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
  27 #include "f2fs.h"
  28 #include "node.h"
  29 #include "segment.h"
  30 #include "xattr.h"
  31 #include "gc.h"
  32 #include "trace.h"
  33
  34 #define CREATE_TRACE_POINTS
  35 #include <trace/events/f2fs.h>
  36
  37 static struct kmem_cache *f2fs_inode_cachep;
  38
  39 #ifdef CONFIG_F2FS_FAULT_INJECTION
  40
  41 const char *f2fs_fault_name[FAULT_MAX] = {
  42         [FAULT_KMALLOC]         = "kmalloc",
  43         [FAULT_KVMALLOC]        = "kvmalloc",
  44         [FAULT_PAGE_ALLOC]      = "page alloc",
  45         [FAULT_PAGE_GET]        = "page get",
  46         [FAULT_ALLOC_BIO]       = "alloc bio",
  47         [FAULT_ALLOC_NID]       = "alloc nid",
  48         [FAULT_ORPHAN]          = "orphan",
  49         [FAULT_BLOCK]           = "no more block",
  50         [FAULT_DIR_DEPTH]       = "too big dir depth",
  51         [FAULT_EVICT_INODE]     = "evict_inode fail",
  52         [FAULT_TRUNCATE]        = "truncate fail",
  53         [FAULT_READ_IO]         = "read IO error",
  54         [FAULT_CHECKPOINT]      = "checkpoint error",
  55         [FAULT_DISCARD]         = "discard error",
  56         [FAULT_WRITE_IO]        = "write IO error",
  57 };
  58
  59 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  60                                                         unsigned int type)
  61 {
  62         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  63
  64         if (rate) {
  65                 atomic_set(&ffi->inject_ops, 0);
  66                 ffi->inject_rate = rate;
  67         }
  68
  69         if (type)
  70                 ffi->inject_type = type;
  71
  72         if (!rate && !type)
  73                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  74 }
  75 #endif
  76
  77 /* f2fs-wide shrinker description */
  78 static struct shrinker f2fs_shrinker_info = {
  79         .scan_objects = f2fs_shrink_scan,
  80         .count_objects = f2fs_shrink_count,
  81         .seeks = DEFAULT_SEEKS,
  82 };
  83
  84 enum {
  85         Opt_gc_background,
  86         Opt_disable_roll_forward,
  87         Opt_norecovery,
  88         Opt_discard,
  89         Opt_nodiscard,
  90         Opt_noheap,
  91         Opt_heap,
  92         Opt_user_xattr,
  93         Opt_nouser_xattr,
  94         Opt_acl,
  95         Opt_noacl,
  96         Opt_active_logs,
  97         Opt_disable_ext_identify,
  98         Opt_inline_xattr,
  99         Opt_noinline_xattr,
 100         Opt_inline_xattr_size,
 101         Opt_inline_data,
 102         Opt_inline_dentry,
 103         Opt_noinline_dentry,
 104         Opt_flush_merge,
 105         Opt_noflush_merge,
 106         Opt_nobarrier,
 107         Opt_fastboot,
 108         Opt_extent_cache,
 109         Opt_noextent_cache,
 110         Opt_noinline_data,
 111         Opt_data_flush,
 112         Opt_reserve_root,
 113         Opt_resgid,
 114         Opt_resuid,
 115         Opt_mode,
 116         Opt_io_size_bits,
 117         Opt_fault_injection,
 118         Opt_fault_type,
 119         Opt_lazytime,
 120         Opt_nolazytime,
 121         Opt_quota,
 122         Opt_noquota,
 123         Opt_usrquota,
 124         Opt_grpquota,
 125         Opt_prjquota,
 126         Opt_usrjquota,
 127         Opt_grpjquota,
 128         Opt_prjjquota,
 129         Opt_offusrjquota,
 130         Opt_offgrpjquota,
 131         Opt_offprjjquota,
 132         Opt_jqfmt_vfsold,
 133         Opt_jqfmt_vfsv0,
 134         Opt_jqfmt_vfsv1,
 135         Opt_whint,
 136         Opt_alloc,
 137         Opt_fsync,
 138         Opt_test_dummy_encryption,
 139         Opt_checkpoint,
 140         Opt_err,
 141 };
 142
 143 static match_table_t f2fs_tokens = {
 144         {Opt_gc_background, "background_gc=%s"},
 145         {Opt_disable_roll_forward, "disable_roll_forward"},
 146         {Opt_norecovery, "norecovery"},
 147         {Opt_discard, "discard"},
 148         {Opt_nodiscard, "nodiscard"},
 149         {Opt_noheap, "no_heap"},
 150         {Opt_heap, "heap"},
 151         {Opt_user_xattr, "user_xattr"},
 152         {Opt_nouser_xattr, "nouser_xattr"},
 153         {Opt_acl, "acl"},
 154         {Opt_noacl, "noacl"},
 155         {Opt_active_logs, "active_logs=%u"},
 156         {Opt_disable_ext_identify, "disable_ext_identify"},
 157         {Opt_inline_xattr, "inline_xattr"},
 158         {Opt_noinline_xattr, "noinline_xattr"},
 159         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 160         {Opt_inline_data, "inline_data"},
 161         {Opt_inline_dentry, "inline_dentry"},
 162         {Opt_noinline_dentry, "noinline_dentry"},
 163         {Opt_flush_merge, "flush_merge"},
 164         {Opt_noflush_merge, "noflush_merge"},
 165         {Opt_nobarrier, "nobarrier"},
 166         {Opt_fastboot, "fastboot"},
 167         {Opt_extent_cache, "extent_cache"},
 168         {Opt_noextent_cache, "noextent_cache"},
 169         {Opt_noinline_data, "noinline_data"},
 170         {Opt_data_flush, "data_flush"},
 171         {Opt_reserve_root, "reserve_root=%u"},
 172         {Opt_resgid, "resgid=%u"},
 173         {Opt_resuid, "resuid=%u"},
 174         {Opt_mode, "mode=%s"},
 175         {Opt_io_size_bits, "io_bits=%u"},
 176         {Opt_fault_injection, "fault_injection=%u"},
 177         {Opt_fault_type, "fault_type=%u"},
 178         {Opt_lazytime, "lazytime"},
 179         {Opt_nolazytime, "nolazytime"},
 180         {Opt_quota, "quota"},
 181         {Opt_noquota, "noquota"},
 182         {Opt_usrquota, "usrquota"},
 183         {Opt_grpquota, "grpquota"},
 184         {Opt_prjquota, "prjquota"},
 185         {Opt_usrjquota, "usrjquota=%s"},
 186         {Opt_grpjquota, "grpjquota=%s"},
 187         {Opt_prjjquota, "prjjquota=%s"},
 188         {Opt_offusrjquota, "usrjquota="},
 189         {Opt_offgrpjquota, "grpjquota="},
 190         {Opt_offprjjquota, "prjjquota="},
 191         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 192         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 193         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 194         {Opt_whint, "whint_mode=%s"},
 195         {Opt_alloc, "alloc_mode=%s"},
 196         {Opt_fsync, "fsync_mode=%s"},
 197         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 198         {Opt_checkpoint, "checkpoint=%s"},
 199         {Opt_err, NULL},
 200 };
 201
 202 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 203 {
 204         struct va_format vaf;
 205         va_list args;
 206
 207         va_start(args, fmt);
 208         vaf.fmt = fmt;
 209         vaf.va = &args;
 210         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 211         va_end(args);
 212 }
 213
 214 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 215 {
 216         block_t limit = (sbi->user_block_count << 1) / 1000;
 217
 218         /* limit is 0.2% */
 219         if (test_opt(sbi, RESERVE_ROOT) &&
 220                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 221                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 222                 f2fs_msg(sbi->sb, KERN_INFO,
 223                         "Reduce reserved blocks for root = %u",
 224                         F2FS_OPTION(sbi).root_reserved_blocks);
 225         }
 226         if (!test_opt(sbi, RESERVE_ROOT) &&
 227                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 228                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 229                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 230                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 231                 f2fs_msg(sbi->sb, KERN_INFO,
 232                         "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 233                                 from_kuid_munged(&init_user_ns,
 234                                         F2FS_OPTION(sbi).s_resuid),
 235                                 from_kgid_munged(&init_user_ns,
 236                                         F2FS_OPTION(sbi).s_resgid));
 237 }
 238
 239 static void init_once(void *foo)
 240 {
 241         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 242
 243         inode_init_once(&fi->vfs_inode);
 244 }
 245
 246 #ifdef CONFIG_QUOTA
 247 static const char * const quotatypes[] = INITQFNAMES;
 248 #define QTYPE2NAME(t) (quotatypes[t])
 249 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 250                                                         substring_t *args)
 251 {
 252         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 253         char *qname;
 254         int ret = -EINVAL;
 255
 256         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 257                 f2fs_msg(sb, KERN_ERR,
 258                         "Cannot change journaled "
 259                         "quota options when quota turned on");
 260                 return -EINVAL;
 261         }
 262         if (f2fs_sb_has_quota_ino(sbi)) {
 263                 f2fs_msg(sb, KERN_INFO,
 264                         "QUOTA feature is enabled, so ignore qf_name");
 265                 return 0;
 266         }
 267
 268         qname = match_strdup(args);
 269         if (!qname) {
 270                 f2fs_msg(sb, KERN_ERR,
 271                         "Not enough memory for storing quotafile name");
 272                 return -ENOMEM;
 273         }
 274         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 275                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 276                         ret = 0;
 277                 else
 278                         f2fs_msg(sb, KERN_ERR,
 279                                  "%s quota file already specified",
 280                                  QTYPE2NAME(qtype));
 281                 goto errout;
 282         }
 283         if (strchr(qname, '/')) {
 284                 f2fs_msg(sb, KERN_ERR,
 285                         "quotafile must be on filesystem root");
 286                 goto errout;
 287         }
 288         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 289         set_opt(sbi, QUOTA);
 290         return 0;
 291 errout:
 292         kvfree(qname);
 293         return ret;
 294 }
 295
 296 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 297 {
 298         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 299
 300         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 301                 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
 302                         " when quota turned on");
 303                 return -EINVAL;
 304         }
 305         kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 306         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 307         return 0;
 308 }
 309
 310 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 311 {
 312         /*
 313          * We do the test below only for project quotas. 'usrquota' and
 314          * 'grpquota' mount options are allowed even without quota feature
 315          * to support legacy quotas in quota files.
 316          */
 317         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 318                 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
 319                          "Cannot enable project quota enforcement.");
 320                 return -1;
 321         }
 322         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 323                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 324                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 325                 if (test_opt(sbi, USRQUOTA) &&
 326                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 327                         clear_opt(sbi, USRQUOTA);
 328
 329                 if (test_opt(sbi, GRPQUOTA) &&
 330                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 331                         clear_opt(sbi, GRPQUOTA);
 332
 333                 if (test_opt(sbi, PRJQUOTA) &&
 334                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 335                         clear_opt(sbi, PRJQUOTA);
 336
 337                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 338                                 test_opt(sbi, PRJQUOTA)) {
 339                         f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
 340                                         "format mixing");
 341                         return -1;
 342                 }
 343
 344                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 345                         f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
 346                                         "not specified");
 347                         return -1;
 348                 }
 349         }
 350
 351         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 352                 f2fs_msg(sbi->sb, KERN_INFO,
 353                         "QUOTA feature is enabled, so ignore jquota_fmt");
 354                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 355         }
 356         return 0;
 357 }
 358 #endif
 359
 360 static int parse_options(struct super_block *sb, char *options)
 361 {
 362         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 363         substring_t args[MAX_OPT_ARGS];
 364         char *p, *name;
 365         int arg = 0;
 366         kuid_t uid;
 367         kgid_t gid;
 368 #ifdef CONFIG_QUOTA
 369         int ret;
 370 #endif
 371
 372         if (!options)
 373                 return 0;
 374
 375         while ((p = strsep(&options, ",")) != NULL) {
 376                 int token;
 377                 if (!*p)
 378                         continue;
 379                 /*
 380                  * Initialize args struct so we know whether arg was
 381                  * found; some options take optional arguments.
