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 <= 0 || 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                 int min_size, max_size;
 825
 826                 if (!f2fs_sb_has_extra_attr(sbi) ||
 827                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 828                         f2fs_msg(sb, KERN_ERR,
 829                                         "extra_attr or flexible_inline_xattr "
 830                                         "feature is off");
 831                         return -EINVAL;
 832                 }
 833                 if (!test_opt(sbi, INLINE_XATTR)) {
 834                         f2fs_msg(sb, KERN_ERR,
 835                                         "inline_xattr_size option should be "
 836                                         "set with inline_xattr option");
 837                         return -EINVAL;
 838                 }
 839
 840                 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
 841                 max_size = DEF_ADDRS_PER_INODE -
 842                         F2FS_TOTAL_EXTRA_ATTR_SIZE / sizeof(__le32) -
 843                         DEF_INLINE_RESERVED_SIZE -
 844                         MIN_INLINE_DENTRY_SIZE / sizeof(__le32);
 845
 846                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
 847                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
 848                         f2fs_msg(sb, KERN_ERR,
 849                                 "inline xattr size is out of range: %d ~ %d",
 850                                 min_size, max_size);
 851                         return -EINVAL;
 852                 }
 853         }
 854
 855         if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
 856                 f2fs_msg(sb, KERN_ERR,
 857                                 "LFS not compatible with checkpoint=disable\n");
 858                 return -EINVAL;
 859         }
 860
 861         /* Not pass down write hints if the number of active logs is lesser
 862          * than NR_CURSEG_TYPE.
 863          */
 864         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 865                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 866         return 0;
 867 }
 868
 869 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 870 {
 871         struct f2fs_inode_info *fi;
 872
 873         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 874         if (!fi)
 875                 return NULL;
 876
 877         init_once((void *) fi);
 878
 879         /* Initialize f2fs-specific inode info */
 880         atomic_set(&fi->dirty_pages, 0);
 881         init_rwsem(&fi->i_sem);
 882         INIT_LIST_HEAD(&fi->dirty_list);
 883         INIT_LIST_HEAD(&fi->gdirty_list);
 884         INIT_LIST_HEAD(&fi->inmem_ilist);
 885         INIT_LIST_HEAD(&fi->inmem_pages);
 886         mutex_init(&fi->inmem_lock);
 887         init_rwsem(&fi->i_gc_rwsem[READ]);
 888         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 889         init_rwsem(&fi->i_mmap_sem);
 890         init_rwsem(&fi->i_xattr_sem);
 891
 892         /* Will be used by directory only */
 893         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 894
 895         return &fi->vfs_inode;
 896 }
 897
 898 static int f2fs_drop_inode(struct inode *inode)
 899 {
 900         int ret;
 901         /*
 902          * This is to avoid a deadlock condition like below.
 903          * writeback_single_inode(inode)
 904          *  - f2fs_write_data_page
 905          *    - f2fs_gc -> iput -> evict
 906          *       - inode_wait_for_writeback(inode)
 907          */
 908         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 909                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 910                         /* to avoid evict_inode call simultaneously */
 911                         atomic_inc(&inode->i_count);
 912                         spin_unlock(&inode->i_lock);
 913
 914                         /* some remained atomic pages should discarded */
 915                         if (f2fs_is_atomic_file(inode))
 916                                 f2fs_drop_inmem_pages(inode);
 917
 918                         /* should remain fi->extent_tree for writepage */
 919                         f2fs_destroy_extent_node(inode);
 920
 921                         sb_start_intwrite(inode->i_sb);
 922                         f2fs_i_size_write(inode, 0);
 923
 924                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
 925                                         inode, NULL, 0, DATA);
 926                         truncate_inode_pages_final(inode->i_mapping);
 927
 928                         if (F2FS_HAS_BLOCKS(inode))
 929                                 f2fs_truncate(inode);
 930
 931                         sb_end_intwrite(inode->i_sb);
 932
 933                         spin_lock(&inode->i_lock);
 934                         atomic_dec(&inode->i_count);
 935                 }
 936                 trace_f2fs_drop_inode(inode, 0);
 937                 return 0;
 938         }
 939         ret = generic_drop_inode(inode);
 940         trace_f2fs_drop_inode(inode, ret);
 941         return ret;
 942 }
 943
 944 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 945 {
 946         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 947         int ret = 0;
 948
 949         spin_lock(&sbi->inode_lock[DIRTY_META]);
 950         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 951                 ret = 1;
 952         } else {
 953                 set_inode_flag(inode, FI_DIRTY_INODE);
 954                 stat_inc_dirty_inode(sbi, DIRTY_META);
 955         }
 956         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 957                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 958                                 &sbi->inode_list[DIRTY_META]);
 959                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 960         }
 961         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 962         return ret;
 963 }
 964
 965 void f2fs_inode_synced(struct inode *inode)
 966 {
 967         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 968
 969         spin_lock(&sbi->inode_lock[DIRTY_META]);
 970         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 971                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 972                 return;
 973         }
 974         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 975                 list_del_init(&F2FS_I(inode)->gdirty_list);
 976                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 977         }
 978         clear_inode_flag(inode, FI_DIRTY_INODE);
 979         clear_inode_flag(inode, FI_AUTO_RECOVER);
 980         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 981         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 982 }
 983
 984 /*
 985  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 986  *
 987  * We should call set_dirty_inode to write the dirty inode through write_inode.
 988  */
 989 static void f2fs_dirty_inode(struct inode *inode, int flags)
 990 {
 991         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 992
 993         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 994                         inode->i_ino == F2FS_META_INO(sbi))
 995                 return;
 996
 997         if (flags == I_DIRTY_TIME)
 998                 return;
 999
1000         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1001                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1002
1003         f2fs_inode_dirtied(inode, false);
1004 }
1005
1006 static void f2fs_i_callback(struct rcu_head *head)
1007 {
1008         struct inode *inode = container_of(head, struct inode, i_rcu);
1009         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1010 }
1011
1012 static void f2fs_destroy_inode(struct inode *inode)
1013 {
1014         call_rcu(&inode->i_rcu, f2fs_i_callback);
1015 }
1016
1017 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1018 {
1019         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1020         percpu_counter_destroy(&sbi->total_valid_inode_count);
1021 }
1022
1023 static void destroy_device_list(struct f2fs_sb_info *sbi)
1024 {
1025         int i;
1026
1027         for (i = 0; i < sbi->s_ndevs; i++) {
1028                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1029 #ifdef CONFIG_BLK_DEV_ZONED
1030                 kvfree(FDEV(i).blkz_type);
1031 #endif
1032         }
1033         kvfree(sbi->devs);
1034 }
1035
1036 static void f2fs_put_super(struct super_block *sb)
1037 {
1038         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1039         int i;
1040         bool dropped;
1041
1042         f2fs_quota_off_umount(sb);
1043
1044         /* prevent remaining shrinker jobs */
1045         mutex_lock(&sbi->umount_mutex);
1046
1047         /*
1048          * We don't need to do checkpoint when superblock is clean.
1049          * But, the previous checkpoint was not done by umount, it needs to do
1050          * clean checkpoint again.
1051          */
1052         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1053                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1054                 struct cp_control cpc = {
1055                         .reason = CP_UMOUNT,
1056                 };
1057                 f2fs_write_checkpoint(sbi, &cpc);
1058         }
1059
1060         /* be sure to wait for any on-going discard commands */
1061         dropped = f2fs_issue_discard_timeout(sbi);
1062
1063         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1064                                         !sbi->discard_blks && !dropped) {
1065                 struct cp_control cpc = {
1066                         .reason = CP_UMOUNT | CP_TRIMMED,
1067                 };
1068                 f2fs_write_checkpoint(sbi, &cpc);
1069         }
1070
1071         /*
1072          * normally superblock is clean, so we need to release this.
1073          * In addition, EIO will skip do checkpoint, we need this as well.
1074          */
1075         f2fs_release_ino_entry(sbi, true);
1076
1077         f2fs_leave_shrinker(sbi);
1078         mutex_unlock(&sbi->umount_mutex);
1079
1080         /* our cp_error case, we can wait for any writeback page */
1081         f2fs_flush_merged_writes(sbi);
1082
1083         f2fs_wait_on_all_pages_writeback(sbi);
1084
1085         f2fs_bug_on(sbi, sbi->fsync_node_num);
1086
1087         iput(sbi->node_inode);
1088         sbi->node_inode = NULL;
1089
1090         iput(sbi->meta_inode);
1091         sbi->meta_inode = NULL;
1092
1093         /*
1094          * iput() can update stat information, if f2fs_write_checkpoint()
1095          * above failed with error.
1096          */
1097         f2fs_destroy_stats(sbi);
1098
1099         /* destroy f2fs internal modules */
1100         f2fs_destroy_node_manager(sbi);
1101         f2fs_destroy_segment_manager(sbi);
1102
1103         kvfree(sbi->ckpt);
1104
1105         f2fs_unregister_sysfs(sbi);
1106
1107         sb->s_fs_info = NULL;
1108         if (sbi->s_chksum_driver)
1109                 crypto_free_shash(sbi->s_chksum_driver);
1110         kvfree(sbi->raw_super);
1111
1112         destroy_device_list(sbi);
1113         mempool_destroy(sbi->write_io_dummy);
1114 #ifdef CONFIG_QUOTA
1115         for (i = 0; i < MAXQUOTAS; i++)
1116                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1117 #endif
1118         destroy_percpu_info(sbi);
1119         for (i = 0; i < NR_PAGE_TYPE; i++)
1120                 kvfree(sbi->write_io[i]);
1121         kvfree(sbi);
1122 }
1123
1124 int f2fs_sync_fs(struct super_block *sb, int sync)
1125 {
1126         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1127         int err = 0;
1128
1129         if (unlikely(f2fs_cp_error(sbi)))
1130                 return 0;
1131         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1132                 return 0;
1133
1134         trace_f2fs_sync_fs(sb, sync);
1135
1136         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1137                 return -EAGAIN;
1138
1139         if (sync) {
1140                 struct cp_control cpc;
1141
1142                 cpc.reason = __get_cp_reason(sbi);
1143
1144                 mutex_lock(&sbi->gc_mutex);
1145                 err = f2fs_write_checkpoint(sbi, &cpc);
1146                 mutex_unlock(&sbi->gc_mutex);
1147         }
1148         f2fs_trace_ios(NULL, 1);
1149
1150         return err;
1151 }
1152
1153 static int f2fs_freeze(struct super_block *sb)
1154 {
1155         if (f2fs_readonly(sb))
1156                 return 0;
1157
1158         /* IO error happened before */
1159         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1160                 return -EIO;
1161
1162         /* must be clean, since sync_filesystem() was already called */
1163         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1164                 return -EINVAL;
1165         return 0;
1166 }
1167
1168 static int f2fs_unfreeze(struct super_block *sb)
1169 {
1170         return 0;
1171 }
1172
1173 #ifdef CONFIG_QUOTA
1174 static int f2fs_statfs_project(struct super_block *sb,
1175                                 kprojid_t projid, struct kstatfs *buf)
1176 {
1177         struct kqid qid;
1178         struct dquot *dquot;
1179         u64 limit;
1180         u64 curblock;
1181
1182         qid = make_kqid_projid(projid);
1183         dquot = dqget(sb, qid);
1184         if (IS_ERR(dquot))
1185                 return PTR_ERR(dquot);
1186         spin_lock(&dquot->dq_dqb_lock);
1187
1188         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1189                  dquot->dq_dqb.dqb_bsoftlimit :
1190                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1191         if (limit && buf->f_blocks > limit) {
1192                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1193                 buf->f_blocks = limit;
1194                 buf->f_bfree = buf->f_bavail =
1195                         (buf->f_blocks > curblock) ?
