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 -EINVAL;
 273         }
 274         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 275                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 276                         ret = 0;
 277                 else
 278                         f2fs_msg(sb, KERN_ERR,
 279                                  "%s quota file already specified",
 280                                  QTYPE2NAME(qtype));
 281                 goto errout;
 282         }
 283         if (strchr(qname, '/')) {
 284                 f2fs_msg(sb, KERN_ERR,
 285                         "quotafile must be on filesystem root");
 286                 goto errout;
 287         }
 288         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 289         set_opt(sbi, QUOTA);
 290         return 0;
 291 errout:
 292         kvfree(qname);
 293         return ret;
 294 }
 295
 296 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 297 {
 298         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 299
 300         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 301                 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
 302                         " when quota turned on");
 303                 return -EINVAL;
 304         }
 305         kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 306         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 307         return 0;
 308 }
 309
 310 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 311 {
 312         /*
 313          * We do the test below only for project quotas. 'usrquota' and
 314          * 'grpquota' mount options are allowed even without quota feature
 315          * to support legacy quotas in quota files.
 316          */
 317         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 318                 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
 319                          "Cannot enable project quota enforcement.");
 320                 return -1;
 321         }
 322         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 323                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 324                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 325                 if (test_opt(sbi, USRQUOTA) &&
 326                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 327                         clear_opt(sbi, USRQUOTA);
 328
 329                 if (test_opt(sbi, GRPQUOTA) &&
 330                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 331                         clear_opt(sbi, GRPQUOTA);
 332
 333                 if (test_opt(sbi, PRJQUOTA) &&
 334                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 335                         clear_opt(sbi, PRJQUOTA);
 336
 337                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 338                                 test_opt(sbi, PRJQUOTA)) {
 339                         f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
 340                                         "format mixing");
 341                         return -1;
 342                 }
 343
 344                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 345                         f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
 346                                         "not specified");
 347                         return -1;
 348                 }
 349         }
 350
 351         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 352                 f2fs_msg(sbi->sb, KERN_INFO,
 353                         "QUOTA feature is enabled, so ignore jquota_fmt");
 354                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 355         }
 356         return 0;
 357 }
 358 #endif
 359
 360 static int parse_options(struct super_block *sb, char *options)
 361 {
 362         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 363         substring_t args[MAX_OPT_ARGS];
 364         char *p, *name;
 365         int arg = 0;
 366         kuid_t uid;
 367         kgid_t gid;
 368 #ifdef CONFIG_QUOTA
 369         int ret;
 370 #endif
 371
 372         if (!options)
 373                 return 0;
 374
 375         while ((p = strsep(&options, ",")) != NULL) {
 376                 int token;
 377                 if (!*p)
 378                         continue;
 379                 /*
 380                  * Initialize args struct so we know whether arg was
 381                  * found; some options take optional arguments.
 382                  */
 383                 args[0].to = args[0].from = NULL;
 384                 token = match_token(p, f2fs_tokens, args);
 385
 386                 switch (token) {
 387                 case Opt_gc_background:
 388                         name = match_strdup(&args[0]);
 389
 390                         if (!name)
 391                                 return -ENOMEM;
 392                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 393                                 set_opt(sbi, BG_GC);
 394                                 clear_opt(sbi, FORCE_FG_GC);
 395                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 396                                 clear_opt(sbi, BG_GC);
 397                                 clear_opt(sbi, FORCE_FG_GC);
 398                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 399                                 set_opt(sbi, BG_GC);
 400                                 set_opt(sbi, FORCE_FG_GC);
 401                         } else {
 402                                 kvfree(name);
 403                                 return -EINVAL;
 404                         }
 405                         kvfree(name);
 406                         break;
 407                 case Opt_disable_roll_forward:
 408                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 409                         break;
 410                 case Opt_norecovery:
 411                         /* this option mounts f2fs with ro */
 412                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 413                         if (!f2fs_readonly(sb))
 414                                 return -EINVAL;
 415                         break;
 416                 case Opt_discard:
 417                         set_opt(sbi, DISCARD);
 418                         break;
 419                 case Opt_nodiscard:
 420                         if (f2fs_sb_has_blkzoned(sbi)) {
 421                                 f2fs_msg(sb, KERN_WARNING,
 422                                         "discard is required for zoned block devices");
 423                                 return -EINVAL;
 424                         }
 425                         clear_opt(sbi, DISCARD);
 426                         break;
 427                 case Opt_noheap:
 428                         set_opt(sbi, NOHEAP);
 429                         break;
 430                 case Opt_heap:
 431                         clear_opt(sbi, NOHEAP);
 432                         break;
 433 #ifdef CONFIG_F2FS_FS_XATTR
 434                 case Opt_user_xattr:
 435                         set_opt(sbi, XATTR_USER);
 436                         break;
 437                 case Opt_nouser_xattr:
 438                         clear_opt(sbi, XATTR_USER);
 439                         break;
 440                 case Opt_inline_xattr:
 441                         set_opt(sbi, INLINE_XATTR);
 442                         break;
 443                 case Opt_noinline_xattr:
 444                         clear_opt(sbi, INLINE_XATTR);
 445                         break;
 446                 case Opt_inline_xattr_size:
 447                         if (args->from && match_int(args, &arg))
 448                                 return -EINVAL;
 449                         set_opt(sbi, INLINE_XATTR_SIZE);
 450                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 451                         break;
 452 #else
 453                 case Opt_user_xattr:
 454                         f2fs_msg(sb, KERN_INFO,
 455                                 "user_xattr options not supported");
 456                         break;
 457                 case Opt_nouser_xattr:
 458                         f2fs_msg(sb, KERN_INFO,
 459                                 "nouser_xattr options not supported");
 460                         break;
 461                 case Opt_inline_xattr:
 462                         f2fs_msg(sb, KERN_INFO,
 463                                 "inline_xattr options not supported");
 464                         break;
 465                 case Opt_noinline_xattr:
 466                         f2fs_msg(sb, KERN_INFO,
 467                                 "noinline_xattr options not supported");
 468                         break;
 469 #endif
 470 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 471                 case Opt_acl:
 472                         set_opt(sbi, POSIX_ACL);
 473                         break;
 474                 case Opt_noacl:
 475                         clear_opt(sbi, POSIX_ACL);
 476                         break;
 477 #else
 478                 case Opt_acl:
 479                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 480                         break;
 481                 case Opt_noacl:
 482                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 483                         break;
 484 #endif
 485                 case Opt_active_logs:
 486                         if (args->from && match_int(args, &arg))
 487                                 return -EINVAL;
 488                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 489                                 return -EINVAL;
 490                         F2FS_OPTION(sbi).active_logs = arg;
 491                         break;
 492                 case Opt_disable_ext_identify:
 493                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 494                         break;
 495                 case Opt_inline_data:
 496                         set_opt(sbi, INLINE_DATA);
 497                         break;
 498                 case Opt_inline_dentry:
 499                         set_opt(sbi, INLINE_DENTRY);
 500                         break;
 501                 case Opt_noinline_dentry:
 502                         clear_opt(sbi, INLINE_DENTRY);
 503                         break;
 504                 case Opt_flush_merge:
 505                         set_opt(sbi, FLUSH_MERGE);
 506                         break;
 507                 case Opt_noflush_merge:
 508                         clear_opt(sbi, FLUSH_MERGE);
 509                         break;
 510                 case Opt_nobarrier:
 511                         set_opt(sbi, NOBARRIER);
 512                         break;
 513                 case Opt_fastboot:
 514                         set_opt(sbi, FASTBOOT);
 515                         break;
 516                 case Opt_extent_cache:
 517                         set_opt(sbi, EXTENT_CACHE);
 518                         break;
 519                 case Opt_noextent_cache:
 520                         clear_opt(sbi, EXTENT_CACHE);
 521                         break;
 522                 case Opt_noinline_data:
 523                         clear_opt(sbi, INLINE_DATA);
 524                         break;
 525                 case Opt_data_flush:
 526                         set_opt(sbi, DATA_FLUSH);
 527                         break;
 528                 case Opt_reserve_root:
 529                         if (args->from && match_int(args, &arg))
 530                                 return -EINVAL;
 531                         if (test_opt(sbi, RESERVE_ROOT)) {
 532                                 f2fs_msg(sb, KERN_INFO,
 533                                         "Preserve previous reserve_root=%u",
 534                                         F2FS_OPTION(sbi).root_reserved_blocks);
 535                         } else {
 536                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 537                                 set_opt(sbi, RESERVE_ROOT);
 538                         }
 539                         break;
 540                 case Opt_resuid:
 541                         if (args->from && match_int(args, &arg))
 542                                 return -EINVAL;
 543                         uid = make_kuid(current_user_ns(), arg);
 544                         if (!uid_valid(uid)) {
 545                                 f2fs_msg(sb, KERN_ERR,
 546                                         "Invalid uid value %d", arg);
 547                                 return -EINVAL;
 548                         }
 549                         F2FS_OPTION(sbi).s_resuid = uid;
 550                         break;
 551                 case Opt_resgid:
 552                         if (args->from && match_int(args, &arg))
 553                                 return -EINVAL;
 554                         gid = make_kgid(current_user_ns(), arg);
 555                         if (!gid_valid(gid)) {
 556                                 f2fs_msg(sb, KERN_ERR,
 557                                         "Invalid gid value %d", arg);
 558                                 return -EINVAL;
 559                         }
 560                         F2FS_OPTION(sbi).s_resgid = gid;
 561                         break;
 562                 case Opt_mode:
 563                         name = match_strdup(&args[0]);
 564
 565                         if (!name)
 566                                 return -ENOMEM;
 567                         if (strlen(name) == 8 &&
 568                                         !strncmp(name, "adaptive", 8)) {
 569                                 if (f2fs_sb_has_blkzoned(sbi)) {
 570                                         f2fs_msg(sb, KERN_WARNING,
 571                                                  "adaptive mode is not allowed with "
 572                                                  "zoned block device feature");
 573                                         kvfree(name);
 574                                         return -EINVAL;
 575                                 }
 576                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 577                         } else if (strlen(name) == 3 &&
 578                                         !strncmp(name, "lfs", 3)) {
 579                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 580                         } else {
 581                                 kvfree(name);
 582                                 return -EINVAL;
 583                         }
 584                         kvfree(name);
 585                         break;
 586                 case Opt_io_size_bits:
 587                         if (args->from && match_int(args, &arg))
 588                                 return -EINVAL;
 589                         if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
 590                                 f2fs_msg(sb, KERN_WARNING,
 591                                         "Not support %d, larger than %d",
 592                                         1 << arg, BIO_MAX_PAGES);
 593                                 return -EINVAL;
 594                         }
 595                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 596                         break;
 597 #ifdef CONFIG_F2FS_FAULT_INJECTION
 598                 case Opt_fault_injection:
 599                         if (args->from && match_int(args, &arg))
 600                                 return -EINVAL;
 601                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 602                         set_opt(sbi, FAULT_INJECTION);
 603                         break;
 604
 605                 case Opt_fault_type:
 606                         if (args->from && match_int(args, &arg))
 607                                 return -EINVAL;
 608                         f2fs_build_fault_attr(sbi, 0, arg);
 609                         set_opt(sbi, FAULT_INJECTION);
 610                         break;
 611 #else
 612                 case Opt_fault_injection:
 613                         f2fs_msg(sb, KERN_INFO,
 614                                 "fault_injection options not supported");
 615                         break;
 616
 617                 case Opt_fault_type:
 618                         f2fs_msg(sb, KERN_INFO,
 619                                 "fault_type options not supported");
 620                         break;
 621 #endif
 622                 case Opt_lazytime:
 623                         sb->s_flags |= SB_LAZYTIME;
 624                         break;
 625                 case Opt_nolazytime:
 626                         sb->s_flags &= ~SB_LAZYTIME;
 627                         break;
 628 #ifdef CONFIG_QUOTA
 629                 case Opt_quota:
 630                 case Opt_usrquota:
 631                         set_opt(sbi, USRQUOTA);
 632                         break;
 633                 case Opt_grpquota:
 634                         set_opt(sbi, GRPQUOTA);
 635                         break;
 636                 case Opt_prjquota:
 637                         set_opt(sbi, PRJQUOTA);
 638                         break;
 639                 case Opt_usrjquota:
 640                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 641                         if (ret)
 642                                 return ret;
 643                         break;
 644                 case Opt_grpjquota:
 645                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 646                         if (ret)
 647                                 return ret;
 648                         break;
 649                 case Opt_prjjquota:
 650                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 651                         if (ret)
 652                                 return ret;
 653                         break;
 654                 case Opt_offusrjquota:
 655                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 656                         if (ret)
 657                                 return ret;
 658                         break;
 659                 case Opt_offgrpjquota:
 660                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 661                         if (ret)
 662                                 return ret;
 663                         break;
 664                 case Opt_offprjjquota:
 665                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 666                         if (ret)
 667                                 return ret;
 668                         break;
 669                 case Opt_jqfmt_vfsold:
 670                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 671                         break;
 672                 case Opt_jqfmt_vfsv0:
 673                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 674                         break;
 675                 case Opt_jqfmt_vfsv1:
 676                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 677                         break;
 678                 case Opt_noquota:
 679                         clear_opt(sbi, QUOTA);
 680                         clear_opt(sbi, USRQUOTA);
 681                         clear_opt(sbi, GRPQUOTA);
 682                         clear_opt(sbi, PRJQUOTA);
 683                         break;
 684 #else
 685                 case Opt_quota:
 686                 case Opt_usrquota:
 687                 case Opt_grpquota:
 688                 case Opt_prjquota:
 689                 case Opt_usrjquota:
 690                 case Opt_grpjquota:
 691                 case Opt_prjjquota:
 692                 case Opt_offusrjquota:
 693                 case Opt_offgrpjquota:
 694                 case Opt_offprjjquota:
 695                 case Opt_jqfmt_vfsold:
 696                 case Opt_jqfmt_vfsv0:
 697                 case Opt_jqfmt_vfsv1:
 698                 case Opt_noquota:
 699                         f2fs_msg(sb, KERN_INFO,
 700                                         "quota operations not supported");
 701                         break;
 702 #endif
 703                 case Opt_whint:
 704                         name = match_strdup(&args[0]);
 705                         if (!name)
 706                                 return -ENOMEM;
 707                         if (strlen(name) == 10 &&
 708                                         !strncmp(name, "user-based", 10)) {
 709                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 710                         } else if (strlen(name) == 3 &&
 711                                         !strncmp(name, "off", 3)) {
 712                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 713                         } else if (strlen(name) == 8 &&
 714                                         !strncmp(name, "fs-based", 8)) {
 715                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 716                         } else {
 717                                 kvfree(name);
 718                                 return -EINVAL;
 719                         }
 720                         kvfree(name);
 721                         break;
 722                 case Opt_alloc:
 723                         name = match_strdup(&args[0]);
 724                         if (!name)
 725                                 return -ENOMEM;
 726
 727                         if (strlen(name) == 7 &&
 728                                         !strncmp(name, "default", 7)) {
 729                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 730                         } else if (strlen(name) == 5 &&
 731                                         !strncmp(name, "reuse", 5)) {
 732                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 733                         } else {
 734                                 kvfree(name);
 735                                 return -EINVAL;
 736                         }
 737                         kvfree(name);
 738                         break;
 739                 case Opt_fsync:
 740                         name = match_strdup(&args[0]);
 741                         if (!name)
 742                                 return -ENOMEM;
 743                         if (strlen(name) == 5 &&
 744                                         !strncmp(name, "posix", 5)) {
 745                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 746                         } else if (strlen(name) == 6 &&
 747                                         !strncmp(name, "strict", 6)) {
 748                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 749                         } else if (strlen(name) == 9 &&
 750                                         !strncmp(name, "nobarrier", 9)) {
 751                                 F2FS_OPTION(sbi).fsync_mode =
 752                                                         FSYNC_MODE_NOBARRIER;
 753                         } else {
 754                                 kvfree(name);
 755                                 return -EINVAL;
 756                         }
 757                         kvfree(name);
 758                         break;
 759                 case Opt_test_dummy_encryption:
 760 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 761                         if (!f2fs_sb_has_encrypt(sbi)) {
 762                                 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
 763                                 return -EINVAL;
 764                         }
 765
 766                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 767                         f2fs_msg(sb, KERN_INFO,
 768                                         "Test dummy encryption mode enabled");
 769 #else
 770                         f2fs_msg(sb, KERN_INFO,
 771                                         "Test dummy encryption mount option ignored");
 772 #endif
 773                         break;
 774                 case Opt_checkpoint:
 775                         name = match_strdup(&args[0]);
 776                         if (!name)
 777                                 return -ENOMEM;
 778
 779                         if (strlen(name) == 6 &&
 780                                         !strncmp(name, "enable", 6)) {
 781                                 clear_opt(sbi, DISABLE_CHECKPOINT);
 782                         } else if (strlen(name) == 7 &&
 783                                         !strncmp(name, "disable", 7)) {
 784                                 set_opt(sbi, DISABLE_CHECKPOINT);
 785                         } else {
 786                                 kvfree(name);
 787                                 return -EINVAL;
 788                         }
 789                         kvfree(name);
 790                         break;
 791                 default:
 792                         f2fs_msg(sb, KERN_ERR,
 793                                 "Unrecognized mount option \"%s\" or missing value",
 794                                 p);
 795                         return -EINVAL;
 796                 }
 797         }
 798 #ifdef CONFIG_QUOTA
 799         if (f2fs_check_quota_options(sbi))
 800                 return -EINVAL;
 801 #else
 802         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
 803                 f2fs_msg(sbi->sb, KERN_INFO,
 804                          "Filesystem with quota feature cannot be mounted RDWR "
 805                          "without CONFIG_QUOTA");
 806                 return -EINVAL;
 807         }
 808         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
 809                 f2fs_msg(sb, KERN_ERR,
 810                         "Filesystem with project quota feature cannot be "
 811                         "mounted RDWR without CONFIG_QUOTA");
 812                 return -EINVAL;
 813         }
 814 #endif
 815
 816         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 817                 f2fs_msg(sb, KERN_ERR,
 818                                 "Should set mode=lfs with %uKB-sized IO",
 819                                 F2FS_IO_SIZE_KB(sbi));
 820                 return -EINVAL;
 821         }
 822
 823         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 824                 if (!f2fs_sb_has_extra_attr(sbi) ||
 825                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 826                         f2fs_msg(sb, KERN_ERR,
 827                                         "extra_attr or flexible_inline_xattr "
 828                                         "feature is off");
 829                         return -EINVAL;
 830                 }
 831                 if (!test_opt(sbi, INLINE_XATTR)) {
 832                         f2fs_msg(sb, KERN_ERR,
 833                                         "inline_xattr_size option should be "
 834                                         "set with inline_xattr option");
 835                         return -EINVAL;
 836                 }
 837                 if (!F2FS_OPTION(sbi).inline_xattr_size ||
 838                         F2FS_OPTION(sbi).inline_xattr_size >=
 839                                         DEF_ADDRS_PER_INODE -
 840                                         F2FS_TOTAL_EXTRA_ATTR_SIZE -
 841                                         DEF_INLINE_RESERVED_SIZE -
 842                                         DEF_MIN_INLINE_SIZE) {
 843                         f2fs_msg(sb, KERN_ERR,
 844                                         "inline xattr size is out of range");
 845                         return -EINVAL;
 846                 }
 847         }
 848
 849         if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
 850                 f2fs_msg(sb, KERN_ERR,
 851                                 "LFS not compatible with checkpoint=disable\n");
 852                 return -EINVAL;
 853         }
 854
 855         /* Not pass down write hints if the number of active logs is lesser
 856          * than NR_CURSEG_TYPE.
 857          */
 858         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 859                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 860         return 0;
 861 }
 862
 863 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 864 {
 865         struct f2fs_inode_info *fi;
 866
 867         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 868         if (!fi)
 869                 return NULL;
 870
 871         init_once((void *) fi);
 872
 873         /* Initialize f2fs-specific inode info */
 874         atomic_set(&fi->dirty_pages, 0);
 875         init_rwsem(&fi->i_sem);
 876         INIT_LIST_HEAD(&fi->dirty_list);
 877         INIT_LIST_HEAD(&fi->gdirty_list);
 878         INIT_LIST_HEAD(&fi->inmem_ilist);
 879         INIT_LIST_HEAD(&fi->inmem_pages);
 880         mutex_init(&fi->inmem_lock);
 881         init_rwsem(&fi->i_gc_rwsem[READ]);
 882         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 883         init_rwsem(&fi->i_mmap_sem);
 884         init_rwsem(&fi->i_xattr_sem);
 885
 886         /* Will be used by directory only */
 887         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 888
 889         return &fi->vfs_inode;
 890 }
 891
 892 static int f2fs_drop_inode(struct inode *inode)
 893 {
 894         int ret;
 895         /*
 896          * This is to avoid a deadlock condition like below.
 897          * writeback_single_inode(inode)
 898          *  - f2fs_write_data_page
 899          *    - f2fs_gc -> iput -> evict
 900          *       - inode_wait_for_writeback(inode)
 901          */
 902         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 903                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 904                         /* to avoid evict_inode call simultaneously */
 905                         atomic_inc(&inode->i_count);
 906                         spin_unlock(&inode->i_lock);
 907
 908                         /* some remained atomic pages should discarded */
 909                         if (f2fs_is_atomic_file(inode))
 910                                 f2fs_drop_inmem_pages(inode);
 911
 912                         /* should remain fi->extent_tree for writepage */
 913                         f2fs_destroy_extent_node(inode);
 914
 915                         sb_start_intwrite(inode->i_sb);
 916                         f2fs_i_size_write(inode, 0);
 917
 918                         if (F2FS_HAS_BLOCKS(inode))
 919                                 f2fs_truncate(inode);
 920
 921                         sb_end_intwrite(inode->i_sb);
 922
 923                         spin_lock(&inode->i_lock);
 924                         atomic_dec(&inode->i_count);
 925                 }
 926                 trace_f2fs_drop_inode(inode, 0);
 927                 return 0;
 928         }
 929         ret = generic_drop_inode(inode);
 930         trace_f2fs_drop_inode(inode, ret);
 931         return ret;
 932 }
 933
 934 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 935 {
 936         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 937         int ret = 0;
 938
 939         spin_lock(&sbi->inode_lock[DIRTY_META]);
 940         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 941                 ret = 1;
 942         } else {
 943                 set_inode_flag(inode, FI_DIRTY_INODE);
 944                 stat_inc_dirty_inode(sbi, DIRTY_META);
 945         }
 946         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 947                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 948                                 &sbi->inode_list[DIRTY_META]);
 949                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 950         }
 951         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 952         return ret;
 953 }
 954
 955 void f2fs_inode_synced(struct inode *inode)
 956 {
 957         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 958
 959         spin_lock(&sbi->inode_lock[DIRTY_META]);
 960         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 961                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 962                 return;
 963         }
 964         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 965                 list_del_init(&F2FS_I(inode)->gdirty_list);
 966                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 967         }
 968         clear_inode_flag(inode, FI_DIRTY_INODE);
 969         clear_inode_flag(inode, FI_AUTO_RECOVER);
 970         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 971         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 972 }
 973
 974 /*
 975  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 976  *
 977  * We should call set_dirty_inode to write the dirty inode through write_inode.
 978  */
 979 static void f2fs_dirty_inode(struct inode *inode, int flags)
 980 {
 981         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 982
 983         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 984                         inode->i_ino == F2FS_META_INO(sbi))
 985                 return;
 986
 987         if (flags == I_DIRTY_TIME)
 988                 return;
 989
 990         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 991                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 992
 993         f2fs_inode_dirtied(inode, false);
 994 }
 995
 996 static void f2fs_i_callback(struct rcu_head *head)
 997 {
 998         struct inode *inode = container_of(head, struct inode, i_rcu);
 999         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1000 }
1001
1002 static void f2fs_destroy_inode(struct inode *inode)
1003 {
1004         call_rcu(&inode->i_rcu, f2fs_i_callback);
1005 }
1006
1007 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1008 {
1009         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1010         percpu_counter_destroy(&sbi->total_valid_inode_count);
1011 }
1012
1013 static void destroy_device_list(struct f2fs_sb_info *sbi)
1014 {
1015         int i;
1016
1017         for (i = 0; i < sbi->s_ndevs; i++) {
1018                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1019 #ifdef CONFIG_BLK_DEV_ZONED
1020                 kvfree(FDEV(i).blkz_type);
1021 #endif
1022         }
1023         kvfree(sbi->devs);
1024 }
1025
1026 static void f2fs_put_super(struct super_block *sb)
1027 {
1028         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1029         int i;
1030         bool dropped;
1031
1032         f2fs_quota_off_umount(sb);
1033
1034         /* prevent remaining shrinker jobs */
1035         mutex_lock(&sbi->umount_mutex);
1036
1037         /*
1038          * We don't need to do checkpoint when superblock is clean.
1039          * But, the previous checkpoint was not done by umount, it needs to do
1040          * clean checkpoint again.
1041          */
1042         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1043                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1044                 struct cp_control cpc = {
1045                         .reason = CP_UMOUNT,
1046                 };
1047                 f2fs_write_checkpoint(sbi, &cpc);
1048         }
1049
1050         /* be sure to wait for any on-going discard commands */
1051         dropped = f2fs_wait_discard_bios(sbi);
1052
1053         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1054                                         !sbi->discard_blks && !dropped) {
1055                 struct cp_control cpc = {
1056                         .reason = CP_UMOUNT | CP_TRIMMED,
1057                 };
1058                 f2fs_write_checkpoint(sbi, &cpc);
1059         }
1060
1061         /*
1062          * normally superblock is clean, so we need to release this.
1063          * In addition, EIO will skip do checkpoint, we need this as well.
1064          */
1065         f2fs_release_ino_entry(sbi, true);
1066
1067         f2fs_leave_shrinker(sbi);
1068         mutex_unlock(&sbi->umount_mutex);
1069
1070         /* our cp_error case, we can wait for any writeback page */
1071         f2fs_flush_merged_writes(sbi);
1072
1073         f2fs_wait_on_all_pages_writeback(sbi);
1074
1075         f2fs_bug_on(sbi, sbi->fsync_node_num);
1076
1077         iput(sbi->node_inode);
1078         iput(sbi->meta_inode);
1079
1080         /*
1081          * iput() can update stat information, if f2fs_write_checkpoint()
1082          * above failed with error.
