fs/f2fs/super.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * fs/f2fs/super.c
   4  *
   5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   6  *             http://www.samsung.com/
   7  */
   8 #include <linux/module.h>
   9 #include <linux/init.h>
  10 #include <linux/fs.h>
  11 #include <linux/statfs.h>
  12 #include <linux/buffer_head.h>
  13 #include <linux/backing-dev.h>
  14 #include <linux/kthread.h>
  15 #include <linux/parser.h>
  16 #include <linux/mount.h>
  17 #include <linux/seq_file.h>
  18 #include <linux/proc_fs.h>
  19 #include <linux/random.h>
  20 #include <linux/exportfs.h>
  21 #include <linux/blkdev.h>
  22 #include <linux/quotaops.h>
  23 #include <linux/f2fs_fs.h>
  24 #include <linux/sysfs.h>
  25 #include <linux/quota.h>
  26
  27 #include "f2fs.h"
  28 #include "node.h"
  29 #include "segment.h"
  30 #include "xattr.h"
  31 #include "gc.h"
  32 #include "trace.h"
  33
  34 #define CREATE_TRACE_POINTS
  35 #include <trace/events/f2fs.h>
  36
  37 static struct kmem_cache *f2fs_inode_cachep;
  38
  39 #ifdef CONFIG_F2FS_FAULT_INJECTION
  40
  41 const char *f2fs_fault_name[FAULT_MAX] = {
  42         [FAULT_KMALLOC]         = "kmalloc",
  43         [FAULT_KVMALLOC]        = "kvmalloc",
  44         [FAULT_PAGE_ALLOC]      = "page alloc",
  45         [FAULT_PAGE_GET]        = "page get",
  46         [FAULT_ALLOC_BIO]       = "alloc bio",
  47         [FAULT_ALLOC_NID]       = "alloc nid",
  48         [FAULT_ORPHAN]          = "orphan",
  49         [FAULT_BLOCK]           = "no more block",
  50         [FAULT_DIR_DEPTH]       = "too big dir depth",
  51         [FAULT_EVICT_INODE]     = "evict_inode fail",
  52         [FAULT_TRUNCATE]        = "truncate fail",
  53         [FAULT_READ_IO]         = "read IO error",
  54         [FAULT_CHECKPOINT]      = "checkpoint error",
  55         [FAULT_DISCARD]         = "discard error",
  56         [FAULT_WRITE_IO]        = "write IO error",
  57 };
  58
  59 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  60                                                         unsigned int type)
  61 {
  62         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  63
  64         if (rate) {
  65                 atomic_set(&ffi->inject_ops, 0);
  66                 ffi->inject_rate = rate;
  67         }
  68
  69         if (type)
  70                 ffi->inject_type = type;
  71
  72         if (!rate && !type)
  73                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  74 }
  75 #endif
  76
  77 /* f2fs-wide shrinker description */
  78 static struct shrinker f2fs_shrinker_info = {
  79         .scan_objects = f2fs_shrink_scan,
  80         .count_objects = f2fs_shrink_count,
  81         .seeks = DEFAULT_SEEKS,
  82 };
  83
  84 enum {
  85         Opt_gc_background,
  86         Opt_disable_roll_forward,
  87         Opt_norecovery,
  88         Opt_discard,
  89         Opt_nodiscard,
  90         Opt_noheap,
  91         Opt_heap,
  92         Opt_user_xattr,
  93         Opt_nouser_xattr,
  94         Opt_acl,
  95         Opt_noacl,
  96         Opt_active_logs,
  97         Opt_disable_ext_identify,
  98         Opt_inline_xattr,
  99         Opt_noinline_xattr,
 100         Opt_inline_xattr_size,
 101         Opt_inline_data,
 102         Opt_inline_dentry,
 103         Opt_noinline_dentry,
 104         Opt_flush_merge,
 105         Opt_noflush_merge,
 106         Opt_nobarrier,
 107         Opt_fastboot,
 108         Opt_extent_cache,
 109         Opt_noextent_cache,
 110         Opt_noinline_data,
 111         Opt_data_flush,
 112         Opt_reserve_root,
 113         Opt_resgid,
 114         Opt_resuid,
 115         Opt_mode,
 116         Opt_io_size_bits,
 117         Opt_fault_injection,
 118         Opt_fault_type,
 119         Opt_lazytime,
 120         Opt_nolazytime,
 121         Opt_quota,
 122         Opt_noquota,
 123         Opt_usrquota,
 124         Opt_grpquota,
 125         Opt_prjquota,
 126         Opt_usrjquota,
 127         Opt_grpjquota,
 128         Opt_prjjquota,
 129         Opt_offusrjquota,
 130         Opt_offgrpjquota,
 131         Opt_offprjjquota,
 132         Opt_jqfmt_vfsold,
 133         Opt_jqfmt_vfsv0,
 134         Opt_jqfmt_vfsv1,
 135         Opt_whint,
 136         Opt_alloc,
 137         Opt_fsync,
 138         Opt_test_dummy_encryption,
 139         Opt_checkpoint,
 140         Opt_err,
 141 };
 142
 143 static match_table_t f2fs_tokens = {
 144         {Opt_gc_background, "background_gc=%s"},
 145         {Opt_disable_roll_forward, "disable_roll_forward"},
 146         {Opt_norecovery, "norecovery"},
 147         {Opt_discard, "discard"},
 148         {Opt_nodiscard, "nodiscard"},
 149         {Opt_noheap, "no_heap"},
 150         {Opt_heap, "heap"},
 151         {Opt_user_xattr, "user_xattr"},
 152         {Opt_nouser_xattr, "nouser_xattr"},
 153         {Opt_acl, "acl"},
 154         {Opt_noacl, "noacl"},
 155         {Opt_active_logs, "active_logs=%u"},
 156         {Opt_disable_ext_identify, "disable_ext_identify"},
 157         {Opt_inline_xattr, "inline_xattr"},
 158         {Opt_noinline_xattr, "noinline_xattr"},
 159         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 160         {Opt_inline_data, "inline_data"},
 161         {Opt_inline_dentry, "inline_dentry"},
 162         {Opt_noinline_dentry, "noinline_dentry"},
 163         {Opt_flush_merge, "flush_merge"},
 164         {Opt_noflush_merge, "noflush_merge"},
 165         {Opt_nobarrier, "nobarrier"},
 166         {Opt_fastboot, "fastboot"},
 167         {Opt_extent_cache, "extent_cache"},
 168         {Opt_noextent_cache, "noextent_cache"},
 169         {Opt_noinline_data, "noinline_data"},
 170         {Opt_data_flush, "data_flush"},
 171         {Opt_reserve_root, "reserve_root=%u"},
 172         {Opt_resgid, "resgid=%u"},
 173         {Opt_resuid, "resuid=%u"},
 174         {Opt_mode, "mode=%s"},
 175         {Opt_io_size_bits, "io_bits=%u"},
 176         {Opt_fault_injection, "fault_injection=%u"},
 177         {Opt_fault_type, "fault_type=%u"},
 178         {Opt_lazytime, "lazytime"},
 179         {Opt_nolazytime, "nolazytime"},
 180         {Opt_quota, "quota"},
 181         {Opt_noquota, "noquota"},
 182         {Opt_usrquota, "usrquota"},
 183         {Opt_grpquota, "grpquota"},
 184         {Opt_prjquota, "prjquota"},
 185         {Opt_usrjquota, "usrjquota=%s"},
 186         {Opt_grpjquota, "grpjquota=%s"},
 187         {Opt_prjjquota, "prjjquota=%s"},
 188         {Opt_offusrjquota, "usrjquota="},
 189         {Opt_offgrpjquota, "grpjquota="},
 190         {Opt_offprjjquota, "prjjquota="},
 191         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 192         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 193         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 194         {Opt_whint, "whint_mode=%s"},
 195         {Opt_alloc, "alloc_mode=%s"},
 196         {Opt_fsync, "fsync_mode=%s"},
 197         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 198         {Opt_checkpoint, "checkpoint=%s"},
 199         {Opt_err, NULL},
 200 };
 201
 202 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 203 {
 204         struct va_format vaf;
 205         va_list args;
 206
 207         va_start(args, fmt);
 208         vaf.fmt = fmt;
 209         vaf.va = &args;
 210         printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 211         va_end(args);
 212 }
 213
 214 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 215 {
 216         block_t limit = (sbi->user_block_count << 1) / 1000;
 217
 218         /* limit is 0.2% */
 219         if (test_opt(sbi, RESERVE_ROOT) &&
 220                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 221                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 222                 f2fs_msg(sbi->sb, KERN_INFO,
 223                         "Reduce reserved blocks for root = %u",
 224                         F2FS_OPTION(sbi).root_reserved_blocks);
 225         }
 226         if (!test_opt(sbi, RESERVE_ROOT) &&
 227                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 228                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 229                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 230                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 231                 f2fs_msg(sbi->sb, KERN_INFO,
 232                         "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 233                                 from_kuid_munged(&init_user_ns,
 234                                         F2FS_OPTION(sbi).s_resuid),
 235                                 from_kgid_munged(&init_user_ns,
 236                                         F2FS_OPTION(sbi).s_resgid));
 237 }
 238
 239 static void init_once(void *foo)
 240 {
 241         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 242
 243         inode_init_once(&fi->vfs_inode);
 244 }
 245
 246 #ifdef CONFIG_QUOTA
 247 static const char * const quotatypes[] = INITQFNAMES;
 248 #define QTYPE2NAME(t) (quotatypes[t])
 249 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 250                                                         substring_t *args)
 251 {
 252         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 253         char *qname;
 254         int ret = -EINVAL;
 255
 256         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 257                 f2fs_msg(sb, KERN_ERR,
 258                         "Cannot change journaled "
 259                         "quota options when quota turned on");
 260                 return -EINVAL;
 261         }
 262         if (f2fs_sb_has_quota_ino(sbi)) {
 263                 f2fs_msg(sb, KERN_INFO,
 264                         "QUOTA feature is enabled, so ignore qf_name");
 265                 return 0;
 266         }
 267
 268         qname = match_strdup(args);
 269         if (!qname) {
 270                 f2fs_msg(sb, KERN_ERR,
 271                         "Not enough memory for storing quotafile name");
 272                 return -ENOMEM;
 273         }
 274         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 275                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 276                         ret = 0;
 277                 else
 278                         f2fs_msg(sb, KERN_ERR,
 279                                  "%s quota file already specified",
 280                                  QTYPE2NAME(qtype));
 281                 goto errout;
 282         }
 283         if (strchr(qname, '/')) {
 284                 f2fs_msg(sb, KERN_ERR,
 285                         "quotafile must be on filesystem root");
 286                 goto errout;
 287         }
 288         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 289         set_opt(sbi, QUOTA);
 290         return 0;
 291 errout:
 292         kvfree(qname);
 293         return ret;
 294 }
 295
 296 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 297 {
 298         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 299
 300         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 301                 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
 302                         " when quota turned on");
 303                 return -EINVAL;
 304         }
 305         kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 306         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 307         return 0;
 308 }
 309
 310 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 311 {
 312         /*
 313          * We do the test below only for project quotas. 'usrquota' and
 314          * 'grpquota' mount options are allowed even without quota feature
 315          * to support legacy quotas in quota files.
 316          */
 317         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 318                 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
 319                          "Cannot enable project quota enforcement.");
 320                 return -1;
 321         }
 322         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 323                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 324                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 325                 if (test_opt(sbi, USRQUOTA) &&
 326                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 327                         clear_opt(sbi, USRQUOTA);
 328
 329                 if (test_opt(sbi, GRPQUOTA) &&
 330                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 331                         clear_opt(sbi, GRPQUOTA);
 332
 333                 if (test_opt(sbi, PRJQUOTA) &&
 334                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 335                         clear_opt(sbi, PRJQUOTA);
 336
 337                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 338                                 test_opt(sbi, PRJQUOTA)) {
 339                         f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
 340                                         "format mixing");
 341                         return -1;
 342                 }
 343
 344                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 345                         f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
 346                                         "not specified");
 347                         return -1;
 348                 }
 349         }
 350
 351         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 352                 f2fs_msg(sbi->sb, KERN_INFO,
 353                         "QUOTA feature is enabled, so ignore jquota_fmt");
 354                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 355         }
 356         return 0;
 357 }
 358 #endif
 359
 360 static int parse_options(struct super_block *sb, char *options)
 361 {
 362         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 363         substring_t args[MAX_OPT_ARGS];
 364         char *p, *name;
 365         int arg = 0;
 366         kuid_t uid;
 367         kgid_t gid;
 368 #ifdef CONFIG_QUOTA
 369         int ret;
 370 #endif
 371
 372         if (!options)
 373                 return 0;
 374
 375         while ((p = strsep(&options, ",")) != NULL) {
 376                 int token;
 377                 if (!*p)
 378                         continue;
 379                 /*
 380                  * Initialize args struct so we know whether arg was
 381                  * found; some options take optional arguments.
