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 = 0;
1217         if (dquot->dq_dqb.dqb_bsoftlimit)
1218                 limit = dquot->dq_dqb.dqb_bsoftlimit;
1219         if (dquot->dq_dqb.dqb_bhardlimit &&
1220                         (!limit || dquot->dq_dqb.dqb_bhardlimit < limit))
1221                 limit = dquot->dq_dqb.dqb_bhardlimit;
1222
1223         if (limit && buf->f_blocks > limit) {
1224                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1225                 buf->f_blocks = limit;
1226                 buf->f_bfree = buf->f_bavail =
1227                         (buf->f_blocks > curblock) ?
1228                          (buf->f_blocks - curblock) : 0;
1229         }
1230
1231         limit = 0;
1232         if (dquot->dq_dqb.dqb_isoftlimit)
1233                 limit = dquot->dq_dqb.dqb_isoftlimit;
1234         if (dquot->dq_dqb.dqb_ihardlimit &&
1235                         (!limit || dquot->dq_dqb.dqb_ihardlimit < limit))
1236                 limit = dquot->dq_dqb.dqb_ihardlimit;
1237
1238         if (limit && buf->f_files > limit) {
1239                 buf->f_files = limit;
1240                 buf->f_ffree =
1241                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1242                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1243         }
1244
1245         spin_unlock(&dquot->dq_dqb_lock);
1246         dqput(dquot);
1247         return 0;
1248 }
1249 #endif
1250
1251 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1252 {
1253         struct super_block *sb = dentry->d_sb;
1254         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1255         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1256         block_t total_count, user_block_count, start_count;
1257         u64 avail_node_count;
1258
1259         total_count = le64_to_cpu(sbi->raw_super->block_count);
1260         user_block_count = sbi->user_block_count;
1261         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1262         buf->f_type = F2FS_SUPER_MAGIC;
1263         buf->f_bsize = sbi->blocksize;
1264
1265         buf->f_blocks = total_count - start_count;
1266         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1267                                                 sbi->current_reserved_blocks;
1268
1269         spin_lock(&sbi->stat_lock);
1270         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1271                 buf->f_bfree = 0;
1272         else
1273                 buf->f_bfree -= sbi->unusable_block_count;
1274         spin_unlock(&sbi->stat_lock);
1275
1276         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1277                 buf->f_bavail = buf->f_bfree -
1278                                 F2FS_OPTION(sbi).root_reserved_blocks;
1279         else
1280                 buf->f_bavail = 0;
1281
1282         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1283
1284         if (avail_node_count > user_block_count) {
1285                 buf->f_files = user_block_count;
1286                 buf->f_ffree = buf->f_bavail;
1287         } else {
1288                 buf->f_files = avail_node_count;
1289                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1290                                         buf->f_bavail);
1291         }
1292
1293         buf->f_namelen = F2FS_NAME_LEN;
1294         buf->f_fsid.val[0] = (u32)id;
1295         buf->f_fsid.val[1] = (u32)(id >> 32);
1296
1297 #ifdef CONFIG_QUOTA
1298         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1299                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1300                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1301         }
1302 #endif
1303         return 0;
1304 }
1305
1306 static inline void f2fs_show_quota_options(struct seq_file *seq,
1307                                            struct super_block *sb)
1308 {
1309 #ifdef CONFIG_QUOTA
1310         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1311
1312         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1313                 char *fmtname = "";
1314
1315                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1316                 case QFMT_VFS_OLD:
1317                         fmtname = "vfsold";
1318                         break;
1319                 case QFMT_VFS_V0:
1320                         fmtname = "vfsv0";
1321                         break;
1322                 case QFMT_VFS_V1:
1323                         fmtname = "vfsv1";
1324                         break;
1325                 }
1326                 seq_printf(seq, ",jqfmt=%s", fmtname);
1327         }
1328
1329         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1330                 seq_show_option(seq, "usrjquota",
1331                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1332
1333         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1334                 seq_show_option(seq, "grpjquota",
1335                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1336
1337         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1338                 seq_show_option(seq, "prjjquota",
1339                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1340 #endif
1341 }
1342
1343 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1344 {
1345         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1346
1347         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1348                 if (test_opt(sbi, FORCE_FG_GC))
1349                         seq_printf(seq, ",background_gc=%s", "sync");
1350                 else
1351                         seq_printf(seq, ",background_gc=%s", "on");
1352         } else {
1353                 seq_printf(seq, ",background_gc=%s", "off");
1354         }
1355         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1356                 seq_puts(seq, ",disable_roll_forward");
1357         if (test_opt(sbi, DISCARD))
1358                 seq_puts(seq, ",discard");
1359         else
1360                 seq_puts(seq, ",nodiscard");
1361         if (test_opt(sbi, NOHEAP))
1362                 seq_puts(seq, ",no_heap");
1363         else
1364                 seq_puts(seq, ",heap");
1365 #ifdef CONFIG_F2FS_FS_XATTR
1366         if (test_opt(sbi, XATTR_USER))
1367                 seq_puts(seq, ",user_xattr");
1368         else
1369                 seq_puts(seq, ",nouser_xattr");
1370         if (test_opt(sbi, INLINE_XATTR))
1371                 seq_puts(seq, ",inline_xattr");
1372         else
1373                 seq_puts(seq, ",noinline_xattr");
1374         if (test_opt(sbi, INLINE_XATTR_SIZE))
1375                 seq_printf(seq, ",inline_xattr_size=%u",
1376                                         F2FS_OPTION(sbi).inline_xattr_size);
1377 #endif
1378 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1379         if (test_opt(sbi, POSIX_ACL))
1380                 seq_puts(seq, ",acl");
1381         else
1382                 seq_puts(seq, ",noacl");
1383 #endif
1384         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1385                 seq_puts(seq, ",disable_ext_identify");
1386         if (test_opt(sbi, INLINE_DATA))
1387                 seq_puts(seq, ",inline_data");
1388         else
1389                 seq_puts(seq, ",noinline_data");
1390         if (test_opt(sbi, INLINE_DENTRY))
1391                 seq_puts(seq, ",inline_dentry");
1392         else
1393                 seq_puts(seq, ",noinline_dentry");
1394         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1395                 seq_puts(seq, ",flush_merge");
1396         if (test_opt(sbi, NOBARRIER))
1397                 seq_puts(seq, ",nobarrier");
1398         if (test_opt(sbi, FASTBOOT))
1399                 seq_puts(seq, ",fastboot");
1400         if (test_opt(sbi, EXTENT_CACHE))
1401                 seq_puts(seq, ",extent_cache");
1402         else
1403                 seq_puts(seq, ",noextent_cache");
1404         if (test_opt(sbi, DATA_FLUSH))
1405                 seq_puts(seq, ",data_flush");
1406
1407         seq_puts(seq, ",mode=");
1408         if (test_opt(sbi, ADAPTIVE))
1409                 seq_puts(seq, "adaptive");
1410         else if (test_opt(sbi, LFS))
1411                 seq_puts(seq, "lfs");
1412         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1413         if (test_opt(sbi, RESERVE_ROOT))
1414                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1415                                 F2FS_OPTION(sbi).root_reserved_blocks,
1416                                 from_kuid_munged(&init_user_ns,
1417                                         F2FS_OPTION(sbi).s_resuid),
1418                                 from_kgid_munged(&init_user_ns,
1419                                         F2FS_OPTION(sbi).s_resgid));
1420         if (F2FS_IO_SIZE_BITS(sbi))
1421                 seq_printf(seq, ",io_bits=%u",
1422                                 F2FS_OPTION(sbi).write_io_size_bits);
1423 #ifdef CONFIG_F2FS_FAULT_INJECTION
1424         if (test_opt(sbi, FAULT_INJECTION)) {
1425                 seq_printf(seq, ",fault_injection=%u",
1426                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1427                 seq_printf(seq, ",fault_type=%u",
1428                                 F2FS_OPTION(sbi).fault_info.inject_type);
1429         }
1430 #endif
1431 #ifdef CONFIG_QUOTA
1432         if (test_opt(sbi, QUOTA))
1433                 seq_puts(seq, ",quota");
1434         if (test_opt(sbi, USRQUOTA))
1435                 seq_puts(seq, ",usrquota");
1436         if (test_opt(sbi, GRPQUOTA))
1437                 seq_puts(seq, ",grpquota");
1438         if (test_opt(sbi, PRJQUOTA))
1439                 seq_puts(seq, ",prjquota");
1440 #endif
1441         f2fs_show_quota_options(seq, sbi->sb);
1442         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1443                 seq_printf(seq, ",whint_mode=%s", "user-based");
1444         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1445                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1446 #ifdef CONFIG_FS_ENCRYPTION
1447         if (F2FS_OPTION(sbi).test_dummy_encryption)
1448                 seq_puts(seq, ",test_dummy_encryption");
1449 #endif
1450
1451         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1452                 seq_printf(seq, ",alloc_mode=%s", "default");
1453         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1454                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1455