1083          */
1084         f2fs_destroy_stats(sbi);
1085
1086         /* destroy f2fs internal modules */
1087         f2fs_destroy_node_manager(sbi);
1088         f2fs_destroy_segment_manager(sbi);
1089
1090         kvfree(sbi->ckpt);
1091
1092         f2fs_unregister_sysfs(sbi);
1093
1094         sb->s_fs_info = NULL;
1095         if (sbi->s_chksum_driver)
1096                 crypto_free_shash(sbi->s_chksum_driver);
1097         kvfree(sbi->raw_super);
1098
1099         destroy_device_list(sbi);
1100         mempool_destroy(sbi->write_io_dummy);
1101 #ifdef CONFIG_QUOTA
1102         for (i = 0; i < MAXQUOTAS; i++)
1103                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1104 #endif
1105         destroy_percpu_info(sbi);
1106         for (i = 0; i < NR_PAGE_TYPE; i++)
1107                 kvfree(sbi->write_io[i]);
1108         kvfree(sbi);
1109 }
1110
1111 int f2fs_sync_fs(struct super_block *sb, int sync)
1112 {
1113         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1114         int err = 0;
1115
1116         if (unlikely(f2fs_cp_error(sbi)))
1117                 return 0;
1118         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1119                 return 0;
1120
1121         trace_f2fs_sync_fs(sb, sync);
1122
1123         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1124                 return -EAGAIN;
1125
1126         if (sync) {
1127                 struct cp_control cpc;
1128
1129                 cpc.reason = __get_cp_reason(sbi);
1130
1131                 mutex_lock(&sbi->gc_mutex);
1132                 err = f2fs_write_checkpoint(sbi, &cpc);
1133                 mutex_unlock(&sbi->gc_mutex);
1134         }
1135         f2fs_trace_ios(NULL, 1);
1136
1137         return err;
1138 }
1139
1140 static int f2fs_freeze(struct super_block *sb)
1141 {
1142         if (f2fs_readonly(sb))
1143                 return 0;
1144
1145         /* IO error happened before */
1146         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1147                 return -EIO;
1148
1149         /* must be clean, since sync_filesystem() was already called */
1150         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1151                 return -EINVAL;
1152         return 0;
1153 }
1154
1155 static int f2fs_unfreeze(struct super_block *sb)
1156 {
1157         return 0;
1158 }
1159
1160 #ifdef CONFIG_QUOTA
1161 static int f2fs_statfs_project(struct super_block *sb,
1162                                 kprojid_t projid, struct kstatfs *buf)
1163 {
1164         struct kqid qid;
1165         struct dquot *dquot;
1166         u64 limit;
1167         u64 curblock;
1168
1169         qid = make_kqid_projid(projid);
1170         dquot = dqget(sb, qid);
1171         if (IS_ERR(dquot))
1172                 return PTR_ERR(dquot);
1173         spin_lock(&dquot->dq_dqb_lock);
1174
1175         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1176                  dquot->dq_dqb.dqb_bsoftlimit :
1177                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1178         if (limit && buf->f_blocks > limit) {
1179                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1180                 buf->f_blocks = limit;
1181                 buf->f_bfree = buf->f_bavail =
1182                         (buf->f_blocks > curblock) ?
1183                          (buf->f_blocks - curblock) : 0;
1184         }
1185
1186         limit = dquot->dq_dqb.dqb_isoftlimit ?
1187                 dquot->dq_dqb.dqb_isoftlimit :
1188                 dquot->dq_dqb.dqb_ihardlimit;
1189         if (limit && buf->f_files > limit) {
1190                 buf->f_files = limit;
1191                 buf->f_ffree =
1192                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1193                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1194         }
1195
1196         spin_unlock(&dquot->dq_dqb_lock);
1197         dqput(dquot);
1198         return 0;
1199 }
1200 #endif
1201
1202 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1203 {
1204         struct super_block *sb = dentry->d_sb;
1205         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1206         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1207         block_t total_count, user_block_count, start_count;
1208         u64 avail_node_count;
1209
1210         total_count = le64_to_cpu(sbi->raw_super->block_count);
1211         user_block_count = sbi->user_block_count;
1212         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1213         buf->f_type = F2FS_SUPER_MAGIC;
1214         buf->f_bsize = sbi->blocksize;
1215
1216         buf->f_blocks = total_count - start_count;
1217         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1218                                                 sbi->current_reserved_blocks;
1219         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1220                 buf->f_bfree = 0;
1221         else
1222                 buf->f_bfree -= sbi->unusable_block_count;
1223
1224         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1225                 buf->f_bavail = buf->f_bfree -
1226                                 F2FS_OPTION(sbi).root_reserved_blocks;
1227         else
1228                 buf->f_bavail = 0;
1229
1230         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1231                                                 F2FS_RESERVED_NODE_NUM;
1232
1233         if (avail_node_count > user_block_count) {
1234                 buf->f_files = user_block_count;
1235                 buf->f_ffree = buf->f_bavail;
1236         } else {
1237                 buf->f_files = avail_node_count;
1238                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1239                                         buf->f_bavail);
1240         }
1241
1242         buf->f_namelen = F2FS_NAME_LEN;
1243         buf->f_fsid.val[0] = (u32)id;
1244         buf->f_fsid.val[1] = (u32)(id >> 32);
1245
1246 #ifdef CONFIG_QUOTA
1247         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1248                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1249                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1250         }
1251 #endif
1252         return 0;
1253 }
1254
1255 static inline void f2fs_show_quota_options(struct seq_file *seq,
1256                                            struct super_block *sb)
1257 {
1258 #ifdef CONFIG_QUOTA
1259         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1260
1261         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1262                 char *fmtname = "";
1263
1264                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1265                 case QFMT_VFS_OLD:
1266                         fmtname = "vfsold";
1267                         break;
1268                 case QFMT_VFS_V0:
1269                         fmtname = "vfsv0";
1270                         break;
1271                 case QFMT_VFS_V1:
1272                         fmtname = "vfsv1";
1273                         break;
1274                 }
1275                 seq_printf(seq, ",jqfmt=%s", fmtname);
1276         }
1277
1278         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1279                 seq_show_option(seq, "usrjquota",
1280                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1281
1282         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1283                 seq_show_option(seq, "grpjquota",
1284                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1285
1286         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1287                 seq_show_option(seq, "prjjquota",
1288                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1289 #endif
1290 }
1291
1292 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1293 {
1294         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1295
1296         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1297                 if (test_opt(sbi, FORCE_FG_GC))
1298                         seq_printf(seq, ",background_gc=%s", "sync");
1299                 else
1300                         seq_printf(seq, ",background_gc=%s", "on");
1301         } else {
1302                 seq_printf(seq, ",background_gc=%s", "off");
1303         }
1304         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1305                 seq_puts(seq, ",disable_roll_forward");
1306         if (test_opt(sbi, DISCARD))
1307                 seq_puts(seq, ",discard");
1308         if (test_opt(sbi, NOHEAP))
1309                 seq_puts(seq, ",no_heap");
1310         else
1311                 seq_puts(seq, ",heap");
1312 #ifdef CONFIG_F2FS_FS_XATTR
1313         if (test_opt(sbi, XATTR_USER))
1314                 seq_puts(seq, ",user_xattr");
1315         else
1316                 seq_puts(seq, ",nouser_xattr");
1317         if (test_opt(sbi, INLINE_XATTR))
1318                 seq_puts(seq, ",inline_xattr");
1319         else
1320                 seq_puts(seq, ",noinline_xattr");
1321         if (test_opt(sbi, INLINE_XATTR_SIZE))
1322                 seq_printf(seq, ",inline_xattr_size=%u",
1323                                         F2FS_OPTION(sbi).inline_xattr_size);
1324 #endif
1325 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1326         if (test_opt(sbi, POSIX_ACL))
1327                 seq_puts(seq, ",acl");
1328         else
1329                 seq_puts(seq, ",noacl");
1330 #endif
1331         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1332                 seq_puts(seq, ",disable_ext_identify");
1333         if (test_opt(sbi, INLINE_DATA))
1334                 seq_puts(seq, ",inline_data");
1335         else
1336                 seq_puts(seq, ",noinline_data");
1337         if (test_opt(sbi, INLINE_DENTRY))
1338                 seq_puts(seq, ",inline_dentry");
1339         else
1340                 seq_puts(seq, ",noinline_dentry");
1341         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1342                 seq_puts(seq, ",flush_merge");
1343         if (test_opt(sbi, NOBARRIER))
1344                 seq_puts(seq, ",nobarrier");
1345         if (test_opt(sbi, FASTBOOT))
1346                 seq_puts(seq, ",fastboot");
1347         if (test_opt(sbi, EXTENT_CACHE))
1348                 seq_puts(seq, ",extent_cache");
1349         else
1350                 seq_puts(seq, ",noextent_cache");
1351         if (test_opt(sbi, DATA_FLUSH))
1352                 seq_puts(seq, ",data_flush");
1353
1354         seq_puts(seq, ",mode=");
1355         if (test_opt(sbi, ADAPTIVE))
1356                 seq_puts(seq, "adaptive");
1357         else if (test_opt(sbi, LFS))
1358                 seq_puts(seq, "lfs");
1359         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1360         if (test_opt(sbi, RESERVE_ROOT))
1361                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1362                                 F2FS_OPTION(sbi).root_reserved_blocks,
1363                                 from_kuid_munged(&init_user_ns,
1364                                         F2FS_OPTION(sbi).s_resuid),
1365                                 from_kgid_munged(&init_user_ns,
1366                                         F2FS_OPTION(sbi).s_resgid));
1367         if (F2FS_IO_SIZE_BITS(sbi))
1368                 seq_printf(seq, ",io_bits=%u",
1369                                 F2FS_OPTION(sbi).write_io_size_bits);
1370 #ifdef CONFIG_F2FS_FAULT_INJECTION
1371         if (test_opt(sbi, FAULT_INJECTION)) {
1372                 seq_printf(seq, ",fault_injection=%u",
1373                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1374                 seq_printf(seq, ",fault_type=%u",
1375                                 F2FS_OPTION(sbi).fault_info.inject_type);
1376         }
1377 #endif
1378 #ifdef CONFIG_QUOTA
1379         if (test_opt(sbi, QUOTA))
1380                 seq_puts(seq, ",quota");
1381         if (test_opt(sbi, USRQUOTA))
1382                 seq_puts(seq, ",usrquota");
1383         if (test_opt(sbi, GRPQUOTA))
1384                 seq_puts(seq, ",grpquota");
1385         if (test_opt(sbi, PRJQUOTA))
1386                 seq_puts(seq, ",prjquota");
1387 #endif
1388         f2fs_show_quota_options(seq, sbi->sb);
1389         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1390                 seq_printf(seq, ",whint_mode=%s", "user-based");
1391         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1392                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1393 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1394         if (F2FS_OPTION(sbi).test_dummy_encryption)
1395                 seq_puts(seq, ",test_dummy_encryption");
1396 #endif
1397
1398         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1399                 seq_printf(seq, ",alloc_mode=%s", "default");
1400         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1401                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1402
1403         if (test_opt(sbi, DISABLE_CHECKPOINT))
1404                 seq_puts(seq, ",checkpoint=disable");
1405
1406         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1407                 seq_printf(seq, ",fsync_mode=%s", "posix");
1408         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1409                 seq_printf(seq, ",fsync_mode=%s", "strict");
1410         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1411                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1412         return 0;
1413 }
1414
1415 static void default_options(struct f2fs_sb_info *sbi)
1416 {
1417         /* init some FS parameters */
1418         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1419         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1420         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1421         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1422         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1423         F2FS_OPTION(sbi).test_dummy_encryption = false;
1424         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1425         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1426
1427         set_opt(sbi, BG_GC);
1428         set_opt(sbi, INLINE_XATTR);
1429         set_opt(sbi, INLINE_DATA);
1430         set_opt(sbi, INLINE_DENTRY);
1431         set_opt(sbi, EXTENT_CACHE);
1432         set_opt(sbi, NOHEAP);
1433         clear_opt(sbi, DISABLE_CHECKPOINT);
1434         sbi->sb->s_flags |= SB_LAZYTIME;
1435         set_opt(sbi, FLUSH_MERGE);
1436         set_opt(sbi, DISCARD);
1437         if (f2fs_sb_has_blkzoned(sbi))
1438                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1439         else
1440                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1441
1442 #ifdef CONFIG_F2FS_FS_XATTR
1443         set_opt(sbi, XATTR_USER);
1444 #endif
1445 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1446         set_opt(sbi, POSIX_ACL);
1447 #endif
1448
1449         f2fs_build_fault_attr(sbi, 0, 0);
1450 }
1451
1452 #ifdef CONFIG_QUOTA
1453 static int f2fs_enable_quotas(struct super_block *sb);
1454 #endif
1455
1456 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1457 {
1458         struct cp_control cpc;
1459         int err;
1460
1461         sbi->sb->s_flags |= SB_ACTIVE;
1462
1463         f2fs_update_time(sbi, DISABLE_TIME);
1464
1465         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1466                 mutex_lock(&sbi->gc_mutex);
1467                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1468                 if (err == -ENODATA)
1469                         break;
1470                 if (err && err != -EAGAIN)
1471                         return err;
1472         }
1473
1474         err = sync_filesystem(sbi->sb);
1475         if (err)
1476                 return err;
1477
1478         if (f2fs_disable_cp_again(sbi))
1479                 return -EAGAIN;
1480
1481         mutex_lock(&sbi->gc_mutex);
1482         cpc.reason = CP_PAUSE;
1483         set_sbi_flag(sbi, SBI_CP_DISABLED);
1484         f2fs_write_checkpoint(sbi, &cpc);
1485
1486         sbi->unusable_block_count = 0;
1487         mutex_unlock(&sbi->gc_mutex);
1488         return 0;
1489 }
1490
1491 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1492 {
1493         mutex_lock(&sbi->gc_mutex);
1494         f2fs_dirty_to_prefree(sbi);
1495
1496         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1497         set_sbi_flag(sbi, SBI_IS_DIRTY);
1498         mutex_unlock(&sbi->gc_mutex);
1499
1500         f2fs_sync_fs(sbi->sb, 1);
1501 }
1502
1503 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1504 {
1505         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1506         struct f2fs_mount_info org_mount_opt;
1507         unsigned long old_sb_flags;
1508         int err;
1509         bool need_restart_gc = false;
1510         bool need_stop_gc = false;
1511         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1512         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1513         bool checkpoint_changed;
1514 #ifdef CONFIG_QUOTA
1515         int i, j;
1516 #endif
1517
1518         /*
1519          * Save the old mount options in case we
1520          * need to restore them.