 382                  */
 383                 args[0].to = args[0].from = NULL;
 384                 token = match_token(p, f2fs_tokens, args);
 385
 386                 switch (token) {
 387                 case Opt_gc_background:
 388                         name = match_strdup(&args[0]);
 389
 390                         if (!name)
 391                                 return -ENOMEM;
 392                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 393                                 set_opt(sbi, BG_GC);
 394                                 clear_opt(sbi, FORCE_FG_GC);
 395                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 396                                 clear_opt(sbi, BG_GC);
 397                                 clear_opt(sbi, FORCE_FG_GC);
 398                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 399                                 set_opt(sbi, BG_GC);
 400                                 set_opt(sbi, FORCE_FG_GC);
 401                         } else {
 402                                 kvfree(name);
 403                                 return -EINVAL;
 404                         }
 405                         kvfree(name);
 406                         break;
 407                 case Opt_disable_roll_forward:
 408                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 409                         break;
 410                 case Opt_norecovery:
 411                         /* this option mounts f2fs with ro */
 412                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 413                         if (!f2fs_readonly(sb))
 414                                 return -EINVAL;
 415                         break;
 416                 case Opt_discard:
 417                         set_opt(sbi, DISCARD);
 418                         break;
 419                 case Opt_nodiscard:
 420                         if (f2fs_sb_has_blkzoned(sbi)) {
 421                                 f2fs_msg(sb, KERN_WARNING,
 422                                         "discard is required for zoned block devices");
 423                                 return -EINVAL;
 424                         }
 425                         clear_opt(sbi, DISCARD);
 426                         break;
 427                 case Opt_noheap:
 428                         set_opt(sbi, NOHEAP);
 429                         break;
 430                 case Opt_heap:
 431                         clear_opt(sbi, NOHEAP);
 432                         break;
 433 #ifdef CONFIG_F2FS_FS_XATTR
 434                 case Opt_user_xattr:
 435                         set_opt(sbi, XATTR_USER);
 436                         break;
 437                 case Opt_nouser_xattr:
 438                         clear_opt(sbi, XATTR_USER);
 439                         break;
 440                 case Opt_inline_xattr:
 441                         set_opt(sbi, INLINE_XATTR);
 442                         break;
 443                 case Opt_noinline_xattr:
 444                         clear_opt(sbi, INLINE_XATTR);
 445                         break;
 446                 case Opt_inline_xattr_size:
 447                         if (args->from && match_int(args, &arg))
 448                                 return -EINVAL;
 449                         set_opt(sbi, INLINE_XATTR_SIZE);
 450                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 451                         break;
 452 #else
 453                 case Opt_user_xattr:
 454                         f2fs_msg(sb, KERN_INFO,
 455                                 "user_xattr options not supported");
 456                         break;
 457                 case Opt_nouser_xattr:
 458                         f2fs_msg(sb, KERN_INFO,
 459                                 "nouser_xattr options not supported");
 460                         break;
 461                 case Opt_inline_xattr:
 462                         f2fs_msg(sb, KERN_INFO,
 463                                 "inline_xattr options not supported");
 464                         break;
 465                 case Opt_noinline_xattr:
 466                         f2fs_msg(sb, KERN_INFO,
 467                                 "noinline_xattr options not supported");
 468                         break;
 469 #endif
 470 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 471                 case Opt_acl:
 472                         set_opt(sbi, POSIX_ACL);
 473                         break;
 474                 case Opt_noacl:
 475                         clear_opt(sbi, POSIX_ACL);
 476                         break;
 477 #else
 478                 case Opt_acl:
 479                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 480                         break;
 481                 case Opt_noacl:
 482                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 483                         break;
 484 #endif
 485                 case Opt_active_logs:
 486                         if (args->from && match_int(args, &arg))
 487                                 return -EINVAL;
 488                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 489                                 return -EINVAL;
 490                         F2FS_OPTION(sbi).active_logs = arg;
 491                         break;
 492                 case Opt_disable_ext_identify:
 493                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 494                         break;
 495                 case Opt_inline_data:
 496                         set_opt(sbi, INLINE_DATA);
 497                         break;
 498                 case Opt_inline_dentry:
 499                         set_opt(sbi, INLINE_DENTRY);
 500                         break;
 501                 case Opt_noinline_dentry:
 502                         clear_opt(sbi, INLINE_DENTRY);
 503                         break;
 504                 case Opt_flush_merge:
 505                         set_opt(sbi, FLUSH_MERGE);
 506                         break;
 507                 case Opt_noflush_merge:
 508                         clear_opt(sbi, FLUSH_MERGE);
 509                         break;
 510                 case Opt_nobarrier:
 511                         set_opt(sbi, NOBARRIER);
 512                         break;
 513                 case Opt_fastboot:
 514                         set_opt(sbi, FASTBOOT);
 515                         break;
 516                 case Opt_extent_cache:
 517                         set_opt(sbi, EXTENT_CACHE);
 518                         break;
 519                 case Opt_noextent_cache:
 520                         clear_opt(sbi, EXTENT_CACHE);
 521                         break;
 522                 case Opt_noinline_data:
 523                         clear_opt(sbi, INLINE_DATA);
 524                         break;
 525                 case Opt_data_flush:
 526                         set_opt(sbi, DATA_FLUSH);
 527                         break;
 528                 case Opt_reserve_root:
 529                         if (args->from && match_int(args, &arg))
 530                                 return -EINVAL;
 531                         if (test_opt(sbi, RESERVE_ROOT)) {
 532                                 f2fs_msg(sb, KERN_INFO,
 533                                         "Preserve previous reserve_root=%u",
 534                                         F2FS_OPTION(sbi).root_reserved_blocks);
 535                         } else {
 536                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 537                                 set_opt(sbi, RESERVE_ROOT);
 538                         }
 539                         break;
 540                 case Opt_resuid:
 541                         if (args->from && match_int(args, &arg))
 542                                 return -EINVAL;
 543                         uid = make_kuid(current_user_ns(), arg);
 544                         if (!uid_valid(uid)) {
 545                                 f2fs_msg(sb, KERN_ERR,
 546                                         "Invalid uid value %d", arg);
 547                                 return -EINVAL;
 548                         }
 549                         F2FS_OPTION(sbi).s_resuid = uid;
 550                         break;
 551                 case Opt_resgid:
 552                         if (args->from && match_int(args, &arg))
 553                                 return -EINVAL;
 554                         gid = make_kgid(current_user_ns(), arg);
 555                         if (!gid_valid(gid)) {
 556                                 f2fs_msg(sb, KERN_ERR,
 557                                         "Invalid gid value %d", arg);
 558                                 return -EINVAL;
 559                         }
 560                         F2FS_OPTION(sbi).s_resgid = gid;
 561                         break;
 562                 case Opt_mode:
 563                         name = match_strdup(&args[0]);
 564
 565                         if (!name)
 566                                 return -ENOMEM;
 567                         if (strlen(name) == 8 &&
 568                                         !strncmp(name, "adaptive", 8)) {
 569                                 if (f2fs_sb_has_blkzoned(sbi)) {
 570                                         f2fs_msg(sb, KERN_WARNING,
 571                                                  "adaptive mode is not allowed with "
 572                                                  "zoned block device feature");
 573                                         kvfree(name);
 574                                         return -EINVAL;
 575                                 }
 576                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 577                         } else if (strlen(name) == 3 &&
 578                                         !strncmp(name, "lfs", 3)) {
 579                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 580                         } else {
 581                                 kvfree(name);
 582                                 return -EINVAL;
 583                         }
 584                         kvfree(name);
 585                         break;
 586                 case Opt_io_size_bits:
 587                         if (args->from && match_int(args, &arg))
 588                                 return -EINVAL;
 589                         if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
 590                                 f2fs_msg(sb, KERN_WARNING,
 591                                         "Not support %d, larger than %d",
 592                                         1 << arg, BIO_MAX_PAGES);
 593                                 return -EINVAL;
 594                         }
 595                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 596                         break;
 597 #ifdef CONFIG_F2FS_FAULT_INJECTION
 598                 case Opt_fault_injection:
 599                         if (args->from && match_int(args, &arg))
 600                                 return -EINVAL;
 601                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 602                         set_opt(sbi, FAULT_INJECTION);
 603                         break;
 604
 605                 case Opt_fault_type:
 606                         if (args->from && match_int(args, &arg))
 607                                 return -EINVAL;
 608                         f2fs_build_fault_attr(sbi, 0, arg);
 609                         set_opt(sbi, FAULT_INJECTION);
 610                         break;
 611 #else
 612                 case Opt_fault_injection:
 613                         f2fs_msg(sb, KERN_INFO,
 614                                 "fault_injection options not supported");
 615                         break;
 616
 617                 case Opt_fault_type:
 618                         f2fs_msg(sb, KERN_INFO,
 619                                 "fault_type options not supported");
 620                         break;
 621 #endif
 622                 case Opt_lazytime:
 623                         sb->s_flags |= SB_LAZYTIME;
 624                         break;
 625                 case Opt_nolazytime:
 626                         sb->s_flags &= ~SB_LAZYTIME;
 627                         break;
 628 #ifdef CONFIG_QUOTA
 629                 case Opt_quota:
 630                 case Opt_usrquota:
 631                         set_opt(sbi, USRQUOTA);
 632                         break;
 633                 case Opt_grpquota:
 634                         set_opt(sbi, GRPQUOTA);
 635                         break;
 636                 case Opt_prjquota:
 637                         set_opt(sbi, PRJQUOTA);
 638                         break;
 639                 case Opt_usrjquota:
 640                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 641                         if (ret)
 642                                 return ret;
 643                         break;
 644                 case Opt_grpjquota:
 645                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 646                         if (ret)
 647                                 return ret;
 648                         break;
 649                 case Opt_prjjquota:
 650                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 651                         if (ret)
 652                                 return ret;
 653                         break;
 654                 case Opt_offusrjquota:
 655                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 656                         if (ret)
 657                                 return ret;
 658                         break;
 659                 case Opt_offgrpjquota:
 660                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 661                         if (ret)
 662                                 return ret;
 663                         break;
 664                 case Opt_offprjjquota:
 665                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 666                         if (ret)
 667                                 return ret;
 668                         break;
 669                 case Opt_jqfmt_vfsold:
 670                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 671                         break;
 672                 case Opt_jqfmt_vfsv0:
 673                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 674                         break;
 675                 case Opt_jqfmt_vfsv1:
 676                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 677                         break;
 678                 case Opt_noquota:
 679                         clear_opt(sbi, QUOTA);
 680                         clear_opt(sbi, USRQUOTA);
 681                         clear_opt(sbi, GRPQUOTA);
 682                         clear_opt(sbi, PRJQUOTA);
 683                         break;
 684 #else
 685                 case Opt_quota:
 686                 case Opt_usrquota:
 687                 case Opt_grpquota:
 688                 case Opt_prjquota:
 689                 case Opt_usrjquota:
 690                 case Opt_grpjquota:
 691                 case Opt_prjjquota:
 692                 case Opt_offusrjquota:
 693                 case Opt_offgrpjquota:
 694                 case Opt_offprjjquota:
 695                 case Opt_jqfmt_vfsold:
 696                 case Opt_jqfmt_vfsv0:
 697                 case Opt_jqfmt_vfsv1:
 698                 case Opt_noquota:
 699                         f2fs_msg(sb, KERN_INFO,
 700                                         "quota operations not supported");
 701                         break;
 702 #endif
 703                 case Opt_whint:
 704                         name = match_strdup(&args[0]);
 705                         if (!name)
 706                                 return -ENOMEM;
 707                         if (strlen(name) == 10 &&
 708                                         !strncmp(name, "user-based", 10)) {
 709                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 710                         } else if (strlen(name) == 3 &&
 711                                         !strncmp(name, "off", 3)) {
 712                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 713                         } else if (strlen(name) == 8 &&
 714                                         !strncmp(name, "fs-based", 8)) {
 715                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 716                         } else {
 717                                 kvfree(name);
 718                                 return -EINVAL;
 719                         }
 720                         kvfree(name);
 721                         break;
 722                 case Opt_alloc:
 723                         name = match_strdup(&args[0]);
 724                         if (!name)
 725                                 return -ENOMEM;
 726
 727                         if (strlen(name) == 7 &&
 728                                         !strncmp(name, "default", 7)) {
 729                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 730                         } else if (strlen(name) == 5 &&
 731                                         !strncmp(name, "reuse", 5)) {
 732                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 733                         } else {
 734                                 kvfree(name);
 735                                 return -EINVAL;
 736                         }
 737                         kvfree(name);
 738                         break;
 739                 case Opt_fsync:
 740                         name = match_strdup(&args[0]);
 741                         if (!name)
 742                                 return -ENOMEM;
 743                         if (strlen(name) == 5 &&
 744                                         !strncmp(name, "posix", 5)) {
 745                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 746                         } else if (strlen(name) == 6 &&
 747                                         !strncmp(name, "strict", 6)) {
 748                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 749                         } else if (strlen(name) == 9 &&
 750                                         !strncmp(name, "nobarrier", 9)) {
 751                                 F2FS_OPTION(sbi).fsync_mode =
 752                                                         FSYNC_MODE_NOBARRIER;
 753                         } else {
 754                                 kvfree(name);
 755                                 return -EINVAL;
 756                         }
 757                         kvfree(name);
 758                         break;
 759                 case Opt_test_dummy_encryption:
 760 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 761                         if (!f2fs_sb_has_encrypt(sbi)) {
 762                                 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
 763                                 return -EINVAL;
 764                         }
 765
 766                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 767                         f2fs_msg(sb, KERN_INFO,
 768                                         "Test dummy encryption mode enabled");
 769 #else
 770                         f2fs_msg(sb, KERN_INFO,
 771                                         "Test dummy encryption mount option ignored");
 772 #endif
 773                         break;
 774                 case Opt_checkpoint:
 775                         name = match_strdup(&args[0]);
 776                         if (!name)
 777                                 return -ENOMEM;
 778
 779                         if (strlen(name) == 6 &&
 780                                         !strncmp(name, "enable", 6)) {
 781                                 clear_opt(sbi, DISABLE_CHECKPOINT);
 782                         } else if (strlen(name) == 7 &&
 783                                         !strncmp(name, "disable", 7)) {
 784                                 set_opt(sbi, DISABLE_CHECKPOINT);
 785                         } else {
 786                                 kvfree(name);
 787                                 return -EINVAL;
 788                         }
 789                         kvfree(name);
 790                         break;
 791                 default:
 792                         f2fs_msg(sb, KERN_ERR,
 793                                 "Unrecognized mount option \"%s\" or missing value",
 794                                 p);
 795                         return -EINVAL;
 796                 }
 797         }
 798 #ifdef CONFIG_QUOTA
 799         if (f2fs_check_quota_options(sbi))
 800                 return -EINVAL;
 801 #else
 802         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
 803                 f2fs_msg(sbi->sb, KERN_INFO,
 804                          "Filesystem with quota feature cannot be mounted RDWR "
 805                          "without CONFIG_QUOTA");
 806                 return -EINVAL;
 807         }
 808         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
 809                 f2fs_msg(sb, KERN_ERR,
 810                         "Filesystem with project quota feature cannot be "
 811                         "mounted RDWR without CONFIG_QUOTA");
 812                 return -EINVAL;
 813         }
 814 #endif
 815
 816         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 817                 f2fs_msg(sb, KERN_ERR,
 818                                 "Should set mode=lfs with %uKB-sized IO",
 819                                 F2FS_IO_SIZE_KB(sbi));
 820                 return -EINVAL;
 821         }
 822
 823         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 824                 if (!f2fs_sb_has_extra_attr(sbi) ||
 825                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 826                         f2fs_msg(sb, KERN_ERR,
 827                                         "extra_attr or flexible_inline_xattr "
 828                                         "feature is off");
 829                         return -EINVAL;
 830                 }
 831                 if (!test_opt(sbi, INLINE_XATTR)) {
 832                         f2fs_msg(sb, KERN_ERR,
 833                                         "inline_xattr_size option should be "
 834                                         "set with inline_xattr option");
 835                         return -EINVAL;
 836                 }
 837                 if (!F2FS_OPTION(sbi).inline_xattr_size ||
 838                         F2FS_OPTION(sbi).inline_xattr_size >=
 839                                         DEF_ADDRS_PER_INODE -
 840                                         F2FS_TOTAL_EXTRA_ATTR_SIZE -
 841                                         DEF_INLINE_RESERVED_SIZE -
 842                                         DEF_MIN_INLINE_SIZE) {
 843                         f2fs_msg(sb, KERN_ERR,
 844                                         "inline xattr size is out of range");
 845                         return -EINVAL;
 846                 }
 847         }
 848
 849         if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
 850                 f2fs_msg(sb, KERN_ERR,
 851                                 "LFS not compatible with checkpoint=disable\n");
 852                 return -EINVAL;
 853         }
 854
 855         /* Not pass down write hints if the number of active logs is lesser
 856          * than NR_CURSEG_TYPE.