1196                          (buf->f_blocks - curblock) : 0;
1197         }
1198
1199         limit = dquot->dq_dqb.dqb_isoftlimit ?
1200                 dquot->dq_dqb.dqb_isoftlimit :
1201                 dquot->dq_dqb.dqb_ihardlimit;
1202         if (limit && buf->f_files > limit) {
1203                 buf->f_files = limit;
1204                 buf->f_ffree =
1205                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1206                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1207         }
1208
1209         spin_unlock(&dquot->dq_dqb_lock);
1210         dqput(dquot);
1211         return 0;
1212 }
1213 #endif
1214
1215 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1216 {
1217         struct super_block *sb = dentry->d_sb;
1218         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1219         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1220         block_t total_count, user_block_count, start_count;
1221         u64 avail_node_count;
1222
1223         total_count = le64_to_cpu(sbi->raw_super->block_count);
1224         user_block_count = sbi->user_block_count;
1225         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1226         buf->f_type = F2FS_SUPER_MAGIC;
1227         buf->f_bsize = sbi->blocksize;
1228
1229         buf->f_blocks = total_count - start_count;
1230         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1231                                                 sbi->current_reserved_blocks;
1232         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1233                 buf->f_bfree = 0;
1234         else
1235                 buf->f_bfree -= sbi->unusable_block_count;
1236
1237         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1238                 buf->f_bavail = buf->f_bfree -
1239                                 F2FS_OPTION(sbi).root_reserved_blocks;
1240         else
1241                 buf->f_bavail = 0;
1242
1243         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1244                                                 F2FS_RESERVED_NODE_NUM;
1245
1246         if (avail_node_count > user_block_count) {
1247                 buf->f_files = user_block_count;
1248                 buf->f_ffree = buf->f_bavail;
1249         } else {
1250                 buf->f_files = avail_node_count;
1251                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1252                                         buf->f_bavail);
1253         }
1254
1255         buf->f_namelen = F2FS_NAME_LEN;
1256         buf->f_fsid.val[0] = (u32)id;
1257         buf->f_fsid.val[1] = (u32)(id >> 32);
1258
1259 #ifdef CONFIG_QUOTA
1260         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1261                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1262                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1263         }
1264 #endif
1265         return 0;
1266 }
1267
1268 static inline void f2fs_show_quota_options(struct seq_file *seq,
1269                                            struct super_block *sb)
1270 {
1271 #ifdef CONFIG_QUOTA
1272         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1273
1274         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1275                 char *fmtname = "";
1276
1277                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1278                 case QFMT_VFS_OLD:
1279                         fmtname = "vfsold";
1280                         break;
1281                 case QFMT_VFS_V0:
1282                         fmtname = "vfsv0";
1283                         break;
1284                 case QFMT_VFS_V1:
1285                         fmtname = "vfsv1";
1286                         break;
1287                 }
1288                 seq_printf(seq, ",jqfmt=%s", fmtname);
1289         }
1290
1291         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1292                 seq_show_option(seq, "usrjquota",
1293                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1294
1295         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1296                 seq_show_option(seq, "grpjquota",
1297                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1298
1299         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1300                 seq_show_option(seq, "prjjquota",
1301                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1302 #endif
1303 }
1304
1305 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1306 {
1307         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1308
1309         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1310                 if (test_opt(sbi, FORCE_FG_GC))
1311                         seq_printf(seq, ",background_gc=%s", "sync");
1312                 else
1313                         seq_printf(seq, ",background_gc=%s", "on");
1314         } else {
1315                 seq_printf(seq, ",background_gc=%s", "off");
1316         }
1317         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1318                 seq_puts(seq, ",disable_roll_forward");
1319         if (test_opt(sbi, DISCARD))
1320                 seq_puts(seq, ",discard");
1321         if (test_opt(sbi, NOHEAP))
1322                 seq_puts(seq, ",no_heap");
1323         else
1324                 seq_puts(seq, ",heap");
1325 #ifdef CONFIG_F2FS_FS_XATTR
1326         if (test_opt(sbi, XATTR_USER))
1327                 seq_puts(seq, ",user_xattr");
1328         else
1329                 seq_puts(seq, ",nouser_xattr");
1330         if (test_opt(sbi, INLINE_XATTR))
1331                 seq_puts(seq, ",inline_xattr");
1332         else
1333                 seq_puts(seq, ",noinline_xattr");
1334         if (test_opt(sbi, INLINE_XATTR_SIZE))
1335                 seq_printf(seq, ",inline_xattr_size=%u",
1336                                         F2FS_OPTION(sbi).inline_xattr_size);
1337 #endif
1338 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1339         if (test_opt(sbi, POSIX_ACL))
1340                 seq_puts(seq, ",acl");
1341         else
1342                 seq_puts(seq, ",noacl");
1343 #endif
1344         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1345                 seq_puts(seq, ",disable_ext_identify");
1346         if (test_opt(sbi, INLINE_DATA))
1347                 seq_puts(seq, ",inline_data");
1348         else
1349                 seq_puts(seq, ",noinline_data");
1350         if (test_opt(sbi, INLINE_DENTRY))
1351                 seq_puts(seq, ",inline_dentry");
1352         else
1353                 seq_puts(seq, ",noinline_dentry");
1354         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1355                 seq_puts(seq, ",flush_merge");
1356         if (test_opt(sbi, NOBARRIER))
1357                 seq_puts(seq, ",nobarrier");
1358         if (test_opt(sbi, FASTBOOT))
1359                 seq_puts(seq, ",fastboot");
1360         if (test_opt(sbi, EXTENT_CACHE))
1361                 seq_puts(seq, ",extent_cache");
1362         else
1363                 seq_puts(seq, ",noextent_cache");
1364         if (test_opt(sbi, DATA_FLUSH))
1365                 seq_puts(seq, ",data_flush");
1366
1367         seq_puts(seq, ",mode=");
1368         if (test_opt(sbi, ADAPTIVE))
1369                 seq_puts(seq, "adaptive");
1370         else if (test_opt(sbi, LFS))
1371                 seq_puts(seq, "lfs");
1372         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1373         if (test_opt(sbi, RESERVE_ROOT))
1374                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1375                                 F2FS_OPTION(sbi).root_reserved_blocks,
1376                                 from_kuid_munged(&init_user_ns,
1377                                         F2FS_OPTION(sbi).s_resuid),
1378                                 from_kgid_munged(&init_user_ns,
1379                                         F2FS_OPTION(sbi).s_resgid));
1380         if (F2FS_IO_SIZE_BITS(sbi))
1381                 seq_printf(seq, ",io_bits=%u",
1382                                 F2FS_OPTION(sbi).write_io_size_bits);
1383 #ifdef CONFIG_F2FS_FAULT_INJECTION
1384         if (test_opt(sbi, FAULT_INJECTION)) {
1385                 seq_printf(seq, ",fault_injection=%u",
1386                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1387                 seq_printf(seq, ",fault_type=%u",
1388                                 F2FS_OPTION(sbi).fault_info.inject_type);
1389         }
1390 #endif
1391 #ifdef CONFIG_QUOTA
1392         if (test_opt(sbi, QUOTA))
1393                 seq_puts(seq, ",quota");
1394         if (test_opt(sbi, USRQUOTA))
1395                 seq_puts(seq, ",usrquota");
1396         if (test_opt(sbi, GRPQUOTA))
1397                 seq_puts(seq, ",grpquota");
1398         if (test_opt(sbi, PRJQUOTA))
1399                 seq_puts(seq, ",prjquota");
1400 #endif
1401         f2fs_show_quota_options(seq, sbi->sb);
1402         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1403                 seq_printf(seq, ",whint_mode=%s", "user-based");
1404         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1405                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1406 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1407         if (F2FS_OPTION(sbi).test_dummy_encryption)
1408                 seq_puts(seq, ",test_dummy_encryption");
1409 #endif
1410
1411         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1412                 seq_printf(seq, ",alloc_mode=%s", "default");
1413         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1414                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1415
1416         if (test_opt(sbi, DISABLE_CHECKPOINT))
1417                 seq_puts(seq, ",checkpoint=disable");
1418
1419         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1420                 seq_printf(seq, ",fsync_mode=%s", "posix");
1421         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1422                 seq_printf(seq, ",fsync_mode=%s", "strict");
1423         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1424                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1425         return 0;
1426 }
1427
1428 static void default_options(struct f2fs_sb_info *sbi)
1429 {
1430         /* init some FS parameters */
1431         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1432         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1433         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1434         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1435         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1436         F2FS_OPTION(sbi).test_dummy_encryption = false;
1437         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1438         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1439
1440         set_opt(sbi, BG_GC);
1441         set_opt(sbi, INLINE_XATTR);
1442         set_opt(sbi, INLINE_DATA);
1443         set_opt(sbi, INLINE_DENTRY);
1444         set_opt(sbi, EXTENT_CACHE);
1445         set_opt(sbi, NOHEAP);
1446         clear_opt(sbi, DISABLE_CHECKPOINT);
1447         sbi->sb->s_flags |= SB_LAZYTIME;
1448         set_opt(sbi, FLUSH_MERGE);
1449         set_opt(sbi, DISCARD);
1450         if (f2fs_sb_has_blkzoned(sbi))
1451                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1452         else
1453                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1454
1455 #ifdef CONFIG_F2FS_FS_XATTR
1456         set_opt(sbi, XATTR_USER);
1457 #endif
1458 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1459         set_opt(sbi, POSIX_ACL);
1460 #endif
1461
1462         f2fs_build_fault_attr(sbi, 0, 0);
1463 }
1464
1465 #ifdef CONFIG_QUOTA
1466 static int f2fs_enable_quotas(struct super_block *sb);
1467 #endif
1468
1469 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1470 {
1471         unsigned int s_flags = sbi->sb->s_flags;
1472         struct cp_control cpc;
1473         int err = 0;
1474         int ret;
1475
1476         if (s_flags & SB_RDONLY) {
1477                 f2fs_msg(sbi->sb, KERN_ERR,
1478                                 "checkpoint=disable on readonly fs");
1479                 return -EINVAL;
1480         }
1481         sbi->sb->s_flags |= SB_ACTIVE;
1482
1483         f2fs_update_time(sbi, DISABLE_TIME);
1484
1485         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1486                 mutex_lock(&sbi->gc_mutex);
1487                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1488                 if (err == -ENODATA) {
1489                         err = 0;
1490                         break;
1491                 }
1492                 if (err && err != -EAGAIN)
1493                         break;
1494         }
1495
1496         ret = sync_filesystem(sbi->sb);
1497         if (ret || err) {
1498                 err = ret ? ret: err;
1499                 goto restore_flag;
1500         }
1501
1502         if (f2fs_disable_cp_again(sbi)) {
1503                 err = -EAGAIN;
1504                 goto restore_flag;
1505         }
1506
1507         mutex_lock(&sbi->gc_mutex);
1508         cpc.reason = CP_PAUSE;
1509         set_sbi_flag(sbi, SBI_CP_DISABLED);
1510         f2fs_write_checkpoint(sbi, &cpc);
1511
1512         sbi->unusable_block_count = 0;
1513         mutex_unlock(&sbi->gc_mutex);
1514 restore_flag:
1515         sbi->sb->s_flags = s_flags;     /* Restore MS_RDONLY status */
1516         return err;
1517 }
1518
1519 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1520 {
1521         mutex_lock(&sbi->gc_mutex);
1522         f2fs_dirty_to_prefree(sbi);
1523
1524         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1525         set_sbi_flag(sbi, SBI_IS_DIRTY);
1526         mutex_unlock(&sbi->gc_mutex);
1527
1528         f2fs_sync_fs(sbi->sb, 1);
1529 }
1530
1531 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1532 {
1533         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1534         struct f2fs_mount_info org_mount_opt;
1535         unsigned long old_sb_flags;
1536         int err;
1537         bool need_restart_gc = false;
1538         bool need_stop_gc = false;
1539         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1540         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1541         bool checkpoint_changed;
1542 #ifdef CONFIG_QUOTA
1543         int i, j;
1544 #endif
1545
1546         /*
1547          * Save the old mount options in case we
1548          * need to restore them.