 382                  */
 383                 args[0].to = args[0].from = NULL;
 384                 token = match_token(p, f2fs_tokens, args);
 385
 386                 switch (token) {
 387                 case Opt_gc_background:
 388                         name = match_strdup(&args[0]);
 389
 390                         if (!name)
 391                                 return -ENOMEM;
 392                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 393                                 set_opt(sbi, BG_GC);
 394                                 clear_opt(sbi, FORCE_FG_GC);
 395                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 396                                 clear_opt(sbi, BG_GC);
 397                                 clear_opt(sbi, FORCE_FG_GC);
 398                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 399                                 set_opt(sbi, BG_GC);
 400                                 set_opt(sbi, FORCE_FG_GC);
 401                         } else {
 402                                 kvfree(name);
 403                                 return -EINVAL;
 404                         }
 405                         kvfree(name);
 406                         break;
 407                 case Opt_disable_roll_forward:
 408                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 409                         break;
 410                 case Opt_norecovery:
 411                         /* this option mounts f2fs with ro */
 412                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 413                         if (!f2fs_readonly(sb))
 414                                 return -EINVAL;
 415                         break;
 416                 case Opt_discard:
 417                         set_opt(sbi, DISCARD);
 418                         break;
 419                 case Opt_nodiscard:
 420                         if (f2fs_sb_has_blkzoned(sbi)) {
 421                                 f2fs_msg(sb, KERN_WARNING,
 422                                         "discard is required for zoned block devices");
 423                                 return -EINVAL;
 424                         }
 425                         clear_opt(sbi, DISCARD);
 426                         break;
 427                 case Opt_noheap:
 428                         set_opt(sbi, NOHEAP);
 429                         break;
 430                 case Opt_heap:
 431                         clear_opt(sbi, NOHEAP);
 432                         break;
 433 #ifdef CONFIG_F2FS_FS_XATTR
 434                 case Opt_user_xattr:
 435                         set_opt(sbi, XATTR_USER);
 436                         break;
 437                 case Opt_nouser_xattr:
 438                         clear_opt(sbi, XATTR_USER);
 439                         break;
 440                 case Opt_inline_xattr:
 441                         set_opt(sbi, INLINE_XATTR);
 442                         break;
 443                 case Opt_noinline_xattr:
 444                         clear_opt(sbi, INLINE_XATTR);
 445                         break;
 446                 case Opt_inline_xattr_size:
 447                         if (args->from && match_int(args, &arg))
 448                                 return -EINVAL;
 449                         set_opt(sbi, INLINE_XATTR_SIZE);
 450                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 451                         break;
 452 #else
 453                 case Opt_user_xattr:
 454                         f2fs_msg(sb, KERN_INFO,
 455                                 "user_xattr options not supported");
 456                         break;
 457                 case Opt_nouser_xattr:
 458                         f2fs_msg(sb, KERN_INFO,
 459                                 "nouser_xattr options not supported");
 460                         break;
 461                 case Opt_inline_xattr:
 462                         f2fs_msg(sb, KERN_INFO,
 463                                 "inline_xattr options not supported");
 464                         break;
 465                 case Opt_noinline_xattr:
 466                         f2fs_msg(sb, KERN_INFO,
 467                                 "noinline_xattr options not supported");
 468                         break;
 469 #endif
 470 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 471                 case Opt_acl:
 472                         set_opt(sbi, POSIX_ACL);
 473                         break;
 474                 case Opt_noacl:
 475                         clear_opt(sbi, POSIX_ACL);
 476                         break;
 477 #else
 478                 case Opt_acl:
 479                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 480                         break;
 481                 case Opt_noacl:
 482                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 483                         break;
 484 #endif
 485                 case Opt_active_logs:
 486                         if (args->from && match_int(args, &arg))
 487                                 return -EINVAL;
 488                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 489                                 return -EINVAL;
 490                         F2FS_OPTION(sbi).active_logs = arg;
 491                         break;
 492                 case Opt_disable_ext_identify:
 493                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 494                         break;
 495                 case Opt_inline_data:
 496                         set_opt(sbi, INLINE_DATA);
 497                         break;
 498                 case Opt_inline_dentry:
 499                         set_opt(sbi, INLINE_DENTRY);
 500                         break;
 501                 case Opt_noinline_dentry:
 502                         clear_opt(sbi, INLINE_DENTRY);
 503                         break;
 504                 case Opt_flush_merge:
 505                         set_opt(sbi, FLUSH_MERGE);
 506                         break;
 507                 case Opt_noflush_merge:
 508                         clear_opt(sbi, FLUSH_MERGE);
 509                         break;
 510                 case Opt_nobarrier:
 511                         set_opt(sbi, NOBARRIER);
 512                         break;
 513                 case Opt_fastboot:
 514                         set_opt(sbi, FASTBOOT);
 515                         break;
 516                 case Opt_extent_cache:
 517                         set_opt(sbi, EXTENT_CACHE);
 518                         break;
 519                 case Opt_noextent_cache:
 520                         clear_opt(sbi, EXTENT_CACHE);
 521                         break;
 522                 case Opt_noinline_data:
 523                         clear_opt(sbi, INLINE_DATA);
 524                         break;
 525                 case Opt_data_flush:
 526                         set_opt(sbi, DATA_FLUSH);
 527                         break;
 528                 case Opt_reserve_root:
 529                         if (args->from && match_int(args, &arg))
 530                                 return -EINVAL;
 531                         if (test_opt(sbi, RESERVE_ROOT)) {
 532                                 f2fs_msg(sb, KERN_INFO,
 533                                         "Preserve previous reserve_root=%u",
 534                                         F2FS_OPTION(sbi).root_reserved_blocks);
 535                         } else {
 536                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 537                                 set_opt(sbi, RESERVE_ROOT);
 538                         }
 539                         break;
 540                 case Opt_resuid:
 541                         if (args->from && match_int(args, &arg))
 542                                 return -EINVAL;
 543                         uid = make_kuid(current_user_ns(), arg);
 544                         if (!uid_valid(uid)) {
 545                                 f2fs_msg(sb, KERN_ERR,
 546                                         "Invalid uid value %d", arg);
 547                                 return -EINVAL;
 548                         }
 549                         F2FS_OPTION(sbi).s_resuid = uid;
 550                         break;
 551                 case Opt_resgid:
 552                         if (args->from && match_int(args, &arg))
 553                                 return -EINVAL;
 554                         gid = make_kgid(current_user_ns(), arg);
 555                         if (!gid_valid(gid)) {
 556                                 f2fs_msg(sb, KERN_ERR,
 557                                         "Invalid gid value %d", arg);
 558                                 return -EINVAL;
 559                         }
 560                         F2FS_OPTION(sbi).s_resgid = gid;
 561                         break;
 562                 case Opt_mode:
 563                         name = match_strdup(&args[0]);
 564
 565                         if (!name)
 566                                 return -ENOMEM;
 567                         if (strlen(name) == 8 &&
 568                                         !strncmp(name, "adaptive", 8)) {
 569                                 if (f2fs_sb_has_blkzoned(sbi)) {
 570                                         f2fs_msg(sb, KERN_WARNING,
 571                                                  "adaptive mode is not allowed with "
 572                                                  "zoned block device feature");
 573                                         kvfree(name);
 574                                         return -EINVAL;
 575                                 }
 576                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 577                         } else if (strlen(name) == 3 &&
 578                                         !strncmp(name, "lfs", 3)) {
 579                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 580                         } else {
 581                                 kvfree(name);
 582                                 return -EINVAL;
 583                         }
 584                         kvfree(name);
 585                         break;
 586                 case Opt_io_size_bits:
 587                         if (args->from && match_int(args, &arg))
 588                                 return -EINVAL;
 589                         if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
 590                                 f2fs_msg(sb, KERN_WARNING,
 591                                         "Not support %d, larger than %d",
 592                                         1 << arg, BIO_MAX_PAGES);
 593                                 return -EINVAL;
 594                         }
 595                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 596                         break;
 597 #ifdef CONFIG_F2FS_FAULT_INJECTION
 598                 case Opt_fault_injection:
 599                         if (args->from && match_int(args, &arg))
 600                                 return -EINVAL;
 601                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 602                         set_opt(sbi, FAULT_INJECTION);
 603                         break;
 604
 605                 case Opt_fault_type:
 606                         if (args->from && match_int(args, &arg))
 607                                 return -EINVAL;
 608                         f2fs_build_fault_attr(sbi, 0, arg);
 609                         set_opt(sbi, FAULT_INJECTION);
 610                         break;
 611 #else
 612                 case Opt_fault_injection:
 613                         f2fs_msg(sb, KERN_INFO,
 614                                 "fault_injection options not supported");
 615                         break;
 616
 617                 case Opt_fault_type:
 618                         f2fs_msg(sb, KERN_INFO,
 619                                 "fault_type options not supported");
 620                         break;
 621 #endif
 622                 case Opt_lazytime:
 623                         sb->s_flags |= SB_LAZYTIME;
 624                         break;
 625                 case Opt_nolazytime:
 626                         sb->s_flags &= ~SB_LAZYTIME;
 627                         break;
 628 #ifdef CONFIG_QUOTA
 629                 case Opt_quota:
 630                 case Opt_usrquota:
 631                         set_opt(sbi, USRQUOTA);
 632                         break;
 633                 case Opt_grpquota:
 634                         set_opt(sbi, GRPQUOTA);
 635                         break;
 636                 case Opt_prjquota:
 637                         set_opt(sbi, PRJQUOTA);
 638                         break;
 639                 case Opt_usrjquota:
 640                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 641                         if (ret)
 642                                 return ret;
 643                         break;
 644                 case Opt_grpjquota:
 645                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 646                         if (ret)
 647                                 return ret;
 648                         break;
 649                 case Opt_prjjquota:
 650                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 651                         if (ret)
 652                                 return ret;
 653                         break;
 654                 case Opt_offusrjquota:
 655                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 656                         if (ret)
 657                                 return ret;
 658                         break;
 659                 case Opt_offgrpjquota:
 660                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 661                         if (ret)
 662                                 return ret;
 663                         break;
 664                 case Opt_offprjjquota:
 665                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 666                         if (ret)
 667                                 return ret;
 668                         break;
 669                 case Opt_jqfmt_vfsold:
 670                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 671                         break;
 672                 case Opt_jqfmt_vfsv0:
 673                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 674                         break;
 675                 case Opt_jqfmt_vfsv1:
 676                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 677                         break;
 678                 case Opt_noquota:
 679                         clear_opt(sbi, QUOTA);
 680                         clear_opt(sbi, USRQUOTA);
 681                         clear_opt(sbi, GRPQUOTA);
 682                         clear_opt(sbi, PRJQUOTA);
 683                         break;
 684 #else
 685                 case Opt_quota:
 686                 case Opt_usrquota:
 687                 case Opt_grpquota:
 688                 case Opt_prjquota:
 689                 case Opt_usrjquota:
 690                 case Opt_grpjquota:
 691                 case Opt_prjjquota:
 692                 case Opt_offusrjquota:
 693                 case Opt_offgrpjquota:
 694                 case Opt_offprjjquota:
 695                 case Opt_jqfmt_vfsold:
 696                 case Opt_jqfmt_vfsv0:
 697                 case Opt_jqfmt_vfsv1:
 698                 case Opt_noquota:
 699                         f2fs_msg(sb, KERN_INFO,
 700                                         "quota operations not supported");
 701                         break;
 702 #endif
 703                 case Opt_whint:
 704                         name = match_strdup(&args[0]);
 705                         if (!name)
 706                                 return -ENOMEM;
 707                         if (strlen(name) == 10 &&
 708                                         !strncmp(name, "user-based", 10)) {
 709                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 710                         } else if (strlen(name) == 3 &&
 711                                         !strncmp(name, "off", 3)) {
 712                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 713                         } else if (strlen(name) == 8 &&
 714                                         !strncmp(name, "fs-based", 8)) {
 715                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 716                         } else {
 717                                 kvfree(name);
 718                                 return -EINVAL;
 719                         }
 720                         kvfree(name);
 721                         break;
 722                 case Opt_alloc:
 723                         name = match_strdup(&args[0]);
 724                         if (!name)
 725                                 return -ENOMEM;
 726
 727                         if (strlen(name) == 7 &&
 728                                         !strncmp(name, "default", 7)) {
 729                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 730                         } else if (strlen(name) == 5 &&
 731                                         !strncmp(name, "reuse", 5)) {
 732                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 733                         } else {
 734                                 kvfree(name);
 735                                 return -EINVAL;
 736                         }
 737                         kvfree(name);
 738                         break;
 739                 case Opt_fsync:
 740                         name = match_strdup(&args[0]);
 741                         if (!name)
 742                                 return -ENOMEM;
 743                         if (strlen(name) == 5 &&
 744                                         !strncmp(name, "posix", 5)) {
 745                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 746                         } else if (strlen(name) == 6 &&
 747                                         !strncmp(name, "strict", 6)) {
 748                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 749                         } else if (strlen(name) == 9 &&
 750                                         !strncmp(name, "nobarrier", 9)) {
 751                                 F2FS_OPTION(sbi).fsync_mode =
 752                                                         FSYNC_MODE_NOBARRIER;
 753                         } else {
 754                                 kvfree(name);
 755                                 return -EINVAL;
 756                         }
 757                         kvfree(name);
 758                         break;
 759                 case Opt_test_dummy_encryption:
 760 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 761                         if (!f2fs_sb_has_encrypt(sbi)) {
 762                                 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
 763                                 return -EINVAL;
 764                         }
 765
 766                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 767                         f2fs_msg(sb, KERN_INFO,
 768                                         "Test dummy encryption mode enabled");
 769 #else
 770                         f2fs_msg(sb, KERN_INFO,
 771                                         "Test dummy encryption mount option ignored");
 772 #endif
 773                         break;
 774                 case Opt_checkpoint:
 775                         name = match_strdup(&args[0]);
 776                         if (!name)
 777                                 return -ENOMEM;
 778
 779                         if (strlen(name) == 6 &&
 780                                         !strncmp(name, "enable", 6)) {
 781                                 clear_opt(sbi, DISABLE_CHECKPOINT);
 782                         } else if (strlen(name) == 7 &&
 783                                         !strncmp(name, "disable", 7)) {
 784                                 set_opt(sbi, DISABLE_CHECKPOINT);
 785                         } else {
 786                                 kvfree(name);
 787                                 return -EINVAL;
 788                         }
 789                         kvfree(name);
 790                         break;
 791                 default:
 792                         f2fs_msg(sb, KERN_ERR,
 793                                 "Unrecognized mount option \"%s\" or missing value",
 794                                 p);
 795                         return -EINVAL;
 796                 }
 797         }
 798 #ifdef CONFIG_QUOTA
 799         if (f2fs_check_quota_options(sbi))
 800                 return -EINVAL;
 801 #else
 802         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
 803                 f2fs_msg(sbi->sb, KERN_INFO,
 804                          "Filesystem with quota feature cannot be mounted RDWR "
 805                          "without CONFIG_QUOTA");
 806                 return -EINVAL;
 807         }
 808         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
 809                 f2fs_msg(sb, KERN_ERR,
 810                         "Filesystem with project quota feature cannot be "
 811                         "mounted RDWR without CONFIG_QUOTA");
 812                 return -EINVAL;
 813         }
 814 #endif
 815
 816         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 817                 f2fs_msg(sb, KERN_ERR,
 818                                 "Should set mode=lfs with %uKB-sized IO",
 819                                 F2FS_IO_SIZE_KB(sbi));
 820                 return -EINVAL;
 821         }
 822
 823         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 824                 if (!f2fs_sb_has_extra_attr(sbi) ||
 825                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 826                         f2fs_msg(sb, KERN_ERR,
 827                                         "extra_attr or flexible_inline_xattr "
 828                                         "feature is off");
 829                         return -EINVAL;
 830                 }
 831                 if (!test_opt(sbi, INLINE_XATTR)) {
 832                         f2fs_msg(sb, KERN_ERR,
 833                                         "inline_xattr_size option should be "
 834                                         "set with inline_xattr option");
 835                         return -EINVAL;
 836                 }
 837                 if (!F2FS_OPTION(sbi).inline_xattr_size ||
 838                         F2FS_OPTION(sbi).inline_xattr_size >=
 839                                         DEF_ADDRS_PER_INODE -
 840                                         F2FS_TOTAL_EXTRA_ATTR_SIZE -
 841                                         DEF_INLINE_RESERVED_SIZE -
 842                                         DEF_MIN_INLINE_SIZE) {
 843                         f2fs_msg(sb, KERN_ERR,
 844                                         "inline xattr size is out of range");
 845                         return -EINVAL;
 846                 }
 847         }
 848
 849         if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
 850                 f2fs_msg(sb, KERN_ERR,
 851                                 "LFS not compatible with checkpoint=disable\n");
 852                 return -EINVAL;
 853         }
 854
 855         /* Not pass down write hints if the number of active logs is lesser
 856          * than NR_CURSEG_TYPE.