 857          */
 858         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 859                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 860         return 0;
 861 }
 862
 863 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 864 {
 865         struct f2fs_inode_info *fi;
 866
 867         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 868         if (!fi)
 869                 return NULL;
 870
 871         init_once((void *) fi);
 872
 873         /* Initialize f2fs-specific inode info */
 874         atomic_set(&fi->dirty_pages, 0);
 875         init_rwsem(&fi->i_sem);
 876         INIT_LIST_HEAD(&fi->dirty_list);
 877         INIT_LIST_HEAD(&fi->gdirty_list);
 878         INIT_LIST_HEAD(&fi->inmem_ilist);
 879         INIT_LIST_HEAD(&fi->inmem_pages);
 880         mutex_init(&fi->inmem_lock);
 881         init_rwsem(&fi->i_gc_rwsem[READ]);
 882         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 883         init_rwsem(&fi->i_mmap_sem);
 884         init_rwsem(&fi->i_xattr_sem);
 885
 886         /* Will be used by directory only */
 887         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 888
 889         return &fi->vfs_inode;
 890 }
 891
 892 static int f2fs_drop_inode(struct inode *inode)
 893 {
 894         int ret;
 895         /*
 896          * This is to avoid a deadlock condition like below.
 897          * writeback_single_inode(inode)
 898          *  - f2fs_write_data_page
 899          *    - f2fs_gc -> iput -> evict
 900          *       - inode_wait_for_writeback(inode)
 901          */
 902         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 903                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 904                         /* to avoid evict_inode call simultaneously */
 905                         atomic_inc(&inode->i_count);
 906                         spin_unlock(&inode->i_lock);
 907
 908                         /* some remained atomic pages should discarded */
 909                         if (f2fs_is_atomic_file(inode))
 910                                 f2fs_drop_inmem_pages(inode);
 911
 912                         /* should remain fi->extent_tree for writepage */
 913                         f2fs_destroy_extent_node(inode);
 914
 915                         sb_start_intwrite(inode->i_sb);
 916                         f2fs_i_size_write(inode, 0);
 917
 918                         if (F2FS_HAS_BLOCKS(inode))
 919                                 f2fs_truncate(inode);
 920
 921                         sb_end_intwrite(inode->i_sb);
 922
 923                         spin_lock(&inode->i_lock);
 924                         atomic_dec(&inode->i_count);
 925                 }
 926                 trace_f2fs_drop_inode(inode, 0);
 927                 return 0;
 928         }
 929         ret = generic_drop_inode(inode);
 930         trace_f2fs_drop_inode(inode, ret);
 931         return ret;
 932 }
 933
 934 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 935 {
 936         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 937         int ret = 0;
 938
 939         spin_lock(&sbi->inode_lock[DIRTY_META]);
 940         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 941                 ret = 1;
 942         } else {
 943                 set_inode_flag(inode, FI_DIRTY_INODE);
 944                 stat_inc_dirty_inode(sbi, DIRTY_META);
 945         }
 946         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 947                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 948                                 &sbi->inode_list[DIRTY_META]);
 949                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 950         }
 951         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 952         return ret;
 953 }
 954
 955 void f2fs_inode_synced(struct inode *inode)
 956 {
 957         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 958
 959         spin_lock(&sbi->inode_lock[DIRTY_META]);
 960         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 961                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 962                 return;
 963         }
 964         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 965                 list_del_init(&F2FS_I(inode)->gdirty_list);
 966                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 967         }
 968         clear_inode_flag(inode, FI_DIRTY_INODE);
 969         clear_inode_flag(inode, FI_AUTO_RECOVER);
 970         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 971         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 972 }
 973
 974 /*
 975  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 976  *
 977  * We should call set_dirty_inode to write the dirty inode through write_inode.
 978  */
 979 static void f2fs_dirty_inode(struct inode *inode, int flags)
 980 {
 981         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 982
 983         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 984                         inode->i_ino == F2FS_META_INO(sbi))
 985                 return;
 986
 987         if (flags == I_DIRTY_TIME)
 988                 return;
 989
 990         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 991                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 992
 993         f2fs_inode_dirtied(inode, false);
 994 }
 995
 996 static void f2fs_i_callback(struct rcu_head *head)
 997 {
 998         struct inode *inode = container_of(head, struct inode, i_rcu);
 999         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1000 }
1001
1002 static void f2fs_destroy_inode(struct inode *inode)
1003 {
1004         call_rcu(&inode->i_rcu, f2fs_i_callback);
1005 }
1006
1007 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1008 {
1009         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1010         percpu_counter_destroy(&sbi->total_valid_inode_count);
1011 }
1012
1013 static void destroy_device_list(struct f2fs_sb_info *sbi)
1014 {
1015         int i;
1016
1017         for (i = 0; i < sbi->s_ndevs; i++) {
1018                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1019 #ifdef CONFIG_BLK_DEV_ZONED
1020                 kvfree(FDEV(i).blkz_type);
1021 #endif
1022         }
1023         kvfree(sbi->devs);
1024 }
1025
1026 static void f2fs_put_super(struct super_block *sb)
1027 {
1028         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1029         int i;
1030         bool dropped;
1031
1032         f2fs_quota_off_umount(sb);
1033
1034         /* prevent remaining shrinker jobs */
1035         mutex_lock(&sbi->umount_mutex);
1036
1037         /*
1038          * We don't need to do checkpoint when superblock is clean.
1039          * But, the previous checkpoint was not done by umount, it needs to do
1040          * clean checkpoint again.
1041          */
1042         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1043                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1044                 struct cp_control cpc = {
1045                         .reason = CP_UMOUNT,
1046                 };
1047                 f2fs_write_checkpoint(sbi, &cpc);
1048         }
1049
1050         /* be sure to wait for any on-going discard commands */
1051         dropped = f2fs_issue_discard_timeout(sbi);
1052
1053         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1054                                         !sbi->discard_blks && !dropped) {
1055                 struct cp_control cpc = {
1056                         .reason = CP_UMOUNT | CP_TRIMMED,
1057                 };
1058                 f2fs_write_checkpoint(sbi, &cpc);
1059         }
1060
1061         /*
1062          * normally superblock is clean, so we need to release this.
1063          * In addition, EIO will skip do checkpoint, we need this as well.
1064          */
1065         f2fs_release_ino_entry(sbi, true);
1066
1067         f2fs_leave_shrinker(sbi);
1068         mutex_unlock(&sbi->umount_mutex);
1069
1070         /* our cp_error case, we can wait for any writeback page */
1071         f2fs_flush_merged_writes(sbi);
1072
1073         f2fs_wait_on_all_pages_writeback(sbi);
1074
1075         f2fs_bug_on(sbi, sbi->fsync_node_num);
1076
1077         iput(sbi->node_inode);
1078         sbi->node_inode = NULL;
1079
1080         iput(sbi->meta_inode);
1081         sbi->meta_inode = NULL;
1082
1083         /*
1084          * iput() can update stat information, if f2fs_write_checkpoint()
1085          * above failed with error.
1086          */
1087         f2fs_destroy_stats(sbi);
1088
1089         /* destroy f2fs internal modules */
1090         f2fs_destroy_node_manager(sbi);
1091         f2fs_destroy_segment_manager(sbi);
1092
1093         kvfree(sbi->ckpt);
1094
1095         f2fs_unregister_sysfs(sbi);
1096
1097         sb->s_fs_info = NULL;
1098         if (sbi->s_chksum_driver)
1099                 crypto_free_shash(sbi->s_chksum_driver);
1100         kvfree(sbi->raw_super);
1101
1102         destroy_device_list(sbi);
1103         mempool_destroy(sbi->write_io_dummy);
1104 #ifdef CONFIG_QUOTA
1105         for (i = 0; i < MAXQUOTAS; i++)
1106                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1107 #endif
1108         destroy_percpu_info(sbi);
1109         for (i = 0; i < NR_PAGE_TYPE; i++)
1110                 kvfree(sbi->write_io[i]);
1111         kvfree(sbi);
1112 }
1113
1114 int f2fs_sync_fs(struct super_block *sb, int sync)
1115 {
1116         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1117         int err = 0;
1118
1119         if (unlikely(f2fs_cp_error(sbi)))
1120                 return 0;
1121         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1122                 return 0;
1123
1124         trace_f2fs_sync_fs(sb, sync);
1125
1126         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1127                 return -EAGAIN;
1128
1129         if (sync) {
1130                 struct cp_control cpc;
1131
1132                 cpc.reason = __get_cp_reason(sbi);
1133
1134                 mutex_lock(&sbi->gc_mutex);
1135                 err = f2fs_write_checkpoint(sbi, &cpc);
1136                 mutex_unlock(&sbi->gc_mutex);
1137         }
1138         f2fs_trace_ios(NULL, 1);
1139
1140         return err;
1141 }
1142
1143 static int f2fs_freeze(struct super_block *sb)
1144 {
1145         if (f2fs_readonly(sb))
1146                 return 0;
1147
1148         /* IO error happened before */
1149         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1150                 return -EIO;
1151
1152         /* must be clean, since sync_filesystem() was already called */
1153         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1154                 return -EINVAL;
1155         return 0;
1156 }
1157
1158 static int f2fs_unfreeze(struct super_block *sb)
1159 {
1160         return 0;
1161 }
1162
1163 #ifdef CONFIG_QUOTA
1164 static int f2fs_statfs_project(struct super_block *sb,
1165                                 kprojid_t projid, struct kstatfs *buf)
1166 {
1167         struct kqid qid;
1168         struct dquot *dquot;
1169         u64 limit;
1170         u64 curblock;
1171
1172         qid = make_kqid_projid(projid);
1173         dquot = dqget(sb, qid);
1174         if (IS_ERR(dquot))
1175                 return PTR_ERR(dquot);
1176         spin_lock(&dquot->dq_dqb_lock);
1177
1178         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1179                  dquot->dq_dqb.dqb_bsoftlimit :
1180                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1181         if (limit && buf->f_blocks > limit) {
1182                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1183                 buf->f_blocks = limit;
1184                 buf->f_bfree = buf->f_bavail =
1185                         (buf->f_blocks > curblock) ?
1186                          (buf->f_blocks - curblock) : 0;
1187         }
1188
1189         limit = dquot->dq_dqb.dqb_isoftlimit ?
