fs/f2fs/super.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * fs/f2fs/super.c
   4  *
   5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   6  *             http://www.samsung.com/
   7  */
   8 #include <linux/module.h>
   9 #include <linux/init.h>
  10 #include <linux/fs.h>
  11 #include <linux/statfs.h>
  12 #include <linux/buffer_head.h>
  13 #include <linux/backing-dev.h>
  14 #include <linux/kthread.h>
  15 #include <linux/parser.h>
  16 #include <linux/mount.h>
  17 #include <linux/seq_file.h>
  18 #include <linux/proc_fs.h>
  19 #include <linux/random.h>
  20 #include <linux/exportfs.h>
  21 #include <linux/blkdev.h>
  22 #include <linux/quotaops.h>
  23 #include <linux/f2fs_fs.h>
  24 #include <linux/sysfs.h>
  25 #include <linux/quota.h>
  26
  27 #include "f2fs.h"
  28 #include "node.h"
  29 #include "segment.h"
  30 #include "xattr.h"
  31 #include "gc.h"
  32 #include "trace.h"
  33
  34 #define CREATE_TRACE_POINTS
  35 #include <trace/events/f2fs.h>
  36
  37 static struct kmem_cache *f2fs_inode_cachep;
  38
  39 #ifdef CONFIG_F2FS_FAULT_INJECTION
  40
  41 const char *f2fs_fault_name[FAULT_MAX] = {
  42         [FAULT_KMALLOC]         = "kmalloc",
  43         [FAULT_KVMALLOC]        = "kvmalloc",
  44         [FAULT_PAGE_ALLOC]      = "page alloc",
  45         [FAULT_PAGE_GET]        = "page get",
  46         [FAULT_ALLOC_BIO]       = "alloc bio",
  47         [FAULT_ALLOC_NID]       = "alloc nid",
  48         [FAULT_ORPHAN]          = "orphan",
  49         [FAULT_BLOCK]           = "no more block",
  50         [FAULT_DIR_DEPTH]       = "too big dir depth",
  51         [FAULT_EVICT_INODE]     = "evict_inode fail",
  52         [FAULT_TRUNCATE]        = "truncate fail",
  53         [FAULT_READ_IO]         = "read IO error",
  54         [FAULT_CHECKPOINT]      = "checkpoint error",
  55         [FAULT_DISCARD]         = "discard error",
  56         [FAULT_WRITE_IO]        = "write IO error",
  57 };
  58
  59 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  60                                                         unsigned int type)
  61 {
  62         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  63
  64         if (rate) {
  65                 atomic_set(&ffi->inject_ops, 0);
  66                 ffi->inject_rate = rate;
  67         }
  68
  69         if (type)
  70                 ffi->inject_type = type;
  71
  72         if (!rate && !type)
  73                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  74 }
  75 #endif
  76
  77 /* f2fs-wide shrinker description */
  78 static struct shrinker f2fs_shrinker_info = {
  79         .scan_objects = f2fs_shrink_scan,
  80         .count_objects = f2fs_shrink_count,
  81         .seeks = DEFAULT_SEEKS,
  82 };
  83
  84 enum {
  85         Opt_gc_background,
  86         Opt_disable_roll_forward,
  87         Opt_norecovery,
  88         Opt_discard,
  89         Opt_nodiscard,
  90         Opt_noheap,
  91         Opt_heap,
  92         Opt_user_xattr,
  93         Opt_nouser_xattr,
  94         Opt_acl,
  95         Opt_noacl,
  96         Opt_active_logs,
  97         Opt_disable_ext_identify,
  98         Opt_inline_xattr,
  99         Opt_noinline_xattr,
 100         Opt_inline_xattr_size,
 101         Opt_inline_data,
 102         Opt_inline_dentry,
 103         Opt_noinline_dentry,
 104         Opt_flush_merge,
 105         Opt_noflush_merge,
 106         Opt_nobarrier,
 107         Opt_fastboot,
 108         Opt_extent_cache,
 109         Opt_noextent_cache,
 110         Opt_noinline_data,
 111         Opt_data_flush,
 112         Opt_reserve_root,
 113         Opt_resgid,
 114         Opt_resuid,
 115         Opt_mode,
 116         Opt_io_size_bits,
 117         Opt_fault_injection,
 118         Opt_fault_type,
 119         Opt_lazytime,
 120         Opt_nolazytime,
 121         Opt_quota,
 122         Opt_noquota,
 123         Opt_usrquota,
 124         Opt_grpquota,
 125         Opt_prjquota,
 126         Opt_usrjquota,
 127         Opt_grpjquota,
 128         Opt_prjjquota,
 129         Opt_offusrjquota,
 130         Opt_offgrpjquota,
 131         Opt_offprjjquota,
 132         Opt_jqfmt_vfsold,
 133         Opt_jqfmt_vfsv0,
 134         Opt_jqfmt_vfsv1,
 135         Opt_whint,
 136         Opt_alloc,
 137         Opt_fsync,
 138         Opt_test_dummy_encryption,
 139         Opt_checkpoint_disable,
 140         Opt_checkpoint_disable_cap,
 141         Opt_checkpoint_disable_cap_perc,
 142         Opt_checkpoint_enable,
 143         Opt_err,
 144 };
 145
 146 static match_table_t f2fs_tokens = {
 147         {Opt_gc_background, "background_gc=%s"},
 148         {Opt_disable_roll_forward, "disable_roll_forward"},
 149         {Opt_norecovery, "norecovery"},
 150         {Opt_discard, "discard"},