1456         if (test_opt(sbi, DISABLE_CHECKPOINT))
1457                 seq_printf(seq, ",checkpoint=disable:%u",
1458                                 F2FS_OPTION(sbi).unusable_cap);
1459         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1460                 seq_printf(seq, ",fsync_mode=%s", "posix");
1461         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1462                 seq_printf(seq, ",fsync_mode=%s", "strict");
1463         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1464                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1465         return 0;
1466 }
1467
1468 static void default_options(struct f2fs_sb_info *sbi)
1469 {
1470         /* init some FS parameters */
1471         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1472         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1473         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1474         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1475         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1476         F2FS_OPTION(sbi).test_dummy_encryption = false;
1477         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1478         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1479
1480         set_opt(sbi, BG_GC);
1481         set_opt(sbi, INLINE_XATTR);
1482         set_opt(sbi, INLINE_DATA);
1483         set_opt(sbi, INLINE_DENTRY);
1484         set_opt(sbi, EXTENT_CACHE);
1485         set_opt(sbi, NOHEAP);
1486         clear_opt(sbi, DISABLE_CHECKPOINT);
1487         F2FS_OPTION(sbi).unusable_cap = 0;
1488         sbi->sb->s_flags |= SB_LAZYTIME;
1489         set_opt(sbi, FLUSH_MERGE);
1490         set_opt(sbi, DISCARD);
1491         if (f2fs_sb_has_blkzoned(sbi))
1492                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1493         else
1494                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1495
1496 #ifdef CONFIG_F2FS_FS_XATTR
1497         set_opt(sbi, XATTR_USER);
1498 #endif
1499 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1500         set_opt(sbi, POSIX_ACL);
1501 #endif
1502
1503         f2fs_build_fault_attr(sbi, 0, 0);
1504 }
1505
1506 #ifdef CONFIG_QUOTA
1507 static int f2fs_enable_quotas(struct super_block *sb);
1508 #endif
1509
1510 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1511 {
1512         unsigned int s_flags = sbi->sb->s_flags;
1513         struct cp_control cpc;
1514         int err = 0;
1515         int ret;
1516         block_t unusable;
1517
1518         if (s_flags & SB_RDONLY) {
1519                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1520                 return -EINVAL;
1521         }
1522         sbi->sb->s_flags |= SB_ACTIVE;
1523
1524         f2fs_update_time(sbi, DISABLE_TIME);
1525
1526         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1527                 mutex_lock(&sbi->gc_mutex);
1528                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1529                 if (err == -ENODATA) {
1530                         err = 0;
1531                         break;
1532                 }
1533                 if (err && err != -EAGAIN)
1534                         break;
1535         }
1536
1537         ret = sync_filesystem(sbi->sb);
1538         if (ret || err) {
1539                 err = ret ? ret: err;
1540                 goto restore_flag;
1541         }
1542
1543         unusable = f2fs_get_unusable_blocks(sbi);
1544         if (f2fs_disable_cp_again(sbi, unusable)) {
1545                 err = -EAGAIN;
1546                 goto restore_flag;
1547         }
1548
1549         mutex_lock(&sbi->gc_mutex);
1550         cpc.reason = CP_PAUSE;
1551         set_sbi_flag(sbi, SBI_CP_DISABLED);
1552         err = f2fs_write_checkpoint(sbi, &cpc);
1553         if (err)
1554                 goto out_unlock;
1555
1556         spin_lock(&sbi->stat_lock);
1557         sbi->unusable_block_count = unusable;
1558         spin_unlock(&sbi->stat_lock);
1559
1560 out_unlock:
1561         mutex_unlock(&sbi->gc_mutex);
1562 restore_flag:
1563         sbi->sb->s_flags = s_flags;     /* Restore MS_RDONLY status */
1564         return err;
1565 }
1566
1567 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1568 {
1569         mutex_lock(&sbi->gc_mutex);
1570         f2fs_dirty_to_prefree(sbi);
1571
1572         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1573         set_sbi_flag(sbi, SBI_IS_DIRTY);
1574         mutex_unlock(&sbi->gc_mutex);
1575
1576         f2fs_sync_fs(sbi->sb, 1);
1577 }
1578
1579 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1580 {
1581         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1582         struct f2fs_mount_info org_mount_opt;
1583         unsigned long old_sb_flags;
1584         int err;
1585         bool need_restart_gc = false;
1586         bool need_stop_gc = false;
1587         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1588         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1589         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1590         bool checkpoint_changed;
1591 #ifdef CONFIG_QUOTA
1592         int i, j;
1593 #endif
1594
1595         /*
1596          * Save the old mount options in case we
1597          * need to restore them.
1598          */
1599         org_mount_opt = sbi->mount_opt;
1600         old_sb_flags = sb->s_flags;
1601
1602 #ifdef CONFIG_QUOTA
1603         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1604         for (i = 0; i < MAXQUOTAS; i++) {
1605                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1606                         org_mount_opt.s_qf_names[i] =
1607                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1608                                 GFP_KERNEL);
1609                         if (!org_mount_opt.s_qf_names[i]) {
1610                                 for (j = 0; j < i; j++)
1611                                         kvfree(org_mount_opt.s_qf_names[j]);
1612                                 return -ENOMEM;
1613                         }
1614                 } else {
1615                         org_mount_opt.s_qf_names[i] = NULL;
1616                 }
1617         }
1618 #endif
1619
1620         /* recover superblocks we couldn't write due to previous RO mount */
1621         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1622                 err = f2fs_commit_super(sbi, false);
1623                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1624                           err);
1625                 if (!err)
1626                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1627         }
1628
1629         default_options(sbi);
1630
1631         /* parse mount options */
1632         err = parse_options(sb, data);
1633         if (err)
1634                 goto restore_opts;
1635         checkpoint_changed =
1636                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1637
1638         /*
1639          * Previous and new state of filesystem is RO,
1640          * so skip checking GC and FLUSH_MERGE conditions.
1641          */
1642         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1643                 goto skip;
1644
1645 #ifdef CONFIG_QUOTA
1646         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1647                 err = dquot_suspend(sb, -1);
1648                 if (err < 0)
1649                         goto restore_opts;
1650         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1651                 /* dquot_resume needs RW */
1652                 sb->s_flags &= ~SB_RDONLY;
1653                 if (sb_any_quota_suspended(sb)) {
1654                         dquot_resume(sb, -1);
1655                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1656                         err = f2fs_enable_quotas(sb);
1657                         if (err)
1658                                 goto restore_opts;
1659                 }
1660         }
1661 #endif
1662         /* disallow enable/disable extent_cache dynamically */
1663         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1664                 err = -EINVAL;
1665                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1666                 goto restore_opts;
1667         }
1668
1669         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1670                 err = -EINVAL;
1671                 f2fs_warn(sbi, "switch io_bits option is not allowed");
1672                 goto restore_opts;
1673         }
1674
1675         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1676                 err = -EINVAL;
1677                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1678                 goto restore_opts;
1679         }
1680
1681         /*
1682          * We stop the GC thread if FS is mounted as RO
1683          * or if background_gc = off is passed in mount
1684          * option. Also sync the filesystem.
1685          */
1686         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1687                 if (sbi->gc_thread) {
1688                         f2fs_stop_gc_thread(sbi);
1689                         need_restart_gc = true;
1690                 }
1691         } else if (!sbi->gc_thread) {
1692                 err = f2fs_start_gc_thread(sbi);
1693                 if (err)
1694                         goto restore_opts;
1695                 need_stop_gc = true;
1696         }
1697
1698         if (*flags & SB_RDONLY ||
1699                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1700                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1701                 sync_inodes_sb(sb);
1702
1703                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1704                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1705                 f2fs_sync_fs(sb, 1);
1706                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1707         }
1708
1709         if (checkpoint_changed) {
1710                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1711                         err = f2fs_disable_checkpoint(sbi);
1712                         if (err)
1713                                 goto restore_gc;
1714                 } else {
1715                         f2fs_enable_checkpoint(sbi);
1716                 }
1717         }
1718
1719         /*
1720          * We stop issue flush thread if FS is mounted as RO
1721          * or if flush_merge is not passed in mount option.