1190                 dquot->dq_dqb.dqb_isoftlimit :
1191                 dquot->dq_dqb.dqb_ihardlimit;
1192         if (limit && buf->f_files > limit) {
1193                 buf->f_files = limit;
1194                 buf->f_ffree =
1195                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1196                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1197         }
1198
1199         spin_unlock(&dquot->dq_dqb_lock);
1200         dqput(dquot);
1201         return 0;
1202 }
1203 #endif
1204
1205 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1206 {
1207         struct super_block *sb = dentry->d_sb;
1208         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1209         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1210         block_t total_count, user_block_count, start_count;
1211         u64 avail_node_count;
1212
1213         total_count = le64_to_cpu(sbi->raw_super->block_count);
1214         user_block_count = sbi->user_block_count;
1215         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1216         buf->f_type = F2FS_SUPER_MAGIC;
1217         buf->f_bsize = sbi->blocksize;
1218
1219         buf->f_blocks = total_count - start_count;
1220         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1221                                                 sbi->current_reserved_blocks;
1222         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1223                 buf->f_bfree = 0;
1224         else
1225                 buf->f_bfree -= sbi->unusable_block_count;
1226
1227         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1228                 buf->f_bavail = buf->f_bfree -
1229                                 F2FS_OPTION(sbi).root_reserved_blocks;
1230         else
1231                 buf->f_bavail = 0;
1232
1233         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1234                                                 F2FS_RESERVED_NODE_NUM;
1235
1236         if (avail_node_count > user_block_count) {
1237                 buf->f_files = user_block_count;
1238                 buf->f_ffree = buf->f_bavail;
1239         } else {
1240                 buf->f_files = avail_node_count;
1241                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1242                                         buf->f_bavail);
1243         }
1244
1245         buf->f_namelen = F2FS_NAME_LEN;
1246         buf->f_fsid.val[0] = (u32)id;
1247         buf->f_fsid.val[1] = (u32)(id >> 32);
1248
1249 #ifdef CONFIG_QUOTA
1250         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1251                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1252                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1253         }
1254 #endif
1255         return 0;
1256 }
1257
1258 static inline void f2fs_show_quota_options(struct seq_file *seq,
1259                                            struct super_block *sb)
1260 {
1261 #ifdef CONFIG_QUOTA
1262         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1263
1264         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1265                 char *fmtname = "";
1266
1267                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1268                 case QFMT_VFS_OLD:
1269                         fmtname = "vfsold";
1270                         break;
1271                 case QFMT_VFS_V0:
1272                         fmtname = "vfsv0";
1273                         break;
1274                 case QFMT_VFS_V1:
1275                         fmtname = "vfsv1";
1276                         break;
1277                 }
1278                 seq_printf(seq, ",jqfmt=%s", fmtname);
1279         }
1280
1281         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1282                 seq_show_option(seq, "usrjquota",
1283                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1284
1285         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1286                 seq_show_option(seq, "grpjquota",
1287                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1288
1289         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1290                 seq_show_option(seq, "prjjquota",
1291                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1292 #endif
1293 }
1294
1295 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1296 {
1297         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1298
1299         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1300                 if (test_opt(sbi, FORCE_FG_GC))
1301                         seq_printf(seq, ",background_gc=%s", "sync");
1302                 else
1303                         seq_printf(seq, ",background_gc=%s", "on");
1304         } else {
1305                 seq_printf(seq, ",background_gc=%s", "off");
1306         }
1307         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1308                 seq_puts(seq, ",disable_roll_forward");
1309         if (test_opt(sbi, DISCARD))
1310                 seq_puts(seq, ",discard");
1311         if (test_opt(sbi, NOHEAP))
1312                 seq_puts(seq, ",no_heap");
1313         else
1314                 seq_puts(seq, ",heap");
1315 #ifdef CONFIG_F2FS_FS_XATTR
1316         if (test_opt(sbi, XATTR_USER))
1317                 seq_puts(seq, ",user_xattr");
1318         else
1319                 seq_puts(seq, ",nouser_xattr");
1320         if (test_opt(sbi, INLINE_XATTR))
1321                 seq_puts(seq, ",inline_xattr");
1322         else
1323                 seq_puts(seq, ",noinline_xattr");
1324         if (test_opt(sbi, INLINE_XATTR_SIZE))
1325                 seq_printf(seq, ",inline_xattr_size=%u",
1326                                         F2FS_OPTION(sbi).inline_xattr_size);
1327 #endif
1328 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1329         if (test_opt(sbi, POSIX_ACL))
1330                 seq_puts(seq, ",acl");
1331         else
1332                 seq_puts(seq, ",noacl");
1333 #endif
1334         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1335                 seq_puts(seq, ",disable_ext_identify");
1336         if (test_opt(sbi, INLINE_DATA))
1337                 seq_puts(seq, ",inline_data");
1338         else
1339                 seq_puts(seq, ",noinline_data");
1340         if (test_opt(sbi, INLINE_DENTRY))
1341                 seq_puts(seq, ",inline_dentry");
1342         else
1343                 seq_puts(seq, ",noinline_dentry");
1344         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1345                 seq_puts(seq, ",flush_merge");
1346         if (test_opt(sbi, NOBARRIER))
1347                 seq_puts(seq, ",nobarrier");
1348         if (test_opt(sbi, FASTBOOT))
1349                 seq_puts(seq, ",fastboot");
1350         if (test_opt(sbi, EXTENT_CACHE))
1351                 seq_puts(seq, ",extent_cache");
1352         else
1353                 seq_puts(seq, ",noextent_cache");
1354         if (test_opt(sbi, DATA_FLUSH))
1355                 seq_puts(seq, ",data_flush");
1356
1357         seq_puts(seq, ",mode=");
1358         if (test_opt(sbi, ADAPTIVE))
1359                 seq_puts(seq, "adaptive");
1360         else if (test_opt(sbi, LFS))
1361                 seq_puts(seq, "lfs");
1362         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1363         if (test_opt(sbi, RESERVE_ROOT))
1364                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1365                                 F2FS_OPTION(sbi).root_reserved_blocks,
1366                                 from_kuid_munged(&init_user_ns,
1367                                         F2FS_OPTION(sbi).s_resuid),
1368                                 from_kgid_munged(&init_user_ns,
1369                                         F2FS_OPTION(sbi).s_resgid));
1370         if (F2FS_IO_SIZE_BITS(sbi))
1371                 seq_printf(seq, ",io_bits=%u",
1372                                 F2FS_OPTION(sbi).write_io_size_bits);
1373 #ifdef CONFIG_F2FS_FAULT_INJECTION
1374         if (test_opt(sbi, FAULT_INJECTION)) {
1375                 seq_printf(seq, ",fault_injection=%u",
1376                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1377                 seq_printf(seq, ",fault_type=%u",
1378                                 F2FS_OPTION(sbi).fault_info.inject_type);
1379         }
1380 #endif
1381 #ifdef CONFIG_QUOTA
1382         if (test_opt(sbi, QUOTA))
1383                 seq_puts(seq, ",quota");
1384         if (test_opt(sbi, USRQUOTA))
1385                 seq_puts(seq, ",usrquota");
1386         if (test_opt(sbi, GRPQUOTA))
1387                 seq_puts(seq, ",grpquota");
1388         if (test_opt(sbi, PRJQUOTA))
1389                 seq_puts(seq, ",prjquota");
1390 #endif
1391         f2fs_show_quota_options(seq, sbi->sb);
1392         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1393                 seq_printf(seq, ",whint_mode=%s", "user-based");
1394         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1395                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1396 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1397         if (F2FS_OPTION(sbi).test_dummy_encryption)
1398                 seq_puts(seq, ",test_dummy_encryption");
1399 #endif
1400
1401         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1402                 seq_printf(seq, ",alloc_mode=%s", "default");
1403         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1404                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1405
1406         if (test_opt(sbi, DISABLE_CHECKPOINT))
1407                 seq_puts(seq, ",checkpoint=disable");
1408
1409         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1410                 seq_printf(seq, ",fsync_mode=%s", "posix");
1411         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1412                 seq_printf(seq, ",fsync_mode=%s", "strict");
1413         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1414                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1415         return 0;
1416 }
1417
1418 static void default_options(struct f2fs_sb_info *sbi)
1419 {
1420         /* init some FS parameters */
1421         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1422         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1423         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1424         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1425         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1426         F2FS_OPTION(sbi).test_dummy_encryption = false;
1427         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1428         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1429
1430         set_opt(sbi, BG_GC);
1431         set_opt(sbi, INLINE_XATTR);
1432         set_opt(sbi, INLINE_DATA);
1433         set_opt(sbi, INLINE_DENTRY);
1434         set_opt(sbi, EXTENT_CACHE);
1435         set_opt(sbi, NOHEAP);
1436         clear_opt(sbi, DISABLE_CHECKPOINT);
1437         sbi->sb->s_flags |= SB_LAZYTIME;
1438         set_opt(sbi, FLUSH_MERGE);
1439         set_opt(sbi, DISCARD);
1440         if (f2fs_sb_has_blkzoned(sbi))
1441                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1442         else
1443                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1444
1445 #ifdef CONFIG_F2FS_FS_XATTR
1446         set_opt(sbi, XATTR_USER);
1447 #endif
1448 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1449         set_opt(sbi, POSIX_ACL);
1450 #endif
1451
1452         f2fs_build_fault_attr(sbi, 0, 0);
1453 }
1454
1455 #ifdef CONFIG_QUOTA
1456 static int f2fs_enable_quotas(struct super_block *sb);
1457 #endif
1458
1459 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1460 {
1461         unsigned int s_flags = sbi->sb->s_flags;
1462         struct cp_control cpc;
1463         int err = 0;
1464         int ret;
1465
1466         if (s_flags & SB_RDONLY) {
1467                 f2fs_msg(sbi->sb, KERN_ERR,
1468                                 "checkpoint=disable on readonly fs");
1469                 return -EINVAL;
1470         }
1471         sbi->sb->s_flags |= SB_ACTIVE;
1472
1473         f2fs_update_time(sbi, DISABLE_TIME);
1474
1475         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1476                 mutex_lock(&sbi->gc_mutex);
1477                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1478                 if (err == -ENODATA) {
1479                         err = 0;
1480                         break;
1481                 }
1482                 if (err && err != -EAGAIN)
1483                         break;
1484         }
1485
1486         ret = sync_filesystem(sbi->sb);
1487         if (ret || err) {
1488                 err = ret ? ret: err;
1489                 goto restore_flag;
1490         }
1491
1492         if (f2fs_disable_cp_again(sbi)) {
1493                 err = -EAGAIN;
1494                 goto restore_flag;
1495         }
1496
1497         mutex_lock(&sbi->gc_mutex);
1498         cpc.reason = CP_PAUSE;
1499         set_sbi_flag(sbi, SBI_CP_DISABLED);
1500         f2fs_write_checkpoint(sbi, &cpc);
1501
1502         sbi->unusable_block_count = 0;
1503         mutex_unlock(&sbi->gc_mutex);
1504 restore_flag:
1505         sbi->sb->s_flags = s_flags;     /* Restore MS_RDONLY status */
1506         return err;
1507 }
1508
1509 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1510 {
1511         mutex_lock(&sbi->gc_mutex);
1512         f2fs_dirty_to_prefree(sbi);
1513
1514         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1515         set_sbi_flag(sbi, SBI_IS_DIRTY);
1516         mutex_unlock(&sbi->gc_mutex);
1517
1518         f2fs_sync_fs(sbi->sb, 1);
1519 }
1520
1521 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1522 {
1523         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1524         struct f2fs_mount_info org_mount_opt;
1525         unsigned long old_sb_flags;
1526         int err;
1527         bool need_restart_gc = false;
1528         bool need_stop_gc = false;
1529         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1530         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1531         bool checkpoint_changed;
1532 #ifdef CONFIG_QUOTA
1533         int i, j;
1534 #endif
1535
1536         /*
1537          * Save the old mount options in case we
1538          * need to restore them.
1539          */
1540         org_mount_opt = sbi->mount_opt;
1541         old_sb_flags = sb->s_flags;
1542
1543 #ifdef CONFIG_QUOTA
1544         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1545         for (i = 0; i < MAXQUOTAS; i++) {
1546                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1547                         org_mount_opt.s_qf_names[i] =
1548                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1549                                 GFP_KERNEL);
1550                         if (!org_mount_opt.s_qf_names[i]) {
1551                                 for (j = 0; j < i; j++)
1552                                         kvfree(org_mount_opt.s_qf_names[j]);
1553                                 return -ENOMEM;
1554                         }
1555                 } else {
1556                         org_mount_opt.s_qf_names[i] = NULL;
1557                 }
1558         }
1559 #endif
1560
1561         /* recover superblocks we couldn't write due to previous RO mount */
1562         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1563                 err = f2fs_commit_super(sbi, false);
1564                 f2fs_msg(sb, KERN_INFO,
1565                         "Try to recover all the superblocks, ret: %d", err);
1566                 if (!err)
1567                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1568         }
1569
1570         default_options(sbi);
1571
1572         /* parse mount options */
1573         err = parse_options(sb, data);
1574         if (err)
1575                 goto restore_opts;
1576         checkpoint_changed =
1577                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1578
1579         /*
1580          * Previous and new state of filesystem is RO,
1581          * so skip checking GC and FLUSH_MERGE conditions.
1582          */
1583         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1584                 goto skip;
1585
1586 #ifdef CONFIG_QUOTA
1587         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1588                 err = dquot_suspend(sb, -1);
1589                 if (err < 0)
1590                         goto restore_opts;
1591         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1592                 /* dquot_resume needs RW */
1593                 sb->s_flags &= ~SB_RDONLY;
1594                 if (sb_any_quota_suspended(sb)) {
1595                         dquot_resume(sb, -1);
1596                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1597                         err = f2fs_enable_quotas(sb);
1598                         if (err)
1599                                 goto restore_opts;
1600                 }
1601         }
1602 #endif
1603         /* disallow enable/disable extent_cache dynamically */
1604         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1605                 err = -EINVAL;
1606                 f2fs_msg(sbi->sb, KERN_WARNING,
1607                                 "switch extent_cache option is not allowed");
1608                 goto restore_opts;
1609         }
1610
1611         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1612                 err = -EINVAL;
1613                 f2fs_msg(sbi->sb, KERN_WARNING,
1614                         "disabling checkpoint not compatible with read-only");
1615                 goto restore_opts;
1616         }
1617
1618         /*
1619          * We stop the GC thread if FS is mounted as RO
1620          * or if background_gc = off is passed in mount
1621          * option. Also sync the filesystem.
1622          */
1623         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1624                 if (sbi->gc_thread) {
1625                         f2fs_stop_gc_thread(sbi);
1626                         need_restart_gc = true;
1627                 }
1628         } else if (!sbi->gc_thread) {
1629                 err = f2fs_start_gc_thread(sbi);
1630                 if (err)
1631                         goto restore_opts;
1632                 need_stop_gc = true;
1633         }
1634
1635         if (*flags & SB_RDONLY ||
1636                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1637                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1638                 sync_inodes_sb(sb);
1639
1640                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1641                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1642                 f2fs_sync_fs(sb, 1);
1643                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1644         }
1645
1646         if (checkpoint_changed) {
1647                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1648                         err = f2fs_disable_checkpoint(sbi);
1649                         if (err)
1650                                 goto restore_gc;
1651                 } else {
1652                         f2fs_enable_checkpoint(sbi);
1653                 }
1654         }
1655
1656         /*
1657          * We stop issue flush thread if FS is mounted as RO
1658          * or if flush_merge is not passed in mount option.
