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