1722          */
1723         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1724                 clear_opt(sbi, FLUSH_MERGE);
1725                 f2fs_destroy_flush_cmd_control(sbi, false);
1726         } else {
1727                 err = f2fs_create_flush_cmd_control(sbi);
1728                 if (err)
1729                         goto restore_gc;
1730         }
1731 skip:
1732 #ifdef CONFIG_QUOTA
1733         /* Release old quota file names */
1734         for (i = 0; i < MAXQUOTAS; i++)
1735                 kvfree(org_mount_opt.s_qf_names[i]);
1736 #endif
1737         /* Update the POSIXACL Flag */
1738         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1739                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1740
1741         limit_reserve_root(sbi);
1742         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1743         return 0;
1744 restore_gc:
1745         if (need_restart_gc) {
1746                 if (f2fs_start_gc_thread(sbi))
1747                         f2fs_warn(sbi, "background gc thread has stopped");
1748         } else if (need_stop_gc) {
1749                 f2fs_stop_gc_thread(sbi);
1750         }
1751 restore_opts:
1752 #ifdef CONFIG_QUOTA
1753         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1754         for (i = 0; i < MAXQUOTAS; i++) {
1755                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1756                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1757         }
1758 #endif
1759         sbi->mount_opt = org_mount_opt;
1760         sb->s_flags = old_sb_flags;
1761         return err;
1762 }
1763
1764 #ifdef CONFIG_QUOTA
1765 /* Read data from quotafile */
1766 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1767                                size_t len, loff_t off)
1768 {
1769         struct inode *inode = sb_dqopt(sb)->files[type];
1770         struct address_space *mapping = inode->i_mapping;
1771         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1772         int offset = off & (sb->s_blocksize - 1);
1773         int tocopy;
1774         size_t toread;
1775         loff_t i_size = i_size_read(inode);
1776         struct page *page;
1777         char *kaddr;
1778
1779         if (off > i_size)
1780                 return 0;
1781
1782         if (off + len > i_size)
1783                 len = i_size - off;
1784         toread = len;
1785         while (toread > 0) {
1786                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1787 repeat:
1788                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1789                 if (IS_ERR(page)) {
1790                         if (PTR_ERR(page) == -ENOMEM) {
1791                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1792                                 goto repeat;
1793                         }
1794                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1795                         return PTR_ERR(page);
1796                 }
1797
1798                 lock_page(page);
1799
1800                 if (unlikely(page->mapping != mapping)) {
1801                         f2fs_put_page(page, 1);
1802                         goto repeat;
1803                 }
1804                 if (unlikely(!PageUptodate(page))) {
1805                         f2fs_put_page(page, 1);
1806                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1807                         return -EIO;
1808                 }
1809
1810                 kaddr = kmap_atomic(page);
1811                 memcpy(data, kaddr + offset, tocopy);
1812                 kunmap_atomic(kaddr);
1813                 f2fs_put_page(page, 1);
1814
1815                 offset = 0;
1816                 toread -= tocopy;
1817                 data += tocopy;
1818                 blkidx++;
1819         }
1820         return len;
1821 }
1822
1823 /* Write to quotafile */
1824 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1825                                 const char *data, size_t len, loff_t off)
1826 {
1827         struct inode *inode = sb_dqopt(sb)->files[type];
1828         struct address_space *mapping = inode->i_mapping;
1829         const struct address_space_operations *a_ops = mapping->a_ops;
1830         int offset = off & (sb->s_blocksize - 1);
1831         size_t towrite = len;
1832         struct page *page;
1833         char *kaddr;
1834         int err = 0;
1835         int tocopy;
1836
1837         while (towrite > 0) {
1838                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1839                                                                 towrite);
1840 retry:
1841                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1842                                                         &page, NULL);
1843                 if (unlikely(err)) {
1844                         if (err == -ENOMEM) {
1845                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1846                                 goto retry;
1847                         }
1848                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1849                         break;
1850                 }
1851
1852                 kaddr = kmap_atomic(page);
1853                 memcpy(kaddr + offset, data, tocopy);
1854                 kunmap_atomic(kaddr);
1855                 flush_dcache_page(page);
1856
1857                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1858                                                 page, NULL);
1859                 offset = 0;
1860                 towrite -= tocopy;
1861                 off += tocopy;
1862                 data += tocopy;
1863                 cond_resched();
1864         }
1865
1866         if (len == towrite)
1867                 return err;
1868         inode->i_mtime = inode->i_ctime = current_time(inode);
1869         f2fs_mark_inode_dirty_sync(inode, false);
1870         return len - towrite;
1871 }
1872
1873 static struct dquot **f2fs_get_dquots(struct inode *inode)
1874 {
1875         return F2FS_I(inode)->i_dquot;
1876 }
1877
1878 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1879 {
1880         return &F2FS_I(inode)->i_reserved_quota;
1881 }
1882
1883 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1884 {
1885         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1886                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1887                 return 0;
1888         }
1889
1890         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1891                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1892 }
1893
1894 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1895 {
1896         int enabled = 0;
1897         int i, err;
1898
1899         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1900                 err = f2fs_enable_quotas(sbi->sb);
1901                 if (err) {
1902                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
1903                         return 0;
1904                 }
1905                 return 1;
1906         }
1907
1908         for (i = 0; i < MAXQUOTAS; i++) {
1909                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1910                         err = f2fs_quota_on_mount(sbi, i);
1911                         if (!err) {
1912                                 enabled = 1;
1913                                 continue;
1914                         }
1915                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
1916                                  err, i);
1917                 }
1918         }
1919         return enabled;
1920 }
1921
1922 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1923                              unsigned int flags)
1924 {
1925         struct inode *qf_inode;
1926         unsigned long qf_inum;
1927         int err;
1928
1929         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1930
1931         qf_inum = f2fs_qf_ino(sb, type);
1932         if (!qf_inum)
1933                 return -EPERM;
1934
1935         qf_inode = f2fs_iget(sb, qf_inum);
1936         if (IS_ERR(qf_inode)) {
1937                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
1938                 return PTR_ERR(qf_inode);
1939         }
1940
1941         /* Don't account quota for quota files to avoid recursion */
1942         qf_inode->i_flags |= S_NOQUOTA;
1943         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
1944         iput(qf_inode);
1945         return err;
1946 }
1947
1948 static int f2fs_enable_quotas(struct super_block *sb)
1949 {
1950         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1951         int type, err = 0;
1952         unsigned long qf_inum;
1953         bool quota_mopt[MAXQUOTAS] = {
1954                 test_opt(sbi, USRQUOTA),
1955                 test_opt(sbi, GRPQUOTA),
1956                 test_opt(sbi, PRJQUOTA),
1957         };
1958
1959         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1960                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
1961                 return 0;
1962         }
1963
1964         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1965
1966         for (type = 0; type < MAXQUOTAS; type++) {
1967                 qf_inum = f2fs_qf_ino(sb, type);
1968                 if (qf_inum) {
1969                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1970                                 DQUOT_USAGE_ENABLED |
1971                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1972                         if (err) {
1973                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
1974                                          type, err);
1975                                 for (type--; type >= 0; type--)
1976                                         dquot_quota_off(sb, type);
1977                                 set_sbi_flag(F2FS_SB(sb),
1978                                                 SBI_QUOTA_NEED_REPAIR);
1979                                 return err;
1980                         }
1981                 }
1982         }
1983         return 0;
1984 }
1985
1986 int f2fs_quota_sync(struct super_block *sb, int type)
1987 {
1988         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1989         struct quota_info *dqopt = sb_dqopt(sb);
1990         int cnt;
1991         int ret;
1992
1993         /*
1994          * do_quotactl
1995          *  f2fs_quota_sync
1996          *  down_read(quota_sem)
1997          *  dquot_writeback_dquots()
1998          *  f2fs_dquot_commit
1999          *                            block_operation
2000          *                            down_read(quota_sem)
2001          */
2002         f2fs_lock_op(sbi);
2003
2004         down_read(&sbi->quota_sem);
2005         ret = dquot_writeback_dquots(sb, type);
2006         if (ret)
2007                 goto out;
2008
2009         /*
2010          * Now when everything is written we can discard the pagecache so
2011          * that userspace sees the changes.
2012          */
2013         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2014                 struct address_space *mapping;
2015
2016                 if (type != -1 && cnt != type)
2017                         continue;
2018                 if (!sb_has_quota_active(sb, cnt))
2019                         continue;
2020
2021                 mapping = dqopt->files[cnt]->i_mapping;
2022
2023                 ret = filemap_fdatawrite(mapping);
2024                 if (ret)
2025                         goto out;
2026
2027                 /* if we are using journalled quota */
2028                 if (is_journalled_quota(sbi))
2029                         continue;
2030
2031                 ret = filemap_fdatawait(mapping);
2032                 if (ret)
2033                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2034
2035                 inode_lock(dqopt->files[cnt]);
2036                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2037                 inode_unlock(dqopt->files[cnt]);
2038         }
2039 out:
2040         if (ret)
2041                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2042         up_read(&sbi->quota_sem);
2043         f2fs_unlock_op(sbi);
2044         return ret;
2045 }
2046
2047 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2048                                                         const struct path *path)
2049 {
2050         struct inode *inode;
2051         int err;
2052
2053         /* if quota sysfile exists, deny enabling quota with specific file */
2054         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2055                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2056                 return -EBUSY;
2057         }
2058
2059         err = f2fs_quota_sync(sb, type);
2060         if (err)
2061                 return err;
2062
2063         err = dquot_quota_on(sb, type, format_id, path);
2064         if (err)
2065                 return err;
2066
2067         inode = d_inode(path->dentry);
2068
2069         inode_lock(inode);
2070         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2071         f2fs_set_inode_flags(inode);
2072         inode_unlock(inode);
2073         f2fs_mark_inode_dirty_sync(inode, false);
2074
2075         return 0;
2076 }
2077
2078 static int __f2fs_quota_off(struct super_block *sb, int type)
2079 {
2080         struct inode *inode = sb_dqopt(sb)->files[type];
2081         int err;
2082
2083         if (!inode || !igrab(inode))
2084                 return dquot_quota_off(sb, type);
2085
2086         err = f2fs_quota_sync(sb, type);
2087         if (err)
2088                 goto out_put;
2089
2090         err = dquot_quota_off(sb, type);
2091         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2092                 goto out_put;
2093
2094         inode_lock(inode);
2095         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2096         f2fs_set_inode_flags(inode);
2097         inode_unlock(inode);
2098         f2fs_mark_inode_dirty_sync(inode, false);
2099 out_put:
2100         iput(inode);
2101         return err;
2102 }
2103
2104 static int f2fs_quota_off(struct super_block *sb, int type)
2105 {
2106         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2107         int err;
2108
2109         err = __f2fs_quota_off(sb, type);
2110
2111         /*
2112          * quotactl can shutdown journalled quota, result in inconsistence
2113          * between quota record and fs data by following updates, tag the
2114          * flag to let fsck be aware of it.
2115          */
2116         if (is_journalled_quota(sbi))
2117                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2118         return err;
2119 }
2120
2121 void f2fs_quota_off_umount(struct super_block *sb)
2122 {
2123         int type;
2124         int err;
2125
2126         for (type = 0; type < MAXQUOTAS; type++) {
2127                 err = __f2fs_quota_off(sb, type);
2128                 if (err) {
2129                         int ret = dquot_quota_off(sb, type);
2130
2131                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2132                                  type, err, ret);
2133                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2134                 }
2135         }
2136         /*
2137          * In case of checkpoint=disable, we must flush quota blocks.
2138          * This can cause NULL exception for node_inode in end_io, since
2139          * put_super already dropped it.