 151         {Opt_nodiscard, "nodiscard"},
 152         {Opt_noheap, "no_heap"},
 153         {Opt_heap, "heap"},
 154         {Opt_user_xattr, "user_xattr"},
 155         {Opt_nouser_xattr, "nouser_xattr"},
 156         {Opt_acl, "acl"},
 157         {Opt_noacl, "noacl"},
 158         {Opt_active_logs, "active_logs=%u"},
 159         {Opt_disable_ext_identify, "disable_ext_identify"},
 160         {Opt_inline_xattr, "inline_xattr"},
 161         {Opt_noinline_xattr, "noinline_xattr"},
 162         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 163         {Opt_inline_data, "inline_data"},
 164         {Opt_inline_dentry, "inline_dentry"},
 165         {Opt_noinline_dentry, "noinline_dentry"},
 166         {Opt_flush_merge, "flush_merge"},
 167         {Opt_noflush_merge, "noflush_merge"},
 168         {Opt_nobarrier, "nobarrier"},
 169         {Opt_fastboot, "fastboot"},
 170         {Opt_extent_cache, "extent_cache"},
 171         {Opt_noextent_cache, "noextent_cache"},
 172         {Opt_noinline_data, "noinline_data"},
 173         {Opt_data_flush, "data_flush"},
 174         {Opt_reserve_root, "reserve_root=%u"},
 175         {Opt_resgid, "resgid=%u"},
 176         {Opt_resuid, "resuid=%u"},
 177         {Opt_mode, "mode=%s"},
 178         {Opt_io_size_bits, "io_bits=%u"},
 179         {Opt_fault_injection, "fault_injection=%u"},
 180         {Opt_fault_type, "fault_type=%u"},
 181         {Opt_lazytime, "lazytime"},
 182         {Opt_nolazytime, "nolazytime"},
 183         {Opt_quota, "quota"},
 184         {Opt_noquota, "noquota"},
 185         {Opt_usrquota, "usrquota"},
 186         {Opt_grpquota, "grpquota"},
 187         {Opt_prjquota, "prjquota"},
 188         {Opt_usrjquota, "usrjquota=%s"},
 189         {Opt_grpjquota, "grpjquota=%s"},
 190         {Opt_prjjquota, "prjjquota=%s"},
 191         {Opt_offusrjquota, "usrjquota="},
 192         {Opt_offgrpjquota, "grpjquota="},
 193         {Opt_offprjjquota, "prjjquota="},
 194         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 195         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 196         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 197         {Opt_whint, "whint_mode=%s"},
 198         {Opt_alloc, "alloc_mode=%s"},
 199         {Opt_fsync, "fsync_mode=%s"},
 200         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 201         {Opt_checkpoint_disable, "checkpoint=disable"},
 202         {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 203         {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 204         {Opt_checkpoint_enable, "checkpoint=enable"},
 205         {Opt_err, NULL},
 206 };
 207
 208 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 209 {
 210         struct va_format vaf;
 211         va_list args;
 212         int level;
 213
 214         va_start(args, fmt);
 215
 216         level = printk_get_level(fmt);
 217         vaf.fmt = printk_skip_level(fmt);
 218         vaf.va = &args;
 219         printk("%c%cF2FS-fs (%s): %pV\n",
 220                KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 221
 222         va_end(args);
 223 }
 224
 225 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 226 {
 227         block_t limit = min((sbi->user_block_count << 1) / 1000,
 228                         sbi->user_block_count - sbi->reserved_blocks);
 229
 230         /* limit is 0.2% */
 231         if (test_opt(sbi, RESERVE_ROOT) &&
 232                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 233                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 234                 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 235                           F2FS_OPTION(sbi).root_reserved_blocks);
 236         }
 237         if (!test_opt(sbi, RESERVE_ROOT) &&
 238                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 239                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 240                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 241                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 242                 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 243                           from_kuid_munged(&init_user_ns,
 244                                            F2FS_OPTION(sbi).s_resuid),
 245                           from_kgid_munged(&init_user_ns,
 246                                            F2FS_OPTION(sbi).s_resgid));
 247 }
 248
 249 static void init_once(void *foo)
 250 {
 251         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 252
 253         inode_init_once(&fi->vfs_inode);
 254 }
 255
 256 #ifdef CONFIG_QUOTA
 257 static const char * const quotatypes[] = INITQFNAMES;
 258 #define QTYPE2NAME(t) (quotatypes[t])
 259 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 260                                                         substring_t *args)