1521          */
1522         org_mount_opt = sbi->mount_opt;
1523         old_sb_flags = sb->s_flags;
1524
1525 #ifdef CONFIG_QUOTA
1526         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1527         for (i = 0; i < MAXQUOTAS; i++) {
1528                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1529                         org_mount_opt.s_qf_names[i] =
1530                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1531                                 GFP_KERNEL);
1532                         if (!org_mount_opt.s_qf_names[i]) {
1533                                 for (j = 0; j < i; j++)
1534                                         kvfree(org_mount_opt.s_qf_names[j]);
1535                                 return -ENOMEM;
1536                         }
1537                 } else {
1538                         org_mount_opt.s_qf_names[i] = NULL;
1539                 }
1540         }
1541 #endif
1542
1543         /* recover superblocks we couldn't write due to previous RO mount */
1544         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1545                 err = f2fs_commit_super(sbi, false);
1546                 f2fs_msg(sb, KERN_INFO,
1547                         "Try to recover all the superblocks, ret: %d", err);
1548                 if (!err)
1549                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1550         }
1551
1552         default_options(sbi);
1553
1554         /* parse mount options */
1555         err = parse_options(sb, data);
1556         if (err)
1557                 goto restore_opts;
1558         checkpoint_changed =
1559                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1560
1561         /*
1562          * Previous and new state of filesystem is RO,
1563          * so skip checking GC and FLUSH_MERGE conditions.
1564          */
1565         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1566                 goto skip;
1567
1568 #ifdef CONFIG_QUOTA
1569         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1570                 err = dquot_suspend(sb, -1);
1571                 if (err < 0)
1572                         goto restore_opts;
1573         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1574                 /* dquot_resume needs RW */
1575                 sb->s_flags &= ~SB_RDONLY;
1576                 if (sb_any_quota_suspended(sb)) {
1577                         dquot_resume(sb, -1);
1578                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1579                         err = f2fs_enable_quotas(sb);
1580                         if (err)
1581                                 goto restore_opts;
1582                 }
1583         }
1584 #endif
1585         /* disallow enable/disable extent_cache dynamically */
1586         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1587                 err = -EINVAL;
1588                 f2fs_msg(sbi->sb, KERN_WARNING,
1589                                 "switch extent_cache option is not allowed");
1590                 goto restore_opts;
1591         }
1592
1593         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1594                 err = -EINVAL;
1595                 f2fs_msg(sbi->sb, KERN_WARNING,
1596                         "disabling checkpoint not compatible with read-only");
1597                 goto restore_opts;
1598         }
1599
1600         /*
1601          * We stop the GC thread if FS is mounted as RO
1602          * or if background_gc = off is passed in mount
1603          * option. Also sync the filesystem.
1604          */
1605         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1606                 if (sbi->gc_thread) {
1607                         f2fs_stop_gc_thread(sbi);
1608                         need_restart_gc = true;
1609                 }
1610         } else if (!sbi->gc_thread) {
1611                 err = f2fs_start_gc_thread(sbi);
1612                 if (err)
1613                         goto restore_opts;
1614                 need_stop_gc = true;
1615         }
1616
1617         if (*flags & SB_RDONLY ||
1618                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1619                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1620                 sync_inodes_sb(sb);
1621
1622                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1623                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1624                 f2fs_sync_fs(sb, 1);
1625                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1626         }
1627
1628         if (checkpoint_changed) {
1629                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1630                         err = f2fs_disable_checkpoint(sbi);
1631                         if (err)
1632                                 goto restore_gc;
1633                 } else {
1634                         f2fs_enable_checkpoint(sbi);
1635                 }
1636         }
1637
1638         /*
1639          * We stop issue flush thread if FS is mounted as RO
1640          * or if flush_merge is not passed in mount option.
1641          */
1642         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1643                 clear_opt(sbi, FLUSH_MERGE);
1644                 f2fs_destroy_flush_cmd_control(sbi, false);
1645         } else {
1646                 err = f2fs_create_flush_cmd_control(sbi);
1647                 if (err)
1648                         goto restore_gc;
1649         }
1650 skip:
1651 #ifdef CONFIG_QUOTA
1652         /* Release old quota file names */
1653         for (i = 0; i < MAXQUOTAS; i++)
1654                 kvfree(org_mount_opt.s_qf_names[i]);
1655 #endif
1656         /* Update the POSIXACL Flag */
1657         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1658                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1659
1660         limit_reserve_root(sbi);
1661         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1662         return 0;
1663 restore_gc:
1664         if (need_restart_gc) {
1665                 if (f2fs_start_gc_thread(sbi))
1666                         f2fs_msg(sbi->sb, KERN_WARNING,
1667                                 "background gc thread has stopped");
1668         } else if (need_stop_gc) {
1669                 f2fs_stop_gc_thread(sbi);
1670         }
1671 restore_opts:
1672 #ifdef CONFIG_QUOTA
1673         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1674         for (i = 0; i < MAXQUOTAS; i++) {
1675                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1676                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1677         }
1678 #endif
1679         sbi->mount_opt = org_mount_opt;
1680         sb->s_flags = old_sb_flags;
1681         return err;
1682 }
1683
1684 #ifdef CONFIG_QUOTA
1685 /* Read data from quotafile */
1686 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1687                                size_t len, loff_t off)
1688 {
1689         struct inode *inode = sb_dqopt(sb)->files[type];
1690         struct address_space *mapping = inode->i_mapping;
1691         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1692         int offset = off & (sb->s_blocksize - 1);
1693         int tocopy;
1694         size_t toread;
1695         loff_t i_size = i_size_read(inode);
1696         struct page *page;
1697         char *kaddr;
1698
1699         if (off > i_size)
1700                 return 0;
1701
1702         if (off + len > i_size)
1703                 len = i_size - off;
1704         toread = len;
1705         while (toread > 0) {
1706                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1707 repeat:
1708                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1709                 if (IS_ERR(page)) {
1710                         if (PTR_ERR(page) == -ENOMEM) {
1711                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1712                                 goto repeat;
1713                         }
1714                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1715                         return PTR_ERR(page);
1716                 }
1717
1718                 lock_page(page);
1719
1720                 if (unlikely(page->mapping != mapping)) {
1721                         f2fs_put_page(page, 1);
1722                         goto repeat;
1723                 }
1724                 if (unlikely(!PageUptodate(page))) {
1725                         f2fs_put_page(page, 1);
1726                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1727                         return -EIO;
1728                 }
1729
1730                 kaddr = kmap_atomic(page);
1731                 memcpy(data, kaddr + offset, tocopy);
1732                 kunmap_atomic(kaddr);
1733                 f2fs_put_page(page, 1);
1734
1735                 offset = 0;
1736                 toread -= tocopy;
1737                 data += tocopy;
1738                 blkidx++;
1739         }
1740         return len;
1741 }
1742
1743 /* Write to quotafile */
1744 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1745                                 const char *data, size_t len, loff_t off)
1746 {
1747         struct inode *inode = sb_dqopt(sb)->files[type];
1748         struct address_space *mapping = inode->i_mapping;
1749         const struct address_space_operations *a_ops = mapping->a_ops;
1750         int offset = off & (sb->s_blocksize - 1);
1751         size_t towrite = len;
1752         struct page *page;
1753         char *kaddr;
1754         int err = 0;
1755         int tocopy;
1756
1757         while (towrite > 0) {
1758                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1759                                                                 towrite);
1760 retry:
1761                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1762                                                         &page, NULL);
1763                 if (unlikely(err)) {
1764                         if (err == -ENOMEM) {
1765                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1766                                 goto retry;
1767                         }
1768                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1769                         break;
1770                 }
1771
1772                 kaddr = kmap_atomic(page);
1773                 memcpy(kaddr + offset, data, tocopy);
1774                 kunmap_atomic(kaddr);
1775                 flush_dcache_page(page);
1776
1777                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1778                                                 page, NULL);
1779                 offset = 0;
1780                 towrite -= tocopy;
1781                 off += tocopy;
1782                 data += tocopy;
1783                 cond_resched();
1784         }
1785
1786         if (len == towrite)
1787                 return err;
1788         inode->i_mtime = inode->i_ctime = current_time(inode);
1789         f2fs_mark_inode_dirty_sync(inode, false);
1790         return len - towrite;
1791 }
1792
1793 static struct dquot **f2fs_get_dquots(struct inode *inode)
1794 {
1795         return F2FS_I(inode)->i_dquot;
1796 }
1797
1798 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1799 {
1800         return &F2FS_I(inode)->i_reserved_quota;
1801 }
1802
1803 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1804 {
1805         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1806                 f2fs_msg(sbi->sb, KERN_ERR,
1807                         "quota sysfile may be corrupted, skip loading it");
1808                 return 0;
1809         }
1810
1811         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1812                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1813 }
1814
1815 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1816 {
1817         int enabled = 0;
1818         int i, err;
1819
1820         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1821                 err = f2fs_enable_quotas(sbi->sb);
1822                 if (err) {
1823                         f2fs_msg(sbi->sb, KERN_ERR,
1824                                         "Cannot turn on quota_ino: %d", err);
1825                         return 0;
1826                 }
1827                 return 1;
1828         }
1829
1830         for (i = 0; i < MAXQUOTAS; i++) {
1831                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1832                         err = f2fs_quota_on_mount(sbi, i);
1833                         if (!err) {
1834                                 enabled = 1;
1835                                 continue;
1836                         }
1837                         f2fs_msg(sbi->sb, KERN_ERR,
1838                                 "Cannot turn on quotas: %d on %d", err, i);
1839                 }
1840         }
1841         return enabled;
1842 }
1843
1844 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1845                              unsigned int flags)
1846 {
1847         struct inode *qf_inode;
1848         unsigned long qf_inum;
1849         int err;
1850
1851         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1852
1853         qf_inum = f2fs_qf_ino(sb, type);
1854         if (!qf_inum)
1855                 return -EPERM;
1856
1857         qf_inode = f2fs_iget(sb, qf_inum);
1858         if (IS_ERR(qf_inode)) {
1859                 f2fs_msg(sb, KERN_ERR,
1860                         "Bad quota inode %u:%lu", type, qf_inum);
1861                 return PTR_ERR(qf_inode);
1862         }
1863
1864         /* Don't account quota for quota files to avoid recursion */
1865         qf_inode->i_flags |= S_NOQUOTA;
1866         err = dquot_enable(qf_inode, type, format_id, flags);
1867         iput(qf_inode);
1868         return err;
1869 }
1870
1871 static int f2fs_enable_quotas(struct super_block *sb)
1872 {
1873         int type, err = 0;
1874         unsigned long qf_inum;
1875         bool quota_mopt[MAXQUOTAS] = {
1876                 test_opt(F2FS_SB(sb), USRQUOTA),
1877                 test_opt(F2FS_SB(sb), GRPQUOTA),
1878                 test_opt(F2FS_SB(sb), PRJQUOTA),
1879         };
1880
1881         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1882                 f2fs_msg(sb, KERN_ERR,
1883                         "quota file may be corrupted, skip loading it");
1884                 return 0;
1885         }
1886
1887         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1888
1889         for (type = 0; type < MAXQUOTAS; type++) {
1890                 qf_inum = f2fs_qf_ino(sb, type);
1891                 if (qf_inum) {
1892                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1893                                 DQUOT_USAGE_ENABLED |
1894                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1895                         if (err) {
1896                                 f2fs_msg(sb, KERN_ERR,
1897                                         "Failed to enable quota tracking "
1898                                         "(type=%d, err=%d). Please run "
1899                                         "fsck to fix.", type, err);
1900                                 for (type--; type >= 0; type--)
1901                                         dquot_quota_off(sb, type);
1902                                 set_sbi_flag(F2FS_SB(sb),
1903                                                 SBI_QUOTA_NEED_REPAIR);
1904                                 return err;
1905                         }
1906                 }
1907         }
1908         return 0;
1909 }
1910
1911 int f2fs_quota_sync(struct super_block *sb, int type)
1912 {
1913         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1914         struct quota_info *dqopt = sb_dqopt(sb);
1915         int cnt;
1916         int ret;
1917
1918         ret = dquot_writeback_dquots(sb, type);
1919         if (ret)
1920                 goto out;
1921
1922         /*
1923          * Now when everything is written we can discard the pagecache so
1924          * that userspace sees the changes.
