sector < nr_sectors) {
2906
2907                 nr_zones = F2FS_REPORT_NR_ZONES;
2908                 err = blkdev_report_zones(bdev, sector, zones, &nr_zones);
2909                 if (err)
2910                         break;
2911                 if (!nr_zones) {
2912                         err = -EIO;
2913                         break;
2914                 }
2915
2916                 for (i = 0; i < nr_zones; i++) {
2917                         if (zones[i].type != BLK_ZONE_TYPE_CONVENTIONAL)
2918                                 set_bit(n, FDEV(devi).blkz_seq);
2919                         sector += zones[i].len;
2920                         n++;
2921                 }
2922         }
2923
2924         kvfree(zones);
2925
2926         return err;
2927 }
2928 #endif
2929
2930 /*
2931  * Read f2fs raw super block.
2932  * Because we have two copies of super block, so read both of them
2933  * to get the first valid one. If any one of them is broken, we pass
2934  * them recovery flag back to the caller.
2935  */
2936 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2937                         struct f2fs_super_block **raw_super,
2938                         int *valid_super_block, int *recovery)
2939 {
2940         struct super_block *sb = sbi->sb;
2941         int block;
2942         struct buffer_head *bh;
2943         struct f2fs_super_block *super;
2944         int err = 0;
2945
2946         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2947         if (!super)
2948                 return -ENOMEM;
2949
2950         for (block = 0; block < 2; block++) {
2951                 bh = sb_bread(sb, block);
2952                 if (!bh) {
2953                         f2fs_err(sbi, "Unable to read %dth superblock",
2954                                  block + 1);
2955                         err = -EIO;
2956                         continue;
2957                 }
2958
2959                 /* sanity checking of raw super */
2960                 err = sanity_check_raw_super(sbi, bh);
2961                 if (err) {
2962                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
2963                                  block + 1);
2964                         brelse(bh);
2965                         continue;
2966                 }
2967
2968                 if (!*raw_super) {
2969                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2970                                                         sizeof(*super));
2971                         *valid_super_block = block;
2972                         *raw_super = super;
2973                 }
2974                 brelse(bh);
2975         }
2976
2977         /* Fail to read any one of the superblocks*/
2978         if (err < 0)
2979                 *recovery = 1;
2980
2981         /* No valid superblock */
2982         if (!*raw_super)
2983                 kvfree(super);
2984         else
2985                 err = 0;
2986
2987         return err;
2988 }
2989
2990 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2991 {
2992         struct buffer_head *bh;
2993         __u32 crc = 0;
2994         int err;
2995
2996         if ((recover && f2fs_readonly(sbi->sb)) ||
2997                                 bdev_read_only(sbi->sb->s_bdev)) {
2998                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2999                 return -EROFS;
3000         }
3001
3002         /* we should update superblock crc here */
3003         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3004                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3005                                 offsetof(struct f2fs_super_block, crc));
3006                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3007         }
3008
3009         /* write back-up superblock first */
3010         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3011         if (!bh)
3012                 return -EIO;
3013         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3014         brelse(bh);
3015
3016         /* if we are in recovery path, skip writing valid superblock */
3017         if (recover || err)
3018                 return err;
3019
3020         /* write current valid superblock */
3021         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3022         if (!bh)
3023                 return -EIO;
3024         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3025         brelse(bh);
3026         return err;
3027 }
3028
3029 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3030 {
3031         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3032         unsigned int max_devices = MAX_DEVICES;
3033         int i;
3034
3035         /* Initialize single device information */
3036         if (!RDEV(0).path[0]) {
3037                 if (!bdev_is_zoned(sbi->sb->s_bdev))
3038                         return 0;
3039                 max_devices = 1;
3040         }
3041
3042         /*
3043          * Initialize multiple devices information, or single
3044          * zoned block device information.