 261 {
 262         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 263         char *qname;
 264         int ret = -EINVAL;
 265
 266         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 267                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 268                 return -EINVAL;
 269         }
 270         if (f2fs_sb_has_quota_ino(sbi)) {
 271                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 272                 return 0;
 273         }
 274
 275         qname = match_strdup(args);
 276         if (!qname) {
 277                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 278                 return -ENOMEM;
 279         }
 280         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 281                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 282                         ret = 0;
 283                 else
 284                         f2fs_err(sbi, "%s quota file already specified",
 285                                  QTYPE2NAME(qtype));
 286                 goto errout;
 287         }
 288         if (strchr(qname, '/')) {
 289                 f2fs_err(sbi, "quotafile must be on filesystem root");
 290                 goto errout;
 291         }
 292         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 293         set_opt(sbi, QUOTA);
 294         return 0;
 295 errout:
 296         kvfree(qname);
 297         return ret;
 298 }
 299
 300 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 301 {
 302         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 303
 304         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 305                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 306                 return -EINVAL;
 307         }
 308         kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 309         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 310         return 0;
 311 }
 312
 313 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 314 {
 315         /*
 316          * We do the test below only for project quotas. 'usrquota' and
 317          * 'grpquota' mount options are allowed even without quota feature
 318          * to support legacy quotas in quota files.
 319          */
 320         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 321                 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 322                 return -1;
 323         }
 324         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 325                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 326                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 327                 if (test_opt(sbi, USRQUOTA) &&
 328                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 329                         clear_opt(sbi, USRQUOTA);
 330
 331                 if (test_opt(sbi, GRPQUOTA) &&
 332                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 333                         clear_opt(sbi, GRPQUOTA);
 334
 335                 if (test_opt(sbi, PRJQUOTA) &&
 336                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 337                         clear_opt(sbi, PRJQUOTA);
 338
 339                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 340                                 test_opt(sbi, PRJQUOTA)) {
 341                         f2fs_err(sbi, "old and new quota format mixing");
 342                         return -1;
 343                 }
 344
 345                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 346                         f2fs_err(sbi, "journaled quota format not specified");
 347                         return -1;
 348                 }
 349         }
 350
 351         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 352                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 353                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 354         }
 355         return 0;
 356 }
 357 #endif
 358
 359 static int parse_options(struct super_block *sb, char *options)
 360 {
 361         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 362         substring_t args[MAX_OPT_ARGS];
 363         char *p, *name;
 364         int arg = 0;
 365         kuid_t uid;
 366         kgid_t gid;
 367 #ifdef CONFIG_QUOTA
 368         int ret;
 369 #endif
 370
 371         if (!options)
 372                 return 0;
 373
 374         while ((p = strsep(&options, ",")) != NULL) {
 375                 int token;
 376                 if (!*p)
 377                         continue;
 378                 /*
 379                  * Initialize args struct so we know whether arg was
 380                  * found; some options take optional arguments.