1925          */
1926         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1927                 struct address_space *mapping;
1928
1929                 if (type != -1 && cnt != type)
1930                         continue;
1931                 if (!sb_has_quota_active(sb, cnt))
1932                         continue;
1933
1934                 mapping = dqopt->files[cnt]->i_mapping;
1935
1936                 ret = filemap_fdatawrite(mapping);
1937                 if (ret)
1938                         goto out;
1939
1940                 /* if we are using journalled quota */
1941                 if (is_journalled_quota(sbi))
1942                         continue;
1943
1944                 ret = filemap_fdatawait(mapping);
1945                 if (ret)
1946                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1947
1948                 inode_lock(dqopt->files[cnt]);
1949                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1950                 inode_unlock(dqopt->files[cnt]);
1951         }
1952 out:
1953         if (ret)
1954                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1955         return ret;
1956 }
1957
1958 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1959                                                         const struct path *path)
1960 {
1961         struct inode *inode;
1962         int err;
1963
1964         err = f2fs_quota_sync(sb, type);
1965         if (err)
1966                 return err;
1967
1968         err = dquot_quota_on(sb, type, format_id, path);
1969         if (err)
1970                 return err;
1971
1972         inode = d_inode(path->dentry);
1973
1974         inode_lock(inode);
1975         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1976         f2fs_set_inode_flags(inode);
1977         inode_unlock(inode);
1978         f2fs_mark_inode_dirty_sync(inode, false);
1979
1980         return 0;
1981 }
1982
1983 static int f2fs_quota_off(struct super_block *sb, int type)
1984 {
1985         struct inode *inode = sb_dqopt(sb)->files[type];
1986         int err;
1987
1988         if (!inode || !igrab(inode))
1989                 return dquot_quota_off(sb, type);
1990
1991         err = f2fs_quota_sync(sb, type);
1992         if (err)
1993                 goto out_put;
1994
1995         err = dquot_quota_off(sb, type);
1996         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
1997                 goto out_put;
1998
1999         inode_lock(inode);
2000         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2001         f2fs_set_inode_flags(inode);
2002         inode_unlock(inode);
2003         f2fs_mark_inode_dirty_sync(inode, false);
2004 out_put:
2005         iput(inode);
2006         return err;
2007 }
2008
2009 void f2fs_quota_off_umount(struct super_block *sb)
2010 {
2011         int type;
2012         int err;
2013
2014         for (type = 0; type < MAXQUOTAS; type++) {
2015                 err = f2fs_quota_off(sb, type);
2016                 if (err) {
2017                         int ret = dquot_quota_off(sb, type);
2018
2019                         f2fs_msg(sb, KERN_ERR,
2020                                 "Fail to turn off disk quota "
2021                                 "(type: %d, err: %d, ret:%d), Please "
2022                                 "run fsck to fix it.", type, err, ret);
2023                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2024                 }
2025         }
2026 }
2027
2028 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2029 {
2030         struct quota_info *dqopt = sb_dqopt(sb);
2031         int type;
2032
2033         for (type = 0; type < MAXQUOTAS; type++) {
2034                 if (!dqopt->files[type])
2035                         continue;
2036                 f2fs_inode_synced(dqopt->files[type]);
2037         }
2038 }
2039
2040 static int f2fs_dquot_commit(struct dquot *dquot)
2041 {
2042         int ret;
2043
2044         ret = dquot_commit(dquot);
2045         if (ret < 0)
2046                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2047         return ret;
2048 }
2049
2050 static int f2fs_dquot_acquire(struct dquot *dquot)
2051 {
2052         int ret;
2053
2054         ret = dquot_acquire(dquot);
2055         if (ret < 0)
2056                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2057
2058         return ret;
2059 }
2060
2061 static int f2fs_dquot_release(struct dquot *dquot)
2062 {
2063         int ret;
2064
2065         ret = dquot_release(dquot);
2066         if (ret < 0)
2067                 set_sbi_flag(F2FS_SB(dquot->dq_sb), SBI_QUOTA_NEED_REPAIR);
2068         return ret;
2069 }
2070
2071 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2072 {
2073         struct super_block *sb = dquot->dq_sb;
2074         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2075         int ret;
2076
2077         ret = dquot_mark_dquot_dirty(dquot);
2078
2079         /* if we are using journalled quota */
2080         if (is_journalled_quota(sbi))
2081                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2082
2083         return ret;
2084 }
2085
2086 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2087 {
2088         int ret;
2089
2090         ret = dquot_commit_info(sb, type);
2091         if (ret < 0)
2092                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2093         return ret;
2094 }
2095
2096 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2097 {
2098         *projid = F2FS_I(inode)->i_projid;
2099         return 0;
2100 }
2101
2102 static const struct dquot_operations f2fs_quota_operations = {
2103         .get_reserved_space = f2fs_get_reserved_space,
2104         .write_dquot    = f2fs_dquot_commit,
2105         .acquire_dquot  = f2fs_dquot_acquire,
2106         .release_dquot  = f2fs_dquot_release,
2107         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2108         .write_info     = f2fs_dquot_commit_info,
2109         .alloc_dquot    = dquot_alloc,
2110         .destroy_dquot  = dquot_destroy,
2111         .get_projid     = f2fs_get_projid,
2112         .get_next_id    = dquot_get_next_id,
2113 };
2114
2115 static const struct quotactl_ops f2fs_quotactl_ops = {
2116         .quota_on       = f2fs_quota_on,
2117         .quota_off      = f2fs_quota_off,
2118         .quota_sync     = f2fs_quota_sync,
2119         .get_state      = dquot_get_state,
2120         .set_info       = dquot_set_dqinfo,
2121         .get_dqblk      = dquot_get_dqblk,
2122         .set_dqblk      = dquot_set_dqblk,
2123         .get_nextdqblk  = dquot_get_next_dqblk,
2124 };
2125 #else
2126 int f2fs_quota_sync(struct super_block *sb, int type)
2127 {
2128         return 0;
2129 }
2130
2131 void f2fs_quota_off_umount(struct super_block *sb)
2132 {
2133 }
2134 #endif
2135
2136 static const struct super_operations f2fs_sops = {
2137         .alloc_inode    = f2fs_alloc_inode,
2138         .drop_inode     = f2fs_drop_inode,
2139         .destroy_inode  = f2fs_destroy_inode,
2140         .write_inode    = f2fs_write_inode,
2141         .dirty_inode    = f2fs_dirty_inode,
2142         .show_options   = f2fs_show_options,
2143 #ifdef CONFIG_QUOTA
2144         .quota_read     = f2fs_quota_read,
2145         .quota_write    = f2fs_quota_write,
2146         .get_dquots     = f2fs_get_dquots,
2147 #endif
2148         .evict_inode    = f2fs_evict_inode,
2149         .put_super      = f2fs_put_super,
2150         .sync_fs        = f2fs_sync_fs,
2151         .freeze_fs      = f2fs_freeze,
2152         .unfreeze_fs    = f2fs_unfreeze,
2153         .statfs         = f2fs_statfs,
2154         .remount_fs     = f2fs_remount,
2155 };
2156
2157 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2158 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2159 {
2160         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2161                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2162                                 ctx, len, NULL);
2163 }
2164
2165 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2166                                                         void *fs_data)
2167 {
2168         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2169
2170         /*
2171          * Encrypting the root directory is not allowed because fsck
2172          * expects lost+found directory to exist and remain unencrypted
2173          * if LOST_FOUND feature is enabled.
2174          *
2175          */
2176         if (f2fs_sb_has_lost_found(sbi) &&
2177                         inode->i_ino == F2FS_ROOT_INO(sbi))
2178                 return -EPERM;
2179
2180         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2181                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2182                                 ctx, len, fs_data, XATTR_CREATE);
2183 }
2184
2185 static bool f2fs_dummy_context(struct inode *inode)
2186 {
2187         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2188 }
2189
2190 static const struct fscrypt_operations f2fs_cryptops = {
2191         .key_prefix     = "f2fs:",
2192         .get_context    = f2fs_get_context,
2193         .set_context    = f2fs_set_context,
2194         .dummy_context  = f2fs_dummy_context,
2195         .empty_dir      = f2fs_empty_dir,
2196         .max_namelen    = F2FS_NAME_LEN,
2197 };
2198 #endif
2199
2200 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2201                 u64 ino, u32 generation)
2202 {
2203         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2204         struct inode *inode;
2205
2206         if (f2fs_check_nid_range(sbi, ino))
2207                 return ERR_PTR(-ESTALE);
2208
2209         /*
2210          * f2fs_iget isn't quite right if the inode is currently unallocated!
2211          * However f2fs_iget currently does appropriate checks to handle stale
2212          * inodes so everything is OK.
2213          */
2214         inode = f2fs_iget(sb, ino);
2215         if (IS_ERR(inode))
2216                 return ERR_CAST(inode);
2217         if (unlikely(generation && inode->i_generation != generation)) {
2218                 /* we didn't find the right inode.. */
2219                 iput(inode);
2220                 return ERR_PTR(-ESTALE);
2221         }
2222         return inode;
2223 }
2224
2225 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2226                 int fh_len, int fh_type)
2227 {
2228         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2229                                     f2fs_nfs_get_inode);
2230 }
2231
2232 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2233                 int fh_len, int fh_type)
2234 {
2235         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2236                                     f2fs_nfs_get_inode);
2237 }
2238
2239 static const struct export_operations f2fs_export_ops = {
2240         .fh_to_dentry = f2fs_fh_to_dentry,
2241         .fh_to_parent = f2fs_fh_to_parent,
2242         .get_parent = f2fs_get_parent,
2243 };
2244
2245 static loff_t max_file_blocks(void)
2246 {
2247         loff_t result = 0;
2248         loff_t leaf_count = ADDRS_PER_BLOCK;
2249
2250         /*
2251          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2252          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2253          * space in inode.i_addr, it will be more safe to reassign
2254          * result as zero.