1549          */
1550         org_mount_opt = sbi->mount_opt;
1551         old_sb_flags = sb->s_flags;
1552
1553 #ifdef CONFIG_QUOTA
1554         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1555         for (i = 0; i < MAXQUOTAS; i++) {
1556                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1557                         org_mount_opt.s_qf_names[i] =
1558                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1559                                 GFP_KERNEL);
1560                         if (!org_mount_opt.s_qf_names[i]) {
1561                                 for (j = 0; j < i; j++)
1562                                         kvfree(org_mount_opt.s_qf_names[j]);
1563                                 return -ENOMEM;
1564                         }
1565                 } else {
1566                         org_mount_opt.s_qf_names[i] = NULL;
1567                 }
1568         }
1569 #endif
1570
1571         /* recover superblocks we couldn't write due to previous RO mount */
1572         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1573                 err = f2fs_commit_super(sbi, false);
1574                 f2fs_msg(sb, KERN_INFO,
1575                         "Try to recover all the superblocks, ret: %d", err);
1576                 if (!err)
1577                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1578         }
1579
1580         default_options(sbi);
1581
1582         /* parse mount options */
1583         err = parse_options(sb, data);
1584         if (err)
1585                 goto restore_opts;
1586         checkpoint_changed =
1587                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1588
1589         /*
1590          * Previous and new state of filesystem is RO,
1591          * so skip checking GC and FLUSH_MERGE conditions.
1592          */
1593         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1594                 goto skip;
1595
1596 #ifdef CONFIG_QUOTA
1597         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1598                 err = dquot_suspend(sb, -1);
1599                 if (err < 0)
1600                         goto restore_opts;
1601         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1602                 /* dquot_resume needs RW */
1603                 sb->s_flags &= ~SB_RDONLY;
1604                 if (sb_any_quota_suspended(sb)) {
1605                         dquot_resume(sb, -1);
1606                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1607                         err = f2fs_enable_quotas(sb);
1608                         if (err)
1609                                 goto restore_opts;
1610                 }
1611         }
1612 #endif
1613         /* disallow enable/disable extent_cache dynamically */
1614         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1615                 err = -EINVAL;
1616                 f2fs_msg(sbi->sb, KERN_WARNING,
1617                                 "switch extent_cache option is not allowed");
1618                 goto restore_opts;
1619         }
1620
1621         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1622                 err = -EINVAL;
1623                 f2fs_msg(sbi->sb, KERN_WARNING,
1624                         "disabling checkpoint not compatible with read-only");
1625                 goto restore_opts;
1626         }
1627
1628         /*
1629          * We stop the GC thread if FS is mounted as RO
1630          * or if background_gc = off is passed in mount
1631          * option. Also sync the filesystem.
1632          */
1633         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1634                 if (sbi->gc_thread) {
1635                         f2fs_stop_gc_thread(sbi);
1636                         need_restart_gc = true;
1637                 }
1638         } else if (!sbi->gc_thread) {
1639                 err = f2fs_start_gc_thread(sbi);
1640                 if (err)
1641                         goto restore_opts;
1642                 need_stop_gc = true;
1643         }
1644
1645         if (*flags & SB_RDONLY ||
1646                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1647                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1648                 sync_inodes_sb(sb);
1649
1650                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1651                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1652                 f2fs_sync_fs(sb, 1);
1653                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1654         }
1655
1656         if (checkpoint_changed) {
1657                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1658                         err = f2fs_disable_checkpoint(sbi);
1659                         if (err)
1660                                 goto restore_gc;
1661                 } else {
1662                         f2fs_enable_checkpoint(sbi);
1663                 }
1664         }
1665
1666         /*
1667          * We stop issue flush thread if FS is mounted as RO
1668          * or if flush_merge is not passed in mount option.
1669          */
1670         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1671                 clear_opt(sbi, FLUSH_MERGE);
1672                 f2fs_destroy_flush_cmd_control(sbi, false);
1673         } else {
1674                 err = f2fs_create_flush_cmd_control(sbi);
1675                 if (err)
1676                         goto restore_gc;
1677         }
1678 skip:
1679 #ifdef CONFIG_QUOTA
1680         /* Release old quota file names */
1681         for (i = 0; i < MAXQUOTAS; i++)
1682                 kvfree(org_mount_opt.s_qf_names[i]);
1683 #endif
1684         /* Update the POSIXACL Flag */
1685         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1686                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1687
1688         limit_reserve_root(sbi);
1689         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1690         return 0;
1691 restore_gc:
1692         if (need_restart_gc) {
1693                 if (f2fs_start_gc_thread(sbi))
1694                         f2fs_msg(sbi->sb, KERN_WARNING,
1695                                 "background gc thread has stopped");
1696         } else if (need_stop_gc) {
1697                 f2fs_stop_gc_thread(sbi);
1698         }
1699 restore_opts:
1700 #ifdef CONFIG_QUOTA
1701         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1702         for (i = 0; i < MAXQUOTAS; i++) {
1703                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1704                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1705         }
1706 #endif
1707         sbi->mount_opt = org_mount_opt;
1708         sb->s_flags = old_sb_flags;
1709         return err;
1710 }
1711
1712 #ifdef CONFIG_QUOTA
1713 /* Read data from quotafile */
1714 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1715                                size_t len, loff_t off)
1716 {
1717         struct inode *inode = sb_dqopt(sb)->files[type];
1718         struct address_space *mapping = inode->i_mapping;
1719         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1720         int offset = off & (sb->s_blocksize - 1);
1721         int tocopy;
1722         size_t toread;
1723         loff_t i_size = i_size_read(inode);
1724         struct page *page;
1725         char *kaddr;
1726
1727         if (off > i_size)
1728                 return 0;
1729
1730         if (off + len > i_size)
1731                 len = i_size - off;
1732         toread = len;
1733         while (toread > 0) {
1734                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1735 repeat:
1736                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1737                 if (IS_ERR(page)) {
1738                         if (PTR_ERR(page) == -ENOMEM) {
1739                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1740                                 goto repeat;
1741                         }
1742                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1743                         return PTR_ERR(page);
1744                 }
1745
1746                 lock_page(page);
1747
1748                 if (unlikely(page->mapping != mapping)) {
1749                         f2fs_put_page(page, 1);
1750                         goto repeat;
1751                 }
1752                 if (unlikely(!PageUptodate(page))) {
1753                         f2fs_put_page(page, 1);
1754                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1755                         return -EIO;
1756                 }
1757
1758                 kaddr = kmap_atomic(page);
1759                 memcpy(data, kaddr + offset, tocopy);
1760                 kunmap_atomic(kaddr);
1761                 f2fs_put_page(page, 1);
1762
1763                 offset = 0;
1764                 toread -= tocopy;
1765                 data += tocopy;
1766                 blkidx++;
1767         }
1768         return len;
1769 }
1770
1771 /* Write to quotafile */
1772 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1773                                 const char *data, size_t len, loff_t off)
1774 {
1775         struct inode *inode = sb_dqopt(sb)->files[type];
1776         struct address_space *mapping = inode->i_mapping;
1777         const struct address_space_operations *a_ops = mapping->a_ops;
1778         int offset = off & (sb->s_blocksize - 1);
1779         size_t towrite = len;
1780         struct page *page;
1781         char *kaddr;
1782         int err = 0;
1783         int tocopy;
1784
1785         while (towrite > 0) {
1786                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1787                                                                 towrite);
1788 retry:
1789                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1790                                                         &page, NULL);
1791                 if (unlikely(err)) {
1792                         if (err == -ENOMEM) {
1793                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1794                                 goto retry;
1795                         }
1796                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1797                         break;
1798                 }
1799
1800                 kaddr = kmap_atomic(page);
1801                 memcpy(kaddr + offset, data, tocopy);
1802                 kunmap_atomic(kaddr);
1803                 flush_dcache_page(page);
1804
1805                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1806                                                 page, NULL);
1807                 offset = 0;
1808                 towrite -= tocopy;
1809                 off += tocopy;
1810                 data += tocopy;
1811                 cond_resched();
1812         }
1813
1814         if (len == towrite)
1815                 return err;
1816         inode->i_mtime = inode->i_ctime = current_time(inode);
1817         f2fs_mark_inode_dirty_sync(inode, false);
1818         return len - towrite;
1819 }
1820
1821 static struct dquot **f2fs_get_dquots(struct inode *inode)
1822 {
1823         return F2FS_I(inode)->i_dquot;
1824 }
1825
1826 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1827 {
1828         return &F2FS_I(inode)->i_reserved_quota;
1829 }
1830
1831 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1832 {
1833         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1834                 f2fs_msg(sbi->sb, KERN_ERR,
1835                         "quota sysfile may be corrupted, skip loading it");
1836                 return 0;
1837         }
1838
1839         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1840                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1841 }
1842
1843 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1844 {
1845         int enabled = 0;
1846         int i, err;
1847
1848         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1849                 err = f2fs_enable_quotas(sbi->sb);
1850                 if (err) {
1851                         f2fs_msg(sbi->sb, KERN_ERR,
1852                                         "Cannot turn on quota_ino: %d", err);
1853                         return 0;
1854                 }
1855                 return 1;
1856         }
1857
1858         for (i = 0; i < MAXQUOTAS; i++) {
1859                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1860                         err = f2fs_quota_on_mount(sbi, i);
1861                         if (!err) {
1862                                 enabled = 1;
1863                                 continue;
1864                         }
1865                         f2fs_msg(sbi->sb, KERN_ERR,
1866                                 "Cannot turn on quotas: %d on %d", err, i);
1867                 }
1868         }
1869         return enabled;
1870 }
1871
1872 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1873                              unsigned int flags)
1874 {
1875         struct inode *qf_inode;
1876         unsigned long qf_inum;
1877         int err;
1878
1879         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1880
1881         qf_inum = f2fs_qf_ino(sb, type);
1882         if (!qf_inum)
1883                 return -EPERM;
1884
1885         qf_inode = f2fs_iget(sb, qf_inum);
1886         if (IS_ERR(qf_inode)) {
1887                 f2fs_msg(sb, KERN_ERR,
1888                         "Bad quota inode %u:%lu", type, qf_inum);
1889                 return PTR_ERR(qf_inode);
1890         }
1891
1892         /* Don't account quota for quota files to avoid recursion */
1893         qf_inode->i_flags |= S_NOQUOTA;
1894         err = dquot_enable(qf_inode, type, format_id, flags);
1895         iput(qf_inode);
1896         return err;
1897 }
1898
1899 static int f2fs_enable_quotas(struct super_block *sb)
1900 {
1901         int type, err = 0;
1902         unsigned long qf_inum;
1903         bool quota_mopt[MAXQUOTAS] = {
1904                 test_opt(F2FS_SB(sb), USRQUOTA),
1905                 test_opt(F2FS_SB(sb), GRPQUOTA),
1906                 test_opt(F2FS_SB(sb), PRJQUOTA),
1907         };
1908
1909         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1910                 f2fs_msg(sb, KERN_ERR,
1911                         "quota file may be corrupted, skip loading it");
1912                 return 0;
1913         }
1914
1915         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1916
1917         for (type = 0; type < MAXQUOTAS; type++) {
1918                 qf_inum = f2fs_qf_ino(sb, type);
1919                 if (qf_inum) {
1920                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1921                                 DQUOT_USAGE_ENABLED |
1922                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1923                         if (err) {
1924                                 f2fs_msg(sb, KERN_ERR,
1925                                         "Failed to enable quota tracking "
1926                                         "(type=%d, err=%d). Please run "
1927                                         "fsck to fix.", type, err);
1928                                 for (type--; type >= 0; type--)
1929                                         dquot_quota_off(sb, type);
1930                                 set_sbi_flag(F2FS_SB(sb),
1931                                                 SBI_QUOTA_NEED_REPAIR);
1932                                 return err;
1933                         }
1934                 }
1935         }
1936         return 0;
1937 }
1938
1939 int f2fs_quota_sync(struct super_block *sb, int type)
1940 {
1941         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1942         struct quota_info *dqopt = sb_dqopt(sb);
1943         int cnt;
1944         int ret;
1945
1946         ret = dquot_writeback_dquots(sb, type);
1947         if (ret)
1948                 goto out;
1949
1950         /*
1951          * Now when everything is written we can discard the pagecache so
1952          * that userspace sees the changes.