 857          */
 858         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 859                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 860         return 0;
 861 }
 862
 863 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 864 {
 865         struct f2fs_inode_info *fi;
 866
 867         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 868         if (!fi)
 869                 return NULL;
 870
 871         init_once((void *) fi);
 872
 873         /* Initialize f2fs-specific inode info */
 874         atomic_set(&fi->dirty_pages, 0);
 875         init_rwsem(&fi->i_sem);
 876         INIT_LIST_HEAD(&fi->dirty_list);
 877         INIT_LIST_HEAD(&fi->gdirty_list);
 878         INIT_LIST_HEAD(&fi->inmem_ilist);
 879         INIT_LIST_HEAD(&fi->inmem_pages);
 880         mutex_init(&fi->inmem_lock);
 881         init_rwsem(&fi->i_gc_rwsem[READ]);
 882         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 883         init_rwsem(&fi->i_mmap_sem);
 884         init_rwsem(&fi->i_xattr_sem);
 885
 886         /* Will be used by directory only */
 887         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 888
 889         return &fi->vfs_inode;
 890 }
 891
 892 static int f2fs_drop_inode(struct inode *inode)
 893 {
 894         int ret;
 895         /*
 896          * This is to avoid a deadlock condition like below.
 897          * writeback_single_inode(inode)
 898          *  - f2fs_write_data_page
 899          *    - f2fs_gc -> iput -> evict
 900          *       - inode_wait_for_writeback(inode)
 901          */
 902         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 903                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 904                         /* to avoid evict_inode call simultaneously */
 905                         atomic_inc(&inode->i_count);
 906                         spin_unlock(&inode->i_lock);
 907
 908                         /* some remained atomic pages should discarded */
 909                         if (f2fs_is_atomic_file(inode))
 910                                 f2fs_drop_inmem_pages(inode);
 911
 912                         /* should remain fi->extent_tree for writepage */
 913                         f2fs_destroy_extent_node(inode);
 914
 915                         sb_start_intwrite(inode->i_sb);
 916                         f2fs_i_size_write(inode, 0);
 917
 918                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
 919                                         inode, NULL, 0, DATA);
 920                         truncate_inode_pages_final(inode->i_mapping);
 921
 922                         if (F2FS_HAS_BLOCKS(inode))
 923                                 f2fs_truncate(inode);
 924
 925                         sb_end_intwrite(inode->i_sb);
 926
 927                         spin_lock(&inode->i_lock);
 928                         atomic_dec(&inode->i_count);
 929                 }
 930                 trace_f2fs_drop_inode(inode, 0);
 931                 return 0;
 932         }
 933         ret = generic_drop_inode(inode);
 934         trace_f2fs_drop_inode(inode, ret);
 935         return ret;
 936 }
 937
 938 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 939 {
 940         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 941         int ret = 0;
 942
 943         spin_lock(&sbi->inode_lock[DIRTY_META]);
 944         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 945                 ret = 1;
 946         } else {
 947                 set_inode_flag(inode, FI_DIRTY_INODE);
 948                 stat_inc_dirty_inode(sbi, DIRTY_META);
 949         }
 950         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 951                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 952                                 &sbi->inode_list[DIRTY_META]);
 953                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 954         }
 955         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 956         return ret;
 957 }
 958
 959 void f2fs_inode_synced(struct inode *inode)
 960 {
 961         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 962
 963         spin_lock(&sbi->inode_lock[DIRTY_META]);
 964         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 965                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 966                 return;
 967         }
 968         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 969                 list_del_init(&F2FS_I(inode)->gdirty_list);
 970                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 971         }
 972         clear_inode_flag(inode, FI_DIRTY_INODE);
 973         clear_inode_flag(inode, FI_AUTO_RECOVER);
 974         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 975         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 976 }
 977
 978 /*
 979  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 980  *
 981  * We should call set_dirty_inode to write the dirty inode through write_inode.
 982  */
 983 static void f2fs_dirty_inode(struct inode *inode, int flags)
 984 {
 985         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 986
 987         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 988                         inode->i_ino == F2FS_META_INO(sbi))
 989                 return;
 990
 991         if (flags == I_DIRTY_TIME)
 992                 return;
 993
 994         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 995                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 996
 997         f2fs_inode_dirtied(inode, false);
 998 }
 999
1000 static void f2fs_i_callback(struct rcu_head *head)
1001 {
1002         struct inode *inode = container_of(head, struct inode, i_rcu);
1003         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1004 }
1005
1006 static void f2fs_destroy_inode(struct inode *inode)
1007 {
1008         call_rcu(&inode->i_rcu, f2fs_i_callback);
1009 }
1010
1011 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1012 {
1013         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1014         percpu_counter_destroy(&sbi->total_valid_inode_count);
1015 }
1016
1017 static void destroy_device_list(struct f2fs_sb_info *sbi)
1018 {
1019         int i;
1020
1021         for (i = 0; i < sbi->s_ndevs; i++) {
1022                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1023 #ifdef CONFIG_BLK_DEV_ZONED
1024                 kvfree(FDEV(i).blkz_type);
1025 #endif
1026         }
1027         kvfree(sbi->devs);
1028 }
1029
1030 static void f2fs_put_super(struct super_block *sb)
1031 {
1032         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1033         int i;
1034         bool dropped;
1035
1036         f2fs_quota_off_umount(sb);
1037
1038         /* prevent remaining shrinker jobs */
1039         mutex_lock(&sbi->umount_mutex);
1040
1041         /*
1042          * We don't need to do checkpoint when superblock is clean.
1043          * But, the previous checkpoint was not done by umount, it needs to do
1044          * clean checkpoint again.
1045          */
1046         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1047                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1048                 struct cp_control cpc = {
1049                         .reason = CP_UMOUNT,
1050                 };
1051                 f2fs_write_checkpoint(sbi, &cpc);
1052         }
1053
1054         /* be sure to wait for any on-going discard commands */
1055         dropped = f2fs_issue_discard_timeout(sbi);
1056
1057         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1058                                         !sbi->discard_blks && !dropped) {
1059                 struct cp_control cpc = {
1060                         .reason = CP_UMOUNT | CP_TRIMMED,
1061                 };
1062                 f2fs_write_checkpoint(sbi, &cpc);
1063         }
1064
1065         /*
1066          * normally superblock is clean, so we need to release this.
1067          * In addition, EIO will skip do checkpoint, we need this as well.
1068          */
1069         f2fs_release_ino_entry(sbi, true);
1070
1071         f2fs_leave_shrinker(sbi);
1072         mutex_unlock(&sbi->umount_mutex);
1073
1074         /* our cp_error case, we can wait for any writeback page */
1075         f2fs_flush_merged_writes(sbi);
1076
1077         f2fs_wait_on_all_pages_writeback(sbi);
1078
1079         f2fs_bug_on(sbi, sbi->fsync_node_num);
1080
1081         iput(sbi->node_inode);
1082         sbi->node_inode = NULL;
1083
1084         iput(sbi->meta_inode);
1085         sbi->meta_inode = NULL;
1086
1087         /*
1088          * iput() can update stat information, if f2fs_write_checkpoint()
1089          * above failed with error.
1090          */
1091         f2fs_destroy_stats(sbi);
1092
1093         /* destroy f2fs internal modules */
1094         f2fs_destroy_node_manager(sbi);
1095         f2fs_destroy_segment_manager(sbi);
1096
1097         kvfree(sbi->ckpt);
1098
1099         f2fs_unregister_sysfs(sbi);
1100
1101         sb->s_fs_info = NULL;
1102         if (sbi->s_chksum_driver)
1103                 crypto_free_shash(sbi->s_chksum_driver);
1104         kvfree(sbi->raw_super);
1105
1106         destroy_device_list(sbi);
1107         mempool_destroy(sbi->write_io_dummy);
1108 #ifdef CONFIG_QUOTA
1109         for (i = 0; i < MAXQUOTAS; i++)
1110                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1111 #endif
1112         destroy_percpu_info(sbi);
1113         for (i = 0; i < NR_PAGE_TYPE; i++)
1114                 kvfree(sbi->write_io[i]);
1115         kvfree(sbi);
1116 }
1117
1118 int f2fs_sync_fs(struct super_block *sb, int sync)
1119 {
1120         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1121         int err = 0;
1122
1123         if (unlikely(f2fs_cp_error(sbi)))
1124                 return 0;
1125         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1126                 return 0;
1127
1128         trace_f2fs_sync_fs(sb, sync);
1129
1130         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1131                 return -EAGAIN;
1132
1133         if (sync) {
1134                 struct cp_control cpc;
1135
1136                 cpc.reason = __get_cp_reason(sbi);
1137
1138                 mutex_lock(&sbi->gc_mutex);
1139                 err = f2fs_write_checkpoint(sbi, &cpc);
1140                 mutex_unlock(&sbi->gc_mutex);
1141         }
1142         f2fs_trace_ios(NULL, 1);
1143
1144         return err;
1145 }
1146
1147 static int f2fs_freeze(struct super_block *sb)
1148 {
1149         if (f2fs_readonly(sb))
1150                 return 0;
1151
1152         /* IO error happened before */
1153         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1154                 return -EIO;
1155
1156         /* must be clean, since sync_filesystem() was already called */
1157         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1158                 return -EINVAL;
1159         return 0;
1160 }
1161
1162 static int f2fs_unfreeze(struct super_block *sb)
1163 {
1164         return 0;
1165 }
1166
1167 #ifdef CONFIG_QUOTA
1168 static int f2fs_statfs_project(struct super_block *sb,
1169                                 kprojid_t projid, struct kstatfs *buf)
1170 {
1171         struct kqid qid;
1172         struct dquot *dquot;
1173         u64 limit;
1174         u64 curblock;
1175
1176         qid = make_kqid_projid(projid);
1177         dquot = dqget(sb, qid);
1178         if (IS_ERR(dquot))
1179                 return PTR_ERR(dquot);
1180         spin_lock(&dquot->dq_dqb_lock);
1181
1182         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1183                  dquot->dq_dqb.dqb_bsoftlimit :
1184                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1185         if (limit && buf->f_blocks > limit) {
1186                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1187                 buf->f_blocks = limit;
1188                 buf->f_bfree = buf->f_bavail =
1189                         (buf->f_blocks > curblock) ?
1190                          (buf->f_blocks - curblock) : 0;
1191         }
1192
1193         limit = dquot->dq_dqb.dqb_isoftlimit ?