3045          */
3046         sbi->devs = f2fs_kzalloc(sbi,
3047                                  array_size(max_devices,
3048                                             sizeof(struct f2fs_dev_info)),
3049                                  GFP_KERNEL);
3050         if (!sbi->devs)
3051                 return -ENOMEM;
3052
3053         for (i = 0; i < max_devices; i++) {
3054
3055                 if (i > 0 && !RDEV(i).path[0])
3056                         break;
3057
3058                 if (max_devices == 1) {
3059                         /* Single zoned block device mount */
3060                         FDEV(0).bdev =
3061                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3062                                         sbi->sb->s_mode, sbi->sb->s_type);
3063                 } else {
3064                         /* Multi-device mount */
3065                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3066                         FDEV(i).total_segments =
3067                                 le32_to_cpu(RDEV(i).total_segments);
3068                         if (i == 0) {
3069                                 FDEV(i).start_blk = 0;
3070                                 FDEV(i).end_blk = FDEV(i).start_blk +
3071                                     (FDEV(i).total_segments <<
3072                                     sbi->log_blocks_per_seg) - 1 +
3073                                     le32_to_cpu(raw_super->segment0_blkaddr);
3074                         } else {
3075                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3076                                 FDEV(i).end_blk = FDEV(i).start_blk +
3077                                         (FDEV(i).total_segments <<
3078                                         sbi->log_blocks_per_seg) - 1;
3079                         }
3080                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3081                                         sbi->sb->s_mode, sbi->sb->s_type);
3082                 }
3083                 if (IS_ERR(FDEV(i).bdev))
3084                         return PTR_ERR(FDEV(i).bdev);
3085
3086                 /* to release errored devices */
3087                 sbi->s_ndevs = i + 1;
3088
3089 #ifdef CONFIG_BLK_DEV_ZONED
3090                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3091                                 !f2fs_sb_has_blkzoned(sbi)) {
3092                         f2fs_err(sbi, "Zoned block device feature not enabled\n");
3093                         return -EINVAL;
3094                 }
3095                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3096                         if (init_blkz_info(sbi, i)) {
3097                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3098                                 return -EINVAL;
3099                         }
3100                         if (max_devices == 1)
3101                                 break;
3102                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3103                                   i, FDEV(i).path,
3104                                   FDEV(i).total_segments,
3105                                   FDEV(i).start_blk, FDEV(i).end_blk,
3106                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3107                                   "Host-aware" : "Host-managed");
3108                         continue;
3109                 }
3110 #endif
3111                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3112                           i, FDEV(i).path,
3113                           FDEV(i).total_segments,
3114                           FDEV(i).start_blk, FDEV(i).end_blk);
3115         }
3116         f2fs_info(sbi,
3117                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3118         return 0;
3119 }
3120
3121 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3122 {
3123 #ifdef CONFIG_UNICODE
3124         if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3125                 const struct f2fs_sb_encodings *encoding_info;
3126                 struct unicode_map *encoding;
3127                 __u16 encoding_flags;
3128
3129                 if (f2fs_sb_has_encrypt(sbi)) {
3130                         f2fs_err(sbi,
3131                                 "Can't mount with encoding and encryption");
3132                         return -EINVAL;
3133                 }
3134
3135                 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3136                                           &encoding_flags)) {
3137                         f2fs_err(sbi,
3138                                  "Encoding requested by superblock is unknown");
3139                         return -EINVAL;
3140                 }
3141
3142                 encoding = utf8_load(encoding_info->version);
3143                 if (IS_ERR(encoding)) {
3144                         f2fs_err(sbi,
3145                                  "can't mount with superblock charset: %s-%s "
3146                                  "not supported by the kernel. flags: 0x%x.",
3147                                  encoding_info->name, encoding_info->version,
3148                                  encoding_flags);
3149                         return PTR_ERR(encoding);