1953          */
1954         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1955                 struct address_space *mapping;
1956
1957                 if (type != -1 && cnt != type)
1958                         continue;
1959                 if (!sb_has_quota_active(sb, cnt))
1960                         continue;
1961
1962                 mapping = dqopt->files[cnt]->i_mapping;
1963
1964                 ret = filemap_fdatawrite(mapping);
1965                 if (ret)
1966                         goto out;
1967
1968                 /* if we are using journalled quota */
1969                 if (is_journalled_quota(sbi))
1970                         continue;
1971
1972                 ret = filemap_fdatawait(mapping);
1973                 if (ret)
1974                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1975
1976                 inode_lock(dqopt->files[cnt]);
1977                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1978                 inode_unlock(dqopt->files[cnt]);
1979         }
1980 out:
1981         if (ret)
1982                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1983         return ret;
1984 }
1985
1986 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1987                                                         const struct path *path)
1988 {
1989         struct inode *inode;
1990         int err;
1991
1992         err = f2fs_quota_sync(sb, type);
1993         if (err)
1994                 return err;
1995
1996         err = dquot_quota_on(sb, type, format_id, path);
1997         if (err)
1998                 return err;
1999
2000         inode = d_inode(path->dentry);
2001
2002         inode_lock(inode);
2003         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2004         f2fs_set_inode_flags(inode);
2005         inode_unlock(inode);
2006         f2fs_mark_inode_dirty_sync(inode, false);
2007
2008         return 0;
2009 }
2010
2011 static int f2fs_quota_off(struct super_block *sb, int type)
2012 {
2013         struct inode *inode = sb_dqopt(sb)->files[type];
2014         int err;
2015
2016         if (!inode || !igrab(inode))
2017                 return dquot_quota_off(sb, type);
2018
2019         err = f2fs_quota_sync(sb, type);
2020         if (err)
2021                 goto out_put;
2022
2023         err = dquot_quota_off(sb, type);
2024         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2025                 goto out_put;
2026
2027         inode_lock(inode);
2028         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2029         f2fs_set_inode_flags(inode);
2030         inode_unlock(inode);
2031         f2fs_mark_inode_dirty_sync(inode, false);
2032 out_put:
2033         iput(inode);
2034         return err;
2035 }
2036
2037 void f2fs_quota_off_umount(struct super_block *sb)
2038 {
2039         int type;
2040         int err;
2041
2042         for (type = 0; type < MAXQUOTAS; type++) {
2043                 err = f2fs_quota_off(sb, type);
2044                 if (err) {
2045                         int ret = dquot_quota_off(sb, type);
2046
2047                         f2fs_msg(sb, KERN_ERR,
2048                                 "Fail to turn off disk quota "
2049                                 "(type: %d, err: %d, ret:%d), Please "
2050                                 "run fsck to fix it.", type, err, ret);
2051                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2052                 }
2053         }
2054         /*
2055          * In case of checkpoint=disable, we must flush quota blocks.
2056          * This can cause NULL exception for node_inode in end_io, since
2057          * put_super already dropped it.
2058          */
2059         sync_filesystem(sb);
2060 }
2061
2062 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2063 {
2064         struct quota_info *dqopt = sb_dqopt(sb);
2065         int type;
2066
2067         for (type = 0; type < MAXQUOTAS; type++) {
2068                 if (!dqopt->files[type])
2069                         continue;
2070                 f2fs_inode_synced(dqopt->files[type]);
2071         }
2072 }
2073
2074 static int f2fs_dquot_commit(struct dquot *dquot)
2075 {
2076         int ret;
2077
2078         ret = dquot_commit(dquot);
2079         if (ret < 0)
2080                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2081         return ret;
2082 }
2083
2084 static int f2fs_dquot_acquire(struct dquot *dquot)
2085 {
2086         int ret;
2087
2088         ret = dquot_acquire(dquot);
2089         if (ret < 0)
2090                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2091
2092         return ret;
2093 }
2094
2095 static int f2fs_dquot_release(struct dquot *dquot)
2096 {
2097         int ret;
2098
2099         ret = dquot_release(dquot);
2100         if (ret < 0)
2101                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2102         return ret;
2103 }
2104
2105 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2106 {
2107         struct super_block *sb = dquot->dq_sb;
2108         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2109         int ret;
2110
2111         ret = dquot_mark_dquot_dirty(dquot);
2112
2113         /* if we are using journalled quota */
2114         if (is_journalled_quota(sbi))
2115                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2116
2117         return ret;
2118 }
2119
2120 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2121 {
2122         int ret;
2123
2124         ret = dquot_commit_info(sb, type);
2125         if (ret < 0)
2126                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2127         return ret;
2128 }
2129
2130 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2131 {
2132         *projid = F2FS_I(inode)->i_projid;
2133         return 0;
2134 }
2135
2136 static const struct dquot_operations f2fs_quota_operations = {
2137         .get_reserved_space = f2fs_get_reserved_space,
2138         .write_dquot    = f2fs_dquot_commit,
2139         .acquire_dquot  = f2fs_dquot_acquire,
2140         .release_dquot  = f2fs_dquot_release,
2141         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2142         .write_info     = f2fs_dquot_commit_info,
2143         .alloc_dquot    = dquot_alloc,
2144         .destroy_dquot  = dquot_destroy,
2145         .get_projid     = f2fs_get_projid,
2146         .get_next_id    = dquot_get_next_id,
2147 };
2148
2149 static const struct quotactl_ops f2fs_quotactl_ops = {
2150         .quota_on       = f2fs_quota_on,
2151         .quota_off      = f2fs_quota_off,
2152         .quota_sync     = f2fs_quota_sync,
2153         .get_state      = dquot_get_state,
2154         .set_info       = dquot_set_dqinfo,
2155         .get_dqblk      = dquot_get_dqblk,
2156         .set_dqblk      = dquot_set_dqblk,
2157         .get_nextdqblk  = dquot_get_next_dqblk,
2158 };
2159 #else
2160 int f2fs_quota_sync(struct super_block *sb, int type)
2161 {
2162         return 0;
2163 }
2164
2165 void f2fs_quota_off_umount(struct super_block *sb)
2166 {
2167 }
2168 #endif
2169
2170 static const struct super_operations f2fs_sops = {
2171         .alloc_inode    = f2fs_alloc_inode,
2172         .drop_inode     = f2fs_drop_inode,
2173         .destroy_inode  = f2fs_destroy_inode,
2174         .write_inode    = f2fs_write_inode,
2175         .dirty_inode    = f2fs_dirty_inode,
2176         .show_options   = f2fs_show_options,
2177 #ifdef CONFIG_QUOTA
2178         .quota_read     = f2fs_quota_read,
2179         .quota_write    = f2fs_quota_write,
2180         .get_dquots     = f2fs_get_dquots,
2181 #endif
2182         .evict_inode    = f2fs_evict_inode,
2183         .put_super      = f2fs_put_super,
2184         .sync_fs        = f2fs_sync_fs,
2185         .freeze_fs      = f2fs_freeze,
2186         .unfreeze_fs    = f2fs_unfreeze,
2187         .statfs         = f2fs_statfs,
2188         .remount_fs     = f2fs_remount,
2189 };
2190
2191 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2192 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2193 {
2194         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2195                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2196                                 ctx, len, NULL);
2197 }
2198
2199 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2200                                                         void *fs_data)
2201 {
2202         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2203
2204         /*
2205          * Encrypting the root directory is not allowed because fsck
2206          * expects lost+found directory to exist and remain unencrypted
2207          * if LOST_FOUND feature is enabled.
2208          *
2209          */
2210         if (f2fs_sb_has_lost_found(sbi) &&
2211                         inode->i_ino == F2FS_ROOT_INO(sbi))
2212                 return -EPERM;
2213
2214         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2215                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2216                                 ctx, len, fs_data, XATTR_CREATE);
2217 }
2218
2219 static bool f2fs_dummy_context(struct inode *inode)
2220 {
2221         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2222 }
2223
2224 static const struct fscrypt_operations f2fs_cryptops = {
2225         .key_prefix     = "f2fs:",
2226         .get_context    = f2fs_get_context,
2227         .set_context    = f2fs_set_context,
2228         .dummy_context  = f2fs_dummy_context,
2229         .empty_dir      = f2fs_empty_dir,
2230         .max_namelen    = F2FS_NAME_LEN,
2231 };
2232 #endif
2233
2234 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2235                 u64 ino, u32 generation)
2236 {
2237         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2238         struct inode *inode;
2239
2240         if (f2fs_check_nid_range(sbi, ino))
2241                 return ERR_PTR(-ESTALE);
2242
2243         /*
2244          * f2fs_iget isn't quite right if the inode is currently unallocated!