1194                 dquot->dq_dqb.dqb_isoftlimit :
1195                 dquot->dq_dqb.dqb_ihardlimit;
1196         if (limit && buf->f_files > limit) {
1197                 buf->f_files = limit;
1198                 buf->f_ffree =
1199                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1200                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1201         }
1202
1203         spin_unlock(&dquot->dq_dqb_lock);
1204         dqput(dquot);
1205         return 0;
1206 }
1207 #endif
1208
1209 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1210 {
1211         struct super_block *sb = dentry->d_sb;
1212         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1213         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1214         block_t total_count, user_block_count, start_count;
1215         u64 avail_node_count;
1216
1217         total_count = le64_to_cpu(sbi->raw_super->block_count);
1218         user_block_count = sbi->user_block_count;
1219         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1220         buf->f_type = F2FS_SUPER_MAGIC;
1221         buf->f_bsize = sbi->blocksize;
1222
1223         buf->f_blocks = total_count - start_count;
1224         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1225                                                 sbi->current_reserved_blocks;
1226         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1227                 buf->f_bfree = 0;
1228         else
1229                 buf->f_bfree -= sbi->unusable_block_count;
1230
1231         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1232                 buf->f_bavail = buf->f_bfree -
1233                                 F2FS_OPTION(sbi).root_reserved_blocks;
1234         else
1235                 buf->f_bavail = 0;
1236
1237         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1238                                                 F2FS_RESERVED_NODE_NUM;
1239
1240         if (avail_node_count > user_block_count) {
1241                 buf->f_files = user_block_count;
1242                 buf->f_ffree = buf->f_bavail;
1243         } else {
1244                 buf->f_files = avail_node_count;
1245                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1246                                         buf->f_bavail);
1247         }
1248
1249         buf->f_namelen = F2FS_NAME_LEN;
1250         buf->f_fsid.val[0] = (u32)id;
1251         buf->f_fsid.val[1] = (u32)(id >> 32);
1252
1253 #ifdef CONFIG_QUOTA
1254         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1255                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1256                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1257         }
1258 #endif
1259         return 0;
1260 }
1261
1262 static inline void f2fs_show_quota_options(struct seq_file *seq,
1263                                            struct super_block *sb)
1264 {
1265 #ifdef CONFIG_QUOTA
1266         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1267
1268         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1269                 char *fmtname = "";
1270
1271                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1272                 case QFMT_VFS_OLD:
1273                         fmtname = "vfsold";
1274                         break;
1275                 case QFMT_VFS_V0:
1276                         fmtname = "vfsv0";
1277                         break;
1278                 case QFMT_VFS_V1:
1279                         fmtname = "vfsv1";
1280                         break;
1281                 }
1282                 seq_printf(seq, ",jqfmt=%s", fmtname);
1283         }
1284
1285         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1286                 seq_show_option(seq, "usrjquota",
1287                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1288
1289         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1290                 seq_show_option(seq, "grpjquota",
1291                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1292
1293         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1294                 seq_show_option(seq, "prjjquota",
1295                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1296 #endif
1297 }
1298
1299 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1300 {
1301         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1302
1303         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1304                 if (test_opt(sbi, FORCE_FG_GC))
1305                         seq_printf(seq, ",background_gc=%s", "sync");
1306                 else
1307                         seq_printf(seq, ",background_gc=%s", "on");
1308         } else {
1309                 seq_printf(seq, ",background_gc=%s", "off");
1310         }
1311         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1312                 seq_puts(seq, ",disable_roll_forward");
1313         if (test_opt(sbi, DISCARD))
1314                 seq_puts(seq, ",discard");
1315         if (test_opt(sbi, NOHEAP))
1316                 seq_puts(seq, ",no_heap");
1317         else
1318                 seq_puts(seq, ",heap");
1319 #ifdef CONFIG_F2FS_FS_XATTR
1320         if (test_opt(sbi, XATTR_USER))
1321                 seq_puts(seq, ",user_xattr");
1322         else
1323                 seq_puts(seq, ",nouser_xattr");
1324         if (test_opt(sbi, INLINE_XATTR))
1325                 seq_puts(seq, ",inline_xattr");
1326         else
1327                 seq_puts(seq, ",noinline_xattr");
1328         if (test_opt(sbi, INLINE_XATTR_SIZE))
1329                 seq_printf(seq, ",inline_xattr_size=%u",
1330                                         F2FS_OPTION(sbi).inline_xattr_size);
1331 #endif
1332 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1333         if (test_opt(sbi, POSIX_ACL))
1334                 seq_puts(seq, ",acl");
1335         else
1336                 seq_puts(seq, ",noacl");
1337 #endif
1338         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1339                 seq_puts(seq, ",disable_ext_identify");
1340         if (test_opt(sbi, INLINE_DATA))
1341                 seq_puts(seq, ",inline_data");
1342         else
1343                 seq_puts(seq, ",noinline_data");
1344         if (test_opt(sbi, INLINE_DENTRY))
1345                 seq_puts(seq, ",inline_dentry");
1346         else
1347                 seq_puts(seq, ",noinline_dentry");
1348         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1349                 seq_puts(seq, ",flush_merge");
1350         if (test_opt(sbi, NOBARRIER))
1351                 seq_puts(seq, ",nobarrier");
1352         if (test_opt(sbi, FASTBOOT))
1353                 seq_puts(seq, ",fastboot");
1354         if (test_opt(sbi, EXTENT_CACHE))
1355                 seq_puts(seq, ",extent_cache");
1356         else
1357                 seq_puts(seq, ",noextent_cache");
1358         if (test_opt(sbi, DATA_FLUSH))
1359                 seq_puts(seq, ",data_flush");
1360
1361         seq_puts(seq, ",mode=");
1362         if (test_opt(sbi, ADAPTIVE))
1363                 seq_puts(seq, "adaptive");
1364         else if (test_opt(sbi, LFS))
1365                 seq_puts(seq, "lfs");
1366         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1367         if (test_opt(sbi, RESERVE_ROOT))
1368                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1369                                 F2FS_OPTION(sbi).root_reserved_blocks,
1370                                 from_kuid_munged(&init_user_ns,
1371                                         F2FS_OPTION(sbi).s_resuid),
1372                                 from_kgid_munged(&init_user_ns,
1373                                         F2FS_OPTION(sbi).s_resgid));
1374         if (F2FS_IO_SIZE_BITS(sbi))
1375                 seq_printf(seq, ",io_bits=%u",
1376                                 F2FS_OPTION(sbi).write_io_size_bits);
1377 #ifdef CONFIG_F2FS_FAULT_INJECTION
1378         if (test_opt(sbi, FAULT_INJECTION)) {
1379                 seq_printf(seq, ",fault_injection=%u",
1380                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1381                 seq_printf(seq, ",fault_type=%u",
1382                                 F2FS_OPTION(sbi).fault_info.inject_type);
1383         }
1384 #endif
1385 #ifdef CONFIG_QUOTA
1386         if (test_opt(sbi, QUOTA))
1387                 seq_puts(seq, ",quota");
1388         if (test_opt(sbi, USRQUOTA))
1389                 seq_puts(seq, ",usrquota");
1390         if (test_opt(sbi, GRPQUOTA))
1391                 seq_puts(seq, ",grpquota");
1392         if (test_opt(sbi, PRJQUOTA))
1393                 seq_puts(seq, ",prjquota");
1394 #endif
1395         f2fs_show_quota_options(seq, sbi->sb);
1396         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1397                 seq_printf(seq, ",whint_mode=%s", "user-based");
1398         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1399                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1400 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1401         if (F2FS_OPTION(sbi).test_dummy_encryption)
1402                 seq_puts(seq, ",test_dummy_encryption");
1403 #endif
1404
1405         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1406                 seq_printf(seq, ",alloc_mode=%s", "default");
1407         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1408                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1409
1410         if (test_opt(sbi, DISABLE_CHECKPOINT))
1411                 seq_puts(seq, ",checkpoint=disable");
1412
1413         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1414                 seq_printf(seq, ",fsync_mode=%s", "posix");
1415         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1416                 seq_printf(seq, ",fsync_mode=%s", "strict");
1417         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1418                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1419         return 0;
1420 }
1421
1422 static void default_options(struct f2fs_sb_info *sbi)
1423 {
1424         /* init some FS parameters */
1425         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1426         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1427         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1428         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1429         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1430         F2FS_OPTION(sbi).test_dummy_encryption = false;
1431         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1432         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1433
1434         set_opt(sbi, BG_GC);
1435         set_opt(sbi, INLINE_XATTR);
1436         set_opt(sbi, INLINE_DATA);
1437         set_opt(sbi, INLINE_DENTRY);
1438         set_opt(sbi, EXTENT_CACHE);
1439         set_opt(sbi, NOHEAP);
1440         clear_opt(sbi, DISABLE_CHECKPOINT);
1441         sbi->sb->s_flags |= SB_LAZYTIME;
1442         set_opt(sbi, FLUSH_MERGE);
1443         set_opt(sbi, DISCARD);
1444         if (f2fs_sb_has_blkzoned(sbi))
1445                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1446         else
1447                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1448
1449 #ifdef CONFIG_F2FS_FS_XATTR
1450         set_opt(sbi, XATTR_USER);
1451 #endif
1452 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1453         set_opt(sbi, POSIX_ACL);
1454 #endif
1455
1456         f2fs_build_fault_attr(sbi, 0, 0);
1457 }
1458
1459 #ifdef CONFIG_QUOTA
1460 static int f2fs_enable_quotas(struct super_block *sb);
1461 #endif
1462
1463 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1464 {
1465         unsigned int s_flags = sbi->sb->s_flags;
1466         struct cp_control cpc;
1467         int err = 0;
1468         int ret;
1469
1470         if (s_flags & SB_RDONLY) {
1471                 f2fs_msg(sbi->sb, KERN_ERR,
1472                                 "checkpoint=disable on readonly fs");
1473                 return -EINVAL;
1474         }
1475         sbi->sb->s_flags |= SB_ACTIVE;
1476
1477         f2fs_update_time(sbi, DISABLE_TIME);
1478
1479         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1480                 mutex_lock(&sbi->gc_mutex);
1481                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1482                 if (err == -ENODATA) {
1483                         err = 0;
1484                         break;
1485                 }
1486                 if (err && err != -EAGAIN)
1487                         break;
1488         }
1489
1490         ret = sync_filesystem(sbi->sb);
1491         if (ret || err) {
1492                 err = ret ? ret: err;
1493                 goto restore_flag;
1494         }
1495
1496         if (f2fs_disable_cp_again(sbi)) {
1497                 err = -EAGAIN;
1498                 goto restore_flag;
1499         }
1500
1501         mutex_lock(&sbi->gc_mutex);
1502         cpc.reason = CP_PAUSE;
1503         set_sbi_flag(sbi, SBI_CP_DISABLED);
1504         f2fs_write_checkpoint(sbi, &cpc);
1505
1506         sbi->unusable_block_count = 0;
1507         mutex_unlock(&sbi->gc_mutex);
1508 restore_flag:
1509         sbi->sb->s_flags = s_flags;     /* Restore MS_RDONLY status */
1510         return err;
1511 }
1512
1513 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1514 {
1515         mutex_lock(&sbi->gc_mutex);
1516         f2fs_dirty_to_prefree(sbi);
1517
1518         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1519         set_sbi_flag(sbi, SBI_IS_DIRTY);
1520         mutex_unlock(&sbi->gc_mutex);
1521
1522         f2fs_sync_fs(sbi->sb, 1);
1523 }
1524
1525 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1526 {
1527         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1528         struct f2fs_mount_info org_mount_opt;
1529         unsigned long old_sb_flags;
1530         int err;
1531         bool need_restart_gc = false;
1532         bool need_stop_gc = false;
1533         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1534         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1535         bool checkpoint_changed;
1536 #ifdef CONFIG_QUOTA
1537         int i, j;
1538 #endif
1539
1540         /*
1541          * Save the old mount options in case we
1542          * need to restore them.
1543          */
1544         org_mount_opt = sbi->mount_opt;
1545         old_sb_flags = sb->s_flags;
1546
1547 #ifdef CONFIG_QUOTA
1548         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1549         for (i = 0; i < MAXQUOTAS; i++) {
1550                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1551                         org_mount_opt.s_qf_names[i] =
1552                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1553                                 GFP_KERNEL);
1554                         if (!org_mount_opt.s_qf_names[i]) {
1555                                 for (j = 0; j < i; j++)
1556                                         kvfree(org_mount_opt.s_qf_names[j]);
1557                                 return -ENOMEM;
1558                         }
1559                 } else {
1560                         org_mount_opt.s_qf_names[i] = NULL;
1561                 }
1562         }
1563 #endif
1564
1565         /* recover superblocks we couldn't write due to previous RO mount */
1566         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1567                 err = f2fs_commit_super(sbi, false);
1568                 f2fs_msg(sb, KERN_INFO,
1569                         "Try to recover all the superblocks, ret: %d", err);
1570                 if (!err)
1571                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1572         }
1573
1574         default_options(sbi);
1575
1576         /* parse mount options */
1577         err = parse_options(sb, data);
1578         if (err)
1579                 goto restore_opts;
1580         checkpoint_changed =
1581                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1582
1583         /*
1584          * Previous and new state of filesystem is RO,
1585          * so skip checking GC and FLUSH_MERGE conditions.
1586          */
1587         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1588                 goto skip;
1589
1590 #ifdef CONFIG_QUOTA
1591         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1592                 err = dquot_suspend(sb, -1);
1593                 if (err < 0)
1594                         goto restore_opts;
1595         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1596                 /* dquot_resume needs RW */
1597                 sb->s_flags &= ~SB_RDONLY;
1598                 if (sb_any_quota_suspended(sb)) {
1599                         dquot_resume(sb, -1);
1600                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1601                         err = f2fs_enable_quotas(sb);
1602                         if (err)
1603                                 goto restore_opts;
1604                 }
1605         }
1606 #endif
1607         /* disallow enable/disable extent_cache dynamically */
1608         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1609                 err = -EINVAL;
1610                 f2fs_msg(sbi->sb, KERN_WARNING,
1611                                 "switch extent_cache option is not allowed");
1612                 goto restore_opts;
1613         }
1614
1615         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1616                 err = -EINVAL;
1617                 f2fs_msg(sbi->sb, KERN_WARNING,
1618                         "disabling checkpoint not compatible with read-only");
1619                 goto restore_opts;
1620         }
1621
1622         /*
1623          * We stop the GC thread if FS is mounted as RO
1624          * or if background_gc = off is passed in mount
1625          * option. Also sync the filesystem.
1626          */
1627         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1628                 if (sbi->gc_thread) {
1629                         f2fs_stop_gc_thread(sbi);
1630                         need_restart_gc = true;
1631                 }
1632         } else if (!sbi->gc_thread) {
1633                 err = f2fs_start_gc_thread(sbi);
1634                 if (err)
1635                         goto restore_opts;
1636                 need_stop_gc = true;
1637         }
1638
1639         if (*flags & SB_RDONLY ||
1640                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1641                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1642                 sync_inodes_sb(sb);
1643
1644                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1645                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1646                 f2fs_sync_fs(sb, 1);
1647                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1648         }
1649
1650         if (checkpoint_changed) {
1651                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1652                         err = f2fs_disable_checkpoint(sbi);
1653                         if (err)
1654                                 goto restore_gc;
1655                 } else {
1656                         f2fs_enable_checkpoint(sbi);
1657                 }
1658         }
1659
1660         /*
1661          * We stop issue flush thread if FS is mounted as RO
1662          * or if flush_merge is not passed in mount option.