2140          */
2141         sync_filesystem(sb);
2142 }
2143
2144 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2145 {
2146         struct quota_info *dqopt = sb_dqopt(sb);
2147         int type;
2148
2149         for (type = 0; type < MAXQUOTAS; type++) {
2150                 if (!dqopt->files[type])
2151                         continue;
2152                 f2fs_inode_synced(dqopt->files[type]);
2153         }
2154 }
2155
2156 static int f2fs_dquot_commit(struct dquot *dquot)
2157 {
2158         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2159         int ret;
2160
2161         down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2162         ret = dquot_commit(dquot);
2163         if (ret < 0)
2164                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2165         up_read(&sbi->quota_sem);
2166         return ret;
2167 }
2168
2169 static int f2fs_dquot_acquire(struct dquot *dquot)
2170 {
2171         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2172         int ret;
2173
2174         down_read(&sbi->quota_sem);
2175         ret = dquot_acquire(dquot);
2176         if (ret < 0)
2177                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2178         up_read(&sbi->quota_sem);
2179         return ret;
2180 }
2181
2182 static int f2fs_dquot_release(struct dquot *dquot)
2183 {
2184         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2185         int ret = dquot_release(dquot);
2186
2187         if (ret < 0)
2188                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2189         return ret;
2190 }
2191
2192 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2193 {
2194         struct super_block *sb = dquot->dq_sb;
2195         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2196         int 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         return ret;
2203 }
2204
2205 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2206 {
2207         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2208         int ret = dquot_commit_info(sb, type);
2209
2210         if (ret < 0)
2211                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2212         return ret;
2213 }
2214
2215 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2216 {
2217         *projid = F2FS_I(inode)->i_projid;
2218         return 0;
2219 }
2220
2221 static const struct dquot_operations f2fs_quota_operations = {
2222         .get_reserved_space = f2fs_get_reserved_space,
2223         .write_dquot    = f2fs_dquot_commit,
2224         .acquire_dquot  = f2fs_dquot_acquire,
2225         .release_dquot  = f2fs_dquot_release,
2226         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2227         .write_info     = f2fs_dquot_commit_info,
2228         .alloc_dquot    = dquot_alloc,
2229         .destroy_dquot  = dquot_destroy,
2230         .get_projid     = f2fs_get_projid,
2231         .get_next_id    = dquot_get_next_id,
2232 };
2233
2234 static const struct quotactl_ops f2fs_quotactl_ops = {
2235         .quota_on       = f2fs_quota_on,
2236         .quota_off      = f2fs_quota_off,
2237         .quota_sync     = f2fs_quota_sync,
2238         .get_state      = dquot_get_state,
2239         .set_info       = dquot_set_dqinfo,
2240         .get_dqblk      = dquot_get_dqblk,
2241         .set_dqblk      = dquot_set_dqblk,
2242         .get_nextdqblk  = dquot_get_next_dqblk,
2243 };
2244 #else
2245 int f2fs_quota_sync(struct super_block *sb, int type)
2246 {
2247         return 0;
2248 }
2249
2250 void f2fs_quota_off_umount(struct super_block *sb)
2251 {
2252 }
2253 #endif
2254
2255 static const struct super_operations f2fs_sops = {
2256         .alloc_inode    = f2fs_alloc_inode,
2257         .free_inode     = f2fs_free_inode,
2258         .drop_inode     = f2fs_drop_inode,
2259         .write_inode    = f2fs_write_inode,
2260         .dirty_inode    = f2fs_dirty_inode,
2261         .show_options   = f2fs_show_options,
2262 #ifdef CONFIG_QUOTA
2263         .quota_read     = f2fs_quota_read,
2264         .quota_write    = f2fs_quota_write,
2265         .get_dquots     = f2fs_get_dquots,
2266 #endif
2267         .evict_inode    = f2fs_evict_inode,
2268         .put_super      = f2fs_put_super,
2269         .sync_fs        = f2fs_sync_fs,
2270         .freeze_fs      = f2fs_freeze,
2271         .unfreeze_fs    = f2fs_unfreeze,
2272         .statfs         = f2fs_statfs,
2273         .remount_fs     = f2fs_remount,
2274 };
2275
2276 #ifdef CONFIG_FS_ENCRYPTION
2277 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2278 {
2279         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2280                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2281                                 ctx, len, NULL);
2282 }
2283
2284 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2285                                                         void *fs_data)
2286 {
2287         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2288
2289         /*
2290          * Encrypting the root directory is not allowed because fsck
2291          * expects lost+found directory to exist and remain unencrypted
2292          * if LOST_FOUND feature is enabled.
2293          *
2294          */
2295         if (f2fs_sb_has_lost_found(sbi) &&
2296                         inode->i_ino == F2FS_ROOT_INO(sbi))
2297                 return -EPERM;
2298
2299         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2300                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2301                                 ctx, len, fs_data, XATTR_CREATE);
2302 }
2303
2304 static bool f2fs_dummy_context(struct inode *inode)
2305 {
2306         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2307 }
2308
2309 static bool f2fs_has_stable_inodes(struct super_block *sb)
2310 {
2311         return true;
2312 }
2313
2314 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2315                                        int *ino_bits_ret, int *lblk_bits_ret)
2316 {
2317         *ino_bits_ret = 8 * sizeof(nid_t);
2318         *lblk_bits_ret = 8 * sizeof(block_t);
2319 }
2320
2321 static const struct fscrypt_operations f2fs_cryptops = {
2322         .key_prefix             = "f2fs:",
2323         .get_context            = f2fs_get_context,
2324         .set_context            = f2fs_set_context,
2325         .dummy_context          = f2fs_dummy_context,
2326         .empty_dir              = f2fs_empty_dir,
2327         .max_namelen            = F2FS_NAME_LEN,
2328         .has_stable_inodes      = f2fs_has_stable_inodes,
2329         .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
2330 };
2331 #endif
2332
2333 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2334                 u64 ino, u32 generation)
2335 {
2336         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2337         struct inode *inode;
2338
2339         if (f2fs_check_nid_range(sbi, ino))
2340                 return ERR_PTR(-ESTALE);
2341
2342         /*
2343          * f2fs_iget isn't quite right if the inode is currently unallocated!
2344          * However f2fs_iget currently does appropriate checks to handle stale
2345          * inodes so everything is OK.
2346          */
2347         inode = f2fs_iget(sb, ino);
2348         if (IS_ERR(inode))
2349                 return ERR_CAST(inode);
2350         if (unlikely(generation && inode->i_generation != generation)) {
2351                 /* we didn't find the right inode.. */
2352                 iput(inode);
2353                 return ERR_PTR(-ESTALE);
2354         }
2355         return inode;
2356 }
2357
2358 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2359                 int fh_len, int fh_type)
2360 {
2361         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2362                                     f2fs_nfs_get_inode);
2363 }
2364
2365 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2366                 int fh_len, int fh_type)
2367 {
2368         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2369                                     f2fs_nfs_get_inode);
2370 }
2371
2372 static const struct export_operations f2fs_export_ops = {
2373         .fh_to_dentry = f2fs_fh_to_dentry,
2374         .fh_to_parent = f2fs_fh_to_parent,
2375         .get_parent = f2fs_get_parent,
2376 };
2377
2378 static loff_t max_file_blocks(void)
2379 {
2380         loff_t result = 0;
2381         loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2382
2383         /*
2384          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2385          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2386          * space in inode.i_addr, it will be more safe to reassign
2387          * result as zero.
2388          */
2389
2390         /* two direct node blocks */
2391         result += (leaf_count * 2);
2392
2393         /* two indirect node blocks */
2394         leaf_count *= NIDS_PER_BLOCK;
2395         result += (leaf_count * 2);
2396
2397         /* one double indirect node block */
2398         leaf_count *= NIDS_PER_BLOCK;
2399         result += leaf_count;
2400
2401         return result;
2402 }
2403
2404 static int __f2fs_commit_super(struct buffer_head *bh,
2405                         struct f2fs_super_block *super)
2406 {
2407         lock_buffer(bh);
2408         if (super)
2409                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2410         set_buffer_dirty(bh);
2411         unlock_buffer(bh);
2412
2413         /* it's rare case, we can do fua all the time */
2414         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2415 }
2416
2417 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2418                                         struct buffer_head *bh)
2419 {
2420         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2421                                         (bh->b_data + F2FS_SUPER_OFFSET);
2422         struct super_block *sb = sbi->sb;
2423         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2424         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2425         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2426         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2427         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2428         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2429         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2430         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2431         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2432         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2433         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2434         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2435         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2436         u64 main_end_blkaddr = main_blkaddr +
2437                                 (segment_count_main << log_blocks_per_seg);
2438         u64 seg_end_blkaddr = segment0_blkaddr +
2439                                 (segment_count << log_blocks_per_seg);
2440
2441         if (segment0_blkaddr != cp_blkaddr) {
2442                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2443                           segment0_blkaddr, cp_blkaddr);
2444                 return true;
2445         }
2446
2447         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2448                                                         sit_blkaddr) {
2449                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2450                           cp_blkaddr, sit_blkaddr,
2451                           segment_count_ckpt << log_blocks_per_seg);
2452                 return true;
2453         }
2454
2455         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2456                                                         nat_blkaddr) {
2457                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2458                           sit_blkaddr, nat_blkaddr,
2459                           segment_count_sit << log_blocks_per_seg);
2460                 return true;
2461         }
2462
2463         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2464                                                         ssa_blkaddr) {
2465                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2466                           nat_blkaddr, ssa_blkaddr,
2467                           segment_count_nat << log_blocks_per_seg);
2468                 return true;
2469         }
2470
2471         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2472                                                         main_blkaddr) {
2473                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2474                           ssa_blkaddr, main_blkaddr,
2475                           segment_count_ssa << log_blocks_per_seg);
2476                 return true;
2477         }
2478
2479         if (main_end_blkaddr > seg_end_blkaddr) {
2480                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2481                           main_blkaddr,
2482                           segment0_blkaddr +
2483                           (segment_count << log_blocks_per_seg),
2484                           segment_count_main << log_blocks_per_seg);
2485                 return true;
2486         } else if (main_end_blkaddr < seg_end_blkaddr) {
2487                 int err = 0;
2488                 char *res;
2489
2490                 /* fix in-memory information all the time */
2491                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2492                                 segment0_blkaddr) >> log_blocks_per_seg);
2493
2494                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2495                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2496                         res = "internally";
2497                 } else {
2498                         err = __f2fs_commit_super(bh, NULL);
2499                         res = err ? "failed" : "done";
2500                 }
2501                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2502                           res, main_blkaddr,
2503                           segment0_blkaddr +
2504                           (segment_count << log_blocks_per_seg),
2505                           segment_count_main << log_blocks_per_seg);
2506                 if (err)
2507                         return true;
2508         }
2509         return false;
2510 }
2511
2512 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2513                                 struct buffer_head *bh)
2514 {
2515         block_t segment_count, segs_per_sec, secs_per_zone;
2516         block_t total_sections, blocks_per_seg;
2517         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2518                                         (bh->b_data + F2FS_SUPER_OFFSET);
2519         unsigned int blocksize;
2520         size_t crc_offset = 0;
2521         __u32 crc = 0;
2522
2523         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2524                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2525                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2526                 return -EINVAL;
2527         }
2528
2529         /* Check checksum_offset and crc in superblock */
2530         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2531                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2532                 if (crc_offset !=
2533                         offsetof(struct f2fs_super_block, crc)) {
2534                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2535                                   crc_offset);
2536                         return -EFSCORRUPTED;
2537                 }
2538                 crc = le32_to_cpu(raw_super->crc);
2539                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2540                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2541                         return -EFSCORRUPTED;
2542                 }
2543         }
2544
2545         /* Currently, support only 4KB page cache size */
2546         if (F2FS_BLKSIZE != PAGE_SIZE) {
2547                 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2548                           PAGE_SIZE);
2549                 return -EFSCORRUPTED;
2550         }
2551
2552         /* Currently, support only 4KB block size */
2553         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2554         if (blocksize != F2FS_BLKSIZE) {
2555                 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2556                           blocksize);
2557                 return -EFSCORRUPTED;
2558         }
2559
2560         /* check log blocks per segment */
2561         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2562                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2563                           le32_to_cpu(raw_super->log_blocks_per_seg));
2564                 return -EFSCORRUPTED;
2565         }
2566