2245          * However f2fs_iget currently does appropriate checks to handle stale
2246          * inodes so everything is OK.
2247          */
2248         inode = f2fs_iget(sb, ino);
2249         if (IS_ERR(inode))
2250                 return ERR_CAST(inode);
2251         if (unlikely(generation && inode->i_generation != generation)) {
2252                 /* we didn't find the right inode.. */
2253                 iput(inode);
2254                 return ERR_PTR(-ESTALE);
2255         }
2256         return inode;
2257 }
2258
2259 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2260                 int fh_len, int fh_type)
2261 {
2262         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2263                                     f2fs_nfs_get_inode);
2264 }
2265
2266 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2267                 int fh_len, int fh_type)
2268 {
2269         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2270                                     f2fs_nfs_get_inode);
2271 }
2272
2273 static const struct export_operations f2fs_export_ops = {
2274         .fh_to_dentry = f2fs_fh_to_dentry,
2275         .fh_to_parent = f2fs_fh_to_parent,
2276         .get_parent = f2fs_get_parent,
2277 };
2278
2279 static loff_t max_file_blocks(void)
2280 {
2281         loff_t result = 0;
2282         loff_t leaf_count = ADDRS_PER_BLOCK;
2283
2284         /*
2285          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2286          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2287          * space in inode.i_addr, it will be more safe to reassign
2288          * result as zero.
2289          */
2290
2291         /* two direct node blocks */
2292         result += (leaf_count * 2);
2293
2294         /* two indirect node blocks */
2295         leaf_count *= NIDS_PER_BLOCK;
2296         result += (leaf_count * 2);
2297
2298         /* one double indirect node block */
2299         leaf_count *= NIDS_PER_BLOCK;
2300         result += leaf_count;
2301
2302         return result;
2303 }
2304
2305 static int __f2fs_commit_super(struct buffer_head *bh,
2306                         struct f2fs_super_block *super)
2307 {
2308         lock_buffer(bh);
2309         if (super)
2310                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2311         set_buffer_dirty(bh);
2312         unlock_buffer(bh);
2313
2314         /* it's rare case, we can do fua all the time */
2315         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2316 }
2317
2318 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2319                                         struct buffer_head *bh)
2320 {
2321         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2322                                         (bh->b_data + F2FS_SUPER_OFFSET);
2323         struct super_block *sb = sbi->sb;
2324         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2325         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2326         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2327         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2328         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2329         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2330         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2331         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2332         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2333         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2334         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2335         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2336         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2337         u64 main_end_blkaddr = main_blkaddr +
2338                                 (segment_count_main << log_blocks_per_seg);
2339         u64 seg_end_blkaddr = segment0_blkaddr +
2340                                 (segment_count << log_blocks_per_seg);
2341
2342         if (segment0_blkaddr != cp_blkaddr) {
2343                 f2fs_msg(sb, KERN_INFO,
2344                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2345                         segment0_blkaddr, cp_blkaddr);
2346                 return true;
2347         }
2348
2349         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2350                                                         sit_blkaddr) {
2351                 f2fs_msg(sb, KERN_INFO,
2352                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2353                         cp_blkaddr, sit_blkaddr,
2354                         segment_count_ckpt << log_blocks_per_seg);
2355                 return true;
2356         }
2357
2358         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2359                                                         nat_blkaddr) {
2360                 f2fs_msg(sb, KERN_INFO,
2361                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2362                         sit_blkaddr, nat_blkaddr,
2363                         segment_count_sit << log_blocks_per_seg);
2364                 return true;
2365         }
2366
2367         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2368                                                         ssa_blkaddr) {
2369                 f2fs_msg(sb, KERN_INFO,
2370                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2371                         nat_blkaddr, ssa_blkaddr,
2372                         segment_count_nat << log_blocks_per_seg);
2373                 return true;
2374         }
2375
2376         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2377                                                         main_blkaddr) {
2378                 f2fs_msg(sb, KERN_INFO,
2379                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2380                         ssa_blkaddr, main_blkaddr,
2381                         segment_count_ssa << log_blocks_per_seg);
2382                 return true;
2383         }
2384
2385         if (main_end_blkaddr > seg_end_blkaddr) {
2386                 f2fs_msg(sb, KERN_INFO,
2387                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2388                         main_blkaddr,
2389                         segment0_blkaddr +
2390                                 (segment_count << log_blocks_per_seg),
2391                         segment_count_main << log_blocks_per_seg);
2392                 return true;
2393         } else if (main_end_blkaddr < seg_end_blkaddr) {
2394                 int err = 0;
2395                 char *res;
2396
2397                 /* fix in-memory information all the time */
2398                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2399                                 segment0_blkaddr) >> log_blocks_per_seg);
2400
2401                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2402                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2403                         res = "internally";
2404                 } else {
2405                         err = __f2fs_commit_super(bh, NULL);
2406                         res = err ? "failed" : "done";
2407                 }
2408                 f2fs_msg(sb, KERN_INFO,
2409                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
2410                         res, main_blkaddr,
2411                         segment0_blkaddr +
2412                                 (segment_count << log_blocks_per_seg),
2413                         segment_count_main << log_blocks_per_seg);
2414                 if (err)
2415                         return true;
2416         }
2417         return false;
2418 }
2419
2420 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2421                                 struct buffer_head *bh)
2422 {
2423         block_t segment_count, segs_per_sec, secs_per_zone;
2424         block_t total_sections, blocks_per_seg;
2425         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2426                                         (bh->b_data + F2FS_SUPER_OFFSET);
2427         struct super_block *sb = sbi->sb;
2428         unsigned int blocksize;
2429         size_t crc_offset = 0;
2430         __u32 crc = 0;
2431
2432         /* Check checksum_offset and crc in superblock */
2433         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2434                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2435                 if (crc_offset !=
2436                         offsetof(struct f2fs_super_block, crc)) {
2437                         f2fs_msg(sb, KERN_INFO,
2438                                 "Invalid SB checksum offset: %zu",
2439                                 crc_offset);
2440                         return 1;
2441                 }
2442                 crc = le32_to_cpu(raw_super->crc);
2443                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2444                         f2fs_msg(sb, KERN_INFO,
2445                                 "Invalid SB checksum value: %u", crc);
2446                         return 1;
2447                 }
2448         }
2449
2450         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2451                 f2fs_msg(sb, KERN_INFO,
2452                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
2453                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2454                 return 1;
2455         }
2456
2457         /* Currently, support only 4KB page cache size */
2458         if (F2FS_BLKSIZE != PAGE_SIZE) {
2459                 f2fs_msg(sb, KERN_INFO,
2460                         "Invalid page_cache_size (%lu), supports only 4KB\n",
2461                         PAGE_SIZE);
2462                 return 1;
2463         }
2464
2465         /* Currently, support only 4KB block size */
2466         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2467         if (blocksize != F2FS_BLKSIZE) {
2468                 f2fs_msg(sb, KERN_INFO,
2469                         "Invalid blocksize (%u), supports only 4KB\n",
2470                         blocksize);
2471                 return 1;
2472         }
2473
2474         /* check log blocks per segment */
2475         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2476                 f2fs_msg(sb, KERN_INFO,
2477                         "Invalid log blocks per segment (%u)\n",
2478                         le32_to_cpu(raw_super->log_blocks_per_seg));
2479                 return 1;
2480         }
2481
2482         /* Currently, support 512/1024/2048/4096 bytes sector size */
2483         if (le32_to_cpu(raw_super->log_sectorsize) >
2484                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2485                 le32_to_cpu(raw_super->log_sectorsize) <
2486                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2487                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2488                         le32_to_cpu(raw_super->log_sectorsize));
2489                 return 1;
2490         }
2491         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2492                 le32_to_cpu(raw_super->log_sectorsize) !=
2493                         F2FS_MAX_LOG_SECTOR_SIZE) {
2494                 f2fs_msg(sb, KERN_INFO,
2495                         "Invalid log sectors per block(%u) log sectorsize(%u)",
2496                         le32_to_cpu(raw_super->log_sectors_per_block),
2497                         le32_to_cpu(raw_super->log_sectorsize));
2498                 return 1;
2499         }
2500
2501         segment_count = le32_to_cpu(raw_super->segment_count);
2502         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2503         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2504         total_sections = le32_to_cpu(raw_super->section_count);
2505
2506         /* blocks_per_seg should be 512, given the above check */
2507         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2508
2509         if (segment_count > F2FS_MAX_SEGMENT ||
2510                                 segment_count < F2FS_MIN_SEGMENTS) {
2511                 f2fs_msg(sb, KERN_INFO,
2512                         "Invalid segment count (%u)",
2513                         segment_count);
2514                 return 1;
2515         }
2516
2517         if (total_sections > segment_count ||
2518                         total_sections < F2FS_MIN_SEGMENTS ||
2519                         segs_per_sec > segment_count || !segs_per_sec) {
2520                 f2fs_msg(sb, KERN_INFO,
2521                         "Invalid segment/section count (%u, %u x %u)",
2522                         segment_count, total_sections, segs_per_sec);
2523                 return 1;
2524         }
2525
2526         if ((segment_count / segs_per_sec) < total_sections) {
2527                 f2fs_msg(sb, KERN_INFO,
2528                         "Small segment_count (%u < %u * %u)",
2529                         segment_count, segs_per_sec, total_sections);
2530                 return 1;
2531         }
2532
2533         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2534                 f2fs_msg(sb, KERN_INFO,
2535                         "Wrong segment_count / block_count (%u > %llu)",
2536                         segment_count, le64_to_cpu(raw_super->block_count));
2537                 return 1;
2538         }
2539
2540         if (secs_per_zone > total_sections || !secs_per_zone) {
2541                 f2fs_msg(sb, KERN_INFO,
2542                         "Wrong secs_per_zone / total_sections (%u, %u)",
2543                         secs_per_zone, total_sections);
2544                 return 1;
2545         }
2546         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2547                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2548                         (le32_to_cpu(raw_super->extension_count) +
2549                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2550                 f2fs_msg(sb, KERN_INFO,
2551                         "Corrupted extension count (%u + %u > %u)",
2552                         le32_to_cpu(raw_super->extension_count),
2553                         raw_super->hot_ext_count,
2554                         F2FS_MAX_EXTENSION);
2555                 return 1;
2556         }
2557
2558         if (le32_to_cpu(raw_super->cp_payload) >
2559                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2560                 f2fs_msg(sb, KERN_INFO,
2561                         "Insane cp_payload (%u > %u)",
2562                         le32_to_cpu(raw_super->cp_payload),
2563                         blocks_per_seg - F2FS_CP_PACKS);
2564                 return 1;
2565         }
2566
2567         /* check reserved ino info */
2568         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2569                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2570                 le32_to_cpu(raw_super->root_ino) != 3) {
2571                 f2fs_msg(sb, KERN_INFO,
2572                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2573                         le32_to_cpu(raw_super->node_ino),
2574                         le32_to_cpu(raw_super->meta_ino),
2575                         le32_to_cpu(raw_super->root_ino));
2576                 return 1;
2577         }
2578
2579         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2580         if (sanity_check_area_boundary(sbi, bh))
2581                 return 1;
2582
2583         return 0;
2584 }
2585
2586 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2587 {
2588         unsigned int total, fsmeta;
2589         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2590         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2591         unsigned int ovp_segments, reserved_segments;