3150                 }
3151                 f2fs_info(sbi, "Using encoding defined by superblock: "
3152                          "%s-%s with flags 0x%hx", encoding_info->name,
3153                          encoding_info->version?:"\b", encoding_flags);
3154
3155                 sbi->s_encoding = encoding;
3156                 sbi->s_encoding_flags = encoding_flags;
3157                 sbi->sb->s_d_op = &f2fs_dentry_ops;
3158         }
3159 #else
3160         if (f2fs_sb_has_casefold(sbi)) {
3161                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3162                 return -EINVAL;
3163         }
3164 #endif
3165         return 0;
3166 }
3167
3168 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3169 {
3170         struct f2fs_sm_info *sm_i = SM_I(sbi);
3171
3172         /* adjust parameters according to the volume size */
3173         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3174                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3175                 sm_i->dcc_info->discard_granularity = 1;
3176                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3177         }
3178
3179         sbi->readdir_ra = 1;
3180 }
3181
3182 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3183 {
3184         struct f2fs_sb_info *sbi;
3185         struct f2fs_super_block *raw_super;
3186         struct inode *root;
3187         int err;
3188         bool skip_recovery = false, need_fsck = false;
3189         char *options = NULL;
3190         int recovery, i, valid_super_block;
3191         struct curseg_info *seg_i;
3192         int retry_cnt = 1;
3193
3194 try_onemore:
3195         err = -EINVAL;
3196         raw_super = NULL;
3197         valid_super_block = -1;
3198         recovery = 0;
3199
3200         /* allocate memory for f2fs-specific super block info */
3201         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3202         if (!sbi)
3203                 return -ENOMEM;
3204
3205         sbi->sb = sb;
3206
3207         /* Load the checksum driver */
3208         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3209         if (IS_ERR(sbi->s_chksum_driver)) {
3210                 f2fs_err(sbi, "Cannot load crc32 driver.");
3211                 err = PTR_ERR(sbi->s_chksum_driver);
3212                 sbi->s_chksum_driver = NULL;
3213                 goto free_sbi;
3214         }
3215
3216         /* set a block size */
3217         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3218                 f2fs_err(sbi, "unable to set blocksize");
3219                 goto free_sbi;
3220         }
3221
3222         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3223                                                                 &recovery);
3224         if (err)
3225                 goto free_sbi;
3226
3227         sb->s_fs_info = sbi;
3228         sbi->raw_super = raw_super;
3229
3230         /* precompute checksum seed for metadata */
3231         if (f2fs_sb_has_inode_chksum(sbi))
3232                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3233                                                 sizeof(raw_super->uuid));
3234
3235         /*
3236          * The BLKZONED feature indicates that the drive was formatted with
3237          * zone alignment optimization. This is optional for host-aware
3238          * devices, but mandatory for host-managed zoned block devices.
3239          */
3240 #ifndef CONFIG_BLK_DEV_ZONED
3241         if (f2fs_sb_has_blkzoned(sbi)) {
3242                 f2fs_err(sbi, "Zoned block device support is not enabled");
3243                 err = -EOPNOTSUPP;
3244                 goto free_sb_buf;
3245         }
3246 #endif
3247         default_options(sbi);
3248         /* parse mount options */
3249         options = kstrdup((const char *)data, GFP_KERNEL);
3250         if (data && !options) {
3251                 err = -ENOMEM;
3252                 goto free_sb_buf;
3253         }
3254
3255         err = parse_options(sb, options);
3256         if (err)
3257                 goto free_options;
3258
3259         sbi->max_file_blocks = max_file_blocks();
3260         sb->s_maxbytes = sbi->max_file_blocks <<
3261                                 le32_to_cpu(raw_super->log_blocksize);
3262         sb->s_max_links = F2FS_LINK_MAX;
3263
3264         err = f2fs_setup_casefold(sbi);
3265         if (err)
3266                 goto free_options;
3267
3268 #ifdef CONFIG_QUOTA
3269         sb->dq_op = &f2fs_quota_operations;
3270         sb->s_qcop = &f2fs_quotactl_ops;
3271         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3272
3273         if (f2fs_sb_has_quota_ino(sbi)) {
3274                 for (i = 0; i < MAXQUOTAS; i++) {
3275                         if (f2fs_qf_ino(sbi->sb, i))
3276                                 sbi->nquota_files++;
3277                 }
3278         }
3279 #endif
3280
3281         sb->s_op = &f2fs_sops;
3282 #ifdef CONFIG_FS_ENCRYPTION
3283         sb->s_cop = &f2fs_cryptops;
3284 #endif
3285 #ifdef CONFIG_FS_VERITY
3286         sb->s_vop = &f2fs_verityops;
3287 #endif
3288         sb->s_xattr = f2fs_xattr_handlers;
3289         sb->s_export_op = &f2fs_export_ops;
3290         sb->s_magic = F2FS_SUPER_MAGIC;