1659          */
1660         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1661                 clear_opt(sbi, FLUSH_MERGE);
1662                 f2fs_destroy_flush_cmd_control(sbi, false);
1663         } else {
1664                 err = f2fs_create_flush_cmd_control(sbi);
1665                 if (err)
1666                         goto restore_gc;
1667         }
1668 skip:
1669 #ifdef CONFIG_QUOTA
1670         /* Release old quota file names */
1671         for (i = 0; i < MAXQUOTAS; i++)
1672                 kvfree(org_mount_opt.s_qf_names[i]);
1673 #endif
1674         /* Update the POSIXACL Flag */
1675         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1676                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1677
1678         limit_reserve_root(sbi);
1679         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1680         return 0;
1681 restore_gc:
1682         if (need_restart_gc) {
1683                 if (f2fs_start_gc_thread(sbi))
1684                         f2fs_msg(sbi->sb, KERN_WARNING,
1685                                 "background gc thread has stopped");
1686         } else if (need_stop_gc) {
1687                 f2fs_stop_gc_thread(sbi);
1688         }
1689 restore_opts:
1690 #ifdef CONFIG_QUOTA
1691         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1692         for (i = 0; i < MAXQUOTAS; i++) {
1693                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1694                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1695         }
1696 #endif
1697         sbi->mount_opt = org_mount_opt;
1698         sb->s_flags = old_sb_flags;
1699         return err;
1700 }
1701
1702 #ifdef CONFIG_QUOTA
1703 /* Read data from quotafile */
1704 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1705                                size_t len, loff_t off)
1706 {
1707         struct inode *inode = sb_dqopt(sb)->files[type];
1708         struct address_space *mapping = inode->i_mapping;
1709         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1710         int offset = off & (sb->s_blocksize - 1);
1711         int tocopy;
1712         size_t toread;
1713         loff_t i_size = i_size_read(inode);
1714         struct page *page;
1715         char *kaddr;
1716
1717         if (off > i_size)
1718                 return 0;
1719
1720         if (off + len > i_size)
1721                 len = i_size - off;
1722         toread = len;
1723         while (toread > 0) {
1724                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1725 repeat:
1726                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1727                 if (IS_ERR(page)) {
1728                         if (PTR_ERR(page) == -ENOMEM) {
1729                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1730                                 goto repeat;
1731                         }
1732                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1733                         return PTR_ERR(page);
1734                 }
1735
1736                 lock_page(page);
1737
1738                 if (unlikely(page->mapping != mapping)) {
1739                         f2fs_put_page(page, 1);
1740                         goto repeat;
1741                 }
1742                 if (unlikely(!PageUptodate(page))) {
1743                         f2fs_put_page(page, 1);
1744                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1745                         return -EIO;
1746                 }
1747
1748                 kaddr = kmap_atomic(page);
1749                 memcpy(data, kaddr + offset, tocopy);
1750                 kunmap_atomic(kaddr);
1751                 f2fs_put_page(page, 1);
1752
1753                 offset = 0;
1754                 toread -= tocopy;
1755                 data += tocopy;
1756                 blkidx++;
1757         }
1758         return len;
1759 }
1760
1761 /* Write to quotafile */
1762 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1763                                 const char *data, size_t len, loff_t off)
1764 {
1765         struct inode *inode = sb_dqopt(sb)->files[type];
1766         struct address_space *mapping = inode->i_mapping;
1767         const struct address_space_operations *a_ops = mapping->a_ops;
1768         int offset = off & (sb->s_blocksize - 1);
1769         size_t towrite = len;
1770         struct page *page;
1771         char *kaddr;
1772         int err = 0;
1773         int tocopy;
1774
1775         while (towrite > 0) {
1776                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1777                                                                 towrite);
1778 retry:
1779                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1780                                                         &page, NULL);
1781                 if (unlikely(err)) {
1782                         if (err == -ENOMEM) {
1783                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1784                                 goto retry;
1785                         }
1786                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1787                         break;
1788                 }
1789
1790                 kaddr = kmap_atomic(page);
1791                 memcpy(kaddr + offset, data, tocopy);
1792                 kunmap_atomic(kaddr);
1793                 flush_dcache_page(page);
1794
1795                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1796                                                 page, NULL);
1797                 offset = 0;
1798                 towrite -= tocopy;
1799                 off += tocopy;
1800                 data += tocopy;
1801                 cond_resched();
1802         }
1803
1804         if (len == towrite)
1805                 return err;
1806         inode->i_mtime = inode->i_ctime = current_time(inode);
1807         f2fs_mark_inode_dirty_sync(inode, false);
1808         return len - towrite;
1809 }
1810
1811 static struct dquot **f2fs_get_dquots(struct inode *inode)
1812 {
1813         return F2FS_I(inode)->i_dquot;
1814 }
1815
1816 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1817 {
1818         return &F2FS_I(inode)->i_reserved_quota;
1819 }
1820
1821 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1822 {
1823         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1824                 f2fs_msg(sbi->sb, KERN_ERR,
1825                         "quota sysfile may be corrupted, skip loading it");
1826                 return 0;
1827         }
1828
1829         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1830                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1831 }
1832
1833 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1834 {
1835         int enabled = 0;
1836         int i, err;
1837
1838         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1839                 err = f2fs_enable_quotas(sbi->sb);
1840                 if (err) {
1841                         f2fs_msg(sbi->sb, KERN_ERR,
1842                                         "Cannot turn on quota_ino: %d", err);
1843                         return 0;
1844                 }
1845                 return 1;
1846         }
1847
1848         for (i = 0; i < MAXQUOTAS; i++) {
1849                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1850                         err = f2fs_quota_on_mount(sbi, i);
1851                         if (!err) {
1852                                 enabled = 1;
1853                                 continue;
1854                         }
1855                         f2fs_msg(sbi->sb, KERN_ERR,
1856                                 "Cannot turn on quotas: %d on %d", err, i);
1857                 }
1858         }
1859         return enabled;
1860 }
1861
1862 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1863                              unsigned int flags)
1864 {
1865         struct inode *qf_inode;
1866         unsigned long qf_inum;
1867         int err;
1868
1869         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1870
1871         qf_inum = f2fs_qf_ino(sb, type);
1872         if (!qf_inum)
1873                 return -EPERM;
1874
1875         qf_inode = f2fs_iget(sb, qf_inum);
1876         if (IS_ERR(qf_inode)) {
1877                 f2fs_msg(sb, KERN_ERR,
1878                         "Bad quota inode %u:%lu", type, qf_inum);
1879                 return PTR_ERR(qf_inode);
1880         }
1881
1882         /* Don't account quota for quota files to avoid recursion */
1883         qf_inode->i_flags |= S_NOQUOTA;
1884         err = dquot_enable(qf_inode, type, format_id, flags);
1885         iput(qf_inode);
1886         return err;
1887 }
1888
1889 static int f2fs_enable_quotas(struct super_block *sb)
1890 {
1891         int type, err = 0;
1892         unsigned long qf_inum;
1893         bool quota_mopt[MAXQUOTAS] = {
1894                 test_opt(F2FS_SB(sb), USRQUOTA),
1895                 test_opt(F2FS_SB(sb), GRPQUOTA),
1896                 test_opt(F2FS_SB(sb), PRJQUOTA),
1897         };
1898
1899         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1900                 f2fs_msg(sb, KERN_ERR,
1901                         "quota file may be corrupted, skip loading it");
1902                 return 0;
1903         }
1904
1905         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1906
1907         for (type = 0; type < MAXQUOTAS; type++) {
1908                 qf_inum = f2fs_qf_ino(sb, type);
1909                 if (qf_inum) {
1910                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1911                                 DQUOT_USAGE_ENABLED |
1912                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1913                         if (err) {
1914                                 f2fs_msg(sb, KERN_ERR,
1915                                         "Failed to enable quota tracking "
1916                                         "(type=%d, err=%d). Please run "
1917                                         "fsck to fix.", type, err);
1918                                 for (type--; type >= 0; type--)
1919                                         dquot_quota_off(sb, type);
1920                                 set_sbi_flag(F2FS_SB(sb),
1921                                                 SBI_QUOTA_NEED_REPAIR);
1922                                 return err;
1923                         }
1924                 }
1925         }
1926         return 0;
1927 }
1928
1929 int f2fs_quota_sync(struct super_block *sb, int type)
1930 {
1931         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1932         struct quota_info *dqopt = sb_dqopt(sb);
1933         int cnt;
1934         int ret;
1935
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         return ret;
1974 }
1975
1976 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1977                                                         const struct path *path)
1978 {
1979         struct inode *inode;
1980         int err;
1981
1982         err = f2fs_quota_sync(sb, type);
1983         if (err)
1984                 return err;
1985
1986         err = dquot_quota_on(sb, type, format_id, path);
1987         if (err)
1988                 return err;
1989
1990         inode = d_inode(path->dentry);
1991
1992         inode_lock(inode);
1993         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1994         f2fs_set_inode_flags(inode);
1995         inode_unlock(inode);
1996         f2fs_mark_inode_dirty_sync(inode, false);
1997
1998         return 0;
1999 }
2000
2001 static int f2fs_quota_off(struct super_block *sb, int type)
2002 {
2003         struct inode *inode = sb_dqopt(sb)->files[type];
2004         int err;
2005
2006         if (!inode || !igrab(inode))
2007                 return dquot_quota_off(sb, type);
2008
2009         err = f2fs_quota_sync(sb, type);
2010         if (err)
2011                 goto out_put;
2012
2013         err = dquot_quota_off(sb, type);
2014         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2015                 goto out_put;
2016
2017         inode_lock(inode);
2018         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2019         f2fs_set_inode_flags(inode);
2020         inode_unlock(inode);
2021         f2fs_mark_inode_dirty_sync(inode, false);
2022 out_put:
2023         iput(inode);
2024         return err;
2025 }
2026
2027 void f2fs_quota_off_umount(struct super_block *sb)
2028 {
2029         int type;
2030         int err;
2031
2032         for (type = 0; type < MAXQUOTAS; type++) {
2033                 err = f2fs_quota_off(sb, type);
2034                 if (err) {
2035                         int ret = dquot_quota_off(sb, type);
2036
2037                         f2fs_msg(sb, KERN_ERR,
2038                                 "Fail to turn off disk quota "
2039                                 "(type: %d, err: %d, ret:%d), Please "
2040                                 "run fsck to fix it.", 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         int ret;
2067
2068         ret = dquot_commit(dquot);
2069         if (ret < 0)
2070                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2071         return ret;
2072 }
2073
2074 static int f2fs_dquot_acquire(struct dquot *dquot)
2075 {
2076         int ret;
2077
2078         ret = dquot_acquire(dquot);
2079         if (ret < 0)
2080                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2081
2082         return ret;
2083 }
2084
2085 static int f2fs_dquot_release(struct dquot *dquot)
2086 {
2087         int ret;
2088
2089         ret = dquot_release(dquot);
2090         if (ret < 0)
2091                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2092         return ret;
2093 }
2094
2095 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2096 {
2097         struct super_block *sb = dquot->dq_sb;
2098         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2099         int ret;
2100
2101         ret = dquot_mark_dquot_dirty(dquot);
2102
2103         /* if we are using journalled quota */
2104         if (is_journalled_quota(sbi))
2105                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2106
2107         return ret;
2108 }
2109
2110 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2111 {
2112         int ret;
2113
2114         ret = dquot_commit_info(sb, type);
2115         if (ret < 0)
2116                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2117         return ret;
2118 }
2119
2120 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2121 {
2122         *projid = F2FS_I(inode)->i_projid;
2123         return 0;
2124 }
2125
2126 static const struct dquot_operations f2fs_quota_operations = {
2127         .get_reserved_space = f2fs_get_reserved_space,
2128         .write_dquot    = f2fs_dquot_commit,
2129         .acquire_dquot  = f2fs_dquot_acquire,
2130         .release_dquot  = f2fs_dquot_release,
2131         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2132         .write_info     = f2fs_dquot_commit_info,
2133         .alloc_dquot    = dquot_alloc,
2134         .destroy_dquot  = dquot_destroy,
2135         .get_projid     = f2fs_get_projid,
2136         .get_next_id    = dquot_get_next_id,
2137 };
2138
2139 static const struct quotactl_ops f2fs_quotactl_ops = {
2140         .quota_on       = f2fs_quota_on,
2141         .quota_off      = f2fs_quota_off,
2142         .quota_sync     = f2fs_quota_sync,
2143         .get_state      = dquot_get_state,
2144         .set_info       = dquot_set_dqinfo,
2145         .get_dqblk      = dquot_get_dqblk,
2146         .set_dqblk      = dquot_set_dqblk,
2147         .get_nextdqblk  = dquot_get_next_dqblk,
2148 };
2149 #else
2150 int f2fs_quota_sync(struct super_block *sb, int type)
2151 {
2152         return 0;
2153 }
2154
2155 void f2fs_quota_off_umount(struct super_block *sb)
2156 {
2157 }
2158 #endif
2159
2160 static const struct super_operations f2fs_sops = {
2161         .alloc_inode    = f2fs_alloc_inode,
2162         .drop_inode     = f2fs_drop_inode,
2163         .destroy_inode  = f2fs_destroy_inode,
2164         .write_inode    = f2fs_write_inode,
2165         .dirty_inode    = f2fs_dirty_inode,
2166         .show_options   = f2fs_show_options,
2167 #ifdef CONFIG_QUOTA
2168         .quota_read     = f2fs_quota_read,
2169         .quota_write    = f2fs_quota_write,
2170         .get_dquots     = f2fs_get_dquots,
2171 #endif
2172         .evict_inode    = f2fs_evict_inode,
2173         .put_super      = f2fs_put_super,
2174         .sync_fs        = f2fs_sync_fs,
2175         .freeze_fs      = f2fs_freeze,
2176         .unfreeze_fs    = f2fs_unfreeze,
2177         .statfs         = f2fs_statfs,
2178         .remount_fs     = f2fs_remount,
2179 };
2180
2181 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2182 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2183 {
2184         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2185                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2186                                 ctx, len, NULL);
2187 }
2188
2189 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2190                                                         void *fs_data)
2191 {
2192         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2193
2194         /*
2195          * Encrypting the root directory is not allowed because fsck
2196          * expects lost+found directory to exist and remain unencrypted
2197          * if LOST_FOUND feature is enabled.
2198          *
2199          */
2200         if (f2fs_sb_has_lost_found(sbi) &&
2201                         inode->i_ino == F2FS_ROOT_INO(sbi))
2202                 return -EPERM;
2203
2204         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2205                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2206                                 ctx, len, fs_data, XATTR_CREATE);
2207 }
2208
2209 static bool f2fs_dummy_context(struct inode *inode)
2210 {
2211         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2212 }
2213
2214 static const struct fscrypt_operations f2fs_cryptops = {
2215         .key_prefix     = "f2fs:",
2216         .get_context    = f2fs_get_context,
2217         .set_context    = f2fs_set_context,
2218         .dummy_context  = f2fs_dummy_context,
2219         .empty_dir      = f2fs_empty_dir,
2220         .max_namelen    = F2FS_NAME_LEN,
2221 };
2222 #endif
2223
2224 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2225                 u64 ino, u32 generation)
2226 {
2227         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2228         struct inode *inode;
2229
2230         if (f2fs_check_nid_range(sbi, ino))
2231                 return ERR_PTR(-ESTALE);
2232
2233         /*
2234          * f2fs_iget isn't quite right if the inode is currently unallocated!
2235          * However f2fs_iget currently does appropriate checks to handle stale
2236          * inodes so everything is OK.
2237          */
2238         inode = f2fs_iget(sb, ino);
2239         if (IS_ERR(inode))
2240                 return ERR_CAST(inode);
2241         if (unlikely(generation && inode->i_generation != generation)) {
2242                 /* we didn't find the right inode.. */
2243                 iput(inode);
2244                 return ERR_PTR(-ESTALE);
2245         }
2246         return inode;
2247 }
2248
2249 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2250                 int fh_len, int fh_type)
2251 {
2252         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2253                                     f2fs_nfs_get_inode);
2254 }
2255
2256 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2257                 int fh_len, int fh_type)
2258 {
2259         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2260                                     f2fs_nfs_get_inode);
2261 }
2262
2263 static const struct export_operations f2fs_export_ops = {
2264         .fh_to_dentry = f2fs_fh_to_dentry,
2265         .fh_to_parent = f2fs_fh_to_parent,
2266         .get_parent = f2fs_get_parent,
2267 };
2268
2269 static loff_t max_file_blocks(void)
2270 {
2271         loff_t result = 0;
2272         loff_t leaf_count = ADDRS_PER_BLOCK;
2273
2274         /*
2275          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2276          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2277          * space in inode.i_addr, it will be more safe to reassign
2278          * result as zero.