1663          */
1664         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1665                 clear_opt(sbi, FLUSH_MERGE);
1666                 f2fs_destroy_flush_cmd_control(sbi, false);
1667         } else {
1668                 err = f2fs_create_flush_cmd_control(sbi);
1669                 if (err)
1670                         goto restore_gc;
1671         }
1672 skip:
1673 #ifdef CONFIG_QUOTA
1674         /* Release old quota file names */
1675         for (i = 0; i < MAXQUOTAS; i++)
1676                 kvfree(org_mount_opt.s_qf_names[i]);
1677 #endif
1678         /* Update the POSIXACL Flag */
1679         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1680                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1681
1682         limit_reserve_root(sbi);
1683         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1684         return 0;
1685 restore_gc:
1686         if (need_restart_gc) {
1687                 if (f2fs_start_gc_thread(sbi))
1688                         f2fs_msg(sbi->sb, KERN_WARNING,
1689                                 "background gc thread has stopped");
1690         } else if (need_stop_gc) {
1691                 f2fs_stop_gc_thread(sbi);
1692         }
1693 restore_opts:
1694 #ifdef CONFIG_QUOTA
1695         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1696         for (i = 0; i < MAXQUOTAS; i++) {
1697                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1698                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1699         }
1700 #endif
1701         sbi->mount_opt = org_mount_opt;
1702         sb->s_flags = old_sb_flags;
1703         return err;
1704 }
1705
1706 #ifdef CONFIG_QUOTA
1707 /* Read data from quotafile */
1708 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1709                                size_t len, loff_t off)
1710 {
1711         struct inode *inode = sb_dqopt(sb)->files[type];
1712         struct address_space *mapping = inode->i_mapping;
1713         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1714         int offset = off & (sb->s_blocksize - 1);
1715         int tocopy;
1716         size_t toread;
1717         loff_t i_size = i_size_read(inode);
1718         struct page *page;
1719         char *kaddr;
1720
1721         if (off > i_size)
1722                 return 0;
1723
1724         if (off + len > i_size)
1725                 len = i_size - off;
1726         toread = len;
1727         while (toread > 0) {
1728                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1729 repeat:
1730                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1731                 if (IS_ERR(page)) {
1732                         if (PTR_ERR(page) == -ENOMEM) {
1733                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1734                                 goto repeat;
1735                         }
1736                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1737                         return PTR_ERR(page);
1738                 }
1739
1740                 lock_page(page);
1741
1742                 if (unlikely(page->mapping != mapping)) {
1743                         f2fs_put_page(page, 1);
1744                         goto repeat;
1745                 }
1746                 if (unlikely(!PageUptodate(page))) {
1747                         f2fs_put_page(page, 1);
1748                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1749                         return -EIO;
1750                 }
1751
1752                 kaddr = kmap_atomic(page);
1753                 memcpy(data, kaddr + offset, tocopy);
1754                 kunmap_atomic(kaddr);
1755                 f2fs_put_page(page, 1);
1756
1757                 offset = 0;
1758                 toread -= tocopy;
1759                 data += tocopy;
1760                 blkidx++;
1761         }
1762         return len;
1763 }
1764
1765 /* Write to quotafile */
1766 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1767                                 const char *data, 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         const struct address_space_operations *a_ops = mapping->a_ops;
1772         int offset = off & (sb->s_blocksize - 1);
1773         size_t towrite = len;
1774         struct page *page;
1775         char *kaddr;
1776         int err = 0;
1777         int tocopy;
1778
1779         while (towrite > 0) {
1780                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1781                                                                 towrite);
1782 retry:
1783                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1784                                                         &page, NULL);
1785                 if (unlikely(err)) {
1786                         if (err == -ENOMEM) {
1787                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1788                                 goto retry;
1789                         }
1790                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1791                         break;
1792                 }
1793
1794                 kaddr = kmap_atomic(page);
1795                 memcpy(kaddr + offset, data, tocopy);
1796                 kunmap_atomic(kaddr);
1797                 flush_dcache_page(page);
1798
1799                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1800                                                 page, NULL);
1801                 offset = 0;
1802                 towrite -= tocopy;
1803                 off += tocopy;
1804                 data += tocopy;
1805                 cond_resched();
1806         }
1807
1808         if (len == towrite)
1809                 return err;
1810         inode->i_mtime = inode->i_ctime = current_time(inode);
1811         f2fs_mark_inode_dirty_sync(inode, false);
1812         return len - towrite;
1813 }
1814
1815 static struct dquot **f2fs_get_dquots(struct inode *inode)
1816 {
1817         return F2FS_I(inode)->i_dquot;
1818 }
1819
1820 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1821 {
1822         return &F2FS_I(inode)->i_reserved_quota;
1823 }
1824
1825 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1826 {
1827         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1828                 f2fs_msg(sbi->sb, KERN_ERR,
1829                         "quota sysfile may be corrupted, skip loading it");
1830                 return 0;
1831         }
1832
1833         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1834                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1835 }
1836
1837 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1838 {
1839         int enabled = 0;
1840         int i, err;
1841
1842         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1843                 err = f2fs_enable_quotas(sbi->sb);
1844                 if (err) {
1845                         f2fs_msg(sbi->sb, KERN_ERR,
1846                                         "Cannot turn on quota_ino: %d", err);
1847                         return 0;
1848                 }
1849                 return 1;
1850         }
1851
1852         for (i = 0; i < MAXQUOTAS; i++) {
1853                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1854                         err = f2fs_quota_on_mount(sbi, i);
1855                         if (!err) {
1856                                 enabled = 1;
1857                                 continue;
1858                         }
1859                         f2fs_msg(sbi->sb, KERN_ERR,
1860                                 "Cannot turn on quotas: %d on %d", err, i);
1861                 }
1862         }
1863         return enabled;
1864 }
1865
1866 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1867                              unsigned int flags)
1868 {
1869         struct inode *qf_inode;
1870         unsigned long qf_inum;
1871         int err;
1872
1873         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1874
1875         qf_inum = f2fs_qf_ino(sb, type);
1876         if (!qf_inum)
1877                 return -EPERM;
1878
1879         qf_inode = f2fs_iget(sb, qf_inum);
1880         if (IS_ERR(qf_inode)) {
1881                 f2fs_msg(sb, KERN_ERR,
1882                         "Bad quota inode %u:%lu", type, qf_inum);
1883                 return PTR_ERR(qf_inode);
1884         }
1885
1886         /* Don't account quota for quota files to avoid recursion */
1887         qf_inode->i_flags |= S_NOQUOTA;
1888         err = dquot_enable(qf_inode, type, format_id, flags);
1889         iput(qf_inode);
1890         return err;
1891 }
1892
1893 static int f2fs_enable_quotas(struct super_block *sb)
1894 {
1895         int type, err = 0;
1896         unsigned long qf_inum;
1897         bool quota_mopt[MAXQUOTAS] = {
1898                 test_opt(F2FS_SB(sb), USRQUOTA),
1899                 test_opt(F2FS_SB(sb), GRPQUOTA),
1900                 test_opt(F2FS_SB(sb), PRJQUOTA),
1901         };
1902
1903         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1904                 f2fs_msg(sb, KERN_ERR,
1905                         "quota file may be corrupted, skip loading it");
1906                 return 0;
1907         }
1908
1909         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1910
1911         for (type = 0; type < MAXQUOTAS; type++) {
1912                 qf_inum = f2fs_qf_ino(sb, type);
1913                 if (qf_inum) {
1914                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1915                                 DQUOT_USAGE_ENABLED |
1916                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1917                         if (err) {
1918                                 f2fs_msg(sb, KERN_ERR,
1919                                         "Failed to enable quota tracking "
1920                                         "(type=%d, err=%d). Please run "
1921                                         "fsck to fix.", type, err);
1922                                 for (type--; type >= 0; type--)
1923                                         dquot_quota_off(sb, type);
1924                                 set_sbi_flag(F2FS_SB(sb),
1925                                                 SBI_QUOTA_NEED_REPAIR);
1926                                 return err;
1927                         }
1928                 }
1929         }
1930         return 0;
1931 }
1932
1933 int f2fs_quota_sync(struct super_block *sb, int type)
1934 {
1935         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1936         struct quota_info *dqopt = sb_dqopt(sb);
1937         int cnt;
1938         int ret;
1939
1940         ret = dquot_writeback_dquots(sb, type);
1941         if (ret)
1942                 goto out;
1943
1944         /*
1945          * Now when everything is written we can discard the pagecache so
1946          * that userspace sees the changes.
1947          */
1948         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1949                 struct address_space *mapping;
1950
1951                 if (type != -1 && cnt != type)
1952                         continue;
1953                 if (!sb_has_quota_active(sb, cnt))
1954                         continue;
1955
1956                 mapping = dqopt->files[cnt]->i_mapping;
1957
1958                 ret = filemap_fdatawrite(mapping);
1959                 if (ret)
1960                         goto out;
1961
1962                 /* if we are using journalled quota */
1963                 if (is_journalled_quota(sbi))
1964                         continue;
1965
1966                 ret = filemap_fdatawait(mapping);
1967                 if (ret)
1968                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1969
1970                 inode_lock(dqopt->files[cnt]);
1971                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1972                 inode_unlock(dqopt->files[cnt]);
1973         }
1974 out:
1975         if (ret)
1976                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1977         return ret;
1978 }
1979
1980 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1981                                                         const struct path *path)
1982 {
1983         struct inode *inode;
1984         int err;
1985
1986         err = f2fs_quota_sync(sb, type);
1987         if (err)
1988                 return err;
1989
1990         err = dquot_quota_on(sb, type, format_id, path);
1991         if (err)
1992                 return err;
1993
1994         inode = d_inode(path->dentry);
1995
1996         inode_lock(inode);
1997         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1998         f2fs_set_inode_flags(inode);
1999         inode_unlock(inode);
2000         f2fs_mark_inode_dirty_sync(inode, false);
2001
2002         return 0;
2003 }
2004
2005 static int f2fs_quota_off(struct super_block *sb, int type)
2006 {
2007         struct inode *inode = sb_dqopt(sb)->files[type];
2008         int err;
2009
2010         if (!inode || !igrab(inode))
2011                 return dquot_quota_off(sb, type);
2012
2013         err = f2fs_quota_sync(sb, type);
2014         if (err)
2015                 goto out_put;
2016
2017         err = dquot_quota_off(sb, type);
2018         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2019                 goto out_put;
2020
2021         inode_lock(inode);
2022         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2023         f2fs_set_inode_flags(inode);
2024         inode_unlock(inode);
2025         f2fs_mark_inode_dirty_sync(inode, false);
2026 out_put:
2027         iput(inode);
2028         return err;
2029 }
2030
2031 void f2fs_quota_off_umount(struct super_block *sb)
2032 {
2033         int type;
2034         int err;
2035
2036         for (type = 0; type < MAXQUOTAS; type++) {
2037                 err = f2fs_quota_off(sb, type);
2038                 if (err) {
2039                         int ret = dquot_quota_off(sb, type);
2040
2041                         f2fs_msg(sb, KERN_ERR,
2042                                 "Fail to turn off disk quota "
2043                                 "(type: %d, err: %d, ret:%d), Please "
2044                                 "run fsck to fix it.", type, err, ret);
2045                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2046                 }
2047         }
2048         /*
2049          * In case of checkpoint=disable, we must flush quota blocks.
2050          * This can cause NULL exception for node_inode in end_io, since
2051          * put_super already dropped it.
2052          */
2053         sync_filesystem(sb);
2054 }
2055
2056 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2057 {
2058         struct quota_info *dqopt = sb_dqopt(sb);
2059         int type;
2060
2061         for (type = 0; type < MAXQUOTAS; type++) {
2062                 if (!dqopt->files[type])
2063                         continue;
2064                 f2fs_inode_synced(dqopt->files[type]);
2065         }
2066 }
2067
2068 static int f2fs_dquot_commit(struct dquot *dquot)
2069 {
2070         int ret;
2071
2072         ret = dquot_commit(dquot);
2073         if (ret < 0)
2074                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2075         return ret;
2076 }
2077
2078 static int f2fs_dquot_acquire(struct dquot *dquot)
2079 {
2080         int ret;
2081
2082         ret = dquot_acquire(dquot);
2083         if (ret < 0)
2084                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2085
2086         return ret;
2087 }
2088
2089 static int f2fs_dquot_release(struct dquot *dquot)
2090 {
2091         int ret;
2092
2093         ret = dquot_release(dquot);
2094         if (ret < 0)
2095                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2096         return ret;
2097 }
2098
2099 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2100 {
2101         struct super_block *sb = dquot->dq_sb;
2102         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2103         int ret;
2104
2105         ret = dquot_mark_dquot_dirty(dquot);
2106
2107         /* if we are using journalled quota */
2108         if (is_journalled_quota(sbi))
2109                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2110
2111         return ret;
2112 }
2113
2114 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2115 {
2116         int ret;
2117
2118         ret = dquot_commit_info(sb, type);
2119         if (ret < 0)
2120                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2121         return ret;
2122 }
2123
2124 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2125 {
2126         *projid = F2FS_I(inode)->i_projid;
2127         return 0;
2128 }
2129
2130 static const struct dquot_operations f2fs_quota_operations = {
2131         .get_reserved_space = f2fs_get_reserved_space,
2132         .write_dquot    = f2fs_dquot_commit,
2133         .acquire_dquot  = f2fs_dquot_acquire,
2134         .release_dquot  = f2fs_dquot_release,
2135         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2136         .write_info     = f2fs_dquot_commit_info,
2137         .alloc_dquot    = dquot_alloc,
2138         .destroy_dquot  = dquot_destroy,
2139         .get_projid     = f2fs_get_projid,
2140         .get_next_id    = dquot_get_next_id,
2141 };
2142
2143 static const struct quotactl_ops f2fs_quotactl_ops = {
2144         .quota_on       = f2fs_quota_on,
2145         .quota_off      = f2fs_quota_off,
2146         .quota_sync     = f2fs_quota_sync,
2147         .get_state      = dquot_get_state,
2148         .set_info       = dquot_set_dqinfo,
2149         .get_dqblk      = dquot_get_dqblk,
2150         .set_dqblk      = dquot_set_dqblk,
2151         .get_nextdqblk  = dquot_get_next_dqblk,
2152 };
2153 #else
2154 int f2fs_quota_sync(struct super_block *sb, int type)
2155 {
2156         return 0;
2157 }
2158
2159 void f2fs_quota_off_umount(struct super_block *sb)
2160 {
2161 }
2162 #endif
2163
2164 static const struct super_operations f2fs_sops = {
2165         .alloc_inode    = f2fs_alloc_inode,
2166         .drop_inode     = f2fs_drop_inode,
2167         .destroy_inode  = f2fs_destroy_inode,
2168         .write_inode    = f2fs_write_inode,
2169         .dirty_inode    = f2fs_dirty_inode,
2170         .show_options   = f2fs_show_options,
2171 #ifdef CONFIG_QUOTA
2172         .quota_read     = f2fs_quota_read,
2173         .quota_write    = f2fs_quota_write,
2174         .get_dquots     = f2fs_get_dquots,
2175 #endif
2176         .evict_inode    = f2fs_evict_inode,
2177         .put_super      = f2fs_put_super,
2178         .sync_fs        = f2fs_sync_fs,
2179         .freeze_fs      = f2fs_freeze,
2180         .unfreeze_fs    = f2fs_unfreeze,
2181         .statfs         = f2fs_statfs,
2182         .remount_fs     = f2fs_remount,
2183 };
2184
2185 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2186 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2187 {
2188         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2189                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2190                                 ctx, len, NULL);
2191 }
2192
2193 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2194                                                         void *fs_data)
2195 {
2196         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2197
2198         /*
2199          * Encrypting the root directory is not allowed because fsck
2200          * expects lost+found directory to exist and remain unencrypted
2201          * if LOST_FOUND feature is enabled.
2202          *
2203          */
2204         if (f2fs_sb_has_lost_found(sbi) &&
2205                         inode->i_ino == F2FS_ROOT_INO(sbi))
2206                 return -EPERM;
2207
2208         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2209                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2210                                 ctx, len, fs_data, XATTR_CREATE);
2211 }
2212
2213 static bool f2fs_dummy_context(struct inode *inode)
2214 {
2215         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2216 }
2217
2218 static const struct fscrypt_operations f2fs_cryptops = {
2219         .key_prefix     = "f2fs:",
2220         .get_context    = f2fs_get_context,
2221         .set_context    = f2fs_set_context,
2222         .dummy_context  = f2fs_dummy_context,
2223         .empty_dir      = f2fs_empty_dir,
2224         .max_namelen    = F2FS_NAME_LEN,
2225 };
2226 #endif
2227
2228 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2229                 u64 ino, u32 generation)
2230 {
2231         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2232         struct inode *inode;
2233
2234         if (f2fs_check_nid_range(sbi, ino))
2235                 return ERR_PTR(-ESTALE);
2236
2237         /*
2238          * f2fs_iget isn't quite right if the inode is currently unallocated!
2239          * However f2fs_iget currently does appropriate checks to handle stale
2240          * inodes so everything is OK.