3291         sb->s_time_gran = 1;
3292         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3293                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3294         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3295         sb->s_iflags |= SB_I_CGROUPWB;
3296
3297         /* init f2fs-specific super block info */
3298         sbi->valid_super_block = valid_super_block;
3299         mutex_init(&sbi->gc_mutex);
3300         mutex_init(&sbi->writepages);
3301         mutex_init(&sbi->cp_mutex);
3302         mutex_init(&sbi->resize_mutex);
3303         init_rwsem(&sbi->node_write);
3304         init_rwsem(&sbi->node_change);
3305
3306         /* disallow all the data/node/meta page writes */
3307         set_sbi_flag(sbi, SBI_POR_DOING);
3308         spin_lock_init(&sbi->stat_lock);
3309
3310         /* init iostat info */
3311         spin_lock_init(&sbi->iostat_lock);
3312         sbi->iostat_enable = false;
3313
3314         for (i = 0; i < NR_PAGE_TYPE; i++) {
3315                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3316                 int j;
3317
3318                 sbi->write_io[i] =
3319                         f2fs_kmalloc(sbi,
3320                                      array_size(n,
3321                                                 sizeof(struct f2fs_bio_info)),
3322                                      GFP_KERNEL);
3323                 if (!sbi->write_io[i]) {
3324                         err = -ENOMEM;
3325                         goto free_bio_info;
3326                 }
3327
3328                 for (j = HOT; j < n; j++) {
3329                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3330                         sbi->write_io[i][j].sbi = sbi;
3331                         sbi->write_io[i][j].bio = NULL;
3332                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3333                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3334                 }
3335         }
3336
3337         init_rwsem(&sbi->cp_rwsem);
3338         init_rwsem(&sbi->quota_sem);
3339         init_waitqueue_head(&sbi->cp_wait);
3340         init_sb_info(sbi);
3341
3342         err = init_percpu_info(sbi);
3343         if (err)
3344                 goto free_bio_info;
3345
3346         if (F2FS_IO_ALIGNED(sbi)) {
3347                 sbi->write_io_dummy =
3348                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3349                 if (!sbi->write_io_dummy) {
3350                         err = -ENOMEM;
3351                         goto free_percpu;
3352                 }
3353         }
3354
3355         /* get an inode for meta space */
3356         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3357         if (IS_ERR(sbi->meta_inode)) {
3358                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3359                 err = PTR_ERR(sbi->meta_inode);
3360                 goto free_io_dummy;
3361         }
3362
3363         err = f2fs_get_valid_checkpoint(sbi);
3364         if (err) {
3365                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3366                 goto free_meta_inode;
3367         }
3368
3369         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3370                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3371         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3372                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3373                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3374         }
3375
3376         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3377                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3378
3379         /* Initialize device list */
3380         err = f2fs_scan_devices(sbi);
3381         if (err) {
3382                 f2fs_err(sbi, "Failed to find devices");
3383                 goto free_devices;
3384         }
3385
3386         sbi->total_valid_node_count =
3387                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3388         percpu_counter_set(&sbi->total_valid_inode_count,
3389                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3390         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3391         sbi->total_valid_block_count =
3392                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3393         sbi->last_valid_block_count = sbi->total_valid_block_count;
3394         sbi->reserved_blocks = 0;
3395         sbi->current_reserved_blocks = 0;
3396         limit_reserve_root(sbi);
3397
3398         for (i = 0; i < NR_INODE_TYPE; i++) {
3399                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3400                 spin_lock_init(&sbi->inode_lock[i]);
3401         }
3402         mutex_init(&sbi->flush_lock);
3403
3404         f2fs_init_extent_cache_info(sbi);
3405
3406         f2fs_init_ino_entry_info(sbi);
3407
3408         f2fs_init_fsync_node_info(sbi);
3409
3410         /* setup f2fs internal modules */
3411         err = f2fs_build_segment_manager(sbi);
3412         if (err) {
3413                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3414                          err);
3415                 goto free_sm;
3416         }