2279          */
2280
2281         /* two direct node blocks */
2282         result += (leaf_count * 2);
2283
2284         /* two indirect node blocks */
2285         leaf_count *= NIDS_PER_BLOCK;
2286         result += (leaf_count * 2);
2287
2288         /* one double indirect node block */
2289         leaf_count *= NIDS_PER_BLOCK;
2290         result += leaf_count;
2291
2292         return result;
2293 }
2294
2295 static int __f2fs_commit_super(struct buffer_head *bh,
2296                         struct f2fs_super_block *super)
2297 {
2298         lock_buffer(bh);
2299         if (super)
2300                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2301         set_buffer_dirty(bh);
2302         unlock_buffer(bh);
2303
2304         /* it's rare case, we can do fua all the time */
2305         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2306 }
2307
2308 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2309                                         struct buffer_head *bh)
2310 {
2311         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2312                                         (bh->b_data + F2FS_SUPER_OFFSET);
2313         struct super_block *sb = sbi->sb;
2314         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2315         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2316         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2317         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2318         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2319         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2320         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2321         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2322         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2323         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2324         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2325         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2326         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2327         u64 main_end_blkaddr = main_blkaddr +
2328                                 (segment_count_main << log_blocks_per_seg);
2329         u64 seg_end_blkaddr = segment0_blkaddr +
2330                                 (segment_count << log_blocks_per_seg);
2331
2332         if (segment0_blkaddr != cp_blkaddr) {
2333                 f2fs_msg(sb, KERN_INFO,
2334                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2335                         segment0_blkaddr, cp_blkaddr);
2336                 return true;
2337         }
2338
2339         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2340                                                         sit_blkaddr) {
2341                 f2fs_msg(sb, KERN_INFO,
2342                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2343                         cp_blkaddr, sit_blkaddr,
2344                         segment_count_ckpt << log_blocks_per_seg);
2345                 return true;
2346         }
2347
2348         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2349                                                         nat_blkaddr) {
2350                 f2fs_msg(sb, KERN_INFO,
2351                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2352                         sit_blkaddr, nat_blkaddr,
2353                         segment_count_sit << log_blocks_per_seg);
2354                 return true;
2355         }
2356
2357         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2358                                                         ssa_blkaddr) {
2359                 f2fs_msg(sb, KERN_INFO,
2360                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2361                         nat_blkaddr, ssa_blkaddr,
2362                         segment_count_nat << log_blocks_per_seg);
2363                 return true;
2364         }
2365
2366         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2367                                                         main_blkaddr) {
2368                 f2fs_msg(sb, KERN_INFO,
2369                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2370                         ssa_blkaddr, main_blkaddr,
2371                         segment_count_ssa << log_blocks_per_seg);
2372                 return true;
2373         }
2374
2375         if (main_end_blkaddr > seg_end_blkaddr) {
2376                 f2fs_msg(sb, KERN_INFO,
2377                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2378                         main_blkaddr,
2379                         segment0_blkaddr +
2380                                 (segment_count << log_blocks_per_seg),
2381                         segment_count_main << log_blocks_per_seg);
2382                 return true;
2383         } else if (main_end_blkaddr < seg_end_blkaddr) {
2384                 int err = 0;
2385                 char *res;
2386
2387                 /* fix in-memory information all the time */
2388                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2389                                 segment0_blkaddr) >> log_blocks_per_seg);
2390
2391                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2392                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2393                         res = "internally";
2394                 } else {
2395                         err = __f2fs_commit_super(bh, NULL);
2396                         res = err ? "failed" : "done";
2397                 }
2398                 f2fs_msg(sb, KERN_INFO,
2399                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
2400                         res, main_blkaddr,
2401                         segment0_blkaddr +
2402                                 (segment_count << log_blocks_per_seg),
2403                         segment_count_main << log_blocks_per_seg);
2404                 if (err)
2405                         return true;
2406         }
2407         return false;
2408 }
2409
2410 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2411                                 struct buffer_head *bh)
2412 {
2413         block_t segment_count, segs_per_sec, secs_per_zone;
2414         block_t total_sections, blocks_per_seg;
2415         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2416                                         (bh->b_data + F2FS_SUPER_OFFSET);
2417         struct super_block *sb = sbi->sb;
2418         unsigned int blocksize;
2419         size_t crc_offset = 0;
2420         __u32 crc = 0;
2421
2422         /* Check checksum_offset and crc in superblock */
2423         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2424                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2425                 if (crc_offset !=
2426                         offsetof(struct f2fs_super_block, crc)) {
2427                         f2fs_msg(sb, KERN_INFO,
2428                                 "Invalid SB checksum offset: %zu",
2429                                 crc_offset);
2430                         return 1;
2431                 }
2432                 crc = le32_to_cpu(raw_super->crc);
2433                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2434                         f2fs_msg(sb, KERN_INFO,
2435                                 "Invalid SB checksum value: %u", crc);
2436                         return 1;
2437                 }
2438         }
2439
2440         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2441                 f2fs_msg(sb, KERN_INFO,
2442                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
2443                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2444                 return 1;
2445         }
2446
2447         /* Currently, support only 4KB page cache size */
2448         if (F2FS_BLKSIZE != PAGE_SIZE) {
2449                 f2fs_msg(sb, KERN_INFO,
2450                         "Invalid page_cache_size (%lu), supports only 4KB\n",
2451                         PAGE_SIZE);
2452                 return 1;
2453         }
2454
2455         /* Currently, support only 4KB block size */
2456         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2457         if (blocksize != F2FS_BLKSIZE) {
2458                 f2fs_msg(sb, KERN_INFO,
2459                         "Invalid blocksize (%u), supports only 4KB\n",
2460                         blocksize);
2461                 return 1;
2462         }
2463
2464         /* check log blocks per segment */
2465         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2466                 f2fs_msg(sb, KERN_INFO,
2467                         "Invalid log blocks per segment (%u)\n",
2468                         le32_to_cpu(raw_super->log_blocks_per_seg));
2469                 return 1;
2470         }
2471
2472         /* Currently, support 512/1024/2048/4096 bytes sector size */
2473         if (le32_to_cpu(raw_super->log_sectorsize) >
2474                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2475                 le32_to_cpu(raw_super->log_sectorsize) <
2476                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2477                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2478                         le32_to_cpu(raw_super->log_sectorsize));
2479                 return 1;
2480         }
2481         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2482                 le32_to_cpu(raw_super->log_sectorsize) !=
2483                         F2FS_MAX_LOG_SECTOR_SIZE) {
2484                 f2fs_msg(sb, KERN_INFO,
2485                         "Invalid log sectors per block(%u) log sectorsize(%u)",
2486                         le32_to_cpu(raw_super->log_sectors_per_block),
2487                         le32_to_cpu(raw_super->log_sectorsize));
2488                 return 1;
2489         }
2490
2491         segment_count = le32_to_cpu(raw_super->segment_count);
2492         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2493         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2494         total_sections = le32_to_cpu(raw_super->section_count);
2495
2496         /* blocks_per_seg should be 512, given the above check */
2497         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2498
2499         if (segment_count > F2FS_MAX_SEGMENT ||
2500                                 segment_count < F2FS_MIN_SEGMENTS) {
2501                 f2fs_msg(sb, KERN_INFO,
2502                         "Invalid segment count (%u)",
2503                         segment_count);
2504                 return 1;
2505         }
2506
2507         if (total_sections > segment_count ||
2508                         total_sections < F2FS_MIN_SEGMENTS ||
2509                         segs_per_sec > segment_count || !segs_per_sec) {
2510                 f2fs_msg(sb, KERN_INFO,
2511                         "Invalid segment/section count (%u, %u x %u)",
2512                         segment_count, total_sections, segs_per_sec);
2513                 return 1;
2514         }
2515
2516         if ((segment_count / segs_per_sec) < total_sections) {
2517                 f2fs_msg(sb, KERN_INFO,
2518                         "Small segment_count (%u < %u * %u)",
2519                         segment_count, segs_per_sec, total_sections);
2520                 return 1;
2521         }
2522
2523         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2524                 f2fs_msg(sb, KERN_INFO,
2525                         "Wrong segment_count / block_count (%u > %llu)",
2526                         segment_count, le64_to_cpu(raw_super->block_count));
2527                 return 1;
2528         }
2529
2530         if (secs_per_zone > total_sections || !secs_per_zone) {
2531                 f2fs_msg(sb, KERN_INFO,
2532                         "Wrong secs_per_zone / total_sections (%u, %u)",
2533                         secs_per_zone, total_sections);
2534                 return 1;
2535         }
2536         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2537                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2538                         (le32_to_cpu(raw_super->extension_count) +
2539                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2540                 f2fs_msg(sb, KERN_INFO,
2541                         "Corrupted extension count (%u + %u > %u)",
2542                         le32_to_cpu(raw_super->extension_count),
2543                         raw_super->hot_ext_count,
2544                         F2FS_MAX_EXTENSION);
2545                 return 1;
2546         }
2547
2548         if (le32_to_cpu(raw_super->cp_payload) >
2549                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2550                 f2fs_msg(sb, KERN_INFO,
2551                         "Insane cp_payload (%u > %u)",
2552                         le32_to_cpu(raw_super->cp_payload),
2553                         blocks_per_seg - F2FS_CP_PACKS);
2554                 return 1;
2555         }
2556
2557         /* check reserved ino info */
2558         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2559                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2560                 le32_to_cpu(raw_super->root_ino) != 3) {
2561                 f2fs_msg(sb, KERN_INFO,
2562                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2563                         le32_to_cpu(raw_super->node_ino),
2564                         le32_to_cpu(raw_super->meta_ino),
2565                         le32_to_cpu(raw_super->root_ino));
2566                 return 1;
2567         }
2568
2569         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2570         if (sanity_check_area_boundary(sbi, bh))
2571                 return 1;
2572
2573         return 0;
2574 }
2575
2576 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2577 {
2578         unsigned int total, fsmeta;
2579         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2580         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2581         unsigned int ovp_segments, reserved_segments;
2582         unsigned int main_segs, blocks_per_seg;
2583         unsigned int sit_segs, nat_segs;
2584         unsigned int sit_bitmap_size, nat_bitmap_size;
2585         unsigned int log_blocks_per_seg;
2586         unsigned int segment_count_main;
2587         unsigned int cp_pack_start_sum, cp_payload;
2588         block_t user_block_count;
2589         int i, j;
2590
2591         total = le32_to_cpu(raw_super->segment_count);
2592         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2593         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2594         fsmeta += sit_segs;
2595         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2596         fsmeta += nat_segs;
2597         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2598         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2599
2600         if (unlikely(fsmeta >= total))
2601                 return 1;
2602
2603         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2604         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2605
2606         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2607                         ovp_segments == 0 || reserved_segments == 0)) {
2608                 f2fs_msg(sbi->sb, KERN_ERR,
2609                         "Wrong layout: check mkfs.f2fs version");
2610                 return 1;
2611         }
2612
2613         user_block_count = le64_to_cpu(ckpt->user_block_count);
2614         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2615         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2616         if (!user_block_count || user_block_count >=
2617                         segment_count_main << log_blocks_per_seg) {
2618                 f2fs_msg(sbi->sb, KERN_ERR,
2619                         "Wrong user_block_count: %u", user_block_count);
2620                 return 1;
2621         }
2622
2623         main_segs = le32_to_cpu(raw_super->segment_count_main);
2624         blocks_per_seg = sbi->blocks_per_seg;
2625
2626         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2627                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2628                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2629                         return 1;
2630                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2631                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2632                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2633                                 f2fs_msg(sbi->sb, KERN_ERR,
2634                                         "Node segment (%u, %u) has the same "
2635                                         "segno: %u", i, j,
2636                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2637                                 return 1;
2638                         }
2639                 }
2640         }
2641         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2642                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2643                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2644                         return 1;
2645                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2646                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2647                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2648                                 f2fs_msg(sbi->sb, KERN_ERR,
2649                                         "Data segment (%u, %u) has the same "
2650                                         "segno: %u", i, j,
2651                                         le32_to_cpu(ckpt->cur_data_segno[i]));
2652                                 return 1;
2653                         }
2654                 }
2655         }
2656         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2657                 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2658                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2659                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2660                                 f2fs_msg(sbi->sb, KERN_ERR,
2661                                         "Data segment (%u) and Data segment (%u)"
2662                                         " has the same segno: %u", i, j,
2663                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2664                                 return 1;
2665                         }
2666                 }
2667         }
2668
2669         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2670         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2671
2672         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2673                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2674                 f2fs_msg(sbi->sb, KERN_ERR,
2675                         "Wrong bitmap size: sit: %u, nat:%u",
2676                         sit_bitmap_size, nat_bitmap_size);
2677                 return 1;
2678         }
2679
2680         cp_pack_start_sum = __start_sum_addr(sbi);
2681         cp_payload = __cp_payload(sbi);
2682         if (cp_pack_start_sum < cp_payload + 1 ||
2683                 cp_pack_start_sum > blocks_per_seg - 1 -
2684                         NR_CURSEG_TYPE) {
2685                 f2fs_msg(sbi->sb, KERN_ERR,
2686                         "Wrong cp_pack_start_sum: %u",
2687                         cp_pack_start_sum);
2688                 return 1;
2689         }
2690
2691         if (unlikely(f2fs_cp_error(sbi))) {
2692                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2693                 return 1;
2694         }
2695         return 0;
2696 }
2697
2698 static void init_sb_info(struct f2fs_sb_info *sbi)
2699 {
2700         struct f2fs_super_block *raw_super = sbi->raw_super;
2701         int i;
2702
2703         sbi->log_sectors_per_block =
2704                 le32_to_cpu(raw_super->log_sectors_per_block);
2705         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2706         sbi->blocksize = 1 << sbi->log_blocksize;
2707         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2708         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2709         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2710         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2711         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2712         sbi->total_node_count =
2713                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2714                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2715         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2716         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2717         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2718         sbi->cur_victim_sec = NULL_SECNO;
2719         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2720         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2721         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2722         sbi->migration_granularity = sbi->segs_per_sec;
2723
2724         sbi->dir_level = DEF_DIR_LEVEL;
2725         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2726         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2727         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2728         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2729         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2730         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2731                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2732         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2733
2734         for (i = 0; i < NR_COUNT_TYPE; i++)
2735                 atomic_set(&sbi->nr_pages[i], 0);
2736
2737         for (i = 0; i < META; i++)
2738                 atomic_set(&sbi->wb_sync_req[i], 0);
2739
2740         INIT_LIST_HEAD(&sbi->s_list);
2741         mutex_init(&sbi->umount_mutex);
2742         init_rwsem(&sbi->io_order_lock);
2743         spin_lock_init(&sbi->cp_lock);
2744
2745         sbi->dirty_device = 0;
2746         spin_lock_init(&sbi->dev_lock);
2747
2748         init_rwsem(&sbi->sb_lock);
2749 }
2750
2751 static int init_percpu_info(struct f2fs_sb_info *sbi)
2752 {
2753         int err;
2754
2755         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2756         if (err)
2757                 return err;
2758
2759         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2760                                                                 GFP_KERNEL);
2761         if (err)
2762                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2763
2764         return err;
2765 }
2766
2767 #ifdef CONFIG_BLK_DEV_ZONED
2768 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2769 {
2770         struct block_device *bdev = FDEV(devi).bdev;
2771         sector_t nr_sectors = bdev->bd_part->nr_sects;
2772         sector_t sector = 0;
2773         struct blk_zone *zones;
2774         unsigned int i, nr_zones;
2775         unsigned int n = 0;
2776         int err = -EIO;
2777
2778         if (!f2fs_sb_has_blkzoned(sbi))
2779                 return 0;
2780
2781         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2782                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2783                 return -EINVAL;
2784         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2785         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2786                                 __ilog2_u32(sbi->blocks_per_blkz))
2787                 return -EINVAL;
2788         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2789         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2790                                         sbi->log_blocks_per_blkz;
2791         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2792                 FDEV(devi).nr_blkz++;
2793
2794         FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2795                                                                 GFP_KERNEL);
2796         if (!FDEV(devi).blkz_type)
2797                 return -ENOMEM;
2798
2799 #define F2FS_REPORT_NR_ZONES   4096
2800
2801         zones = f2fs_kzalloc(sbi,
2802                              array_size(F2FS_REPORT_NR_ZONES,
2803                                         sizeof(struct blk_zone)),
2804                              GFP_KERNEL);
2805         if (!zones)
2806                 return -ENOMEM;
2807
2808         /* Get block zones type */
2809         while (zones && sector < nr_sectors) {
2810
2811                 nr_zones = F2FS_REPORT_NR_ZONES;
2812                 err = blkdev_report_zones(bdev, sector,
2813                                           zones, &nr_zones,
2814                                           GFP_KERNEL);
2815                 if (err)
2816                         break;
2817                 if (!nr_zones) {
2818                         err = -EIO;
2819                         break;
2820                 }
2821
2822                 for (i = 0; i < nr_zones; i++) {
2823                         FDEV(devi).blkz_type[n] = zones[i].type;
2824                         sector += zones[i].len;
2825                         n++;
2826                 }
2827         }
2828
2829         kvfree(zones);
2830
2831         return err;
2832 }
2833 #endif
2834
2835 /*
2836  * Read f2fs raw super block.