2255          */
2256
2257         /* two direct node blocks */
2258         result += (leaf_count * 2);
2259
2260         /* two indirect node blocks */
2261         leaf_count *= NIDS_PER_BLOCK;
2262         result += (leaf_count * 2);
2263
2264         /* one double indirect node block */
2265         leaf_count *= NIDS_PER_BLOCK;
2266         result += leaf_count;
2267
2268         return result;
2269 }
2270
2271 static int __f2fs_commit_super(struct buffer_head *bh,
2272                         struct f2fs_super_block *super)
2273 {
2274         lock_buffer(bh);
2275         if (super)
2276                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2277         set_buffer_dirty(bh);
2278         unlock_buffer(bh);
2279
2280         /* it's rare case, we can do fua all the time */
2281         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2282 }
2283
2284 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2285                                         struct buffer_head *bh)
2286 {
2287         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2288                                         (bh->b_data + F2FS_SUPER_OFFSET);
2289         struct super_block *sb = sbi->sb;
2290         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2291         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2292         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2293         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2294         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2295         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2296         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2297         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2298         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2299         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2300         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2301         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2302         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2303         u64 main_end_blkaddr = main_blkaddr +
2304                                 (segment_count_main << log_blocks_per_seg);
2305         u64 seg_end_blkaddr = segment0_blkaddr +
2306                                 (segment_count << log_blocks_per_seg);
2307
2308         if (segment0_blkaddr != cp_blkaddr) {
2309                 f2fs_msg(sb, KERN_INFO,
2310                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2311                         segment0_blkaddr, cp_blkaddr);
2312                 return true;
2313         }
2314
2315         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2316                                                         sit_blkaddr) {
2317                 f2fs_msg(sb, KERN_INFO,
2318                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2319                         cp_blkaddr, sit_blkaddr,
2320                         segment_count_ckpt << log_blocks_per_seg);
2321                 return true;
2322         }
2323
2324         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2325                                                         nat_blkaddr) {
2326                 f2fs_msg(sb, KERN_INFO,
2327                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2328                         sit_blkaddr, nat_blkaddr,
2329                         segment_count_sit << log_blocks_per_seg);
2330                 return true;
2331         }
2332
2333         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2334                                                         ssa_blkaddr) {
2335                 f2fs_msg(sb, KERN_INFO,
2336                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2337                         nat_blkaddr, ssa_blkaddr,
2338                         segment_count_nat << log_blocks_per_seg);
2339                 return true;
2340         }
2341
2342         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2343                                                         main_blkaddr) {
2344                 f2fs_msg(sb, KERN_INFO,
2345                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2346                         ssa_blkaddr, main_blkaddr,
2347                         segment_count_ssa << log_blocks_per_seg);
2348                 return true;
2349         }
2350
2351         if (main_end_blkaddr > seg_end_blkaddr) {
2352                 f2fs_msg(sb, KERN_INFO,
2353                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2354                         main_blkaddr,
2355                         segment0_blkaddr +
2356                                 (segment_count << log_blocks_per_seg),
2357                         segment_count_main << log_blocks_per_seg);
2358                 return true;
2359         } else if (main_end_blkaddr < seg_end_blkaddr) {
2360                 int err = 0;
2361                 char *res;
2362
2363                 /* fix in-memory information all the time */
2364                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2365                                 segment0_blkaddr) >> log_blocks_per_seg);
2366
2367                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2368                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2369                         res = "internally";
2370                 } else {
2371                         err = __f2fs_commit_super(bh, NULL);
2372                         res = err ? "failed" : "done";
2373                 }
2374                 f2fs_msg(sb, KERN_INFO,
2375                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
2376                         res, main_blkaddr,
2377                         segment0_blkaddr +
2378                                 (segment_count << log_blocks_per_seg),
2379                         segment_count_main << log_blocks_per_seg);
2380                 if (err)
2381                         return true;
2382         }
2383         return false;
2384 }
2385
2386 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2387                                 struct buffer_head *bh)
2388 {
2389         block_t segment_count, segs_per_sec, secs_per_zone;
2390         block_t total_sections, blocks_per_seg;
2391         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2392                                         (bh->b_data + F2FS_SUPER_OFFSET);
2393         struct super_block *sb = sbi->sb;
2394         unsigned int blocksize;
2395         size_t crc_offset = 0;
2396         __u32 crc = 0;
2397
2398         /* Check checksum_offset and crc in superblock */
2399         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2400                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2401                 if (crc_offset !=
2402                         offsetof(struct f2fs_super_block, crc)) {
2403                         f2fs_msg(sb, KERN_INFO,
2404                                 "Invalid SB checksum offset: %zu",
2405                                 crc_offset);
2406                         return 1;
2407                 }
2408                 crc = le32_to_cpu(raw_super->crc);
2409                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2410                         f2fs_msg(sb, KERN_INFO,
2411                                 "Invalid SB checksum value: %u", crc);
2412                         return 1;
2413                 }
2414         }
2415
2416         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2417                 f2fs_msg(sb, KERN_INFO,
2418                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
2419                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2420                 return 1;
2421         }
2422
2423         /* Currently, support only 4KB page cache size */
2424         if (F2FS_BLKSIZE != PAGE_SIZE) {
2425                 f2fs_msg(sb, KERN_INFO,
2426                         "Invalid page_cache_size (%lu), supports only 4KB\n",
2427                         PAGE_SIZE);
2428                 return 1;
2429         }
2430
2431         /* Currently, support only 4KB block size */
2432         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2433         if (blocksize != F2FS_BLKSIZE) {
2434                 f2fs_msg(sb, KERN_INFO,
2435                         "Invalid blocksize (%u), supports only 4KB\n",
2436                         blocksize);
2437                 return 1;
2438         }
2439
2440         /* check log blocks per segment */
2441         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2442                 f2fs_msg(sb, KERN_INFO,
2443                         "Invalid log blocks per segment (%u)\n",
2444                         le32_to_cpu(raw_super->log_blocks_per_seg));
2445                 return 1;
2446         }
2447
2448         /* Currently, support 512/1024/2048/4096 bytes sector size */
2449         if (le32_to_cpu(raw_super->log_sectorsize) >
2450                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2451                 le32_to_cpu(raw_super->log_sectorsize) <
2452                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2453                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2454                         le32_to_cpu(raw_super->log_sectorsize));
2455                 return 1;
2456         }
2457         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2458                 le32_to_cpu(raw_super->log_sectorsize) !=
2459                         F2FS_MAX_LOG_SECTOR_SIZE) {
2460                 f2fs_msg(sb, KERN_INFO,
2461                         "Invalid log sectors per block(%u) log sectorsize(%u)",
2462                         le32_to_cpu(raw_super->log_sectors_per_block),
2463                         le32_to_cpu(raw_super->log_sectorsize));
2464                 return 1;
2465         }
2466
2467         segment_count = le32_to_cpu(raw_super->segment_count);
2468         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2469         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2470         total_sections = le32_to_cpu(raw_super->section_count);
2471
2472         /* blocks_per_seg should be 512, given the above check */
2473         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2474
2475         if (segment_count > F2FS_MAX_SEGMENT ||
2476                                 segment_count < F2FS_MIN_SEGMENTS) {
2477                 f2fs_msg(sb, KERN_INFO,
2478                         "Invalid segment count (%u)",
2479                         segment_count);
2480                 return 1;
2481         }
2482
2483         if (total_sections > segment_count ||
2484                         total_sections < F2FS_MIN_SEGMENTS ||
2485                         segs_per_sec > segment_count || !segs_per_sec) {
2486                 f2fs_msg(sb, KERN_INFO,
2487                         "Invalid segment/section count (%u, %u x %u)",
2488                         segment_count, total_sections, segs_per_sec);
2489                 return 1;
2490         }
2491
2492         if ((segment_count / segs_per_sec) < total_sections) {
2493                 f2fs_msg(sb, KERN_INFO,
2494                         "Small segment_count (%u < %u * %u)",
2495                         segment_count, segs_per_sec, total_sections);
2496                 return 1;
2497         }
2498
2499         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2500                 f2fs_msg(sb, KERN_INFO,
2501                         "Wrong segment_count / block_count (%u > %llu)",
2502                         segment_count, le64_to_cpu(raw_super->block_count));
2503                 return 1;
2504         }
2505
2506         if (secs_per_zone > total_sections || !secs_per_zone) {
2507                 f2fs_msg(sb, KERN_INFO,
2508                         "Wrong secs_per_zone / total_sections (%u, %u)",
2509                         secs_per_zone, total_sections);
2510                 return 1;
2511         }
2512         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2513                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2514                         (le32_to_cpu(raw_super->extension_count) +
2515                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2516                 f2fs_msg(sb, KERN_INFO,
2517                         "Corrupted extension count (%u + %u > %u)",
2518                         le32_to_cpu(raw_super->extension_count),
2519                         raw_super->hot_ext_count,
2520                         F2FS_MAX_EXTENSION);
2521                 return 1;
2522         }
2523
2524         if (le32_to_cpu(raw_super->cp_payload) >
2525                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2526                 f2fs_msg(sb, KERN_INFO,
2527                         "Insane cp_payload (%u > %u)",
2528                         le32_to_cpu(raw_super->cp_payload),
2529                         blocks_per_seg - F2FS_CP_PACKS);
2530                 return 1;
2531         }
2532
2533         /* check reserved ino info */
2534         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2535                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2536                 le32_to_cpu(raw_super->root_ino) != 3) {
2537                 f2fs_msg(sb, KERN_INFO,
2538                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2539                         le32_to_cpu(raw_super->node_ino),
2540                         le32_to_cpu(raw_super->meta_ino),
2541                         le32_to_cpu(raw_super->root_ino));
2542                 return 1;
2543         }
2544
2545         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2546         if (sanity_check_area_boundary(sbi, bh))
2547                 return 1;
2548
2549         return 0;
2550 }
2551
2552 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2553 {
2554         unsigned int total, fsmeta;
2555         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2556         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2557         unsigned int ovp_segments, reserved_segments;
2558         unsigned int main_segs, blocks_per_seg;
2559         unsigned int sit_segs, nat_segs;
2560         unsigned int sit_bitmap_size, nat_bitmap_size;
2561         unsigned int log_blocks_per_seg;
2562         unsigned int segment_count_main;
2563         unsigned int cp_pack_start_sum, cp_payload;
2564         block_t user_block_count;
2565         int i, j;
2566
2567         total = le32_to_cpu(raw_super->segment_count);
2568         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2569         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2570         fsmeta += sit_segs;
2571         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2572         fsmeta += nat_segs;
2573         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2574         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2575
2576         if (unlikely(fsmeta >= total))
2577                 return 1;
2578
2579         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2580         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2581
2582         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2583                         ovp_segments == 0 || reserved_segments == 0)) {
2584                 f2fs_msg(sbi->sb, KERN_ERR,
2585                         "Wrong layout: check mkfs.f2fs version");
2586                 return 1;
2587         }
2588
2589         user_block_count = le64_to_cpu(ckpt->user_block_count);
2590         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2591         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2592         if (!user_block_count || user_block_count >=
2593                         segment_count_main << log_blocks_per_seg) {
2594                 f2fs_msg(sbi->sb, KERN_ERR,
2595                         "Wrong user_block_count: %u", user_block_count);
2596                 return 1;
2597         }
2598
2599         main_segs = le32_to_cpu(raw_super->segment_count_main);
2600         blocks_per_seg = sbi->blocks_per_seg;
2601
2602         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2603                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2604                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2605                         return 1;
2606                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2607                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2608                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2609                                 f2fs_msg(sbi->sb, KERN_ERR,
2610                                         "Node segment (%u, %u) has the same "
2611                                         "segno: %u", i, j,
2612                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2613                                 return 1;
2614                         }
2615                 }
2616         }
2617         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2618                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2619                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2620                         return 1;
2621                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2622                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2623                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2624                                 f2fs_msg(sbi->sb, KERN_ERR,
2625                                         "Data segment (%u, %u) has the same "
2626                                         "segno: %u", i, j,
2627                                         le32_to_cpu(ckpt->cur_data_segno[i]));
2628                                 return 1;
2629                         }
2630                 }
2631         }
2632         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2633                 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2634                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2635                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2636                                 f2fs_msg(sbi->sb, KERN_ERR,
2637                                         "Data segment (%u) and Data segment (%u)"
2638                                         " has the same segno: %u", i, j,
2639                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2640                                 return 1;
2641                         }
2642                 }
2643         }
2644
2645         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2646         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2647
2648         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2649                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2650                 f2fs_msg(sbi->sb, KERN_ERR,
2651                         "Wrong bitmap size: sit: %u, nat:%u",
2652                         sit_bitmap_size, nat_bitmap_size);
2653                 return 1;
2654         }
2655
2656         cp_pack_start_sum = __start_sum_addr(sbi);
2657         cp_payload = __cp_payload(sbi);
2658         if (cp_pack_start_sum < cp_payload + 1 ||
2659                 cp_pack_start_sum > blocks_per_seg - 1 -
2660                         NR_CURSEG_TYPE) {
2661                 f2fs_msg(sbi->sb, KERN_ERR,
2662                         "Wrong cp_pack_start_sum: %u",
2663                         cp_pack_start_sum);
2664                 return 1;
2665         }
2666
2667         if (unlikely(f2fs_cp_error(sbi))) {
2668                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2669                 return 1;
2670         }
2671         return 0;
2672 }
2673
2674 static void init_sb_info(struct f2fs_sb_info *sbi)
2675 {
2676         struct f2fs_super_block *raw_super = sbi->raw_super;
2677         int i;
2678
2679         sbi->log_sectors_per_block =
2680                 le32_to_cpu(raw_super->log_sectors_per_block);
2681         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2682         sbi->blocksize = 1 << sbi->log_blocksize;
2683         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2684         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2685         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2686         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2687         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2688         sbi->total_node_count =
2689                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2690                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2691         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2692         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2693         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2694         sbi->cur_victim_sec = NULL_SECNO;
2695         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2696         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2697         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2698         sbi->migration_granularity = sbi->segs_per_sec;
2699
2700         sbi->dir_level = DEF_DIR_LEVEL;
2701         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2702         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2703         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2704         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2705         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2706         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2707
2708         for (i = 0; i < NR_COUNT_TYPE; i++)
2709                 atomic_set(&sbi->nr_pages[i], 0);
2710
2711         for (i = 0; i < META; i++)
2712                 atomic_set(&sbi->wb_sync_req[i], 0);
2713
2714         INIT_LIST_HEAD(&sbi->s_list);
2715         mutex_init(&sbi->umount_mutex);
2716         init_rwsem(&sbi->io_order_lock);
2717         spin_lock_init(&sbi->cp_lock);
2718
2719         sbi->dirty_device = 0;
2720         spin_lock_init(&sbi->dev_lock);
2721
2722         init_rwsem(&sbi->sb_lock);
2723 }
2724
2725 static int init_percpu_info(struct f2fs_sb_info *sbi)
2726 {
2727         int err;
2728
2729         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2730         if (err)
2731                 return err;
2732
2733         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2734                                                                 GFP_KERNEL);
2735         if (err)
2736                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2737
2738         return err;
2739 }
2740
2741 #ifdef CONFIG_BLK_DEV_ZONED
2742 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2743 {
2744         struct block_device *bdev = FDEV(devi).bdev;
2745         sector_t nr_sectors = bdev->bd_part->nr_sects;
2746         sector_t sector = 0;
2747         struct blk_zone *zones;
2748         unsigned int i, nr_zones;
2749         unsigned int n = 0;
2750         int err = -EIO;
2751
2752         if (!f2fs_sb_has_blkzoned(sbi))
2753                 return 0;
2754
2755         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2756                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2757                 return -EINVAL;
2758         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2759         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2760                                 __ilog2_u32(sbi->blocks_per_blkz))
2761                 return -EINVAL;
2762         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2763         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2764                                         sbi->log_blocks_per_blkz;
2765         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2766                 FDEV(devi).nr_blkz++;
2767
2768         FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2769                                                                 GFP_KERNEL);
2770         if (!FDEV(devi).blkz_type)
2771                 return -ENOMEM;
2772
2773 #define F2FS_REPORT_NR_ZONES   4096
2774
2775         zones = f2fs_kzalloc(sbi,
2776                              array_size(F2FS_REPORT_NR_ZONES,
2777                                         sizeof(struct blk_zone)),
2778                              GFP_KERNEL);
2779         if (!zones)
2780                 return -ENOMEM;
2781
2782         /* Get block zones type */
2783         while (zones && sector < nr_sectors) {
2784
2785                 nr_zones = F2FS_REPORT_NR_ZONES;
2786                 err = blkdev_report_zones(bdev, sector,
2787                                           zones, &nr_zones,
2788                                           GFP_KERNEL);
2789                 if (err)
2790                         break;
2791                 if (!nr_zones) {
2792                         err = -EIO;
2793                         break;
2794                 }
2795
2796                 for (i = 0; i < nr_zones; i++) {
2797                         FDEV(devi).blkz_type[n] = zones[i].type;
2798                         sector += zones[i].len;
2799                         n++;
2800                 }
2801         }
2802
2803         kvfree(zones);
2804
2805         return err;
2806 }
2807 #endif
2808
2809 /*
2810  * Read f2fs raw super block.