2567         /* Currently, support 512/1024/2048/4096 bytes sector size */
2568         if (le32_to_cpu(raw_super->log_sectorsize) >
2569                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2570                 le32_to_cpu(raw_super->log_sectorsize) <
2571                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2572                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2573                           le32_to_cpu(raw_super->log_sectorsize));
2574                 return -EFSCORRUPTED;
2575         }
2576         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2577                 le32_to_cpu(raw_super->log_sectorsize) !=
2578                         F2FS_MAX_LOG_SECTOR_SIZE) {
2579                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2580                           le32_to_cpu(raw_super->log_sectors_per_block),
2581                           le32_to_cpu(raw_super->log_sectorsize));
2582                 return -EFSCORRUPTED;
2583         }
2584
2585         segment_count = le32_to_cpu(raw_super->segment_count);
2586         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2587         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2588         total_sections = le32_to_cpu(raw_super->section_count);
2589
2590         /* blocks_per_seg should be 512, given the above check */
2591         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2592
2593         if (segment_count > F2FS_MAX_SEGMENT ||
2594                                 segment_count < F2FS_MIN_SEGMENTS) {
2595                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2596                 return -EFSCORRUPTED;
2597         }
2598
2599         if (total_sections > segment_count ||
2600                         total_sections < F2FS_MIN_SEGMENTS ||
2601                         segs_per_sec > segment_count || !segs_per_sec) {
2602                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2603                           segment_count, total_sections, segs_per_sec);
2604                 return -EFSCORRUPTED;
2605         }
2606
2607         if ((segment_count / segs_per_sec) < total_sections) {
2608                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2609                           segment_count, segs_per_sec, total_sections);
2610                 return -EFSCORRUPTED;
2611         }
2612
2613         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2614                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2615                           segment_count, le64_to_cpu(raw_super->block_count));
2616                 return -EFSCORRUPTED;
2617         }
2618
2619         if (RDEV(0).path[0]) {
2620                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2621                 int i = 1;
2622
2623                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2624                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2625                         i++;
2626                 }
2627                 if (segment_count != dev_seg_count) {
2628                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2629                                         segment_count, dev_seg_count);
2630                         return -EFSCORRUPTED;
2631                 }
2632         }
2633
2634         if (secs_per_zone > total_sections || !secs_per_zone) {
2635                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2636                           secs_per_zone, total_sections);
2637                 return -EFSCORRUPTED;
2638         }
2639         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2640                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2641                         (le32_to_cpu(raw_super->extension_count) +
2642                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2643                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2644                           le32_to_cpu(raw_super->extension_count),
2645                           raw_super->hot_ext_count,
2646                           F2FS_MAX_EXTENSION);
2647                 return -EFSCORRUPTED;
2648         }
2649
2650         if (le32_to_cpu(raw_super->cp_payload) >
2651                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2652                 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2653                           le32_to_cpu(raw_super->cp_payload),
2654                           blocks_per_seg - F2FS_CP_PACKS);
2655                 return -EFSCORRUPTED;
2656         }
2657
2658         /* check reserved ino info */
2659         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2660                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2661                 le32_to_cpu(raw_super->root_ino) != 3) {
2662                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2663                           le32_to_cpu(raw_super->node_ino),
2664                           le32_to_cpu(raw_super->meta_ino),
2665                           le32_to_cpu(raw_super->root_ino));
2666                 return -EFSCORRUPTED;
2667         }
2668
2669         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2670         if (sanity_check_area_boundary(sbi, bh))
2671                 return -EFSCORRUPTED;
2672
2673         return 0;
2674 }
2675
2676 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2677 {
2678         unsigned int total, fsmeta;
2679         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2680         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2681         unsigned int ovp_segments, reserved_segments;
2682         unsigned int main_segs, blocks_per_seg;
2683         unsigned int sit_segs, nat_segs;
2684         unsigned int sit_bitmap_size, nat_bitmap_size;
2685         unsigned int log_blocks_per_seg;
2686         unsigned int segment_count_main;
2687         unsigned int cp_pack_start_sum, cp_payload;
2688         block_t user_block_count, valid_user_blocks;
2689         block_t avail_node_count, valid_node_count;
2690         int i, j;
2691
2692         total = le32_to_cpu(raw_super->segment_count);
2693         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2694         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2695         fsmeta += sit_segs;
2696         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2697         fsmeta += nat_segs;
2698         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2699         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2700
2701         if (unlikely(fsmeta >= total))
2702                 return 1;
2703
2704         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2705         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2706
2707         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2708                         ovp_segments == 0 || reserved_segments == 0)) {
2709                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2710                 return 1;
2711         }
2712
2713         user_block_count = le64_to_cpu(ckpt->user_block_count);
2714         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2715         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2716         if (!user_block_count || user_block_count >=
2717                         segment_count_main << log_blocks_per_seg) {
2718                 f2fs_err(sbi, "Wrong user_block_count: %u",
2719                          user_block_count);
2720                 return 1;
2721         }
2722
2723         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2724         if (valid_user_blocks > user_block_count) {
2725                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2726                          valid_user_blocks, user_block_count);
2727                 return 1;
2728         }
2729
2730         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2731         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2732         if (valid_node_count > avail_node_count) {
2733                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2734                          valid_node_count, avail_node_count);
2735                 return 1;
2736         }
2737
2738         main_segs = le32_to_cpu(raw_super->segment_count_main);
2739         blocks_per_seg = sbi->blocks_per_seg;
2740
2741         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2742                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2743                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2744                         return 1;
2745                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2746                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2747                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2748                                 f2fs_err(sbi, "Node segment (%u, %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         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2756                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2757                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2758                         return 1;
2759                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2760                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2761                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2762                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2763                                          i, j,
2764                                          le32_to_cpu(ckpt->cur_data_segno[i]));
2765                                 return 1;
2766                         }
2767                 }
2768         }
2769         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2770                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2771                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2772                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2773                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2774                                          i, j,
2775                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2776                                 return 1;
2777                         }
2778                 }
2779         }
2780
2781         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2782         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2783
2784         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2785                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2786                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2787                          sit_bitmap_size, nat_bitmap_size);
2788                 return 1;
2789         }
2790
2791         cp_pack_start_sum = __start_sum_addr(sbi);
2792         cp_payload = __cp_payload(sbi);
2793         if (cp_pack_start_sum < cp_payload + 1 ||
2794                 cp_pack_start_sum > blocks_per_seg - 1 -
2795                         NR_CURSEG_TYPE) {
2796                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2797                          cp_pack_start_sum);
2798                 return 1;
2799         }
2800
2801         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2802                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2803                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2804                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2805                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2806                           le32_to_cpu(ckpt->checksum_offset));
2807                 return 1;
2808         }
2809
2810         if (unlikely(f2fs_cp_error(sbi))) {
2811                 f2fs_err(sbi, "A bug case: need to run fsck");
2812                 return 1;
2813         }
2814         return 0;
2815 }
2816
2817 static void init_sb_info(struct f2fs_sb_info *sbi)
2818 {
2819         struct f2fs_super_block *raw_super = sbi->raw_super;
2820         int i;
2821
2822         sbi->log_sectors_per_block =
2823                 le32_to_cpu(raw_super->log_sectors_per_block);
2824         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2825         sbi->blocksize = 1 << sbi->log_blocksize;
2826         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2827         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2828         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2829         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2830         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2831         sbi->total_node_count =
2832                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2833                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2834         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2835         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2836         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2837         sbi->cur_victim_sec = NULL_SECNO;
2838         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2839         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2840         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2841         sbi->migration_granularity = sbi->segs_per_sec;
2842
2843         sbi->dir_level = DEF_DIR_LEVEL;
2844         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2845         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2846         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2847         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2848         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2849         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2850                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2851         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2852
2853         for (i = 0; i < NR_COUNT_TYPE; i++)
2854                 atomic_set(&sbi->nr_pages[i], 0);
2855
2856         for (i = 0; i < META; i++)
2857                 atomic_set(&sbi->wb_sync_req[i], 0);
2858
2859         INIT_LIST_HEAD(&sbi->s_list);
2860         mutex_init(&sbi->umount_mutex);
2861         init_rwsem(&sbi->io_order_lock);
2862         spin_lock_init(&sbi->cp_lock);
2863
2864         sbi->dirty_device = 0;
2865         spin_lock_init(&sbi->dev_lock);
2866
2867         init_rwsem(&sbi->sb_lock);
2868         init_rwsem(&sbi->pin_sem);
2869 }
2870
2871 static int init_percpu_info(struct f2fs_sb_info *sbi)
2872 {
2873         int err;
2874
2875         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2876         if (err)
2877                 return err;
2878
2879         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2880                                                                 GFP_KERNEL);
2881         if (err)
2882                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2883
2884         return err;
2885 }
2886
2887 #ifdef CONFIG_BLK_DEV_ZONED
2888 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
2889                                void *data)
2890 {
2891         struct f2fs_dev_info *dev = data;
2892
2893         if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
2894                 set_bit(idx, dev->blkz_seq);
2895         return 0;
2896 }
2897
2898 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2899 {
2900         struct block_device *bdev = FDEV(devi).bdev;
2901         sector_t nr_sectors = bdev->bd_part->nr_sects;
2902         int ret;
2903
2904         if (!f2fs_sb_has_blkzoned(sbi))
2905                 return 0;
2906
2907         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2908                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2909                 return -EINVAL;
2910         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2911         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2912                                 __ilog2_u32(sbi->blocks_per_blkz))
2913                 return -EINVAL;
2914         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2915         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2916                                         sbi->log_blocks_per_blkz;
2917         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2918                 FDEV(devi).nr_blkz++;
2919
2920         FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
2921                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
2922                                         * sizeof(unsigned long),
2923                                         GFP_KERNEL);
2924         if (!FDEV(devi).blkz_seq)
2925                 return -ENOMEM;
2926
2927         /* Get block zones type */
2928         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
2929                                   &FDEV(devi));
2930         if (ret < 0)
2931                 return ret;
2932
2933         return 0;
2934 }
2935 #endif
2936
2937 /*
2938  * Read f2fs raw super block.