3417         err = f2fs_build_node_manager(sbi);
3418         if (err) {
3419                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3420                          err);
3421                 goto free_nm;
3422         }
3423
3424         /* For write statistics */
3425         if (sb->s_bdev->bd_part)
3426                 sbi->sectors_written_start =
3427                         (u64)part_stat_read(sb->s_bdev->bd_part,
3428                                             sectors[STAT_WRITE]);
3429
3430         /* Read accumulated write IO statistics if exists */
3431         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3432         if (__exist_node_summaries(sbi))
3433                 sbi->kbytes_written =
3434                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3435
3436         f2fs_build_gc_manager(sbi);
3437
3438         err = f2fs_build_stats(sbi);
3439         if (err)
3440                 goto free_nm;
3441
3442         /* get an inode for node space */
3443         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3444         if (IS_ERR(sbi->node_inode)) {
3445                 f2fs_err(sbi, "Failed to read node inode");
3446                 err = PTR_ERR(sbi->node_inode);
3447                 goto free_stats;
3448         }
3449
3450         /* read root inode and dentry */
3451         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3452         if (IS_ERR(root)) {
3453                 f2fs_err(sbi, "Failed to read root inode");
3454                 err = PTR_ERR(root);
3455                 goto free_node_inode;
3456         }
3457         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3458                         !root->i_size || !root->i_nlink) {
3459                 iput(root);
3460                 err = -EINVAL;
3461                 goto free_node_inode;
3462         }
3463
3464         sb->s_root = d_make_root(root); /* allocate root dentry */
3465         if (!sb->s_root) {
3466                 err = -ENOMEM;
3467                 goto free_node_inode;
3468         }
3469
3470         err = f2fs_register_sysfs(sbi);
3471         if (err)
3472                 goto free_root_inode;
3473
3474 #ifdef CONFIG_QUOTA
3475         /* Enable quota usage during mount */
3476         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3477                 err = f2fs_enable_quotas(sb);
3478                 if (err)
3479                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3480         }
3481 #endif
3482         /* if there are nt orphan nodes free them */
3483         err = f2fs_recover_orphan_inodes(sbi);
3484         if (err)
3485                 goto free_meta;
3486
3487         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3488                 goto reset_checkpoint;
3489
3490         /* recover fsynced data */
3491         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3492                 /*
3493                  * mount should be failed, when device has readonly mode, and
3494                  * previous checkpoint was not done by clean system shutdown.
3495                  */
3496                 if (f2fs_hw_is_readonly(sbi)) {
3497                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3498                                 err = -EROFS;
3499                                 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3500                                 goto free_meta;
3501                         }
3502                         f2fs_info(sbi, "write access unavailable, skipping recovery");
3503                         goto reset_checkpoint;
3504                 }
3505
3506                 if (need_fsck)
3507                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3508
3509                 if (skip_recovery)
3510                         goto reset_checkpoint;
3511
3512                 err = f2fs_recover_fsync_data(sbi, false);
3513                 if (err < 0) {
3514                         if (err != -ENOMEM)
3515                                 skip_recovery = true;
3516                         need_fsck = true;
3517                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3518                                  err);
3519                         goto free_meta;
3520                 }
3521         } else {
3522                 err = f2fs_recover_fsync_data(sbi, true);
3523
3524                 if (!f2fs_readonly(sb) && err > 0) {
3525                         err = -EINVAL;
3526                         f2fs_err(sbi, "Need to recover fsync data");
3527                         goto free_meta;
3528                 }
3529         }
3530 reset_checkpoint:
3531         /* f2fs_recover_fsync_data() cleared this already */
3532         clear_sbi_flag(sbi, SBI_POR_DOING);
3533
3534         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3535                 err = f2fs_disable_checkpoint(sbi);
3536                 if (err)
3537                         goto sync_free_meta;
3538         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3539                 f2fs_enable_checkpoint(sbi);
3540         }
3541
3542         /*
3543          * If filesystem is not mounted as read-only then
3544          * do start the gc_thread.