2811  * Because we have two copies of super block, so read both of them
2812  * to get the first valid one. If any one of them is broken, we pass
2813  * them recovery flag back to the caller.
2814  */
2815 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2816                         struct f2fs_super_block **raw_super,
2817                         int *valid_super_block, int *recovery)
2818 {
2819         struct super_block *sb = sbi->sb;
2820         int block;
2821         struct buffer_head *bh;
2822         struct f2fs_super_block *super;
2823         int err = 0;
2824
2825         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2826         if (!super)
2827                 return -ENOMEM;
2828
2829         for (block = 0; block < 2; block++) {
2830                 bh = sb_bread(sb, block);
2831                 if (!bh) {
2832                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2833                                 block + 1);
2834                         err = -EIO;
2835                         continue;
2836                 }
2837
2838                 /* sanity checking of raw super */
2839                 if (sanity_check_raw_super(sbi, bh)) {
2840                         f2fs_msg(sb, KERN_ERR,
2841                                 "Can't find valid F2FS filesystem in %dth superblock",
2842                                 block + 1);
2843                         err = -EINVAL;
2844                         brelse(bh);
2845                         continue;
2846                 }
2847
2848                 if (!*raw_super) {
2849                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2850                                                         sizeof(*super));
2851                         *valid_super_block = block;
2852                         *raw_super = super;
2853                 }
2854                 brelse(bh);
2855         }
2856
2857         /* Fail to read any one of the superblocks*/
2858         if (err < 0)
2859                 *recovery = 1;
2860
2861         /* No valid superblock */
2862         if (!*raw_super)
2863                 kvfree(super);
2864         else
2865                 err = 0;
2866
2867         return err;
2868 }
2869
2870 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2871 {
2872         struct buffer_head *bh;
2873         __u32 crc = 0;
2874         int err;
2875
2876         if ((recover && f2fs_readonly(sbi->sb)) ||
2877                                 bdev_read_only(sbi->sb->s_bdev)) {
2878                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2879                 return -EROFS;
2880         }
2881
2882         /* we should update superblock crc here */
2883         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2884                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2885                                 offsetof(struct f2fs_super_block, crc));
2886                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2887         }
2888
2889         /* write back-up superblock first */
2890         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2891         if (!bh)
2892                 return -EIO;
2893         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2894         brelse(bh);
2895
2896         /* if we are in recovery path, skip writing valid superblock */
2897         if (recover || err)
2898                 return err;
2899
2900         /* write current valid superblock */
2901         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2902         if (!bh)
2903                 return -EIO;
2904         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2905         brelse(bh);
2906         return err;
2907 }
2908
2909 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2910 {
2911         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2912         unsigned int max_devices = MAX_DEVICES;
2913         int i;
2914
2915         /* Initialize single device information */
2916         if (!RDEV(0).path[0]) {
2917                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2918                         return 0;
2919                 max_devices = 1;
2920         }
2921
2922         /*
2923          * Initialize multiple devices information, or single
2924          * zoned block device information.
2925          */
2926         sbi->devs = f2fs_kzalloc(sbi,
2927                                  array_size(max_devices,
2928                                             sizeof(struct f2fs_dev_info)),
2929                                  GFP_KERNEL);
2930         if (!sbi->devs)
2931                 return -ENOMEM;
2932
2933         for (i = 0; i < max_devices; i++) {
2934
2935                 if (i > 0 && !RDEV(i).path[0])
2936                         break;
2937
2938                 if (max_devices == 1) {
2939                         /* Single zoned block device mount */
2940                         FDEV(0).bdev =
2941                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2942                                         sbi->sb->s_mode, sbi->sb->s_type);
2943                 } else {
2944                         /* Multi-device mount */
2945                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2946                         FDEV(i).total_segments =
2947                                 le32_to_cpu(RDEV(i).total_segments);
2948                         if (i == 0) {
2949                                 FDEV(i).start_blk = 0;
2950                                 FDEV(i).end_blk = FDEV(i).start_blk +
2951                                     (FDEV(i).total_segments <<
2952                                     sbi->log_blocks_per_seg) - 1 +
2953                                     le32_to_cpu(raw_super->segment0_blkaddr);
2954                         } else {
2955                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2956                                 FDEV(i).end_blk = FDEV(i).start_blk +
2957                                         (FDEV(i).total_segments <<
2958                                         sbi->log_blocks_per_seg) - 1;
2959                         }
2960                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2961                                         sbi->sb->s_mode, sbi->sb->s_type);
2962                 }
2963                 if (IS_ERR(FDEV(i).bdev))
2964                         return PTR_ERR(FDEV(i).bdev);
2965
2966                 /* to release errored devices */
2967                 sbi->s_ndevs = i + 1;
2968
2969 #ifdef CONFIG_BLK_DEV_ZONED
2970                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2971                                 !f2fs_sb_has_blkzoned(sbi)) {
2972                         f2fs_msg(sbi->sb, KERN_ERR,
2973                                 "Zoned block device feature not enabled\n");
2974                         return -EINVAL;
2975                 }
2976                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2977                         if (init_blkz_info(sbi, i)) {
2978                                 f2fs_msg(sbi->sb, KERN_ERR,
2979                                         "Failed to initialize F2FS blkzone information");
2980                                 return -EINVAL;
2981                         }
2982                         if (max_devices == 1)
2983                                 break;
2984                         f2fs_msg(sbi->sb, KERN_INFO,
2985                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2986                                 i, FDEV(i).path,
2987                                 FDEV(i).total_segments,
2988                                 FDEV(i).start_blk, FDEV(i).end_blk,
2989                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2990                                 "Host-aware" : "Host-managed");
2991                         continue;
2992                 }
2993 #endif
2994                 f2fs_msg(sbi->sb, KERN_INFO,
2995                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2996                                 i, FDEV(i).path,
2997                                 FDEV(i).total_segments,
2998                                 FDEV(i).start_blk, FDEV(i).end_blk);
2999         }
3000         f2fs_msg(sbi->sb, KERN_INFO,
3001                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3002         return 0;
3003 }
3004
3005 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3006 {
3007         struct f2fs_sm_info *sm_i = SM_I(sbi);
3008
3009         /* adjust parameters according to the volume size */
3010         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3011                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3012                 sm_i->dcc_info->discard_granularity = 1;
3013                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3014         }
3015
3016         sbi->readdir_ra = 1;
3017 }
3018
3019 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3020 {
3021         struct f2fs_sb_info *sbi;
3022         struct f2fs_super_block *raw_super;
3023         struct inode *root;
3024         int err;
3025         bool retry = true, need_fsck = false;
3026         char *options = NULL;
3027         int recovery, i, valid_super_block;
3028         struct curseg_info *seg_i;
3029
3030 try_onemore:
3031         err = -EINVAL;
3032         raw_super = NULL;
3033         valid_super_block = -1;
3034         recovery = 0;
3035
3036         /* allocate memory for f2fs-specific super block info */
3037         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3038         if (!sbi)
3039                 return -ENOMEM;
3040
3041         sbi->sb = sb;
3042
3043         /* Load the checksum driver */
3044         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3045         if (IS_ERR(sbi->s_chksum_driver)) {
3046                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
3047                 err = PTR_ERR(sbi->s_chksum_driver);
3048                 sbi->s_chksum_driver = NULL;
3049                 goto free_sbi;
3050         }
3051
3052         /* set a block size */
3053         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3054                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
3055                 goto free_sbi;
3056         }
3057
3058         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3059                                                                 &recovery);
3060         if (err)
3061                 goto free_sbi;
3062
3063         sb->s_fs_info = sbi;
3064         sbi->raw_super = raw_super;
3065
3066         /* precompute checksum seed for metadata */
3067         if (f2fs_sb_has_inode_chksum(sbi))
3068                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3069                                                 sizeof(raw_super->uuid));
3070
3071         /*
3072          * The BLKZONED feature indicates that the drive was formatted with
3073          * zone alignment optimization. This is optional for host-aware
3074          * devices, but mandatory for host-managed zoned block devices.