2837  * Because we have two copies of super block, so read both of them
2838  * to get the first valid one. If any one of them is broken, we pass
2839  * them recovery flag back to the caller.
2840  */
2841 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2842                         struct f2fs_super_block **raw_super,
2843                         int *valid_super_block, int *recovery)
2844 {
2845         struct super_block *sb = sbi->sb;
2846         int block;
2847         struct buffer_head *bh;
2848         struct f2fs_super_block *super;
2849         int err = 0;
2850
2851         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2852         if (!super)
2853                 return -ENOMEM;
2854
2855         for (block = 0; block < 2; block++) {
2856                 bh = sb_bread(sb, block);
2857                 if (!bh) {
2858                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2859                                 block + 1);
2860                         err = -EIO;
2861                         continue;
2862                 }
2863
2864                 /* sanity checking of raw super */
2865                 if (sanity_check_raw_super(sbi, bh)) {
2866                         f2fs_msg(sb, KERN_ERR,
2867                                 "Can't find valid F2FS filesystem in %dth superblock",
2868                                 block + 1);
2869                         err = -EINVAL;
2870                         brelse(bh);
2871                         continue;
2872                 }
2873
2874                 if (!*raw_super) {
2875                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2876                                                         sizeof(*super));
2877                         *valid_super_block = block;
2878                         *raw_super = super;
2879                 }
2880                 brelse(bh);
2881         }
2882
2883         /* Fail to read any one of the superblocks*/
2884         if (err < 0)
2885                 *recovery = 1;
2886
2887         /* No valid superblock */
2888         if (!*raw_super)
2889                 kvfree(super);
2890         else
2891                 err = 0;
2892
2893         return err;
2894 }
2895
2896 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2897 {
2898         struct buffer_head *bh;
2899         __u32 crc = 0;
2900         int err;
2901
2902         if ((recover && f2fs_readonly(sbi->sb)) ||
2903                                 bdev_read_only(sbi->sb->s_bdev)) {
2904                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2905                 return -EROFS;
2906         }
2907
2908         /* we should update superblock crc here */
2909         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2910                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2911                                 offsetof(struct f2fs_super_block, crc));
2912                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2913         }
2914
2915         /* write back-up superblock first */
2916         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2917         if (!bh)
2918                 return -EIO;
2919         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2920         brelse(bh);
2921
2922         /* if we are in recovery path, skip writing valid superblock */
2923         if (recover || err)
2924                 return err;
2925
2926         /* write current valid superblock */
2927         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2928         if (!bh)
2929                 return -EIO;
2930         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2931         brelse(bh);
2932         return err;
2933 }
2934
2935 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2936 {
2937         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2938         unsigned int max_devices = MAX_DEVICES;
2939         int i;
2940
2941         /* Initialize single device information */
2942         if (!RDEV(0).path[0]) {
2943                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2944                         return 0;
2945                 max_devices = 1;
2946         }
2947
2948         /*
2949          * Initialize multiple devices information, or single
2950          * zoned block device information.
2951          */
2952         sbi->devs = f2fs_kzalloc(sbi,
2953                                  array_size(max_devices,
2954                                             sizeof(struct f2fs_dev_info)),
2955                                  GFP_KERNEL);
2956         if (!sbi->devs)
2957                 return -ENOMEM;
2958
2959         for (i = 0; i < max_devices; i++) {
2960
2961                 if (i > 0 && !RDEV(i).path[0])
2962                         break;
2963
2964                 if (max_devices == 1) {
2965                         /* Single zoned block device mount */
2966                         FDEV(0).bdev =
2967                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2968                                         sbi->sb->s_mode, sbi->sb->s_type);
2969                 } else {
2970                         /* Multi-device mount */
2971                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2972                         FDEV(i).total_segments =
2973                                 le32_to_cpu(RDEV(i).total_segments);
2974                         if (i == 0) {
2975                                 FDEV(i).start_blk = 0;
2976                                 FDEV(i).end_blk = FDEV(i).start_blk +
2977                                     (FDEV(i).total_segments <<
2978                                     sbi->log_blocks_per_seg) - 1 +
2979                                     le32_to_cpu(raw_super->segment0_blkaddr);
2980                         } else {
2981                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2982                                 FDEV(i).end_blk = FDEV(i).start_blk +
2983                                         (FDEV(i).total_segments <<
2984                                         sbi->log_blocks_per_seg) - 1;
2985                         }
2986                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2987                                         sbi->sb->s_mode, sbi->sb->s_type);
2988                 }
2989                 if (IS_ERR(FDEV(i).bdev))
2990                         return PTR_ERR(FDEV(i).bdev);
2991
2992                 /* to release errored devices */
2993                 sbi->s_ndevs = i + 1;
2994
2995 #ifdef CONFIG_BLK_DEV_ZONED
2996                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2997                                 !f2fs_sb_has_blkzoned(sbi)) {
2998                         f2fs_msg(sbi->sb, KERN_ERR,
2999                                 "Zoned block device feature not enabled\n");
3000                         return -EINVAL;
3001                 }
3002                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3003                         if (init_blkz_info(sbi, i)) {
3004                                 f2fs_msg(sbi->sb, KERN_ERR,
3005                                         "Failed to initialize F2FS blkzone information");
3006                                 return -EINVAL;
3007                         }
3008                         if (max_devices == 1)
3009                                 break;
3010                         f2fs_msg(sbi->sb, KERN_INFO,
3011                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3012                                 i, FDEV(i).path,
3013                                 FDEV(i).total_segments,
3014                                 FDEV(i).start_blk, FDEV(i).end_blk,
3015                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3016                                 "Host-aware" : "Host-managed");
3017                         continue;
3018                 }
3019 #endif
3020                 f2fs_msg(sbi->sb, KERN_INFO,
3021                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3022                                 i, FDEV(i).path,
3023                                 FDEV(i).total_segments,
3024                                 FDEV(i).start_blk, FDEV(i).end_blk);
3025         }
3026         f2fs_msg(sbi->sb, KERN_INFO,
3027                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3028         return 0;
3029 }
3030
3031 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3032 {
3033         struct f2fs_sm_info *sm_i = SM_I(sbi);
3034
3035         /* adjust parameters according to the volume size */
3036         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3037                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3038                 sm_i->dcc_info->discard_granularity = 1;
3039                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3040         }
3041
3042         sbi->readdir_ra = 1;
3043 }
3044
3045 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3046 {
3047         struct f2fs_sb_info *sbi;
3048         struct f2fs_super_block *raw_super;
3049         struct inode *root;
3050         int err;
3051         bool retry = true, need_fsck = false;
3052         char *options = NULL;
3053         int recovery, i, valid_super_block;
3054         struct curseg_info *seg_i;
3055
3056 try_onemore:
3057         err = -EINVAL;
3058         raw_super = NULL;
3059         valid_super_block = -1;
3060         recovery = 0;
3061
3062         /* allocate memory for f2fs-specific super block info */
3063         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3064         if (!sbi)
3065                 return -ENOMEM;
3066
3067         sbi->sb = sb;
3068
3069         /* Load the checksum driver */
3070         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3071         if (IS_ERR(sbi->s_chksum_driver)) {
3072                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3073                 err = PTR_ERR(sbi->s_chksum_driver);
3074                 sbi->s_chksum_driver = NULL;
3075                 goto free_sbi;
3076         }
3077
3078         /* set a block size */
3079         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3080                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3081                 goto free_sbi;
3082         }
3083
3084         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3085                                                                 &recovery);
3086         if (err)
3087                 goto free_sbi;
3088
3089         sb->s_fs_info = sbi;
3090         sbi->raw_super = raw_super;
3091
3092         /* precompute checksum seed for metadata */
3093         if (f2fs_sb_has_inode_chksum(sbi))
3094                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3095                                                 sizeof(raw_super->uuid));
3096
3097         /*
3098          * The BLKZONED feature indicates that the drive was formatted with
3099          * zone alignment optimization. This is optional for host-aware
3100          * devices, but mandatory for host-managed zoned block devices.