2939  * Because we have two copies of super block, so read both of them
2940  * to get the first valid one. If any one of them is broken, we pass
2941  * them recovery flag back to the caller.
2942  */
2943 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2944                         struct f2fs_super_block **raw_super,
2945                         int *valid_super_block, int *recovery)
2946 {
2947         struct super_block *sb = sbi->sb;
2948         int block;
2949         struct buffer_head *bh;
2950         struct f2fs_super_block *super;
2951         int err = 0;
2952
2953         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2954         if (!super)
2955                 return -ENOMEM;
2956
2957         for (block = 0; block < 2; block++) {
2958                 bh = sb_bread(sb, block);
2959                 if (!bh) {
2960                         f2fs_err(sbi, "Unable to read %dth superblock",
2961                                  block + 1);
2962                         err = -EIO;
2963                         *recovery = 1;
2964                         continue;
2965                 }
2966
2967                 /* sanity checking of raw super */
2968                 err = sanity_check_raw_super(sbi, bh);
2969                 if (err) {
2970                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
2971                                  block + 1);
2972                         brelse(bh);
2973                         *recovery = 1;
2974                         continue;
2975                 }
2976
2977                 if (!*raw_super) {
2978                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2979                                                         sizeof(*super));
2980                         *valid_super_block = block;
2981                         *raw_super = super;
2982                 }
2983                 brelse(bh);
2984         }
2985
2986         /* No valid superblock */
2987         if (!*raw_super)
2988                 kvfree(super);
2989         else
2990                 err = 0;
2991
2992         return err;
2993 }
2994
2995 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2996 {
2997         struct buffer_head *bh;
2998         __u32 crc = 0;
2999         int err;
3000
3001         if ((recover && f2fs_readonly(sbi->sb)) ||
3002                                 bdev_read_only(sbi->sb->s_bdev)) {
3003                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3004                 return -EROFS;
3005         }
3006
3007         /* we should update superblock crc here */
3008         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3009                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3010                                 offsetof(struct f2fs_super_block, crc));
3011                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3012         }
3013
3014         /* write back-up superblock first */
3015         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3016         if (!bh)
3017                 return -EIO;
3018         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3019         brelse(bh);
3020
3021         /* if we are in recovery path, skip writing valid superblock */
3022         if (recover || err)
3023                 return err;
3024
3025         /* write current valid superblock */
3026         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3027         if (!bh)
3028                 return -EIO;
3029         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3030         brelse(bh);
3031         return err;
3032 }
3033
3034 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3035 {
3036         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3037         unsigned int max_devices = MAX_DEVICES;
3038         int i;
3039
3040         /* Initialize single device information */
3041         if (!RDEV(0).path[0]) {
3042                 if (!bdev_is_zoned(sbi->sb->s_bdev))
3043                         return 0;
3044                 max_devices = 1;
3045         }
3046
3047         /*
3048          * Initialize multiple devices information, or single
3049          * zoned block device information.
3050          */
3051         sbi->devs = f2fs_kzalloc(sbi,
3052                                  array_size(max_devices,
3053                                             sizeof(struct f2fs_dev_info)),
3054                                  GFP_KERNEL);
3055         if (!sbi->devs)
3056                 return -ENOMEM;
3057
3058         for (i = 0; i < max_devices; i++) {
3059
3060                 if (i > 0 && !RDEV(i).path[0])
3061                         break;
3062
3063                 if (max_devices == 1) {
3064                         /* Single zoned block device mount */
3065                         FDEV(0).bdev =
3066                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3067                                         sbi->sb->s_mode, sbi->sb->s_type);
3068                 } else {
3069                         /* Multi-device mount */
3070                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3071                         FDEV(i).total_segments =
3072                                 le32_to_cpu(RDEV(i).total_segments);
3073                         if (i == 0) {
3074                                 FDEV(i).start_blk = 0;
3075                                 FDEV(i).end_blk = FDEV(i).start_blk +
3076                                     (FDEV(i).total_segments <<
3077                                     sbi->log_blocks_per_seg) - 1 +
3078                                     le32_to_cpu(raw_super->segment0_blkaddr);
3079                         } else {
3080                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3081                                 FDEV(i).end_blk = FDEV(i).start_blk +
3082                                         (FDEV(i).total_segments <<
3083                                         sbi->log_blocks_per_seg) - 1;
3084                         }
3085                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3086                                         sbi->sb->s_mode, sbi->sb->s_type);
3087                 }
3088                 if (IS_ERR(FDEV(i).bdev))
3089                         return PTR_ERR(FDEV(i).bdev);
3090
3091                 /* to release errored devices */
3092                 sbi->s_ndevs = i + 1;
3093
3094 #ifdef CONFIG_BLK_DEV_ZONED
3095                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3096                                 !f2fs_sb_has_blkzoned(sbi)) {
3097                         f2fs_err(sbi, "Zoned block device feature not enabled\n");
3098                         return -EINVAL;
3099                 }
3100                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3101                         if (init_blkz_info(sbi, i)) {
3102                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3103                                 return -EINVAL;
3104                         }
3105                         if (max_devices == 1)
3106                                 break;
3107                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3108                                   i, FDEV(i).path,
3109                                   FDEV(i).total_segments,
3110                                   FDEV(i).start_blk, FDEV(i).end_blk,
3111                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3112                                   "Host-aware" : "Host-managed");
3113                         continue;
3114                 }
3115 #endif
3116                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3117                           i, FDEV(i).path,
3118                           FDEV(i).total_segments,
3119                           FDEV(i).start_blk, FDEV(i).end_blk);
3120         }
3121         f2fs_info(sbi,
3122                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3123         return 0;
3124 }
3125
3126 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3127 {
3128 #ifdef CONFIG_UNICODE
3129         if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3130                 const struct f2fs_sb_encodings *encoding_info;
3131                 struct unicode_map *encoding;
3132                 __u16 encoding_flags;
3133
3134                 if (f2fs_sb_has_encrypt(sbi)) {
3135                         f2fs_err(sbi,
3136                                 "Can't mount with encoding and encryption");
3137                         return -EINVAL;
3138                 }
3139
3140                 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3141                                           &encoding_flags)) {
3142                         f2fs_err(sbi,
3143                                  "Encoding requested by superblock is unknown");
3144                         return -EINVAL;
3145                 }
3146
3147                 encoding = utf8_load(encoding_info->version);
3148                 if (IS_ERR(encoding)) {
3149                         f2fs_err(sbi,
3150                                  "can't mount with superblock charset: %s-%s "
3151                                  "not supported by the kernel. flags: 0x%x.",
3152                                  encoding_info->name, encoding_info->version,
3153                                  encoding_flags);
3154                         return PTR_ERR(encoding);
3155                 }
3156                 f2fs_info(sbi, "Using encoding defined by superblock: "
3157                          "%s-%s with flags 0x%hx", encoding_info->name,
3158                          encoding_info->version?:"\b", encoding_flags);
3159
3160                 sbi->s_encoding = encoding;
3161                 sbi->s_encoding_flags = encoding_flags;
3162                 sbi->sb->s_d_op = &f2fs_dentry_ops;
3163         }
3164 #else
3165         if (f2fs_sb_has_casefold(sbi)) {
3166                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3167                 return -EINVAL;
3168         }
3169 #endif
3170         return 0;
3171 }
3172
3173 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3174 {
3175         struct f2fs_sm_info *sm_i = SM_I(sbi);
3176
3177         /* adjust parameters according to the volume size */
3178         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3179                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3180                 sm_i->dcc_info->discard_granularity = 1;
3181                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3182         }
3183
3184         sbi->readdir_ra = 1;
3185 }
3186
3187 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3188 {
3189         struct f2fs_sb_info *sbi;
3190         struct f2fs_super_block *raw_super;
3191         struct inode *root;
3192         int err;
3193         bool skip_recovery = false, need_fsck = false;
3194         char *options = NULL;
3195         int recovery, i, valid_super_block;
3196         struct curseg_info *seg_i;
3197         int retry_cnt = 1;
3198
3199 try_onemore:
3200         err = -EINVAL;
3201         raw_super = NULL;
3202         valid_super_block = -1;
3203         recovery = 0;
3204
3205         /* allocate memory for f2fs-specific super block info */
3206         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3207         if (!sbi)
3208                 return -ENOMEM;
3209
3210         sbi->sb = sb;
3211
3212         /* Load the checksum driver */
3213         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3214         if (IS_ERR(sbi->s_chksum_driver)) {
3215                 f2fs_err(sbi, "Cannot load crc32 driver.");
3216                 err = PTR_ERR(sbi->s_chksum_driver);
3217                 sbi->s_chksum_driver = NULL;
3218                 goto free_sbi;
3219         }
3220
3221         /* set a block size */
3222         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3223                 f2fs_err(sbi, "unable to set blocksize");
3224                 goto free_sbi;
3225         }
3226
3227         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3228                                                                 &recovery);
3229         if (err)
3230                 goto free_sbi;
3231
3232         sb->s_fs_info = sbi;
3233         sbi->raw_super = raw_super;
3234
3235         /* precompute checksum seed for metadata */
3236         if (f2fs_sb_has_inode_chksum(sbi))
3237                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3238                                                 sizeof(raw_super->uuid));
3239
3240         /*
3241          * The BLKZONED feature indicates that the drive was formatted with
3242          * zone alignment optimization. This is optional for host-aware
3243          * devices, but mandatory for host-managed zoned block devices.