2592         unsigned int main_segs, blocks_per_seg;
2593         unsigned int sit_segs, nat_segs;
2594         unsigned int sit_bitmap_size, nat_bitmap_size;
2595         unsigned int log_blocks_per_seg;
2596         unsigned int segment_count_main;
2597         unsigned int cp_pack_start_sum, cp_payload;
2598         block_t user_block_count;
2599         int i, j;
2600
2601         total = le32_to_cpu(raw_super->segment_count);
2602         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2603         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2604         fsmeta += sit_segs;
2605         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2606         fsmeta += nat_segs;
2607         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2608         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2609
2610         if (unlikely(fsmeta >= total))
2611                 return 1;
2612
2613         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2614         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2615
2616         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2617                         ovp_segments == 0 || reserved_segments == 0)) {
2618                 f2fs_msg(sbi->sb, KERN_ERR,
2619                         "Wrong layout: check mkfs.f2fs version");
2620                 return 1;
2621         }
2622
2623         user_block_count = le64_to_cpu(ckpt->user_block_count);
2624         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2625         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2626         if (!user_block_count || user_block_count >=
2627                         segment_count_main << log_blocks_per_seg) {
2628                 f2fs_msg(sbi->sb, KERN_ERR,
2629                         "Wrong user_block_count: %u", user_block_count);
2630                 return 1;
2631         }
2632
2633         main_segs = le32_to_cpu(raw_super->segment_count_main);
2634         blocks_per_seg = sbi->blocks_per_seg;
2635
2636         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2637                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2638                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2639                         return 1;
2640                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2641                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2642                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2643                                 f2fs_msg(sbi->sb, KERN_ERR,
2644                                         "Node segment (%u, %u) has the same "
2645                                         "segno: %u", i, j,
2646                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2647                                 return 1;
2648                         }
2649                 }
2650         }
2651         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2652                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2653                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2654                         return 1;
2655                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2656                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2657                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2658                                 f2fs_msg(sbi->sb, KERN_ERR,
2659                                         "Data segment (%u, %u) has the same "
2660                                         "segno: %u", i, j,
2661                                         le32_to_cpu(ckpt->cur_data_segno[i]));
2662                                 return 1;
2663                         }
2664                 }
2665         }
2666         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2667                 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2668                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2669                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2670                                 f2fs_msg(sbi->sb, KERN_ERR,
2671                                         "Data segment (%u) and Data segment (%u)"
2672                                         " has the same segno: %u", i, j,
2673                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2674                                 return 1;
2675                         }
2676                 }
2677         }
2678
2679         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2680         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2681
2682         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2683                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2684                 f2fs_msg(sbi->sb, KERN_ERR,
2685                         "Wrong bitmap size: sit: %u, nat:%u",
2686                         sit_bitmap_size, nat_bitmap_size);
2687                 return 1;
2688         }
2689
2690         cp_pack_start_sum = __start_sum_addr(sbi);
2691         cp_payload = __cp_payload(sbi);
2692         if (cp_pack_start_sum < cp_payload + 1 ||
2693                 cp_pack_start_sum > blocks_per_seg - 1 -
2694                         NR_CURSEG_TYPE) {
2695                 f2fs_msg(sbi->sb, KERN_ERR,
2696                         "Wrong cp_pack_start_sum: %u",
2697                         cp_pack_start_sum);
2698                 return 1;
2699         }
2700
2701         if (unlikely(f2fs_cp_error(sbi))) {
2702                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2703                 return 1;
2704         }
2705         return 0;
2706 }
2707
2708 static void init_sb_info(struct f2fs_sb_info *sbi)
2709 {
2710         struct f2fs_super_block *raw_super = sbi->raw_super;
2711         int i;
2712
2713         sbi->log_sectors_per_block =
2714                 le32_to_cpu(raw_super->log_sectors_per_block);
2715         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2716         sbi->blocksize = 1 << sbi->log_blocksize;
2717         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2718         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2719         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2720         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2721         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2722         sbi->total_node_count =
2723                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2724                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2725         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2726         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2727         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2728         sbi->cur_victim_sec = NULL_SECNO;
2729         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2730         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2731         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2732         sbi->migration_granularity = sbi->segs_per_sec;
2733
2734         sbi->dir_level = DEF_DIR_LEVEL;
2735         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2736         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2737         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2738         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2739         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2740         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2741                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2742         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2743
2744         for (i = 0; i < NR_COUNT_TYPE; i++)
2745                 atomic_set(&sbi->nr_pages[i], 0);
2746
2747         for (i = 0; i < META; i++)
2748                 atomic_set(&sbi->wb_sync_req[i], 0);
2749
2750         INIT_LIST_HEAD(&sbi->s_list);
2751         mutex_init(&sbi->umount_mutex);
2752         init_rwsem(&sbi->io_order_lock);
2753         spin_lock_init(&sbi->cp_lock);
2754
2755         sbi->dirty_device = 0;
2756         spin_lock_init(&sbi->dev_lock);
2757
2758         init_rwsem(&sbi->sb_lock);
2759 }
2760
2761 static int init_percpu_info(struct f2fs_sb_info *sbi)
2762 {
2763         int err;
2764
2765         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2766         if (err)
2767                 return err;
2768
2769         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2770                                                                 GFP_KERNEL);
2771         if (err)
2772                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2773
2774         return err;
2775 }
2776
2777 #ifdef CONFIG_BLK_DEV_ZONED
2778 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2779 {
2780         struct block_device *bdev = FDEV(devi).bdev;
2781         sector_t nr_sectors = bdev->bd_part->nr_sects;
2782         sector_t sector = 0;
2783         struct blk_zone *zones;
2784         unsigned int i, nr_zones;
2785         unsigned int n = 0;
2786         int err = -EIO;
2787
2788         if (!f2fs_sb_has_blkzoned(sbi))
2789                 return 0;
2790
2791         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2792                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2793                 return -EINVAL;
2794         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2795         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2796                                 __ilog2_u32(sbi->blocks_per_blkz))
2797                 return -EINVAL;
2798         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2799         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2800                                         sbi->log_blocks_per_blkz;
2801         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2802                 FDEV(devi).nr_blkz++;
2803
2804         FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2805                                                                 GFP_KERNEL);
2806         if (!FDEV(devi).blkz_type)
2807                 return -ENOMEM;
2808
2809 #define F2FS_REPORT_NR_ZONES   4096
2810
2811         zones = f2fs_kzalloc(sbi,
2812                              array_size(F2FS_REPORT_NR_ZONES,
2813                                         sizeof(struct blk_zone)),
2814                              GFP_KERNEL);
2815         if (!zones)
2816                 return -ENOMEM;
2817
2818         /* Get block zones type */
2819         while (zones && sector < nr_sectors) {
2820
2821                 nr_zones = F2FS_REPORT_NR_ZONES;
2822                 err = blkdev_report_zones(bdev, sector,
2823                                           zones, &nr_zones,
2824                                           GFP_KERNEL);
2825                 if (err)
2826                         break;
2827                 if (!nr_zones) {
2828                         err = -EIO;
2829                         break;
2830                 }
2831
2832                 for (i = 0; i < nr_zones; i++) {
2833                         FDEV(devi).blkz_type[n] = zones[i].type;
2834                         sector += zones[i].len;
2835                         n++;
2836                 }
2837         }
2838
2839         kvfree(zones);
2840
2841         return err;
2842 }
2843 #endif
2844
2845 /*
2846  * Read f2fs raw super block.
2847  * Because we have two copies of super block, so read both of them
2848  * to get the first valid one. If any one of them is broken, we pass
2849  * them recovery flag back to the caller.
2850  */
2851 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2852                         struct f2fs_super_block **raw_super,
2853                         int *valid_super_block, int *recovery)
2854 {
2855         struct super_block *sb = sbi->sb;
2856         int block;
2857         struct buffer_head *bh;
2858         struct f2fs_super_block *super;
2859         int err = 0;
2860
2861         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2862         if (!super)
2863                 return -ENOMEM;
2864
2865         for (block = 0; block < 2; block++) {
2866                 bh = sb_bread(sb, block);
2867                 if (!bh) {
2868                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2869                                 block + 1);
2870                         err = -EIO;
2871                         continue;
2872                 }
2873
2874                 /* sanity checking of raw super */
2875                 if (sanity_check_raw_super(sbi, bh)) {
2876                         f2fs_msg(sb, KERN_ERR,
2877                                 "Can't find valid F2FS filesystem in %dth superblock",
2878                                 block + 1);
2879                         err = -EINVAL;
2880                         brelse(bh);
2881                         continue;
2882                 }
2883
2884                 if (!*raw_super) {
2885                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2886                                                         sizeof(*super));
2887                         *valid_super_block = block;
2888                         *raw_super = super;
2889                 }
2890                 brelse(bh);
2891         }
2892
2893         /* Fail to read any one of the superblocks*/
2894         if (err < 0)
2895                 *recovery = 1;
2896
2897         /* No valid superblock */
2898         if (!*raw_super)
2899                 kvfree(super);
2900         else
2901                 err = 0;
2902
2903         return err;
2904 }
2905
2906 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2907 {
2908         struct buffer_head *bh;
2909         __u32 crc = 0;
2910         int err;
2911
2912         if ((recover && f2fs_readonly(sbi->sb)) ||
2913                                 bdev_read_only(sbi->sb->s_bdev)) {
2914                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2915                 return -EROFS;
2916         }
2917
2918         /* we should update superblock crc here */
2919         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2920                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2921                                 offsetof(struct f2fs_super_block, crc));
2922                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2923         }
2924
2925         /* write back-up superblock first */
2926         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2927         if (!bh)
2928                 return -EIO;
2929         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2930         brelse(bh);
2931
2932         /* if we are in recovery path, skip writing valid superblock */
2933         if (recover || err)
2934                 return err;
2935
2936         /* write current valid superblock */
2937         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2938         if (!bh)
2939                 return -EIO;
2940         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2941         brelse(bh);
2942         return err;
2943 }
2944
2945 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2946 {
2947         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2948         unsigned int max_devices = MAX_DEVICES;
2949         int i;
2950
2951         /* Initialize single device information */
2952         if (!RDEV(0).path[0]) {
2953                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2954                         return 0;
2955                 max_devices = 1;
2956         }
2957
2958         /*
2959          * Initialize multiple devices information, or single
2960          * zoned block device information.