3545          */
3546         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3547                 /* After POR, we can run background GC thread.*/
3548                 err = f2fs_start_gc_thread(sbi);
3549                 if (err)
3550                         goto sync_free_meta;
3551         }
3552         kvfree(options);
3553
3554         /* recover broken superblock */
3555         if (recovery) {
3556                 err = f2fs_commit_super(sbi, true);
3557                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3558                           sbi->valid_super_block ? 1 : 2, err);
3559         }
3560
3561         f2fs_join_shrinker(sbi);
3562
3563         f2fs_tuning_parameters(sbi);
3564
3565         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3566                     cur_cp_version(F2FS_CKPT(sbi)));
3567         f2fs_update_time(sbi, CP_TIME);
3568         f2fs_update_time(sbi, REQ_TIME);
3569         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3570         return 0;
3571
3572 sync_free_meta:
3573         /* safe to flush all the data */
3574         sync_filesystem(sbi->sb);
3575         retry_cnt = 0;
3576
3577 free_meta:
3578 #ifdef CONFIG_QUOTA
3579         f2fs_truncate_quota_inode_pages(sb);
3580         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3581                 f2fs_quota_off_umount(sbi->sb);
3582 #endif
3583         /*
3584          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3585          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3586          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3587          * falls into an infinite loop in f2fs_sync_meta_pages().
3588          */
3589         truncate_inode_pages_final(META_MAPPING(sbi));
3590         /* evict some inodes being cached by GC */
3591         evict_inodes(sb);
3592         f2fs_unregister_sysfs(sbi);
3593 free_root_inode:
3594         dput(sb->s_root);
3595         sb->s_root = NULL;
3596 free_node_inode:
3597         f2fs_release_ino_entry(sbi, true);
3598         truncate_inode_pages_final(NODE_MAPPING(sbi));
3599         iput(sbi->node_inode);
3600         sbi->node_inode = NULL;
3601 free_stats:
3602         f2fs_destroy_stats(sbi);
3603 free_nm:
3604         f2fs_destroy_node_manager(sbi);
3605 free_sm:
3606         f2fs_destroy_segment_manager(sbi);
3607 free_devices:
3608         destroy_device_list(sbi);
3609         kvfree(sbi->ckpt);
3610 free_meta_inode:
3611         make_bad_inode(sbi->meta_inode);
3612         iput(sbi->meta_inode);
3613         sbi->meta_inode = NULL;
3614 free_io_dummy:
3615         mempool_destroy(sbi->write_io_dummy);
3616 free_percpu:
3617         destroy_percpu_info(sbi);
3618 free_bio_info:
3619         for (i = 0; i < NR_PAGE_TYPE; i++)
3620                 kvfree(sbi->write_io[i]);
3621
3622 #ifdef CONFIG_UNICODE
3623         utf8_unload(sbi->s_encoding);
3624 #endif
3625 free_options:
3626 #ifdef CONFIG_QUOTA
3627         for (i = 0; i < MAXQUOTAS; i++)
3628                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3629 #endif
3630         kvfree(options);
3631 free_sb_buf:
3632         kvfree(raw_super);
3633 free_sbi:
3634         if (sbi->s_chksum_driver)
3635                 crypto_free_shash(sbi->s_chksum_driver);
3636         kvfree(sbi);
3637
3638         /* give only one another chance */
3639         if (retry_cnt > 0 && skip_recovery) {
3640                 retry_cnt--;
3641                 shrink_dcache_sb(sb);
3642                 goto try_onemore;
3643         }
3644         return err;
3645 }
3646
3647 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3648                         const char *dev_name, void *data)
3649 {
3650         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3651 }
3652
3653 static void kill_f2fs_super(struct super_block *sb)
3654 {
3655         if (sb->s_root) {
3656                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3657
3658                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3659                 f2fs_stop_gc_thread(sbi);
3660                 f2fs_stop_discard_thread(sbi);
3661
3662                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3663                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3664                         struct cp_control cpc = {
3665                                 .reason = CP_UMOUNT,
3666                         };
3667                         f2fs_write_checkpoint(sbi, &cpc);
3668                 }
3669
3670                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3671                         sb->s_flags &= ~SB_RDONLY;
3672         }
3673         kill_block_super(sb);
3674 }
3675
3676 static struct file_system_type f2fs_fs_type = {
3677         .owner          = THIS_MODULE,
3678         .name           = "f2fs",
3679         .mount          = f2fs_mount,
3680         .kill_sb        = kill_f2fs_super,
3681         .fs_flags       = FS_REQUIRES_DEV,
3682 };
3683 MODULE_ALIAS_FS("f2fs");
3684
3685 static int __init init_inodecache(void)
3686 {
3687         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3688                         sizeof(struct f2fs_inode_info), 0,
3689                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3690         if (!f2fs_inode_cachep)
3691                 return -ENOMEM;
3692         return 0;
3693 }
3694
3695 static void destroy_inodecache(void)
3696 {
3697         /*
3698          * Make sure all delayed rcu free inodes are flushed before we
3699          * destroy cache.