2241          */
2242         inode = f2fs_iget(sb, ino);
2243         if (IS_ERR(inode))
2244                 return ERR_CAST(inode);
2245         if (unlikely(generation && inode->i_generation != generation)) {
2246                 /* we didn't find the right inode.. */
2247                 iput(inode);
2248                 return ERR_PTR(-ESTALE);
2249         }
2250         return inode;
2251 }
2252
2253 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2254                 int fh_len, int fh_type)
2255 {
2256         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2257                                     f2fs_nfs_get_inode);
2258 }
2259
2260 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2261                 int fh_len, int fh_type)
2262 {
2263         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2264                                     f2fs_nfs_get_inode);
2265 }
2266
2267 static const struct export_operations f2fs_export_ops = {
2268         .fh_to_dentry = f2fs_fh_to_dentry,
2269         .fh_to_parent = f2fs_fh_to_parent,
2270         .get_parent = f2fs_get_parent,
2271 };
2272
2273 static loff_t max_file_blocks(void)
2274 {
2275         loff_t result = 0;
2276         loff_t leaf_count = ADDRS_PER_BLOCK;
2277
2278         /*
2279          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2280          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2281          * space in inode.i_addr, it will be more safe to reassign
2282          * result as zero.
2283          */
2284
2285         /* two direct node blocks */
2286         result += (leaf_count * 2);
2287
2288         /* two indirect node blocks */
2289         leaf_count *= NIDS_PER_BLOCK;
2290         result += (leaf_count * 2);
2291
2292         /* one double indirect node block */
2293         leaf_count *= NIDS_PER_BLOCK;
2294         result += leaf_count;
2295
2296         return result;
2297 }
2298
2299 static int __f2fs_commit_super(struct buffer_head *bh,
2300                         struct f2fs_super_block *super)
2301 {
2302         lock_buffer(bh);
2303         if (super)
2304                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2305         set_buffer_dirty(bh);
2306         unlock_buffer(bh);
2307
2308         /* it's rare case, we can do fua all the time */
2309         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2310 }
2311
2312 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2313                                         struct buffer_head *bh)
2314 {
2315         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2316                                         (bh->b_data + F2FS_SUPER_OFFSET);
2317         struct super_block *sb = sbi->sb;
2318         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2319         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2320         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2321         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2322         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2323         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2324         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2325         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2326         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2327         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2328         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2329         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2330         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2331         u64 main_end_blkaddr = main_blkaddr +
2332                                 (segment_count_main << log_blocks_per_seg);
2333         u64 seg_end_blkaddr = segment0_blkaddr +
2334                                 (segment_count << log_blocks_per_seg);
2335
2336         if (segment0_blkaddr != cp_blkaddr) {
2337                 f2fs_msg(sb, KERN_INFO,
2338                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2339                         segment0_blkaddr, cp_blkaddr);
2340                 return true;
2341         }
2342
2343         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2344                                                         sit_blkaddr) {
2345                 f2fs_msg(sb, KERN_INFO,
2346                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2347                         cp_blkaddr, sit_blkaddr,
2348                         segment_count_ckpt << log_blocks_per_seg);
2349                 return true;
2350         }
2351
2352         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2353                                                         nat_blkaddr) {
2354                 f2fs_msg(sb, KERN_INFO,
2355                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2356                         sit_blkaddr, nat_blkaddr,
2357                         segment_count_sit << log_blocks_per_seg);
2358                 return true;
2359         }
2360
2361         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2362                                                         ssa_blkaddr) {
2363                 f2fs_msg(sb, KERN_INFO,
2364                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2365                         nat_blkaddr, ssa_blkaddr,
2366                         segment_count_nat << log_blocks_per_seg);
2367                 return true;
2368         }
2369
2370         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2371                                                         main_blkaddr) {
2372                 f2fs_msg(sb, KERN_INFO,
2373                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2374                         ssa_blkaddr, main_blkaddr,
2375                         segment_count_ssa << log_blocks_per_seg);
2376                 return true;
2377         }
2378
2379         if (main_end_blkaddr > seg_end_blkaddr) {
2380                 f2fs_msg(sb, KERN_INFO,
2381                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2382                         main_blkaddr,
2383                         segment0_blkaddr +
2384                                 (segment_count << log_blocks_per_seg),
2385                         segment_count_main << log_blocks_per_seg);
2386                 return true;
2387         } else if (main_end_blkaddr < seg_end_blkaddr) {
2388                 int err = 0;
2389                 char *res;
2390
2391                 /* fix in-memory information all the time */
2392                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2393                                 segment0_blkaddr) >> log_blocks_per_seg);
2394
2395                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2396                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2397                         res = "internally";
2398                 } else {
2399                         err = __f2fs_commit_super(bh, NULL);
2400                         res = err ? "failed" : "done";
2401                 }
2402                 f2fs_msg(sb, KERN_INFO,
2403                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
2404                         res, main_blkaddr,
2405                         segment0_blkaddr +
2406                                 (segment_count << log_blocks_per_seg),
2407                         segment_count_main << log_blocks_per_seg);
2408                 if (err)
2409                         return true;
2410         }
2411         return false;
2412 }
2413
2414 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2415                                 struct buffer_head *bh)
2416 {
2417         block_t segment_count, segs_per_sec, secs_per_zone;
2418         block_t total_sections, blocks_per_seg;
2419         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2420                                         (bh->b_data + F2FS_SUPER_OFFSET);
2421         struct super_block *sb = sbi->sb;
2422         unsigned int blocksize;
2423         size_t crc_offset = 0;
2424         __u32 crc = 0;
2425
2426         /* Check checksum_offset and crc in superblock */
2427         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2428                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2429                 if (crc_offset !=
2430                         offsetof(struct f2fs_super_block, crc)) {
2431                         f2fs_msg(sb, KERN_INFO,
2432                                 "Invalid SB checksum offset: %zu",
2433                                 crc_offset);
2434                         return 1;
2435                 }
2436                 crc = le32_to_cpu(raw_super->crc);
2437                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2438                         f2fs_msg(sb, KERN_INFO,
2439                                 "Invalid SB checksum value: %u", crc);
2440                         return 1;
2441                 }
2442         }
2443
2444         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2445                 f2fs_msg(sb, KERN_INFO,
2446                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
2447                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2448                 return 1;
2449         }
2450
2451         /* Currently, support only 4KB page cache size */
2452         if (F2FS_BLKSIZE != PAGE_SIZE) {
2453                 f2fs_msg(sb, KERN_INFO,
2454                         "Invalid page_cache_size (%lu), supports only 4KB\n",
2455                         PAGE_SIZE);
2456                 return 1;
2457         }
2458
2459         /* Currently, support only 4KB block size */
2460         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2461         if (blocksize != F2FS_BLKSIZE) {
2462                 f2fs_msg(sb, KERN_INFO,
2463                         "Invalid blocksize (%u), supports only 4KB\n",
2464                         blocksize);
2465                 return 1;
2466         }
2467
2468         /* check log blocks per segment */
2469         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2470                 f2fs_msg(sb, KERN_INFO,
2471                         "Invalid log blocks per segment (%u)\n",
2472                         le32_to_cpu(raw_super->log_blocks_per_seg));
2473                 return 1;
2474         }
2475
2476         /* Currently, support 512/1024/2048/4096 bytes sector size */
2477         if (le32_to_cpu(raw_super->log_sectorsize) >
2478                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2479                 le32_to_cpu(raw_super->log_sectorsize) <
2480                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2481                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2482                         le32_to_cpu(raw_super->log_sectorsize));
2483                 return 1;
2484         }
2485         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2486                 le32_to_cpu(raw_super->log_sectorsize) !=
2487                         F2FS_MAX_LOG_SECTOR_SIZE) {
2488                 f2fs_msg(sb, KERN_INFO,
2489                         "Invalid log sectors per block(%u) log sectorsize(%u)",
2490                         le32_to_cpu(raw_super->log_sectors_per_block),
2491                         le32_to_cpu(raw_super->log_sectorsize));
2492                 return 1;
2493         }
2494
2495         segment_count = le32_to_cpu(raw_super->segment_count);
2496         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2497         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2498         total_sections = le32_to_cpu(raw_super->section_count);
2499
2500         /* blocks_per_seg should be 512, given the above check */
2501         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2502
2503         if (segment_count > F2FS_MAX_SEGMENT ||
2504                                 segment_count < F2FS_MIN_SEGMENTS) {
2505                 f2fs_msg(sb, KERN_INFO,
2506                         "Invalid segment count (%u)",
2507                         segment_count);
2508                 return 1;
2509         }
2510
2511         if (total_sections > segment_count ||
2512                         total_sections < F2FS_MIN_SEGMENTS ||
2513                         segs_per_sec > segment_count || !segs_per_sec) {
2514                 f2fs_msg(sb, KERN_INFO,
2515                         "Invalid segment/section count (%u, %u x %u)",
2516                         segment_count, total_sections, segs_per_sec);
2517                 return 1;
2518         }
2519
2520         if ((segment_count / segs_per_sec) < total_sections) {
2521                 f2fs_msg(sb, KERN_INFO,
2522                         "Small segment_count (%u < %u * %u)",
2523                         segment_count, segs_per_sec, total_sections);
2524                 return 1;
2525         }
2526
2527         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2528                 f2fs_msg(sb, KERN_INFO,
2529                         "Wrong segment_count / block_count (%u > %llu)",
2530                         segment_count, le64_to_cpu(raw_super->block_count));
2531                 return 1;
2532         }
2533
2534         if (secs_per_zone > total_sections || !secs_per_zone) {
2535                 f2fs_msg(sb, KERN_INFO,
2536                         "Wrong secs_per_zone / total_sections (%u, %u)",
2537                         secs_per_zone, total_sections);
2538                 return 1;
2539         }
2540         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2541                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2542                         (le32_to_cpu(raw_super->extension_count) +
2543                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2544                 f2fs_msg(sb, KERN_INFO,
2545                         "Corrupted extension count (%u + %u > %u)",
2546                         le32_to_cpu(raw_super->extension_count),
2547                         raw_super->hot_ext_count,
2548                         F2FS_MAX_EXTENSION);
2549                 return 1;
2550         }
2551
2552         if (le32_to_cpu(raw_super->cp_payload) >
2553                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2554                 f2fs_msg(sb, KERN_INFO,
2555                         "Insane cp_payload (%u > %u)",
2556                         le32_to_cpu(raw_super->cp_payload),
2557                         blocks_per_seg - F2FS_CP_PACKS);
2558                 return 1;
2559         }
2560
2561         /* check reserved ino info */
2562         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2563                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2564                 le32_to_cpu(raw_super->root_ino) != 3) {
2565                 f2fs_msg(sb, KERN_INFO,
2566                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2567                         le32_to_cpu(raw_super->node_ino),
2568                         le32_to_cpu(raw_super->meta_ino),
2569                         le32_to_cpu(raw_super->root_ino));
2570                 return 1;
2571         }
2572
2573         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2574         if (sanity_check_area_boundary(sbi, bh))
2575                 return 1;
2576
2577         return 0;
2578 }
2579
2580 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2581 {
2582         unsigned int total, fsmeta;
2583         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2584         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2585         unsigned int ovp_segments, reserved_segments;
2586         unsigned int main_segs, blocks_per_seg;
2587         unsigned int sit_segs, nat_segs;
2588         unsigned int sit_bitmap_size, nat_bitmap_size;
2589         unsigned int log_blocks_per_seg;
2590         unsigned int segment_count_main;
2591         unsigned int cp_pack_start_sum, cp_payload;
2592         block_t user_block_count;
2593         int i, j;
2594
2595         total = le32_to_cpu(raw_super->segment_count);
2596         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2597         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2598         fsmeta += sit_segs;
2599         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2600         fsmeta += nat_segs;
2601         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2602         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2603
2604         if (unlikely(fsmeta >= total))
2605                 return 1;
2606
2607         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2608         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2609
2610         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2611                         ovp_segments == 0 || reserved_segments == 0)) {
2612                 f2fs_msg(sbi->sb, KERN_ERR,
2613                         "Wrong layout: check mkfs.f2fs version");
2614                 return 1;
2615         }
2616
2617         user_block_count = le64_to_cpu(ckpt->user_block_count);
2618         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2619         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2620         if (!user_block_count || user_block_count >=
2621                         segment_count_main << log_blocks_per_seg) {
2622                 f2fs_msg(sbi->sb, KERN_ERR,
2623                         "Wrong user_block_count: %u", user_block_count);
2624                 return 1;
2625         }
2626
2627         main_segs = le32_to_cpu(raw_super->segment_count_main);
2628         blocks_per_seg = sbi->blocks_per_seg;
2629
2630         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2631                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2632                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2633                         return 1;
2634                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2635                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2636                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2637                                 f2fs_msg(sbi->sb, KERN_ERR,
2638                                         "Node segment (%u, %u) has the same "
2639                                         "segno: %u", i, j,
2640                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2641                                 return 1;
2642                         }
2643                 }
2644         }
2645         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2646                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2647                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2648                         return 1;
2649                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2650                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2651                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2652                                 f2fs_msg(sbi->sb, KERN_ERR,
2653                                         "Data segment (%u, %u) has the same "
2654                                         "segno: %u", i, j,
2655                                         le32_to_cpu(ckpt->cur_data_segno[i]));
2656                                 return 1;
2657                         }
2658                 }
2659         }
2660         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2661                 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2662                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2663                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2664                                 f2fs_msg(sbi->sb, KERN_ERR,
2665                                         "Data segment (%u) and Data segment (%u)"
2666                                         " has the same segno: %u", i, j,
2667                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2668                                 return 1;
2669                         }
2670                 }
2671         }
2672
2673         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2674         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2675
2676         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2677                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2678                 f2fs_msg(sbi->sb, KERN_ERR,
2679                         "Wrong bitmap size: sit: %u, nat:%u",
2680                         sit_bitmap_size, nat_bitmap_size);
2681                 return 1;
2682         }
2683
2684         cp_pack_start_sum = __start_sum_addr(sbi);
2685         cp_payload = __cp_payload(sbi);
2686         if (cp_pack_start_sum < cp_payload + 1 ||
2687                 cp_pack_start_sum > blocks_per_seg - 1 -
2688                         NR_CURSEG_TYPE) {
2689                 f2fs_msg(sbi->sb, KERN_ERR,
2690                         "Wrong cp_pack_start_sum: %u",
2691                         cp_pack_start_sum);
2692                 return 1;
2693         }
2694
2695         if (unlikely(f2fs_cp_error(sbi))) {
2696                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2697                 return 1;
2698         }
2699         return 0;
2700 }
2701
2702 static void init_sb_info(struct f2fs_sb_info *sbi)
2703 {
2704         struct f2fs_super_block *raw_super = sbi->raw_super;
2705         int i;
2706
2707         sbi->log_sectors_per_block =
2708                 le32_to_cpu(raw_super->log_sectors_per_block);
2709         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2710         sbi->blocksize = 1 << sbi->log_blocksize;
2711         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2712         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2713         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2714         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2715         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2716         sbi->total_node_count =
2717                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2718                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2719         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2720         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2721         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2722         sbi->cur_victim_sec = NULL_SECNO;
2723         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2724         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2725         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2726         sbi->migration_granularity = sbi->segs_per_sec;
2727
2728         sbi->dir_level = DEF_DIR_LEVEL;
2729         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2730         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2731         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2732         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2733         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2734         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2735                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2736         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2737
2738         for (i = 0; i < NR_COUNT_TYPE; i++)
2739                 atomic_set(&sbi->nr_pages[i], 0);
2740
2741         for (i = 0; i < META; i++)
2742                 atomic_set(&sbi->wb_sync_req[i], 0);
2743
2744         INIT_LIST_HEAD(&sbi->s_list);
2745         mutex_init(&sbi->umount_mutex);
2746         init_rwsem(&sbi->io_order_lock);
2747         spin_lock_init(&sbi->cp_lock);
2748
2749         sbi->dirty_device = 0;
2750         spin_lock_init(&sbi->dev_lock);
2751
2752         init_rwsem(&sbi->sb_lock);
2753 }
2754
2755 static int init_percpu_info(struct f2fs_sb_info *sbi)
2756 {
2757         int err;
2758
2759         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2760         if (err)
2761                 return err;
2762
2763         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2764                                                                 GFP_KERNEL);
2765         if (err)
2766                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2767
2768         return err;
2769 }
2770
2771 #ifdef CONFIG_BLK_DEV_ZONED
2772 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2773 {
2774         struct block_device *bdev = FDEV(devi).bdev;
2775         sector_t nr_sectors = bdev->bd_part->nr_sects;
2776         sector_t sector = 0;
2777         struct blk_zone *zones;
2778         unsigned int i, nr_zones;
2779         unsigned int n = 0;
2780         int err = -EIO;
2781
2782         if (!f2fs_sb_has_blkzoned(sbi))
2783                 return 0;
2784
2785         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2786                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2787                 return -EINVAL;
2788         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2789         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2790                                 __ilog2_u32(sbi->blocks_per_blkz))
2791                 return -EINVAL;
2792         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2793         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2794                                         sbi->log_blocks_per_blkz;
2795         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2796                 FDEV(devi).nr_blkz++;
2797
2798         FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2799                                                                 GFP_KERNEL);
2800         if (!FDEV(devi).blkz_type)
2801                 return -ENOMEM;
2802
2803 #define F2FS_REPORT_NR_ZONES   4096
2804
2805         zones = f2fs_kzalloc(sbi,
2806                              array_size(F2FS_REPORT_NR_ZONES,
2807                                         sizeof(struct blk_zone)),
2808                              GFP_KERNEL);
2809         if (!zones)
2810                 return -ENOMEM;
2811
2812         /* Get block zones type */
2813         while (zones && sector < nr_sectors) {
2814
2815                 nr_zones = F2FS_REPORT_NR_ZONES;
2816                 err = blkdev_report_zones(bdev, sector,
2817                                           zones, &nr_zones,
2818                                           GFP_KERNEL);
2819                 if (err)
2820                         break;
2821                 if (!nr_zones) {
2822                         err = -EIO;
2823                         break;
2824                 }
2825
2826                 for (i = 0; i < nr_zones; i++) {
2827                         FDEV(devi).blkz_type[n] = zones[i].type;
2828                         sector += zones[i].len;
2829                         n++;
2830                 }
2831         }
2832
2833         kvfree(zones);
2834
2835         return err;
2836 }
2837 #endif
2838
2839 /*
2840  * Read f2fs raw super block.