3244          */
3245 #ifndef CONFIG_BLK_DEV_ZONED
3246         if (f2fs_sb_has_blkzoned(sbi)) {
3247                 f2fs_err(sbi, "Zoned block device support is not enabled");
3248                 err = -EOPNOTSUPP;
3249                 goto free_sb_buf;
3250         }
3251 #endif
3252         default_options(sbi);
3253         /* parse mount options */
3254         options = kstrdup((const char *)data, GFP_KERNEL);
3255         if (data && !options) {
3256                 err = -ENOMEM;
3257                 goto free_sb_buf;
3258         }
3259
3260         err = parse_options(sb, options);
3261         if (err)
3262                 goto free_options;
3263
3264         sbi->max_file_blocks = max_file_blocks();
3265         sb->s_maxbytes = sbi->max_file_blocks <<
3266                                 le32_to_cpu(raw_super->log_blocksize);
3267         sb->s_max_links = F2FS_LINK_MAX;
3268
3269         err = f2fs_setup_casefold(sbi);
3270         if (err)
3271                 goto free_options;
3272
3273 #ifdef CONFIG_QUOTA
3274         sb->dq_op = &f2fs_quota_operations;
3275         sb->s_qcop = &f2fs_quotactl_ops;
3276         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3277
3278         if (f2fs_sb_has_quota_ino(sbi)) {
3279                 for (i = 0; i < MAXQUOTAS; i++) {
3280                         if (f2fs_qf_ino(sbi->sb, i))
3281                                 sbi->nquota_files++;
3282                 }
3283         }
3284 #endif
3285
3286         sb->s_op = &f2fs_sops;
3287 #ifdef CONFIG_FS_ENCRYPTION
3288         sb->s_cop = &f2fs_cryptops;
3289 #endif
3290 #ifdef CONFIG_FS_VERITY
3291         sb->s_vop = &f2fs_verityops;
3292 #endif
3293         sb->s_xattr = f2fs_xattr_handlers;
3294         sb->s_export_op = &f2fs_export_ops;
3295         sb->s_magic = F2FS_SUPER_MAGIC;
3296         sb->s_time_gran = 1;
3297         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3298                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3299         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3300         sb->s_iflags |= SB_I_CGROUPWB;
3301
3302         /* init f2fs-specific super block info */
3303         sbi->valid_super_block = valid_super_block;
3304         mutex_init(&sbi->gc_mutex);
3305         mutex_init(&sbi->writepages);
3306         mutex_init(&sbi->cp_mutex);
3307         mutex_init(&sbi->resize_mutex);
3308         init_rwsem(&sbi->node_write);
3309         init_rwsem(&sbi->node_change);
3310
3311         /* disallow all the data/node/meta page writes */
3312         set_sbi_flag(sbi, SBI_POR_DOING);
3313         spin_lock_init(&sbi->stat_lock);
3314
3315         /* init iostat info */
3316         spin_lock_init(&sbi->iostat_lock);
3317         sbi->iostat_enable = false;
3318
3319         for (i = 0; i < NR_PAGE_TYPE; i++) {
3320                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3321                 int j;
3322
3323                 sbi->write_io[i] =
3324                         f2fs_kmalloc(sbi,
3325                                      array_size(n,
3326                                                 sizeof(struct f2fs_bio_info)),
3327                                      GFP_KERNEL);
3328                 if (!sbi->write_io[i]) {
3329                         err = -ENOMEM;
3330                         goto free_bio_info;
3331                 }
3332
3333                 for (j = HOT; j < n; j++) {
3334                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3335                         sbi->write_io[i][j].sbi = sbi;
3336                         sbi->write_io[i][j].bio = NULL;
3337                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3338                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3339                         INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3340                         init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3341                 }
3342         }
3343
3344         init_rwsem(&sbi->cp_rwsem);
3345         init_rwsem(&sbi->quota_sem);
3346         init_waitqueue_head(&sbi->cp_wait);
3347         init_sb_info(sbi);
3348
3349         err = init_percpu_info(sbi);
3350         if (err)
3351                 goto free_bio_info;
3352
3353         if (F2FS_IO_ALIGNED(sbi)) {
3354                 sbi->write_io_dummy =
3355                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3356                 if (!sbi->write_io_dummy) {
3357                         err = -ENOMEM;
3358                         goto free_percpu;
3359                 }
3360         }
3361
3362         /* get an inode for meta space */
3363         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3364         if (IS_ERR(sbi->meta_inode)) {
3365                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3366                 err = PTR_ERR(sbi->meta_inode);
3367                 goto free_io_dummy;
3368         }
3369
3370         err = f2fs_get_valid_checkpoint(sbi);
3371         if (err) {
3372                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3373                 goto free_meta_inode;
3374         }
3375
3376         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3377                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3378         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3379                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3380                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3381         }
3382
3383         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3384                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3385
3386         /* Initialize device list */
3387         err = f2fs_scan_devices(sbi);
3388         if (err) {
3389                 f2fs_err(sbi, "Failed to find devices");
3390                 goto free_devices;
3391         }
3392
3393         sbi->total_valid_node_count =
3394                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3395         percpu_counter_set(&sbi->total_valid_inode_count,
3396                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3397         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3398         sbi->total_valid_block_count =
3399                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3400         sbi->last_valid_block_count = sbi->total_valid_block_count;
3401         sbi->reserved_blocks = 0;
3402         sbi->current_reserved_blocks = 0;
3403         limit_reserve_root(sbi);
3404
3405         for (i = 0; i < NR_INODE_TYPE; i++) {
3406                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3407                 spin_lock_init(&sbi->inode_lock[i]);
3408         }
3409         mutex_init(&sbi->flush_lock);
3410
3411         f2fs_init_extent_cache_info(sbi);
3412
3413         f2fs_init_ino_entry_info(sbi);
3414
3415         f2fs_init_fsync_node_info(sbi);
3416
3417         /* setup f2fs internal modules */
3418         err = f2fs_build_segment_manager(sbi);
3419         if (err) {
3420                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3421                          err);
3422                 goto free_sm;
3423         }
3424         err = f2fs_build_node_manager(sbi);
3425         if (err) {
3426                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3427                          err);
3428                 goto free_nm;
3429         }
3430
3431         /* For write statistics */
3432         if (sb->s_bdev->bd_part)
3433                 sbi->sectors_written_start =
3434                         (u64)part_stat_read(sb->s_bdev->bd_part,
3435                                             sectors[STAT_WRITE]);
3436
3437         /* Read accumulated write IO statistics if exists */
3438         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3439         if (__exist_node_summaries(sbi))
3440                 sbi->kbytes_written =
3441                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3442
3443         f2fs_build_gc_manager(sbi);
3444
3445         err = f2fs_build_stats(sbi);
3446         if (err)
3447                 goto free_nm;
3448
3449         /* get an inode for node space */
3450         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3451         if (IS_ERR(sbi->node_inode)) {
3452                 f2fs_err(sbi, "Failed to read node inode");
3453                 err = PTR_ERR(sbi->node_inode);
3454                 goto free_stats;
3455         }
3456
3457         /* read root inode and dentry */
3458         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3459         if (IS_ERR(root)) {
3460                 f2fs_err(sbi, "Failed to read root inode");
3461                 err = PTR_ERR(root);
3462                 goto free_node_inode;
3463         }
3464         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3465                         !root->i_size || !root->i_nlink) {
3466                 iput(root);
3467                 err = -EINVAL;
3468                 goto free_node_inode;
3469         }
3470
3471         sb->s_root = d_make_root(root); /* allocate root dentry */
3472         if (!sb->s_root) {
3473                 err = -ENOMEM;
3474                 goto free_node_inode;
3475         }
3476
3477         err = f2fs_register_sysfs(sbi);
3478         if (err)
3479                 goto free_root_inode;
3480
3481 #ifdef CONFIG_QUOTA
3482         /* Enable quota usage during mount */
3483         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3484                 err = f2fs_enable_quotas(sb);
3485                 if (err)
3486                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3487         }
3488 #endif
3489         /* if there are nt orphan nodes free them */
3490         err = f2fs_recover_orphan_inodes(sbi);
3491         if (err)
3492                 goto free_meta;
3493
3494         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3495                 goto reset_checkpoint;
3496
3497         /* recover fsynced data */
3498         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3499                 /*
3500                  * mount should be failed, when device has readonly mode, and
3501                  * previous checkpoint was not done by clean system shutdown.