3075          */
3076 #ifndef CONFIG_BLK_DEV_ZONED
3077         if (f2fs_sb_has_blkzoned(sbi)) {
3078                 f2fs_msg(sb, KERN_ERR,
3079                          "Zoned block device support is not enabled\n");
3080                 err = -EOPNOTSUPP;
3081                 goto free_sb_buf;
3082         }
3083 #endif
3084         default_options(sbi);
3085         /* parse mount options */
3086         options = kstrdup((const char *)data, GFP_KERNEL);
3087         if (data && !options) {
3088                 err = -ENOMEM;
3089                 goto free_sb_buf;
3090         }
3091
3092         err = parse_options(sb, options);
3093         if (err)
3094                 goto free_options;
3095
3096         sbi->max_file_blocks = max_file_blocks();
3097         sb->s_maxbytes = sbi->max_file_blocks <<
3098                                 le32_to_cpu(raw_super->log_blocksize);
3099         sb->s_max_links = F2FS_LINK_MAX;
3100         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3101
3102 #ifdef CONFIG_QUOTA
3103         sb->dq_op = &f2fs_quota_operations;
3104         if (f2fs_sb_has_quota_ino(sbi))
3105                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3106         else
3107                 sb->s_qcop = &f2fs_quotactl_ops;
3108         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3109
3110         if (f2fs_sb_has_quota_ino(sbi)) {
3111                 for (i = 0; i < MAXQUOTAS; i++) {
3112                         if (f2fs_qf_ino(sbi->sb, i))
3113                                 sbi->nquota_files++;
3114                 }
3115         }
3116 #endif
3117
3118         sb->s_op = &f2fs_sops;
3119 #ifdef CONFIG_F2FS_FS_ENCRYPTION
3120         sb->s_cop = &f2fs_cryptops;
3121 #endif
3122         sb->s_xattr = f2fs_xattr_handlers;
3123         sb->s_export_op = &f2fs_export_ops;
3124         sb->s_magic = F2FS_SUPER_MAGIC;
3125         sb->s_time_gran = 1;
3126         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3127                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3128         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3129         sb->s_iflags |= SB_I_CGROUPWB;
3130
3131         /* init f2fs-specific super block info */
3132         sbi->valid_super_block = valid_super_block;
3133         mutex_init(&sbi->gc_mutex);
3134         mutex_init(&sbi->writepages);
3135         mutex_init(&sbi->cp_mutex);
3136         init_rwsem(&sbi->node_write);
3137         init_rwsem(&sbi->node_change);
3138
3139         /* disallow all the data/node/meta page writes */
3140         set_sbi_flag(sbi, SBI_POR_DOING);
3141         spin_lock_init(&sbi->stat_lock);
3142
3143         /* init iostat info */
3144         spin_lock_init(&sbi->iostat_lock);
3145         sbi->iostat_enable = false;
3146
3147         for (i = 0; i < NR_PAGE_TYPE; i++) {
3148                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3149                 int j;
3150
3151                 sbi->write_io[i] =
3152                         f2fs_kmalloc(sbi,
3153                                      array_size(n,
3154                                                 sizeof(struct f2fs_bio_info)),
3155                                      GFP_KERNEL);
3156                 if (!sbi->write_io[i]) {
3157                         err = -ENOMEM;
3158                         goto free_bio_info;
3159                 }
3160
3161                 for (j = HOT; j < n; j++) {
3162                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3163                         sbi->write_io[i][j].sbi = sbi;
3164                         sbi->write_io[i][j].bio = NULL;
3165                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3166                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3167                 }
3168         }
3169
3170         init_rwsem(&sbi->cp_rwsem);
3171         init_waitqueue_head(&sbi->cp_wait);
3172         init_sb_info(sbi);
3173
3174         err = init_percpu_info(sbi);
3175         if (err)
3176                 goto free_bio_info;
3177
3178         if (F2FS_IO_SIZE(sbi) > 1) {
3179                 sbi->write_io_dummy =
3180                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3181                 if (!sbi->write_io_dummy) {
3182                         err = -ENOMEM;
3183                         goto free_percpu;
3184                 }
3185         }
3186
3187         /* get an inode for meta space */
3188         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3189         if (IS_ERR(sbi->meta_inode)) {
3190                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
3191                 err = PTR_ERR(sbi->meta_inode);
3192                 goto free_io_dummy;
3193         }
3194
3195         err = f2fs_get_valid_checkpoint(sbi);
3196         if (err) {
3197                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
3198                 goto free_meta_inode;
3199         }
3200
3201         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3202                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3203
3204         /* Initialize device list */
3205         err = f2fs_scan_devices(sbi);
3206         if (err) {
3207                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
3208                 goto free_devices;
3209         }
3210
3211         sbi->total_valid_node_count =
3212                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3213         percpu_counter_set(&sbi->total_valid_inode_count,
3214                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3215         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3216         sbi->total_valid_block_count =
3217                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3218         sbi->last_valid_block_count = sbi->total_valid_block_count;
3219         sbi->reserved_blocks = 0;
3220         sbi->current_reserved_blocks = 0;
3221         limit_reserve_root(sbi);
3222
3223         for (i = 0; i < NR_INODE_TYPE; i++) {
3224                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3225                 spin_lock_init(&sbi->inode_lock[i]);
3226         }
3227
3228         f2fs_init_extent_cache_info(sbi);
3229
3230         f2fs_init_ino_entry_info(sbi);
3231
3232         f2fs_init_fsync_node_info(sbi);
3233
3234         /* setup f2fs internal modules */
3235         err = f2fs_build_segment_manager(sbi);
3236         if (err) {
3237                 f2fs_msg(sb, KERN_ERR,
3238                         "Failed to initialize F2FS segment manager");
3239                 goto free_sm;
3240         }
3241         err = f2fs_build_node_manager(sbi);
3242         if (err) {
3243                 f2fs_msg(sb, KERN_ERR,
3244                         "Failed to initialize F2FS node manager");
3245                 goto free_nm;
3246         }
3247
3248         /* For write statistics */
3249         if (sb->s_bdev->bd_part)
3250                 sbi->sectors_written_start =
3251                         (u64)part_stat_read(sb->s_bdev->bd_part,
3252                                             sectors[STAT_WRITE]);
3253
3254         /* Read accumulated write IO statistics if exists */
3255         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3256         if (__exist_node_summaries(sbi))
3257                 sbi->kbytes_written =
3258                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3259
3260         f2fs_build_gc_manager(sbi);
3261
3262         err = f2fs_build_stats(sbi);
3263         if (err)
3264                 goto free_nm;
3265
3266         /* get an inode for node space */
3267         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3268         if (IS_ERR(sbi->node_inode)) {
3269                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3270                 err = PTR_ERR(sbi->node_inode);
3271                 goto free_stats;
3272         }
3273
3274         /* read root inode and dentry */
3275         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3276         if (IS_ERR(root)) {
3277                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3278                 err = PTR_ERR(root);
3279                 goto free_node_inode;
3280         }
3281         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3282                         !root->i_size || !root->i_nlink) {
3283                 iput(root);
3284                 err = -EINVAL;
3285                 goto free_node_inode;
3286         }
3287
3288         sb->s_root = d_make_root(root); /* allocate root dentry */
3289         if (!sb->s_root) {
3290                 err = -ENOMEM;
3291                 goto free_root_inode;
3292         }
3293
3294         err = f2fs_register_sysfs(sbi);
3295         if (err)
3296                 goto free_root_inode;
3297
3298 #ifdef CONFIG_QUOTA
3299         /* Enable quota usage during mount */
3300         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3301                 err = f2fs_enable_quotas(sb);
3302                 if (err)
3303                         f2fs_msg(sb, KERN_ERR,
3304                                 "Cannot turn on quotas: error %d", err);
3305         }
3306 #endif
3307         /* if there are nt orphan nodes free them */
3308         err = f2fs_recover_orphan_inodes(sbi);
3309         if (err)
3310                 goto free_meta;
3311
3312         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3313                 goto skip_recovery;
3314
3315         /* recover fsynced data */
3316         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3317                 /*
3318                  * mount should be failed, when device has readonly mode, and
3319                  * previous checkpoint was not done by clean system shutdown.
3320                  */
3321                 if (bdev_read_only(sb->s_bdev) &&
3322                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3323                         err = -EROFS;
3324                         goto free_meta;
3325                 }
3326
3327                 if (need_fsck)
3328                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3329
3330                 if (!retry)
3331                         goto skip_recovery;
3332
3333                 err = f2fs_recover_fsync_data(sbi, false);
3334                 if (err < 0) {
3335                         need_fsck = true;
3336                         f2fs_msg(sb, KERN_ERR,
3337                                 "Cannot recover all fsync data errno=%d", err);
3338                         goto free_meta;
3339                 }
3340         } else {
3341                 err = f2fs_recover_fsync_data(sbi, true);
3342
3343                 if (!f2fs_readonly(sb) && err > 0) {
3344                         err = -EINVAL;
3345                         f2fs_msg(sb, KERN_ERR,
3346                                 "Need to recover fsync data");
3347                         goto free_meta;
3348                 }
3349         }
3350 skip_recovery:
3351         /* f2fs_recover_fsync_data() cleared this already */
3352         clear_sbi_flag(sbi, SBI_POR_DOING);
3353
3354         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3355                 err = f2fs_disable_checkpoint(sbi);
3356                 if (err)
3357                         goto free_meta;
3358         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3359                 f2fs_enable_checkpoint(sbi);
3360         }
3361
3362         /*
3363          * If filesystem is not mounted as read-only then
3364          * do start the gc_thread.
3365          */
3366         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3367                 /* After POR, we can run background GC thread.*/
3368                 err = f2fs_start_gc_thread(sbi);
3369                 if (err)
3370                         goto free_meta;
3371         }
3372         kvfree(options);
3373
3374         /* recover broken superblock */
3375         if (recovery) {
3376                 err = f2fs_commit_super(sbi, true);
3377                 f2fs_msg(sb, KERN_INFO,
3378                         "Try to recover %dth superblock, ret: %d",
3379                         sbi->valid_super_block ? 1 : 2, err);
3380         }
3381
3382         f2fs_join_shrinker(sbi);
3383
3384         f2fs_tuning_parameters(sbi);
3385
3386         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3387                                 cur_cp_version(F2FS_CKPT(sbi)));
3388         f2fs_update_time(sbi, CP_TIME);
3389         f2fs_update_time(sbi, REQ_TIME);
3390         return 0;
3391
3392 free_meta:
3393 #ifdef CONFIG_QUOTA
3394         f2fs_truncate_quota_inode_pages(sb);
3395         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3396                 f2fs_quota_off_umount(sbi->sb);
3397 #endif
3398         /*
3399          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3400          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3401          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3402          * falls into an infinite loop in f2fs_sync_meta_pages().
3403          */
3404         truncate_inode_pages_final(META_MAPPING(sbi));
3405         f2fs_unregister_sysfs(sbi);
3406 free_root_inode:
3407         dput(sb->s_root);
3408         sb->s_root = NULL;
3409 free_node_inode:
3410         f2fs_release_ino_entry(sbi, true);
3411         truncate_inode_pages_final(NODE_MAPPING(sbi));
3412         iput(sbi->node_inode);
3413 free_stats:
3414         f2fs_destroy_stats(sbi);
3415 free_nm:
3416         f2fs_destroy_node_manager(sbi);
3417 free_sm:
3418         f2fs_destroy_segment_manager(sbi);
3419 free_devices:
3420         destroy_device_list(sbi);
3421         kvfree(sbi->ckpt);
3422 free_meta_inode:
3423         make_bad_inode(sbi->meta_inode);
3424         iput(sbi->meta_inode);
3425 free_io_dummy:
3426         mempool_destroy(sbi->write_io_dummy);
3427 free_percpu:
3428         destroy_percpu_info(sbi);
3429 free_bio_info:
3430         for (i = 0; i < NR_PAGE_TYPE; i++)
3431                 kvfree(sbi->write_io[i]);
3432 free_options:
3433 #ifdef CONFIG_QUOTA
3434         for (i = 0; i < MAXQUOTAS; i++)
3435                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3436 #endif
3437         kvfree(options);
3438 free_sb_buf:
3439         kvfree(raw_super);
3440 free_sbi:
3441         if (sbi->s_chksum_driver)
3442                 crypto_free_shash(sbi->s_chksum_driver);
3443         kvfree(sbi);
3444
3445         /* give only one another chance */
3446         if (retry) {
3447                 retry = false;
3448                 shrink_dcache_sb(sb);
3449                 goto try_onemore;
3450         }
3451         return err;
3452 }
3453
3454 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3455                         const char *dev_name, void *data)
3456 {
3457         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3458 }
3459
3460 static void kill_f2fs_super(struct super_block *sb)
3461 {
3462         if (sb->s_root) {
3463                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3464
3465                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3466                 f2fs_stop_gc_thread(sbi);
3467                 f2fs_stop_discard_thread(sbi);
3468
3469                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3470                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3471                         struct cp_control cpc = {
3472                                 .reason = CP_UMOUNT,
3473                         };
3474                         f2fs_write_checkpoint(sbi, &cpc);
3475                 }
3476
3477                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3478                         sb->s_flags &= ~SB_RDONLY;
3479         }
3480         kill_block_super(sb);
3481 }
3482
3483 static struct file_system_type f2fs_fs_type = {
3484         .owner          = THIS_MODULE,
3485         .name           = "f2fs",
3486         .mount          = f2fs_mount,
3487         .kill_sb        = kill_f2fs_super,
3488         .fs_flags       = FS_REQUIRES_DEV,
3489 };
3490 MODULE_ALIAS_FS("f2fs");
3491
3492 static int __init init_inodecache(void)
3493 {
3494         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3495                         sizeof(struct f2fs_inode_info), 0,
3496                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3497         if (!f2fs_inode_cachep)
3498                 return -ENOMEM;
3499         return 0;
3500 }
3501
3502 static void destroy_inodecache(void)
3503 {
3504         /*
3505          * Make sure all delayed rcu free inodes are flushed before we
3506          * destroy cache.
3507          */
3508         rcu_barrier();
3509         kmem_cache_destroy(f2fs_inode_cachep);
3510 }
3511
3512 static int __init init_f2fs_fs(void)
3513 {
3514         int err;
3515
3516         if (PAGE_SIZE != F2FS_BLKSIZE) {
3517                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3518                                 PAGE_SIZE, F2FS_BLKSIZE);
3519                 return -EINVAL;
3520         }
3521
3522         f2fs_build_trace_ios();
3523
3524         err = init_inodecache();
3525         if (err)
3526                 goto fail;
3527         err = f2fs_create_node_manager_caches();
3528         if (err)
3529                 goto free_inodecache;
3530         err = f2fs_create_segment_manager_caches();
3531         if (err)
3532                 goto free_node_manager_caches;
3533         err = f2fs_create_checkpoint_caches();
3534         if (err)
3535                 goto free_segment_manager_caches;
3536         err = f2fs_create_extent_cache();
3537         if (err)
3538                 goto free_checkpoint_caches;
3539         err = f2fs_init_sysfs();
3540         if (err)
3541                 goto free_extent_cache;
3542         err = register_shrinker(&f2fs_shrinker_info);
3543         if (err)
3544                 goto free_sysfs;
3545         err = register_filesystem(&f2fs_fs_type);
3546         if (err)
3547                 goto free_shrinker;
3548         f2fs_create_root_stats();
3549         err = f2fs_init_post_read_processing();
3550         if (err)
3551                 goto free_root_stats;
3552         return 0;
3553
3554 free_root_stats:
3555         f2fs_destroy_root_stats();
3556         unregister_filesystem(&f2fs_fs_type);
3557 free_shrinker:
3558         unregister_shrinker(&f2fs_shrinker_info);
3559 free_sysfs:
3560         f2fs_exit_sysfs();
3561 free_extent_cache:
3562         f2fs_destroy_extent_cache();
3563 free_checkpoint_caches:
3564         f2fs_destroy_checkpoint_caches();
3565 free_segment_manager_caches:
3566         f2fs_destroy_segment_manager_caches();
3567 free_node_manager_caches:
3568         f2fs_destroy_node_manager_caches();
3569 free_inodecache:
3570         destroy_inodecache();
3571 fail:
3572         return err;
3573 }
3574
3575 static void __exit exit_f2fs_fs(void)
3576 {
3577         f2fs_destroy_post_read_processing();
3578         f2fs_destroy_root_stats();
3579         unregister_filesystem(&f2fs_fs_type);
3580         unregister_shrinker(&f2fs_shrinker_info);
3581         f2fs_exit_sysfs();
3582         f2fs_destroy_extent_cache();
3583         f2fs_destroy_checkpoint_caches();
3584         f2fs_destroy_segment_manager_caches();
3585         f2fs_destroy_node_manager_caches();
3586         destroy_inodecache();
3587         f2fs_destroy_trace_ios();
3588 }
3589
3590 module_init(init_f2fs_fs)
3591 module_exit(exit_f2fs_fs)
3592
3593 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3594 MODULE_DESCRIPTION("Flash Friendly File System");
3595 MODULE_LICENSE("GPL");
3596