2841  * Because we have two copies of super block, so read both of them
2842  * to get the first valid one. If any one of them is broken, we pass
2843  * them recovery flag back to the caller.
2844  */
2845 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2846                         struct f2fs_super_block **raw_super,
2847                         int *valid_super_block, int *recovery)
2848 {
2849         struct super_block *sb = sbi->sb;
2850         int block;
2851         struct buffer_head *bh;
2852         struct f2fs_super_block *super;
2853         int err = 0;
2854
2855         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2856         if (!super)
2857                 return -ENOMEM;
2858
2859         for (block = 0; block < 2; block++) {
2860                 bh = sb_bread(sb, block);
2861                 if (!bh) {
2862                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2863                                 block + 1);
2864                         err = -EIO;
2865                         continue;
2866                 }
2867
2868                 /* sanity checking of raw super */
2869                 if (sanity_check_raw_super(sbi, bh)) {
2870                         f2fs_msg(sb, KERN_ERR,
2871                                 "Can't find valid F2FS filesystem in %dth superblock",
2872                                 block + 1);
2873                         err = -EINVAL;
2874                         brelse(bh);
2875                         continue;
2876                 }
2877
2878                 if (!*raw_super) {
2879                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2880                                                         sizeof(*super));
2881                         *valid_super_block = block;
2882                         *raw_super = super;
2883                 }
2884                 brelse(bh);
2885         }
2886
2887         /* Fail to read any one of the superblocks*/
2888         if (err < 0)
2889                 *recovery = 1;
2890
2891         /* No valid superblock */
2892         if (!*raw_super)
2893                 kvfree(super);
2894         else
2895                 err = 0;
2896
2897         return err;
2898 }
2899
2900 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2901 {
2902         struct buffer_head *bh;
2903         __u32 crc = 0;
2904         int err;
2905
2906         if ((recover && f2fs_readonly(sbi->sb)) ||
2907                                 bdev_read_only(sbi->sb->s_bdev)) {
2908                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2909                 return -EROFS;
2910         }
2911
2912         /* we should update superblock crc here */
2913         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2914                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2915                                 offsetof(struct f2fs_super_block, crc));
2916                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2917         }
2918
2919         /* write back-up superblock first */
2920         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2921         if (!bh)
2922                 return -EIO;
2923         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2924         brelse(bh);
2925
2926         /* if we are in recovery path, skip writing valid superblock */
2927         if (recover || err)
2928                 return err;
2929
2930         /* write current valid superblock */
2931         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2932         if (!bh)
2933                 return -EIO;
2934         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2935         brelse(bh);
2936         return err;
2937 }
2938
2939 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2940 {
2941         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2942         unsigned int max_devices = MAX_DEVICES;
2943         int i;
2944
2945         /* Initialize single device information */
2946         if (!RDEV(0).path[0]) {
2947                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2948                         return 0;
2949                 max_devices = 1;
2950         }
2951
2952         /*
2953          * Initialize multiple devices information, or single
2954          * zoned block device information.
2955          */
2956         sbi->devs = f2fs_kzalloc(sbi,
2957                                  array_size(max_devices,
2958                                             sizeof(struct f2fs_dev_info)),
2959                                  GFP_KERNEL);
2960         if (!sbi->devs)
2961                 return -ENOMEM;
2962
2963         for (i = 0; i < max_devices; i++) {
2964
2965                 if (i > 0 && !RDEV(i).path[0])
2966                         break;
2967
2968                 if (max_devices == 1) {
2969                         /* Single zoned block device mount */
2970                         FDEV(0).bdev =
2971                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2972                                         sbi->sb->s_mode, sbi->sb->s_type);
2973                 } else {
2974                         /* Multi-device mount */
2975                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2976                         FDEV(i).total_segments =
2977                                 le32_to_cpu(RDEV(i).total_segments);
2978                         if (i == 0) {
2979                                 FDEV(i).start_blk = 0;
2980                                 FDEV(i).end_blk = FDEV(i).start_blk +
2981                                     (FDEV(i).total_segments <<
2982                                     sbi->log_blocks_per_seg) - 1 +
2983                                     le32_to_cpu(raw_super->segment0_blkaddr);
2984                         } else {
2985                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2986                                 FDEV(i).end_blk = FDEV(i).start_blk +
2987                                         (FDEV(i).total_segments <<
2988                                         sbi->log_blocks_per_seg) - 1;
2989                         }
2990                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2991                                         sbi->sb->s_mode, sbi->sb->s_type);
2992                 }
2993                 if (IS_ERR(FDEV(i).bdev))
2994                         return PTR_ERR(FDEV(i).bdev);
2995
2996                 /* to release errored devices */
2997                 sbi->s_ndevs = i + 1;
2998
2999 #ifdef CONFIG_BLK_DEV_ZONED
3000                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3001                                 !f2fs_sb_has_blkzoned(sbi)) {
3002                         f2fs_msg(sbi->sb, KERN_ERR,
3003                                 "Zoned block device feature not enabled\n");
3004                         return -EINVAL;
3005                 }
3006                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3007                         if (init_blkz_info(sbi, i)) {
3008                                 f2fs_msg(sbi->sb, KERN_ERR,
3009                                         "Failed to initialize F2FS blkzone information");
3010                                 return -EINVAL;
3011                         }
3012                         if (max_devices == 1)
3013                                 break;
3014                         f2fs_msg(sbi->sb, KERN_INFO,
3015                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3016                                 i, FDEV(i).path,
3017                                 FDEV(i).total_segments,
3018                                 FDEV(i).start_blk, FDEV(i).end_blk,
3019                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3020                                 "Host-aware" : "Host-managed");
3021                         continue;
3022                 }
3023 #endif
3024                 f2fs_msg(sbi->sb, KERN_INFO,
3025                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3026                                 i, FDEV(i).path,
3027                                 FDEV(i).total_segments,
3028                                 FDEV(i).start_blk, FDEV(i).end_blk);
3029         }
3030         f2fs_msg(sbi->sb, KERN_INFO,
3031                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3032         return 0;
3033 }
3034
3035 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3036 {
3037         struct f2fs_sm_info *sm_i = SM_I(sbi);
3038
3039         /* adjust parameters according to the volume size */
3040         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3041                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3042                 sm_i->dcc_info->discard_granularity = 1;
3043                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3044         }
3045
3046         sbi->readdir_ra = 1;
3047 }
3048
3049 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3050 {
3051         struct f2fs_sb_info *sbi;
3052         struct f2fs_super_block *raw_super;
3053         struct inode *root;
3054         int err;
3055         bool retry = true, need_fsck = false;
3056         char *options = NULL;
3057         int recovery, i, valid_super_block;
3058         struct curseg_info *seg_i;
3059
3060 try_onemore:
3061         err = -EINVAL;
3062         raw_super = NULL;
3063         valid_super_block = -1;
3064         recovery = 0;
3065
3066         /* allocate memory for f2fs-specific super block info */
3067         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3068         if (!sbi)
3069                 return -ENOMEM;
3070
3071         sbi->sb = sb;
3072
3073         /* Load the checksum driver */
3074         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3075         if (IS_ERR(sbi->s_chksum_driver)) {
3076                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3077                 err = PTR_ERR(sbi->s_chksum_driver);
3078                 sbi->s_chksum_driver = NULL;
3079                 goto free_sbi;
3080         }
3081
3082         /* set a block size */
3083         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3084                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3085                 goto free_sbi;
3086         }
3087
3088         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3089                                                                 &recovery);
3090         if (err)
3091                 goto free_sbi;
3092
3093         sb->s_fs_info = sbi;
3094         sbi->raw_super = raw_super;
3095
3096         /* precompute checksum seed for metadata */
3097         if (f2fs_sb_has_inode_chksum(sbi))
3098                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3099                                                 sizeof(raw_super->uuid));
3100
3101         /*
3102          * The BLKZONED feature indicates that the drive was formatted with
3103          * zone alignment optimization. This is optional for host-aware
3104          * devices, but mandatory for host-managed zoned block devices.