2961          */
2962         sbi->devs = f2fs_kzalloc(sbi,
2963                                  array_size(max_devices,
2964                                             sizeof(struct f2fs_dev_info)),
2965                                  GFP_KERNEL);
2966         if (!sbi->devs)
2967                 return -ENOMEM;
2968
2969         for (i = 0; i < max_devices; i++) {
2970
2971                 if (i > 0 && !RDEV(i).path[0])
2972                         break;
2973
2974                 if (max_devices == 1) {
2975                         /* Single zoned block device mount */
2976                         FDEV(0).bdev =
2977                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2978                                         sbi->sb->s_mode, sbi->sb->s_type);
2979                 } else {
2980                         /* Multi-device mount */
2981                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2982                         FDEV(i).total_segments =
2983                                 le32_to_cpu(RDEV(i).total_segments);
2984                         if (i == 0) {
2985                                 FDEV(i).start_blk = 0;
2986                                 FDEV(i).end_blk = FDEV(i).start_blk +
2987                                     (FDEV(i).total_segments <<
2988                                     sbi->log_blocks_per_seg) - 1 +
2989                                     le32_to_cpu(raw_super->segment0_blkaddr);
2990                         } else {
2991                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2992                                 FDEV(i).end_blk = FDEV(i).start_blk +
2993                                         (FDEV(i).total_segments <<
2994                                         sbi->log_blocks_per_seg) - 1;
2995                         }
2996                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2997                                         sbi->sb->s_mode, sbi->sb->s_type);
2998                 }
2999                 if (IS_ERR(FDEV(i).bdev))
3000                         return PTR_ERR(FDEV(i).bdev);
3001
3002                 /* to release errored devices */
3003                 sbi->s_ndevs = i + 1;
3004
3005 #ifdef CONFIG_BLK_DEV_ZONED
3006                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3007                                 !f2fs_sb_has_blkzoned(sbi)) {
3008                         f2fs_msg(sbi->sb, KERN_ERR,
3009                                 "Zoned block device feature not enabled\n");
3010                         return -EINVAL;
3011                 }
3012                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3013                         if (init_blkz_info(sbi, i)) {
3014                                 f2fs_msg(sbi->sb, KERN_ERR,
3015                                         "Failed to initialize F2FS blkzone information");
3016                                 return -EINVAL;
3017                         }
3018                         if (max_devices == 1)
3019                                 break;
3020                         f2fs_msg(sbi->sb, KERN_INFO,
3021                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3022                                 i, FDEV(i).path,
3023                                 FDEV(i).total_segments,
3024                                 FDEV(i).start_blk, FDEV(i).end_blk,
3025                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3026                                 "Host-aware" : "Host-managed");
3027                         continue;
3028                 }
3029 #endif
3030                 f2fs_msg(sbi->sb, KERN_INFO,
3031                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3032                                 i, FDEV(i).path,
3033                                 FDEV(i).total_segments,
3034                                 FDEV(i).start_blk, FDEV(i).end_blk);
3035         }
3036         f2fs_msg(sbi->sb, KERN_INFO,
3037                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3038         return 0;
3039 }
3040
3041 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3042 {
3043         struct f2fs_sm_info *sm_i = SM_I(sbi);
3044
3045         /* adjust parameters according to the volume size */
3046         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3047                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3048                 sm_i->dcc_info->discard_granularity = 1;
3049                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3050         }
3051
3052         sbi->readdir_ra = 1;
3053 }
3054
3055 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3056 {
3057         struct f2fs_sb_info *sbi;
3058         struct f2fs_super_block *raw_super;
3059         struct inode *root;
3060         int err;
3061         bool skip_recovery = false, need_fsck = false;
3062         char *options = NULL;
3063         int recovery, i, valid_super_block;
3064         struct curseg_info *seg_i;
3065         int retry_cnt = 1;
3066
3067 try_onemore:
3068         err = -EINVAL;
3069         raw_super = NULL;
3070         valid_super_block = -1;
3071         recovery = 0;
3072
3073         /* allocate memory for f2fs-specific super block info */
3074         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3075         if (!sbi)
3076                 return -ENOMEM;
3077
3078         sbi->sb = sb;
3079
3080         /* Load the checksum driver */
3081         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3082         if (IS_ERR(sbi->s_chksum_driver)) {
3083                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3084                 err = PTR_ERR(sbi->s_chksum_driver);
3085                 sbi->s_chksum_driver = NULL;
3086                 goto free_sbi;
3087         }
3088
3089         /* set a block size */
3090         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3091                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3092                 goto free_sbi;
3093         }
3094
3095         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3096                                                                 &recovery);
3097         if (err)
3098                 goto free_sbi;
3099
3100         sb->s_fs_info = sbi;
3101         sbi->raw_super = raw_super;
3102
3103         /* precompute checksum seed for metadata */
3104         if (f2fs_sb_has_inode_chksum(sbi))
3105                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3106                                                 sizeof(raw_super->uuid));
3107
3108         /*
3109          * The BLKZONED feature indicates that the drive was formatted with
3110          * zone alignment optimization. This is optional for host-aware
3111          * devices, but mandatory for host-managed zoned block devices.
3112          */
3113 #ifndef CONFIG_BLK_DEV_ZONED
3114         if (f2fs_sb_has_blkzoned(sbi)) {
3115                 f2fs_msg(sb, KERN_ERR,
3116                          "Zoned block device support is not enabled\n");
3117                 err = -EOPNOTSUPP;
3118                 goto free_sb_buf;
3119         }
3120 #endif
3121         default_options(sbi);
3122         /* parse mount options */
3123         options = kstrdup((const char *)data, GFP_KERNEL);
3124         if (data && !options) {
3125                 err = -ENOMEM;
3126                 goto free_sb_buf;
3127         }
3128
3129         err = parse_options(sb, options);
3130         if (err)
3131                 goto free_options;
3132
3133         sbi->max_file_blocks = max_file_blocks();
3134         sb->s_maxbytes = sbi->max_file_blocks <<
3135                                 le32_to_cpu(raw_super->log_blocksize);
3136         sb->s_max_links = F2FS_LINK_MAX;
3137
3138 #ifdef CONFIG_QUOTA
3139         sb->dq_op = &f2fs_quota_operations;
3140         if (f2fs_sb_has_quota_ino(sbi))
3141                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3142         else
3143                 sb->s_qcop = &f2fs_quotactl_ops;
3144         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3145
3146         if (f2fs_sb_has_quota_ino(sbi)) {
3147                 for (i = 0; i < MAXQUOTAS; i++) {
3148                         if (f2fs_qf_ino(sbi->sb, i))
3149                                 sbi->nquota_files++;
3150                 }
3151         }
3152 #endif
3153
3154         sb->s_op = &f2fs_sops;
3155 #ifdef CONFIG_F2FS_FS_ENCRYPTION
3156         sb->s_cop = &f2fs_cryptops;
3157 #endif
3158         sb->s_xattr = f2fs_xattr_handlers;
3159         sb->s_export_op = &f2fs_export_ops;
3160         sb->s_magic = F2FS_SUPER_MAGIC;
3161         sb->s_time_gran = 1;
3162         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3163                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3164         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3165         sb->s_iflags |= SB_I_CGROUPWB;
3166
3167         /* init f2fs-specific super block info */
3168         sbi->valid_super_block = valid_super_block;
3169         mutex_init(&sbi->gc_mutex);
3170         mutex_init(&sbi->writepages);
3171         mutex_init(&sbi->cp_mutex);
3172         init_rwsem(&sbi->node_write);
3173         init_rwsem(&sbi->node_change);
3174
3175         /* disallow all the data/node/meta page writes */
3176         set_sbi_flag(sbi, SBI_POR_DOING);
3177         spin_lock_init(&sbi->stat_lock);
3178
3179         /* init iostat info */
3180         spin_lock_init(&sbi->iostat_lock);
3181         sbi->iostat_enable = false;
3182
3183         for (i = 0; i < NR_PAGE_TYPE; i++) {
3184                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3185                 int j;
3186
3187                 sbi->write_io[i] =
3188                         f2fs_kmalloc(sbi,
3189                                      array_size(n,
3190                                                 sizeof(struct f2fs_bio_info)),
3191                                      GFP_KERNEL);
3192                 if (!sbi->write_io[i]) {
3193                         err = -ENOMEM;
3194                         goto free_bio_info;
3195                 }
3196
3197                 for (j = HOT; j < n; j++) {
3198                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3199                         sbi->write_io[i][j].sbi = sbi;
3200                         sbi->write_io[i][j].bio = NULL;
3201                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3202                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3203                 }
3204         }
3205
3206         init_rwsem(&sbi->cp_rwsem);
3207         init_waitqueue_head(&sbi->cp_wait);
3208         init_sb_info(sbi);
3209
3210         err = init_percpu_info(sbi);
3211         if (err)
3212                 goto free_bio_info;
3213
3214         if (F2FS_IO_SIZE(sbi) > 1) {
3215                 sbi->write_io_dummy =
3216                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3217                 if (!sbi->write_io_dummy) {
3218                         err = -ENOMEM;
3219                         goto free_percpu;
3220                 }
3221         }
3222
3223         /* get an inode for meta space */
3224         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3225         if (IS_ERR(sbi->meta_inode)) {
3226                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3227                 err = PTR_ERR(sbi->meta_inode);
3228                 goto free_io_dummy;
3229         }
3230
3231         err = f2fs_get_valid_checkpoint(sbi);
3232         if (err) {
3233                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3234                 goto free_meta_inode;
3235         }
3236
3237         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3238                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3239         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3240                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3241                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3242         }
3243
3244         /* Initialize device list */
3245         err = f2fs_scan_devices(sbi);
3246         if (err) {
3247                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3248                 goto free_devices;
3249         }
3250
3251         sbi->total_valid_node_count =
3252                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3253         percpu_counter_set(&sbi->total_valid_inode_count,
3254                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3255         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3256         sbi->total_valid_block_count =
3257                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3258         sbi->last_valid_block_count = sbi->total_valid_block_count;
3259         sbi->reserved_blocks = 0;
3260         sbi->current_reserved_blocks = 0;
3261         limit_reserve_root(sbi);
3262
3263         for (i = 0; i < NR_INODE_TYPE; i++) {
3264                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3265                 spin_lock_init(&sbi->inode_lock[i]);
3266         }
3267
3268         f2fs_init_extent_cache_info(sbi);
3269
3270         f2fs_init_ino_entry_info(sbi);
3271
3272         f2fs_init_fsync_node_info(sbi);
3273
3274         /* setup f2fs internal modules */
3275         err = f2fs_build_segment_manager(sbi);
3276         if (err) {
3277                 f2fs_msg(sb, KERN_ERR,
3278                         "Failed to initialize F2FS segment manager");
3279                 goto free_sm;
3280         }
3281         err = f2fs_build_node_manager(sbi);
3282         if (err) {
3283                 f2fs_msg(sb, KERN_ERR,
3284                         "Failed to initialize F2FS node manager");
3285                 goto free_nm;
3286         }
3287
3288         /* For write statistics */
3289         if (sb->s_bdev->bd_part)
3290                 sbi->sectors_written_start =
3291                         (u64)part_stat_read(sb->s_bdev->bd_part,
3292                                             sectors[STAT_WRITE]);
3293
3294         /* Read accumulated write IO statistics if exists */
3295         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3296         if (__exist_node_summaries(sbi))
3297                 sbi->kbytes_written =
3298                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3299
3300         f2fs_build_gc_manager(sbi);
3301
3302         err = f2fs_build_stats(sbi);
3303         if (err)
3304                 goto free_nm;
3305
3306         /* get an inode for node space */
3307         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3308         if (IS_ERR(sbi->node_inode)) {
3309                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3310                 err = PTR_ERR(sbi->node_inode);
3311                 goto free_stats;
3312         }
3313
3314         /* read root inode and dentry */
3315         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3316         if (IS_ERR(root)) {
3317                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3318                 err = PTR_ERR(root);
3319                 goto free_node_inode;
3320         }
3321         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3322                         !root->i_size || !root->i_nlink) {
3323                 iput(root);
3324                 err = -EINVAL;
3325                 goto free_node_inode;
3326         }
3327
3328         sb->s_root = d_make_root(root); /* allocate root dentry */
3329         if (!sb->s_root) {
3330                 err = -ENOMEM;
3331                 goto free_node_inode;
3332         }
3333
3334         err = f2fs_register_sysfs(sbi);
3335         if (err)
3336                 goto free_root_inode;
3337
3338 #ifdef CONFIG_QUOTA
3339         /* Enable quota usage during mount */
3340         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3341                 err = f2fs_enable_quotas(sb);
3342                 if (err)
3343                         f2fs_msg(sb, KERN_ERR,
3344                                 "Cannot turn on quotas: error %d", err);
3345         }
3346 #endif
3347         /* if there are nt orphan nodes free them */
3348         err = f2fs_recover_orphan_inodes(sbi);
3349         if (err)
3350                 goto free_meta;
3351
3352         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3353                 goto reset_checkpoint;
3354
3355         /* recover fsynced data */
3356         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3357                 /*
3358                  * mount should be failed, when device has readonly mode, and
3359                  * previous checkpoint was not done by clean system shutdown.