3101          */
3102 #ifndef CONFIG_BLK_DEV_ZONED
3103         if (f2fs_sb_has_blkzoned(sbi)) {
3104                 f2fs_msg(sb, KERN_ERR,
3105                          "Zoned block device support is not enabled\n");
3106                 err = -EOPNOTSUPP;
3107                 goto free_sb_buf;
3108         }
3109 #endif
3110         default_options(sbi);
3111         /* parse mount options */
3112         options = kstrdup((const char *)data, GFP_KERNEL);
3113         if (data && !options) {
3114                 err = -ENOMEM;
3115                 goto free_sb_buf;
3116         }
3117
3118         err = parse_options(sb, options);
3119         if (err)
3120                 goto free_options;
3121
3122         sbi->max_file_blocks = max_file_blocks();
3123         sb->s_maxbytes = sbi->max_file_blocks <<
3124                                 le32_to_cpu(raw_super->log_blocksize);
3125         sb->s_max_links = F2FS_LINK_MAX;
3126         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3127
3128 #ifdef CONFIG_QUOTA
3129         sb->dq_op = &f2fs_quota_operations;
3130         if (f2fs_sb_has_quota_ino(sbi))
3131                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3132         else
3133                 sb->s_qcop = &f2fs_quotactl_ops;
3134         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3135
3136         if (f2fs_sb_has_quota_ino(sbi)) {
3137                 for (i = 0; i < MAXQUOTAS; i++) {
3138                         if (f2fs_qf_ino(sbi->sb, i))
3139                                 sbi->nquota_files++;
3140                 }
3141         }
3142 #endif
3143
3144         sb->s_op = &f2fs_sops;
3145 #ifdef CONFIG_F2FS_FS_ENCRYPTION
3146         sb->s_cop = &f2fs_cryptops;
3147 #endif
3148         sb->s_xattr = f2fs_xattr_handlers;
3149         sb->s_export_op = &f2fs_export_ops;
3150         sb->s_magic = F2FS_SUPER_MAGIC;
3151         sb->s_time_gran = 1;
3152         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3153                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3154         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3155         sb->s_iflags |= SB_I_CGROUPWB;
3156
3157         /* init f2fs-specific super block info */
3158         sbi->valid_super_block = valid_super_block;
3159         mutex_init(&sbi->gc_mutex);
3160         mutex_init(&sbi->writepages);
3161         mutex_init(&sbi->cp_mutex);
3162         init_rwsem(&sbi->node_write);
3163         init_rwsem(&sbi->node_change);
3164
3165         /* disallow all the data/node/meta page writes */
3166         set_sbi_flag(sbi, SBI_POR_DOING);
3167         spin_lock_init(&sbi->stat_lock);
3168
3169         /* init iostat info */
3170         spin_lock_init(&sbi->iostat_lock);
3171         sbi->iostat_enable = false;
3172
3173         for (i = 0; i < NR_PAGE_TYPE; i++) {
3174                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3175                 int j;
3176
3177                 sbi->write_io[i] =
3178                         f2fs_kmalloc(sbi,
3179                                      array_size(n,
3180                                                 sizeof(struct f2fs_bio_info)),
3181                                      GFP_KERNEL);
3182                 if (!sbi->write_io[i]) {
3183                         err = -ENOMEM;
3184                         goto free_bio_info;
3185                 }
3186
3187                 for (j = HOT; j < n; j++) {
3188                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3189                         sbi->write_io[i][j].sbi = sbi;
3190                         sbi->write_io[i][j].bio = NULL;
3191                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3192                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3193                 }
3194         }
3195
3196         init_rwsem(&sbi->cp_rwsem);
3197         init_waitqueue_head(&sbi->cp_wait);
3198         init_sb_info(sbi);
3199
3200         err = init_percpu_info(sbi);
3201         if (err)
3202                 goto free_bio_info;
3203
3204         if (F2FS_IO_SIZE(sbi) > 1) {
3205                 sbi->write_io_dummy =
3206                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3207                 if (!sbi->write_io_dummy) {
3208                         err = -ENOMEM;
3209                         goto free_percpu;
3210                 }
3211         }
3212
3213         /* get an inode for meta space */
3214         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3215         if (IS_ERR(sbi->meta_inode)) {
3216                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3217                 err = PTR_ERR(sbi->meta_inode);
3218                 goto free_io_dummy;
3219         }
3220
3221         err = f2fs_get_valid_checkpoint(sbi);
3222         if (err) {
3223                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3224                 goto free_meta_inode;
3225         }
3226
3227         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3228                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3229         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3230                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3231                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3232         }
3233
3234         /* Initialize device list */
3235         err = f2fs_scan_devices(sbi);
3236         if (err) {
3237                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3238                 goto free_devices;
3239         }
3240
3241         sbi->total_valid_node_count =
3242                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3243         percpu_counter_set(&sbi->total_valid_inode_count,
3244                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3245         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3246         sbi->total_valid_block_count =
3247                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3248         sbi->last_valid_block_count = sbi->total_valid_block_count;
3249         sbi->reserved_blocks = 0;
3250         sbi->current_reserved_blocks = 0;
3251         limit_reserve_root(sbi);
3252
3253         for (i = 0; i < NR_INODE_TYPE; i++) {
3254                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3255                 spin_lock_init(&sbi->inode_lock[i]);
3256         }
3257
3258         f2fs_init_extent_cache_info(sbi);
3259
3260         f2fs_init_ino_entry_info(sbi);
3261
3262         f2fs_init_fsync_node_info(sbi);
3263
3264         /* setup f2fs internal modules */
3265         err = f2fs_build_segment_manager(sbi);
3266         if (err) {
3267                 f2fs_msg(sb, KERN_ERR,
3268                         "Failed to initialize F2FS segment manager");
3269                 goto free_sm;
3270         }
3271         err = f2fs_build_node_manager(sbi);
3272         if (err) {
3273                 f2fs_msg(sb, KERN_ERR,
3274                         "Failed to initialize F2FS node manager");
3275                 goto free_nm;
3276         }
3277
3278         /* For write statistics */
3279         if (sb->s_bdev->bd_part)
3280                 sbi->sectors_written_start =
3281                         (u64)part_stat_read(sb->s_bdev->bd_part,
3282                                             sectors[STAT_WRITE]);
3283
3284         /* Read accumulated write IO statistics if exists */
3285         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3286         if (__exist_node_summaries(sbi))
3287                 sbi->kbytes_written =
3288                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3289
3290         f2fs_build_gc_manager(sbi);
3291
3292         err = f2fs_build_stats(sbi);
3293         if (err)
3294                 goto free_nm;
3295
3296         /* get an inode for node space */
3297         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3298         if (IS_ERR(sbi->node_inode)) {
3299                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3300                 err = PTR_ERR(sbi->node_inode);
3301                 goto free_stats;
3302         }
3303
3304         /* read root inode and dentry */
3305         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3306         if (IS_ERR(root)) {
3307                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3308                 err = PTR_ERR(root);
3309                 goto free_node_inode;
3310         }
3311         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3312                         !root->i_size || !root->i_nlink) {
3313                 iput(root);
3314                 err = -EINVAL;
3315                 goto free_node_inode;
3316         }
3317
3318         sb->s_root = d_make_root(root); /* allocate root dentry */
3319         if (!sb->s_root) {
3320                 err = -ENOMEM;
3321                 goto free_root_inode;
3322         }
3323
3324         err = f2fs_register_sysfs(sbi);
3325         if (err)
3326                 goto free_root_inode;
3327
3328 #ifdef CONFIG_QUOTA
3329         /* Enable quota usage during mount */
3330         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3331                 err = f2fs_enable_quotas(sb);
3332                 if (err)
3333                         f2fs_msg(sb, KERN_ERR,
3334                                 "Cannot turn on quotas: error %d", err);
3335         }
3336 #endif
3337         /* if there are nt orphan nodes free them */
3338         err = f2fs_recover_orphan_inodes(sbi);
3339         if (err)
3340                 goto free_meta;
3341
3342         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3343                 goto skip_recovery;
3344
3345         /* recover fsynced data */
3346         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3347                 /*
3348                  * mount should be failed, when device has readonly mode, and
3349                  * previous checkpoint was not done by clean system shutdown.
3350                  */
3351                 if (bdev_read_only(sb->s_bdev) &&
3352                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3353                         err = -EROFS;
3354                         goto free_meta;
3355                 }
3356
3357                 if (need_fsck)
3358                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3359
3360                 if (!retry)
3361                         goto skip_recovery;
3362
3363                 err = f2fs_recover_fsync_data(sbi, false);
3364                 if (err < 0) {
3365                         need_fsck = true;
3366                         f2fs_msg(sb, KERN_ERR,
3367                                 "Cannot recover all fsync data errno=%d", err);
3368                         goto free_meta;
3369                 }
3370         } else {
3371                 err = f2fs_recover_fsync_data(sbi, true);
3372
3373                 if (!f2fs_readonly(sb) && err > 0) {
3374                         err = -EINVAL;
3375                         f2fs_msg(sb, KERN_ERR,
3376                                 "Need to recover fsync data");
3377                         goto free_meta;
3378                 }
3379         }
3380 skip_recovery:
3381         /* f2fs_recover_fsync_data() cleared this already */
3382         clear_sbi_flag(sbi, SBI_POR_DOING);
3383
3384         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3385                 err = f2fs_disable_checkpoint(sbi);
3386                 if (err)
3387                         goto sync_free_meta;
3388         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3389                 f2fs_enable_checkpoint(sbi);
3390         }
3391
3392         /*
3393          * If filesystem is not mounted as read-only then
3394          * do start the gc_thread.
3395          */
3396         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3397                 /* After POR, we can run background GC thread.*/
3398                 err = f2fs_start_gc_thread(sbi);
3399                 if (err)
3400                         goto sync_free_meta;
3401         }
3402         kvfree(options);
3403
3404         /* recover broken superblock */
3405         if (recovery) {
3406                 err = f2fs_commit_super(sbi, true);
3407                 f2fs_msg(sb, KERN_INFO,
3408                         "Try to recover %dth superblock, ret: %d",
3409                         sbi->valid_super_block ? 1 : 2, err);
3410         }
3411
3412         f2fs_join_shrinker(sbi);
3413
3414         f2fs_tuning_parameters(sbi);
3415
3416         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3417                                 cur_cp_version(F2FS_CKPT(sbi)));
3418         f2fs_update_time(sbi, CP_TIME);
3419         f2fs_update_time(sbi, REQ_TIME);
3420         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3421         return 0;
3422
3423 sync_free_meta:
3424         /* safe to flush all the data */
3425         sync_filesystem(sbi->sb);
3426         retry = false;
3427
3428 free_meta:
3429 #ifdef CONFIG_QUOTA
3430         f2fs_truncate_quota_inode_pages(sb);
3431         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3432                 f2fs_quota_off_umount(sbi->sb);
3433 #endif
3434         /*
3435          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3436          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3437          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3438          * falls into an infinite loop in f2fs_sync_meta_pages().
3439          */
3440         truncate_inode_pages_final(META_MAPPING(sbi));
3441         /* evict some inodes being cached by GC */
3442         evict_inodes(sb);
3443         f2fs_unregister_sysfs(sbi);
3444 free_root_inode:
3445         dput(sb->s_root);
3446         sb->s_root = NULL;
3447 free_node_inode:
3448         f2fs_release_ino_entry(sbi, true);
3449         truncate_inode_pages_final(NODE_MAPPING(sbi));
3450         iput(sbi->node_inode);
3451         sbi->node_inode = NULL;
3452 free_stats:
3453         f2fs_destroy_stats(sbi);
3454 free_nm:
3455         f2fs_destroy_node_manager(sbi);
3456 free_sm:
3457         f2fs_destroy_segment_manager(sbi);
3458 free_devices:
3459         destroy_device_list(sbi);
3460         kvfree(sbi->ckpt);
3461 free_meta_inode:
3462         make_bad_inode(sbi->meta_inode);
3463         iput(sbi->meta_inode);
3464         sbi->meta_inode = NULL;
3465 free_io_dummy:
3466         mempool_destroy(sbi->write_io_dummy);
3467 free_percpu:
3468         destroy_percpu_info(sbi);
3469 free_bio_info:
3470         for (i = 0; i < NR_PAGE_TYPE; i++)
3471                 kvfree(sbi->write_io[i]);
3472 free_options:
3473 #ifdef CONFIG_QUOTA
3474         for (i = 0; i < MAXQUOTAS; i++)
3475                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3476 #endif
3477         kvfree(options);
3478 free_sb_buf:
3479         kvfree(raw_super);
3480 free_sbi:
3481         if (sbi->s_chksum_driver)
3482                 crypto_free_shash(sbi->s_chksum_driver);
3483         kvfree(sbi);
3484
3485         /* give only one another chance */
3486         if (retry) {
3487                 retry = false;
3488                 shrink_dcache_sb(sb);
3489                 goto try_onemore;
3490         }
3491         return err;
3492 }
3493
3494 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3495                         const char *dev_name, void *data)
3496 {
3497         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3498 }
3499
3500 static void kill_f2fs_super(struct super_block *sb)
3501 {
3502         if (sb->s_root) {
3503                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3504
3505                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3506                 f2fs_stop_gc_thread(sbi);
3507                 f2fs_stop_discard_thread(sbi);
3508
3509                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3510                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3511                         struct cp_control cpc = {
3512                                 .reason = CP_UMOUNT,
3513                         };
3514                         f2fs_write_checkpoint(sbi, &cpc);
3515                 }
3516
3517                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3518                         sb->s_flags &= ~SB_RDONLY;
3519         }
3520         kill_block_super(sb);
3521 }
3522
3523 static struct file_system_type f2fs_fs_type = {
3524         .owner          = THIS_MODULE,
3525         .name           = "f2fs",
3526         .mount          = f2fs_mount,
3527         .kill_sb        = kill_f2fs_super,
3528         .fs_flags       = FS_REQUIRES_DEV,
3529 };
3530 MODULE_ALIAS_FS("f2fs");
3531
3532 static int __init init_inodecache(void)
3533 {
3534         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3535                         sizeof(struct f2fs_inode_info), 0,
3536                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3537         if (!f2fs_inode_cachep)
3538                 return -ENOMEM;
3539         return 0;
3540 }
3541
3542 static void destroy_inodecache(void)
3543 {
3544         /*
3545          * Make sure all delayed rcu free inodes are flushed before we
3546          * destroy cache.
3547          */
3548         rcu_barrier();
3549         kmem_cache_destroy(f2fs_inode_cachep);
3550 }
3551
3552 static int __init init_f2fs_fs(void)
3553 {
3554         int err;
3555
3556         if (PAGE_SIZE != F2FS_BLKSIZE) {
3557                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3558                                 PAGE_SIZE, F2FS_BLKSIZE);
3559                 return -EINVAL;
3560         }
3561
3562         f2fs_build_trace_ios();
3563
3564         err = init_inodecache();
3565         if (err)
3566                 goto fail;
3567         err = f2fs_create_node_manager_caches();
3568         if (err)
3569                 goto free_inodecache;
3570         err = f2fs_create_segment_manager_caches();
3571         if (err)
3572                 goto free_node_manager_caches;
3573         err = f2fs_create_checkpoint_caches();
3574         if (err)
3575                 goto free_segment_manager_caches;
3576         err = f2fs_create_extent_cache();
3577         if (err)
3578                 goto free_checkpoint_caches;
3579         err = f2fs_init_sysfs();
3580         if (err)
3581                 goto free_extent_cache;
3582         err = register_shrinker(&f2fs_shrinker_info);
3583         if (err)
3584                 goto free_sysfs;
3585         err = register_filesystem(&f2fs_fs_type);
3586         if (err)
3587                 goto free_shrinker;
3588         f2fs_create_root_stats();
3589         err = f2fs_init_post_read_processing();
3590         if (err)
3591                 goto free_root_stats;
3592         return 0;
3593
3594 free_root_stats:
3595         f2fs_destroy_root_stats();
3596         unregister_filesystem(&f2fs_fs_type);
3597 free_shrinker:
3598         unregister_shrinker(&f2fs_shrinker_info);
3599 free_sysfs:
3600         f2fs_exit_sysfs();
3601 free_extent_cache:
3602         f2fs_destroy_extent_cache();
3603 free_checkpoint_caches:
3604         f2fs_destroy_checkpoint_caches();
3605 free_segment_manager_caches:
3606         f2fs_destroy_segment_manager_caches();
3607 free_node_manager_caches:
3608         f2fs_destroy_node_manager_caches();
3609 free_inodecache:
3610         destroy_inodecache();
3611 fail:
3612         return err;
3613 }
3614
3615 static void __exit exit_f2fs_fs(void)
3616 {
3617         f2fs_destroy_post_read_processing();
3618         f2fs_destroy_root_stats();
3619         unregister_filesystem(&f2fs_fs_type);
3620         unregister_shrinker(&f2fs_shrinker_info);
3621         f2fs_exit_sysfs();
3622         f2fs_destroy_extent_cache();
3623         f2fs_destroy_checkpoint_caches();
3624         f2fs_destroy_segment_manager_caches();
3625         f2fs_destroy_node_manager_caches();
3626         destroy_inodecache();
3627         f2fs_destroy_trace_ios();
3628 }
3629
3630 module_init(init_f2fs_fs)
3631 module_exit(exit_f2fs_fs)
3632
3633 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3634 MODULE_DESCRIPTION("Flash Friendly File System");
3635 MODULE_LICENSE("GPL");
3636