3502                  */
3503                 if (f2fs_hw_is_readonly(sbi)) {
3504                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3505                                 err = -EROFS;
3506                                 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3507                                 goto free_meta;
3508                         }
3509                         f2fs_info(sbi, "write access unavailable, skipping recovery");
3510                         goto reset_checkpoint;
3511                 }
3512
3513                 if (need_fsck)
3514                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3515
3516                 if (skip_recovery)
3517                         goto reset_checkpoint;
3518
3519                 err = f2fs_recover_fsync_data(sbi, false);
3520                 if (err < 0) {
3521                         if (err != -ENOMEM)
3522                                 skip_recovery = true;
3523                         need_fsck = true;
3524                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3525                                  err);
3526                         goto free_meta;
3527                 }
3528         } else {
3529                 err = f2fs_recover_fsync_data(sbi, true);
3530
3531                 if (!f2fs_readonly(sb) && err > 0) {
3532                         err = -EINVAL;
3533                         f2fs_err(sbi, "Need to recover fsync data");
3534                         goto free_meta;
3535                 }
3536         }
3537
3538         /*
3539          * If the f2fs is not readonly and fsync data recovery succeeds,
3540          * check zoned block devices' write pointer consistency.
3541          */
3542         if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3543                 err = f2fs_check_write_pointer(sbi);
3544                 if (err)
3545                         goto free_meta;
3546         }
3547
3548 reset_checkpoint:
3549         /* f2fs_recover_fsync_data() cleared this already */
3550         clear_sbi_flag(sbi, SBI_POR_DOING);
3551
3552         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3553                 err = f2fs_disable_checkpoint(sbi);
3554                 if (err)
3555                         goto sync_free_meta;
3556         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3557                 f2fs_enable_checkpoint(sbi);
3558         }
3559
3560         /*
3561          * If filesystem is not mounted as read-only then
3562          * do start the gc_thread.
3563          */
3564         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3565                 /* After POR, we can run background GC thread.*/
3566                 err = f2fs_start_gc_thread(sbi);
3567                 if (err)
3568                         goto sync_free_meta;
3569         }
3570         kvfree(options);
3571
3572         /* recover broken superblock */
3573         if (recovery) {
3574                 err = f2fs_commit_super(sbi, true);
3575                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3576                           sbi->valid_super_block ? 1 : 2, err);
3577         }
3578
3579         f2fs_join_shrinker(sbi);
3580
3581         f2fs_tuning_parameters(sbi);
3582
3583         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3584                     cur_cp_version(F2FS_CKPT(sbi)));
3585         f2fs_update_time(sbi, CP_TIME);
3586         f2fs_update_time(sbi, REQ_TIME);
3587         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3588         return 0;
3589
3590 sync_free_meta:
3591         /* safe to flush all the data */
3592         sync_filesystem(sbi->sb);
3593         retry_cnt = 0;
3594
3595 free_meta:
3596 #ifdef CONFIG_QUOTA
3597         f2fs_truncate_quota_inode_pages(sb);
3598         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3599                 f2fs_quota_off_umount(sbi->sb);
3600 #endif
3601         /*
3602          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3603          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3604          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3605          * falls into an infinite loop in f2fs_sync_meta_pages().
3606          */
3607         truncate_inode_pages_final(META_MAPPING(sbi));
3608         /* evict some inodes being cached by GC */
3609         evict_inodes(sb);
3610         f2fs_unregister_sysfs(sbi);
3611 free_root_inode:
3612         dput(sb->s_root);
3613         sb->s_root = NULL;
3614 free_node_inode:
3615         f2fs_release_ino_entry(sbi, true);
3616         truncate_inode_pages_final(NODE_MAPPING(sbi));
3617         iput(sbi->node_inode);
3618         sbi->node_inode = NULL;
3619 free_stats:
3620         f2fs_destroy_stats(sbi);
3621 free_nm:
3622         f2fs_destroy_node_manager(sbi);
3623 free_sm:
3624         f2fs_destroy_segment_manager(sbi);
3625 free_devices:
3626         destroy_device_list(sbi);
3627         kvfree(sbi->ckpt);
3628 free_meta_inode:
3629         make_bad_inode(sbi->meta_inode);
3630         iput(sbi->meta_inode);
3631         sbi->meta_inode = NULL;
3632 free_io_dummy:
3633         mempool_destroy(sbi->write_io_dummy);
3634 free_percpu:
3635         destroy_percpu_info(sbi);
3636 free_bio_info:
3637         for (i = 0; i < NR_PAGE_TYPE; i++)
3638                 kvfree(sbi->write_io[i]);
3639
3640 #ifdef CONFIG_UNICODE
3641         utf8_unload(sbi->s_encoding);
3642 #endif
3643 free_options:
3644 #ifdef CONFIG_QUOTA
3645         for (i = 0; i < MAXQUOTAS; i++)
3646                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3647 #endif
3648         kvfree(options);
3649 free_sb_buf:
3650         kvfree(raw_super);
3651 free_sbi:
3652         if (sbi->s_chksum_driver)
3653                 crypto_free_shash(sbi->s_chksum_driver);
3654         kvfree(sbi);
3655
3656         /* give only one another chance */
3657         if (retry_cnt > 0 && skip_recovery) {
3658                 retry_cnt--;
3659                 shrink_dcache_sb(sb);
3660                 goto try_onemore;
3661         }
3662         return err;
3663 }
3664
3665 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3666                         const char *dev_name, void *data)
3667 {
3668         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3669 }
3670
3671 static void kill_f2fs_super(struct super_block *sb)
3672 {
3673         if (sb->s_root) {
3674                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3675
3676                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3677                 f2fs_stop_gc_thread(sbi);
3678                 f2fs_stop_discard_thread(sbi);
3679
3680                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3681                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3682                         struct cp_control cpc = {
3683                                 .reason = CP_UMOUNT,
3684                         };
3685                         f2fs_write_checkpoint(sbi, &cpc);
3686                 }
3687
3688                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3689                         sb->s_flags &= ~SB_RDONLY;
3690         }
3691         kill_block_super(sb);
3692 }
3693
3694 static struct file_system_type f2fs_fs_type = {
3695         .owner          = THIS_MODULE,
3696         .name           = "f2fs",
3697         .mount          = f2fs_mount,
3698         .kill_sb        = kill_f2fs_super,
3699         .fs_flags       = FS_REQUIRES_DEV,
3700 };
3701 MODULE_ALIAS_FS("f2fs");
3702
3703 static int __init init_inodecache(void)
3704 {
3705         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3706                         sizeof(struct f2fs_inode_info), 0,
3707                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3708         if (!f2fs_inode_cachep)
3709                 return -ENOMEM;
3710         return 0;
3711 }
3712
3713 static void destroy_inodecache(void)
3714 {
3715         /*
3716          * Make sure all delayed rcu free inodes are flushed before we
3717          * destroy cache.
3718          */
3719         rcu_barrier();
3720         kmem_cache_destroy(f2fs_inode_cachep);
3721 }
3722
3723 static int __init init_f2fs_fs(void)
3724 {
3725         int err;
3726
3727         if (PAGE_SIZE != F2FS_BLKSIZE) {
3728                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3729                                 PAGE_SIZE, F2FS_BLKSIZE);
3730                 return -EINVAL;
3731         }
3732
3733         f2fs_build_trace_ios();
3734
3735         err = init_inodecache();
3736         if (err)
3737                 goto fail;
3738         err = f2fs_create_node_manager_caches();
3739         if (err)
3740                 goto free_inodecache;
3741         err = f2fs_create_segment_manager_caches();
3742         if (err)
3743                 goto free_node_manager_caches;
3744         err = f2fs_create_checkpoint_caches();
3745         if (err)
3746                 goto free_segment_manager_caches;
3747         err = f2fs_create_extent_cache();
3748         if (err)
3749                 goto free_checkpoint_caches;
3750         err = f2fs_init_sysfs();
3751         if (err)
3752                 goto free_extent_cache;
3753         err = register_shrinker(&f2fs_shrinker_info);
3754         if (err)
3755                 goto free_sysfs;
3756         err = register_filesystem(&f2fs_fs_type);
3757         if (err)
3758                 goto free_shrinker;
3759         f2fs_create_root_stats();
3760         err = f2fs_init_post_read_processing();
3761         if (err)
3762                 goto free_root_stats;
3763         err = f2fs_init_bio_entry_cache();
3764         if (err)
3765                 goto free_post_read;
3766         err = f2fs_init_bioset();
3767         if (err)
3768                 goto free_bio_enrty_cache;
3769         return 0;
3770 free_bio_enrty_cache:
3771         f2fs_destroy_bio_entry_cache();
3772 free_post_read:
3773         f2fs_destroy_post_read_processing();
3774 free_root_stats:
3775         f2fs_destroy_root_stats();
3776         unregister_filesystem(&f2fs_fs_type);
3777 free_shrinker:
3778         unregister_shrinker(&f2fs_shrinker_info);
3779 free_sysfs:
3780         f2fs_exit_sysfs();
3781 free_extent_cache:
3782         f2fs_destroy_extent_cache();
3783 free_checkpoint_caches:
3784         f2fs_destroy_checkpoint_caches();
3785 free_segment_manager_caches:
3786         f2fs_destroy_segment_manager_caches();
3787 free_node_manager_caches:
3788         f2fs_destroy_node_manager_caches();
3789 free_inodecache:
3790         destroy_inodecache();
3791 fail:
3792         return err;
3793 }
3794
3795 static void __exit exit_f2fs_fs(void)
3796 {
3797         f2fs_destroy_bioset();
3798         f2fs_destroy_bio_entry_cache();
3799         f2fs_destroy_post_read_processing();
3800         f2fs_destroy_root_stats();
3801         unregister_filesystem(&f2fs_fs_type);
3802         unregister_shrinker(&f2fs_shrinker_info);
3803         f2fs_exit_sysfs();
3804         f2fs_destroy_extent_cache();
3805         f2fs_destroy_checkpoint_caches();
3806         f2fs_destroy_segment_manager_caches();
3807         f2fs_destroy_node_manager_caches();
3808         destroy_inodecache();
3809         f2fs_destroy_trace_ios();
3810 }
3811
3812 module_init(init_f2fs_fs)
3813 module_exit(exit_f2fs_fs)
3814
3815 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3816 MODULE_DESCRIPTION("Flash Friendly File System");
3817 MODULE_LICENSE("GPL");
3818