3700          */
3701         rcu_barrier();
3702         kmem_cache_destroy(f2fs_inode_cachep);
3703 }
3704
3705 static int __init init_f2fs_fs(void)
3706 {
3707         int err;
3708
3709         if (PAGE_SIZE != F2FS_BLKSIZE) {
3710                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3711                                 PAGE_SIZE, F2FS_BLKSIZE);
3712                 return -EINVAL;
3713         }
3714
3715         f2fs_build_trace_ios();
3716
3717         err = init_inodecache();
3718         if (err)
3719                 goto fail;
3720         err = f2fs_create_node_manager_caches();
3721         if (err)
3722                 goto free_inodecache;
3723         err = f2fs_create_segment_manager_caches();
3724         if (err)
3725                 goto free_node_manager_caches;
3726         err = f2fs_create_checkpoint_caches();
3727         if (err)
3728                 goto free_segment_manager_caches;
3729         err = f2fs_create_extent_cache();
3730         if (err)
3731                 goto free_checkpoint_caches;
3732         err = f2fs_init_sysfs();
3733         if (err)
3734                 goto free_extent_cache;
3735         err = register_shrinker(&f2fs_shrinker_info);
3736         if (err)
3737                 goto free_sysfs;
3738         err = register_filesystem(&f2fs_fs_type);
3739         if (err)
3740                 goto free_shrinker;
3741         f2fs_create_root_stats();
3742         err = f2fs_init_post_read_processing();
3743         if (err)
3744                 goto free_root_stats;
3745         return 0;
3746
3747 free_root_stats:
3748         f2fs_destroy_root_stats();
3749         unregister_filesystem(&f2fs_fs_type);
3750 free_shrinker:
3751         unregister_shrinker(&f2fs_shrinker_info);
3752 free_sysfs:
3753         f2fs_exit_sysfs();
3754 free_extent_cache:
3755         f2fs_destroy_extent_cache();
3756 free_checkpoint_caches:
3757         f2fs_destroy_checkpoint_caches();
3758 free_segment_manager_caches:
3759         f2fs_destroy_segment_manager_caches();
3760 free_node_manager_caches:
3761         f2fs_destroy_node_manager_caches();
3762 free_inodecache:
3763         destroy_inodecache();
3764 fail:
3765         return err;
3766 }
3767
3768 static void __exit exit_f2fs_fs(void)
3769 {
3770         f2fs_destroy_post_read_processing();
3771         f2fs_destroy_root_stats();
3772         unregister_filesystem(&f2fs_fs_type);
3773         unregister_shrinker(&f2fs_shrinker_info);
3774         f2fs_exit_sysfs();
3775         f2fs_destroy_extent_cache();
3776         f2fs_destroy_checkpoint_caches();
3777         f2fs_destroy_segment_manager_caches();
3778         f2fs_destroy_node_manager_caches();
3779         destroy_inodecache();
3780         f2fs_destroy_trace_ios();
3781 }
3782
3783 module_init(init_f2fs_fs)
3784 module_exit(exit_f2fs_fs)
3785
3786 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3787 MODULE_DESCRIPTION("Flash Friendly File System");
3788 MODULE_LICENSE("GPL");
3789