3105          */
3106 #ifndef CONFIG_BLK_DEV_ZONED
3107         if (f2fs_sb_has_blkzoned(sbi)) {
3108                 f2fs_msg(sb, KERN_ERR,
3109                          "Zoned block device support is not enabled\n");
3110                 err = -EOPNOTSUPP;
3111                 goto free_sb_buf;
3112         }
3113 #endif
3114         default_options(sbi);
3115         /* parse mount options */
3116         options = kstrdup((const char *)data, GFP_KERNEL);
3117         if (data && !options) {
3118                 err = -ENOMEM;
3119                 goto free_sb_buf;
3120         }
3121
3122         err = parse_options(sb, options);
3123         if (err)
3124                 goto free_options;
3125
3126         sbi->max_file_blocks = max_file_blocks();
3127         sb->s_maxbytes = sbi->max_file_blocks <<
3128                                 le32_to_cpu(raw_super->log_blocksize);
3129         sb->s_max_links = F2FS_LINK_MAX;
3130         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3131
3132 #ifdef CONFIG_QUOTA
3133         sb->dq_op = &f2fs_quota_operations;
3134         if (f2fs_sb_has_quota_ino(sbi))
3135                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3136         else
3137                 sb->s_qcop = &f2fs_quotactl_ops;
3138         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3139
3140         if (f2fs_sb_has_quota_ino(sbi)) {
3141                 for (i = 0; i < MAXQUOTAS; i++) {
3142                         if (f2fs_qf_ino(sbi->sb, i))
3143                                 sbi->nquota_files++;
3144                 }
3145         }
3146 #endif
3147
3148         sb->s_op = &f2fs_sops;
3149 #ifdef CONFIG_F2FS_FS_ENCRYPTION
3150         sb->s_cop = &f2fs_cryptops;
3151 #endif
3152         sb->s_xattr = f2fs_xattr_handlers;
3153         sb->s_export_op = &f2fs_export_ops;
3154         sb->s_magic = F2FS_SUPER_MAGIC;
3155         sb->s_time_gran = 1;
3156         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3157                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3158         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3159         sb->s_iflags |= SB_I_CGROUPWB;
3160
3161         /* init f2fs-specific super block info */
3162         sbi->valid_super_block = valid_super_block;
3163         mutex_init(&sbi->gc_mutex);
3164         mutex_init(&sbi->writepages);
3165         mutex_init(&sbi->cp_mutex);
3166         init_rwsem(&sbi->node_write);
3167         init_rwsem(&sbi->node_change);
3168
3169         /* disallow all the data/node/meta page writes */
3170         set_sbi_flag(sbi, SBI_POR_DOING);
3171         spin_lock_init(&sbi->stat_lock);
3172
3173         /* init iostat info */
3174         spin_lock_init(&sbi->iostat_lock);
3175         sbi->iostat_enable = false;
3176
3177         for (i = 0; i < NR_PAGE_TYPE; i++) {
3178                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3179                 int j;
3180
3181                 sbi->write_io[i] =
3182                         f2fs_kmalloc(sbi,
3183                                      array_size(n,
3184                                                 sizeof(struct f2fs_bio_info)),
3185                                      GFP_KERNEL);
3186                 if (!sbi->write_io[i]) {
3187                         err = -ENOMEM;
3188                         goto free_bio_info;
3189                 }
3190
3191                 for (j = HOT; j < n; j++) {
3192                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3193                         sbi->write_io[i][j].sbi = sbi;
3194                         sbi->write_io[i][j].bio = NULL;
3195                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3196                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3197                 }
3198         }
3199
3200         init_rwsem(&sbi->cp_rwsem);
3201         init_waitqueue_head(&sbi->cp_wait);
3202         init_sb_info(sbi);
3203
3204         err = init_percpu_info(sbi);
3205         if (err)
3206                 goto free_bio_info;
3207
3208         if (F2FS_IO_SIZE(sbi) > 1) {
3209                 sbi->write_io_dummy =
3210                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3211                 if (!sbi->write_io_dummy) {
3212                         err = -ENOMEM;
3213                         goto free_percpu;
3214                 }
3215         }
3216
3217         /* get an inode for meta space */
3218         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3219         if (IS_ERR(sbi->meta_inode)) {
3220                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3221                 err = PTR_ERR(sbi->meta_inode);
3222                 goto free_io_dummy;
3223         }
3224
3225         err = f2fs_get_valid_checkpoint(sbi);
3226         if (err) {
3227                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3228                 goto free_meta_inode;
3229         }
3230
3231         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3232                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3233         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3234                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3235                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3236         }
3237
3238         /* Initialize device list */
3239         err = f2fs_scan_devices(sbi);
3240         if (err) {
3241                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3242                 goto free_devices;
3243         }
3244
3245         sbi->total_valid_node_count =
3246                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3247         percpu_counter_set(&sbi->total_valid_inode_count,
3248                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3249         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3250         sbi->total_valid_block_count =
3251                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3252         sbi->last_valid_block_count = sbi->total_valid_block_count;
3253         sbi->reserved_blocks = 0;
3254         sbi->current_reserved_blocks = 0;
3255         limit_reserve_root(sbi);
3256
3257         for (i = 0; i < NR_INODE_TYPE; i++) {
3258                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3259                 spin_lock_init(&sbi->inode_lock[i]);
3260         }
3261
3262         f2fs_init_extent_cache_info(sbi);
3263
3264         f2fs_init_ino_entry_info(sbi);
3265
3266         f2fs_init_fsync_node_info(sbi);
3267
3268         /* setup f2fs internal modules */
3269         err = f2fs_build_segment_manager(sbi);
3270         if (err) {
3271                 f2fs_msg(sb, KERN_ERR,
3272                         "Failed to initialize F2FS segment manager");
3273                 goto free_sm;
3274         }
3275         err = f2fs_build_node_manager(sbi);
3276         if (err) {
3277                 f2fs_msg(sb, KERN_ERR,
3278                         "Failed to initialize F2FS node manager");
3279                 goto free_nm;
3280         }
3281
3282         /* For write statistics */
3283         if (sb->s_bdev->bd_part)
3284                 sbi->sectors_written_start =
3285                         (u64)part_stat_read(sb->s_bdev->bd_part,
3286                                             sectors[STAT_WRITE]);
3287
3288         /* Read accumulated write IO statistics if exists */
3289         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3290         if (__exist_node_summaries(sbi))
3291                 sbi->kbytes_written =
3292                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3293
3294         f2fs_build_gc_manager(sbi);
3295
3296         err = f2fs_build_stats(sbi);
3297         if (err)
3298                 goto free_nm;
3299
3300         /* get an inode for node space */
3301         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3302         if (IS_ERR(sbi->node_inode)) {
3303                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3304                 err = PTR_ERR(sbi->node_inode);
3305                 goto free_stats;
3306         }
3307
3308         /* read root inode and dentry */
3309         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3310         if (IS_ERR(root)) {
3311                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3312                 err = PTR_ERR(root);
3313                 goto free_node_inode;
3314         }
3315         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3316                         !root->i_size || !root->i_nlink) {
3317                 iput(root);
3318                 err = -EINVAL;
3319                 goto free_node_inode;
3320         }
3321
3322         sb->s_root = d_make_root(root); /* allocate root dentry */
3323         if (!sb->s_root) {
3324                 err = -ENOMEM;
3325                 goto free_root_inode;
3326         }
3327
3328         err = f2fs_register_sysfs(sbi);
3329         if (err)
3330                 goto free_root_inode;
3331
3332 #ifdef CONFIG_QUOTA
3333         /* Enable quota usage during mount */
3334         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3335                 err = f2fs_enable_quotas(sb);
3336                 if (err)
3337                         f2fs_msg(sb, KERN_ERR,
3338                                 "Cannot turn on quotas: error %d", err);
3339         }
3340 #endif
3341         /* if there are nt orphan nodes free them */
3342         err = f2fs_recover_orphan_inodes(sbi);
3343         if (err)
3344                 goto free_meta;
3345
3346         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3347                 goto skip_recovery;
3348
3349         /* recover fsynced data */
3350         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3351                 /*
3352                  * mount should be failed, when device has readonly mode, and
3353                  * previous checkpoint was not done by clean system shutdown.
3354                  */
3355                 if (bdev_read_only(sb->s_bdev) &&
3356                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3357                         err = -EROFS;
3358                         goto free_meta;
3359                 }
3360
3361                 if (need_fsck)
3362                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3363
3364                 if (!retry)
3365                         goto skip_recovery;
3366
3367                 err = f2fs_recover_fsync_data(sbi, false);
3368                 if (err < 0) {
3369                         need_fsck = true;
3370                         f2fs_msg(sb, KERN_ERR,
3371                                 "Cannot recover all fsync data errno=%d", err);
3372                         goto free_meta;
3373                 }
3374         } else {
3375                 err = f2fs_recover_fsync_data(sbi, true);
3376
3377                 if (!f2fs_readonly(sb) && err > 0) {
3378                         err = -EINVAL;
3379                         f2fs_msg(sb, KERN_ERR,
3380                                 "Need to recover fsync data");
3381                         goto free_meta;
3382                 }
3383         }
3384 skip_recovery:
3385         /* f2fs_recover_fsync_data() cleared this already */
3386         clear_sbi_flag(sbi, SBI_POR_DOING);
3387
3388         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3389                 err = f2fs_disable_checkpoint(sbi);
3390                 if (err)
3391                         goto sync_free_meta;
3392         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3393                 f2fs_enable_checkpoint(sbi);
3394         }
3395
3396         /*
3397          * If filesystem is not mounted as read-only then
3398          * do start the gc_thread.
3399          */
3400         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3401                 /* After POR, we can run background GC thread.*/
3402                 err = f2fs_start_gc_thread(sbi);
3403                 if (err)
3404                         goto sync_free_meta;
3405         }
3406         kvfree(options);
3407
3408         /* recover broken superblock */
3409         if (recovery) {
3410                 err = f2fs_commit_super(sbi, true);
3411                 f2fs_msg(sb, KERN_INFO,
3412                         "Try to recover %dth superblock, ret: %d",
3413                         sbi->valid_super_block ? 1 : 2, err);
3414         }
3415
3416         f2fs_join_shrinker(sbi);
3417
3418         f2fs_tuning_parameters(sbi);
3419
3420         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3421                                 cur_cp_version(F2FS_CKPT(sbi)));
3422         f2fs_update_time(sbi, CP_TIME);
3423         f2fs_update_time(sbi, REQ_TIME);
3424         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3425         return 0;
3426
3427 sync_free_meta:
3428         /* safe to flush all the data */
3429         sync_filesystem(sbi->sb);
3430         retry = false;
3431
3432 free_meta:
3433 #ifdef CONFIG_QUOTA
3434         f2fs_truncate_quota_inode_pages(sb);
3435         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3436                 f2fs_quota_off_umount(sbi->sb);
3437 #endif
3438         /*
3439          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3440          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3441          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3442          * falls into an infinite loop in f2fs_sync_meta_pages().
3443          */
3444         truncate_inode_pages_final(META_MAPPING(sbi));
3445         /* evict some inodes being cached by GC */
3446         evict_inodes(sb);
3447         f2fs_unregister_sysfs(sbi);
3448 free_root_inode:
3449         dput(sb->s_root);
3450         sb->s_root = NULL;
3451 free_node_inode:
3452         f2fs_release_ino_entry(sbi, true);
3453         truncate_inode_pages_final(NODE_MAPPING(sbi));
3454         iput(sbi->node_inode);
3455         sbi->node_inode = NULL;
3456 free_stats:
3457         f2fs_destroy_stats(sbi);
3458 free_nm:
3459         f2fs_destroy_node_manager(sbi);
3460 free_sm:
3461         f2fs_destroy_segment_manager(sbi);
3462 free_devices:
3463         destroy_device_list(sbi);
3464         kvfree(sbi->ckpt);
3465 free_meta_inode:
3466         make_bad_inode(sbi->meta_inode);
3467         iput(sbi->meta_inode);
3468         sbi->meta_inode = NULL;
3469 free_io_dummy:
3470         mempool_destroy(sbi->write_io_dummy);
3471 free_percpu:
3472         destroy_percpu_info(sbi);
3473 free_bio_info:
3474         for (i = 0; i < NR_PAGE_TYPE; i++)
3475                 kvfree(sbi->write_io[i]);
3476 free_options:
3477 #ifdef CONFIG_QUOTA
3478         for (i = 0; i < MAXQUOTAS; i++)
3479                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3480 #endif
3481         kvfree(options);
3482 free_sb_buf:
3483         kvfree(raw_super);
3484 free_sbi:
3485         if (sbi->s_chksum_driver)
3486                 crypto_free_shash(sbi->s_chksum_driver);
3487         kvfree(sbi);
3488
3489         /* give only one another chance */
3490         if (retry) {
3491                 retry = false;
3492                 shrink_dcache_sb(sb);
3493                 goto try_onemore;
3494         }
3495         return err;
3496 }
3497
3498 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3499                         const char *dev_name, void *data)
3500 {
3501         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3502 }
3503
3504 static void kill_f2fs_super(struct super_block *sb)
3505 {
3506         if (sb->s_root) {
3507                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3508
3509                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3510                 f2fs_stop_gc_thread(sbi);
3511                 f2fs_stop_discard_thread(sbi);
3512
3513                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3514                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3515                         struct cp_control cpc = {
3516                                 .reason = CP_UMOUNT,
3517                         };
3518                         f2fs_write_checkpoint(sbi, &cpc);
3519                 }
3520
3521                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3522                         sb->s_flags &= ~SB_RDONLY;
3523         }
3524         kill_block_super(sb);
3525 }
3526
3527 static struct file_system_type f2fs_fs_type = {
3528         .owner          = THIS_MODULE,
3529         .name           = "f2fs",
3530         .mount          = f2fs_mount,
3531         .kill_sb        = kill_f2fs_super,
3532         .fs_flags       = FS_REQUIRES_DEV,
3533 };
3534 MODULE_ALIAS_FS("f2fs");
3535
3536 static int __init init_inodecache(void)
3537 {
3538         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3539                         sizeof(struct f2fs_inode_info), 0,
3540                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3541         if (!f2fs_inode_cachep)
3542                 return -ENOMEM;
3543         return 0;
3544 }
3545
3546 static void destroy_inodecache(void)
3547 {
3548         /*
3549          * Make sure all delayed rcu free inodes are flushed before we
3550          * destroy cache.
3551          */
3552         rcu_barrier();
3553         kmem_cache_destroy(f2fs_inode_cachep);
3554 }
3555
3556 static int __init init_f2fs_fs(void)
3557 {
3558         int err;
3559
3560         if (PAGE_SIZE != F2FS_BLKSIZE) {
3561                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3562                                 PAGE_SIZE, F2FS_BLKSIZE);
3563                 return -EINVAL;
3564         }
3565
3566         f2fs_build_trace_ios();
3567
3568         err = init_inodecache();
3569         if (err)
3570                 goto fail;
3571         err = f2fs_create_node_manager_caches();
3572         if (err)
3573                 goto free_inodecache;
3574         err = f2fs_create_segment_manager_caches();
3575         if (err)
3576                 goto free_node_manager_caches;
3577         err = f2fs_create_checkpoint_caches();
3578         if (err)
3579                 goto free_segment_manager_caches;
3580         err = f2fs_create_extent_cache();
3581         if (err)
3582                 goto free_checkpoint_caches;
3583         err = f2fs_init_sysfs();
3584         if (err)
3585                 goto free_extent_cache;
3586         err = register_shrinker(&f2fs_shrinker_info);
3587         if (err)
3588                 goto free_sysfs;
3589         err = register_filesystem(&f2fs_fs_type);
3590         if (err)
3591                 goto free_shrinker;
3592         f2fs_create_root_stats();
3593         err = f2fs_init_post_read_processing();
3594         if (err)
3595                 goto free_root_stats;
3596         return 0;
3597
3598 free_root_stats:
3599         f2fs_destroy_root_stats();
3600         unregister_filesystem(&f2fs_fs_type);
3601 free_shrinker:
3602         unregister_shrinker(&f2fs_shrinker_info);
3603 free_sysfs:
3604         f2fs_exit_sysfs();
3605 free_extent_cache:
3606         f2fs_destroy_extent_cache();
3607 free_checkpoint_caches:
3608         f2fs_destroy_checkpoint_caches();
3609 free_segment_manager_caches:
3610         f2fs_destroy_segment_manager_caches();
3611 free_node_manager_caches:
3612         f2fs_destroy_node_manager_caches();
3613 free_inodecache:
3614         destroy_inodecache();
3615 fail:
3616         return err;
3617 }
3618
3619 static void __exit exit_f2fs_fs(void)
3620 {
3621         f2fs_destroy_post_read_processing();
3622         f2fs_destroy_root_stats();
3623         unregister_filesystem(&f2fs_fs_type);
3624         unregister_shrinker(&f2fs_shrinker_info);
3625         f2fs_exit_sysfs();
3626         f2fs_destroy_extent_cache();
3627         f2fs_destroy_checkpoint_caches();
3628         f2fs_destroy_segment_manager_caches();
3629         f2fs_destroy_node_manager_caches();
3630         destroy_inodecache();
3631         f2fs_destroy_trace_ios();
3632 }
3633
3634 module_init(init_f2fs_fs)
3635 module_exit(exit_f2fs_fs)
3636
3637 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3638 MODULE_DESCRIPTION("Flash Friendly File System");
3639 MODULE_LICENSE("GPL");
3640