3360                  */
3361                 if (bdev_read_only(sb->s_bdev) &&
3362                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3363                         err = -EROFS;
3364                         goto free_meta;
3365                 }
3366
3367                 if (need_fsck)
3368                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3369
3370                 if (skip_recovery)
3371                         goto reset_checkpoint;
3372
3373                 err = f2fs_recover_fsync_data(sbi, false);
3374                 if (err < 0) {
3375                         if (err != -ENOMEM)
3376                                 skip_recovery = true;
3377                         need_fsck = true;
3378                         f2fs_msg(sb, KERN_ERR,
3379                                 "Cannot recover all fsync data errno=%d", err);
3380                         goto free_meta;
3381                 }
3382         } else {
3383                 err = f2fs_recover_fsync_data(sbi, true);
3384
3385                 if (!f2fs_readonly(sb) && err > 0) {
3386                         err = -EINVAL;
3387                         f2fs_msg(sb, KERN_ERR,
3388                                 "Need to recover fsync data");
3389                         goto free_meta;
3390                 }
3391         }
3392 reset_checkpoint:
3393         /* f2fs_recover_fsync_data() cleared this already */
3394         clear_sbi_flag(sbi, SBI_POR_DOING);
3395
3396         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3397                 err = f2fs_disable_checkpoint(sbi);
3398                 if (err)
3399                         goto sync_free_meta;
3400         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3401                 f2fs_enable_checkpoint(sbi);
3402         }
3403
3404         /*
3405          * If filesystem is not mounted as read-only then
3406          * do start the gc_thread.
3407          */
3408         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3409                 /* After POR, we can run background GC thread.*/
3410                 err = f2fs_start_gc_thread(sbi);
3411                 if (err)
3412                         goto sync_free_meta;
3413         }
3414         kvfree(options);
3415
3416         /* recover broken superblock */
3417         if (recovery) {
3418                 err = f2fs_commit_super(sbi, true);
3419                 f2fs_msg(sb, KERN_INFO,
3420                         "Try to recover %dth superblock, ret: %d",
3421                         sbi->valid_super_block ? 1 : 2, err);
3422         }
3423
3424         f2fs_join_shrinker(sbi);
3425
3426         f2fs_tuning_parameters(sbi);
3427
3428         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3429                                 cur_cp_version(F2FS_CKPT(sbi)));
3430         f2fs_update_time(sbi, CP_TIME);
3431         f2fs_update_time(sbi, REQ_TIME);
3432         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3433         return 0;
3434
3435 sync_free_meta:
3436         /* safe to flush all the data */
3437         sync_filesystem(sbi->sb);
3438         retry_cnt = 0;
3439
3440 free_meta:
3441 #ifdef CONFIG_QUOTA
3442         f2fs_truncate_quota_inode_pages(sb);
3443         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3444                 f2fs_quota_off_umount(sbi->sb);
3445 #endif
3446         /*
3447          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3448          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3449          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3450          * falls into an infinite loop in f2fs_sync_meta_pages().
3451          */
3452         truncate_inode_pages_final(META_MAPPING(sbi));
3453         /* evict some inodes being cached by GC */
3454         evict_inodes(sb);
3455         f2fs_unregister_sysfs(sbi);
3456 free_root_inode:
3457         dput(sb->s_root);
3458         sb->s_root = NULL;
3459 free_node_inode:
3460         f2fs_release_ino_entry(sbi, true);
3461         truncate_inode_pages_final(NODE_MAPPING(sbi));
3462         iput(sbi->node_inode);
3463         sbi->node_inode = NULL;
3464 free_stats:
3465         f2fs_destroy_stats(sbi);
3466 free_nm:
3467         f2fs_destroy_node_manager(sbi);
3468 free_sm:
3469         f2fs_destroy_segment_manager(sbi);
3470 free_devices:
3471         destroy_device_list(sbi);
3472         kvfree(sbi->ckpt);
3473 free_meta_inode:
3474         make_bad_inode(sbi->meta_inode);
3475         iput(sbi->meta_inode);
3476         sbi->meta_inode = NULL;
3477 free_io_dummy:
3478         mempool_destroy(sbi->write_io_dummy);
3479 free_percpu:
3480         destroy_percpu_info(sbi);
3481 free_bio_info:
3482         for (i = 0; i < NR_PAGE_TYPE; i++)
3483                 kvfree(sbi->write_io[i]);
3484 free_options:
3485 #ifdef CONFIG_QUOTA
3486         for (i = 0; i < MAXQUOTAS; i++)
3487                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3488 #endif
3489         kvfree(options);
3490 free_sb_buf:
3491         kvfree(raw_super);
3492 free_sbi:
3493         if (sbi->s_chksum_driver)
3494                 crypto_free_shash(sbi->s_chksum_driver);
3495         kvfree(sbi);
3496
3497         /* give only one another chance */
3498         if (retry_cnt > 0 && skip_recovery) {
3499                 retry_cnt--;
3500                 shrink_dcache_sb(sb);
3501                 goto try_onemore;
3502         }
3503         return err;
3504 }
3505
3506 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3507                         const char *dev_name, void *data)
3508 {
3509         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3510 }
3511
3512 static void kill_f2fs_super(struct super_block *sb)
3513 {
3514         if (sb->s_root) {
3515                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3516
3517                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3518                 f2fs_stop_gc_thread(sbi);
3519                 f2fs_stop_discard_thread(sbi);
3520
3521                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3522                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3523                         struct cp_control cpc = {
3524                                 .reason = CP_UMOUNT,
3525                         };
3526                         f2fs_write_checkpoint(sbi, &cpc);
3527                 }
3528
3529                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3530                         sb->s_flags &= ~SB_RDONLY;
3531         }
3532         kill_block_super(sb);
3533 }
3534
3535 static struct file_system_type f2fs_fs_type = {
3536         .owner          = THIS_MODULE,
3537         .name           = "f2fs",
3538         .mount          = f2fs_mount,
3539         .kill_sb        = kill_f2fs_super,
3540         .fs_flags       = FS_REQUIRES_DEV,
3541 };
3542 MODULE_ALIAS_FS("f2fs");
3543
3544 static int __init init_inodecache(void)
3545 {
3546         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3547                         sizeof(struct f2fs_inode_info), 0,
3548                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3549         if (!f2fs_inode_cachep)
3550                 return -ENOMEM;
3551         return 0;
3552 }
3553
3554 static void destroy_inodecache(void)
3555 {
3556         /*
3557          * Make sure all delayed rcu free inodes are flushed before we
3558          * destroy cache.
3559          */
3560         rcu_barrier();
3561         kmem_cache_destroy(f2fs_inode_cachep);
3562 }
3563
3564 static int __init init_f2fs_fs(void)
3565 {
3566         int err;
3567
3568         if (PAGE_SIZE != F2FS_BLKSIZE) {
3569                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3570                                 PAGE_SIZE, F2FS_BLKSIZE);
3571                 return -EINVAL;
3572         }
3573
3574         f2fs_build_trace_ios();
3575
3576         err = init_inodecache();
3577         if (err)
3578                 goto fail;
3579         err = f2fs_create_node_manager_caches();
3580         if (err)
3581                 goto free_inodecache;
3582         err = f2fs_create_segment_manager_caches();
3583         if (err)
3584                 goto free_node_manager_caches;
3585         err = f2fs_create_checkpoint_caches();
3586         if (err)
3587                 goto free_segment_manager_caches;
3588         err = f2fs_create_extent_cache();
3589         if (err)
3590                 goto free_checkpoint_caches;
3591         err = f2fs_init_sysfs();
3592         if (err)
3593                 goto free_extent_cache;
3594         err = register_shrinker(&f2fs_shrinker_info);
3595         if (err)
3596                 goto free_sysfs;
3597         err = register_filesystem(&f2fs_fs_type);
3598         if (err)
3599                 goto free_shrinker;
3600         f2fs_create_root_stats();
3601         err = f2fs_init_post_read_processing();
3602         if (err)
3603                 goto free_root_stats;
3604         return 0;
3605
3606 free_root_stats:
3607         f2fs_destroy_root_stats();
3608         unregister_filesystem(&f2fs_fs_type);
3609 free_shrinker:
3610         unregister_shrinker(&f2fs_shrinker_info);
3611 free_sysfs:
3612         f2fs_exit_sysfs();
3613 free_extent_cache:
3614         f2fs_destroy_extent_cache();
3615 free_checkpoint_caches:
3616         f2fs_destroy_checkpoint_caches();
3617 free_segment_manager_caches:
3618         f2fs_destroy_segment_manager_caches();
3619 free_node_manager_caches:
3620         f2fs_destroy_node_manager_caches();
3621 free_inodecache:
3622         destroy_inodecache();
3623 fail:
3624         return err;
3625 }
3626
3627 static void __exit exit_f2fs_fs(void)
3628 {
3629         f2fs_destroy_post_read_processing();
3630         f2fs_destroy_root_stats();
3631         unregister_filesystem(&f2fs_fs_type);
3632         unregister_shrinker(&f2fs_shrinker_info);
3633         f2fs_exit_sysfs();
3634         f2fs_destroy_extent_cache();
3635         f2fs_destroy_checkpoint_caches();
3636         f2fs_destroy_segment_manager_caches();
3637         f2fs_destroy_node_manager_caches();
3638         destroy_inodecache();
3639         f2fs_destroy_trace_ios();
3640 }
3641
3642 module_init(init_f2fs_fs)
3643 module_exit(exit_f2fs_fs)
3644
3645 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3646 MODULE_DESCRIPTION("Flash Friendly File System");
3647 MODULE_LICENSE("GPL");
3648