 381                  */
 382                 args[0].to = args[0].from = NULL;
 383                 token = match_token(p, f2fs_tokens, args);
 384
 385                 switch (token) {
 386                 case Opt_gc_background:
 387                         name = match_strdup(&args[0]);
 388
 389                         if (!name)
 390                                 return -ENOMEM;
 391                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 392                                 set_opt(sbi, BG_GC);
 393                                 clear_opt(sbi, FORCE_FG_GC);
 394                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 395                                 clear_opt(sbi, BG_GC);
 396                                 clear_opt(sbi, FORCE_FG_GC);
 397                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 398                                 set_opt(sbi, BG_GC);
 399                                 set_opt(sbi, FORCE_FG_GC);
 400                         } else {
 401                                 kvfree(name);
 402                                 return -EINVAL;
 403                         }
 404                         kvfree(name);
 405                         break;
 406                 case Opt_disable_roll_forward:
 407                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 408                         break;
 409                 case Opt_norecovery:
 410                         /* this option mounts f2fs with ro */
 411                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 412                         if (!f2fs_readonly(sb))
 413                                 return -EINVAL;
 414                         break;
 415                 case Opt_discard:
 416                         set_opt(sbi, DISCARD);
 417                         break;
 418                 case Opt_nodiscard:
 419                         if (f2fs_sb_has_blkzoned(sbi)) {
 420                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 421                                 return -EINVAL;
 422                         }
 423                         clear_opt(sbi, DISCARD);
 424                         break;
 425                 case Opt_noheap:
 426                         set_opt(sbi, NOHEAP);
 427                         break;
 428                 case Opt_heap:
 429                         clear_opt(sbi, NOHEAP);
 430                         break;
 431 #ifdef CONFIG_F2FS_FS_XATTR
 432                 case Opt_user_xattr:
 433                         set_opt(sbi, XATTR_USER);
 434                         break;
 435                 case Opt_nouser_xattr:
 436                         clear_opt(sbi, XATTR_USER);
 437                         break;
 438                 case Opt_inline_xattr:
 439                         set_opt(sbi, INLINE_XATTR);
 440                         break;
 441                 case Opt_noinline_xattr:
 442                         clear_opt(sbi, INLINE_XATTR);
 443                         break;
 444                 case Opt_inline_xattr_size:
 445                         if (args->from && match_int(args, &arg))
 446                                 return -EINVAL;
 447                         set_opt(sbi, INLINE_XATTR_SIZE);
 448                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 449                         break;
 450 #else
 451                 case Opt_user_xattr:
 452                         f2fs_info(sbi, "user_xattr options not supported");
 453                         break;
 454                 case Opt_nouser_xattr:
 455                         f2fs_info(sbi, "nouser_xattr options not supported");
 456                         break;
 457                 case Opt_inline_xattr:
 458                         f2fs_info(sbi, "inline_xattr options not supported");
 459                         break;
 460                 case Opt_noinline_xattr:
 461                         f2fs_info(sbi, "noinline_xattr options not supported");
 462                         break;
 463 #endif
 464 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 465                 case Opt_acl:
 466                         set_opt(sbi, POSIX_ACL);
 467                         break;
 468                 case Opt_noacl:
 469                         clear_opt(sbi, POSIX_ACL);
 470                         break;
 471 #else
 472                 case Opt_acl:
 473                         f2fs_info(sbi, "acl options not supported");
 474                         break;
 475                 case Opt_noacl:
 476                         f2fs_info(sbi, "noacl options not supported");
 477                         break;
 478 #endif
 479                 case Opt_active_logs:
 480                         if (args->from && match_int(args, &arg))
 481                                 return -EINVAL;
 482                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 483                                 return -EINVAL;
 484                         F2FS_OPTION(sbi).active_logs = arg;
 485                         break;
 486                 case Opt_disable_ext_identify:
 487                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 488                         break;
 489                 case Opt_inline_data:
 490                         set_opt(sbi, INLINE_DATA);
 491                         break;
 492                 case Opt_inline_dentry:
 493                         set_opt(sbi, INLINE_DENTRY);
 494                         break;
 495                 case Opt_noinline_dentry:
 496                         clear_opt(sbi, INLINE_DENTRY);
 497                         break;
 498                 case Opt_flush_merge:
 499                         set_opt(sbi, FLUSH_MERGE);
 500                         break;
 501                 case Opt_noflush_merge:
 502                         clear_opt(sbi, FLUSH_MERGE);
 503                         break;
 504                 case Opt_nobarrier:
 505                         set_opt(sbi, NOBARRIER);
 506                         break;
 507                 case Opt_fastboot:
 508                         set_opt(sbi, FASTBOOT);
 509                         break;
 510                 case Opt_extent_cache:
 511                         set_opt(sbi, EXTENT_CACHE);
 512                         break;
 513                 case Opt_noextent_cache:
 514                         clear_opt(sbi, EXTENT_CACHE);
 515                         break;
 516                 case Opt_noinline_data:
 517                         clear_opt(sbi, INLINE_DATA);
 518                         break;
 519                 case Opt_data_flush:
 520                         set_opt(sbi, DATA_FLUSH);
 521                         break;
 522                 case Opt_reserve_root:
 523                         if (args->from && match_int(args, &arg))
 524                                 return -EINVAL;
 525                         if (test_opt(sbi, RESERVE_ROOT)) {
 526                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 527                                           F2FS_OPTION(sbi).root_reserved_blocks);
 528                         } else {
 529                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 530                                 set_opt(sbi, RESERVE_ROOT);
 531                         }
 532                         break;
 533                 case Opt_resuid:
 534                         if (args->from && match_int(args, &arg))
 535                                 return -EINVAL;
 536                         uid = make_kuid(current_user_ns(), arg);
 537                         if (!uid_valid(uid)) {
 538                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 539                                 return -EINVAL;
 540                         }
 541                         F2FS_OPTION(sbi).s_resuid = uid;
 542                         break;
 543                 case Opt_resgid:
 544                         if (args->from && match_int(args, &arg))
 545                                 return -EINVAL;
 546                         gid = make_kgid(current_user_ns(), arg);
 547                         if (!gid_valid(gid)) {
 548                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 549                                 return -EINVAL;
 550                         }
 551                         F2FS_OPTION(sbi).s_resgid = gid;
 552                         break;
 553                 case Opt_mode:
 554                         name = match_strdup(&args[0]);
 555
 556                         if (!name)
 557                                 return -ENOMEM;
 558                         if (strlen(name) == 8 &&
 559                                         !strncmp(name, "adaptive", 8)) {
 560                                 if (f2fs_sb_has_blkzoned(sbi)) {
 561                                         f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
 562                                         kvfree(name);
 563                                         return -EINVAL;
 564                                 }
 565                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 566                         } else if (strlen(name) == 3 &&
 567                                         !strncmp(name, "lfs", 3)) {
 568                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 569                         } else {
 570                                 kvfree(name);
 571                                 return -EINVAL;
 572                         }
 573                         kvfree(name);
 574                         break;
 575                 case Opt_io_size_bits:
 576                         if (args->from && match_int(args, &arg))
 577                                 return -EINVAL;
 578                         if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
 579                                 f2fs_warn(sbi, "Not support %d, larger than %d",
 580                                           1 << arg, BIO_MAX_PAGES);
 581                                 return -EINVAL;
 582                         }
 583                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 584                         break;
 585 #ifdef CONFIG_F2FS_FAULT_INJECTION
 586                 case Opt_fault_injection:
 587                         if (args->from && match_int(args, &arg))
 588                                 return -EINVAL;
 589                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 590                         set_opt(sbi, FAULT_INJECTION);
 591                         break;
 592
 593                 case Opt_fault_type:
 594                         if (args->from && match_int(args, &arg))
 595                                 return -EINVAL;
 596                         f2fs_build_fault_attr(sbi, 0, arg);
 597                         set_opt(sbi, FAULT_INJECTION);
 598                         break;
 599 #else
 600                 case Opt_fault_injection:
 601                         f2fs_info(sbi, "fault_injection options not supported");
 602                         break;
 603
 604                 case Opt_fault_type:
 605                         f2fs_info(sbi, "fault_type options not supported");
 606                         break;
 607 #endif
 608                 case Opt_lazytime:
 609                         sb->s_flags |= SB_LAZYTIME;
 610                         break;
 611                 case Opt_nolazytime:
 612                         sb->s_flags &= ~SB_LAZYTIME;
 613                         break;
 614 #ifdef CONFIG_QUOTA
 615                 case Opt_quota:
 616                 case Opt_usrquota:
 617                         set_opt(sbi, USRQUOTA);
 618                         break;
 619                 case Opt_grpquota:
 620                         set_opt(sbi, GRPQUOTA);
 621                         break;
 622                 case Opt_prjquota:
 623                         set_opt(sbi, PRJQUOTA);
 624                         break;
 625                 case Opt_usrjquota:
 626                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 627                         if (ret)
 628                                 return ret;
 629                         break;
 630                 case Opt_grpjquota:
 631                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 632                         if (ret)
 633                                 return ret;
 634                         break;
 635                 case Opt_prjjquota:
 636                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 637                         if (ret)
 638                                 return ret;
 639                         break;
 640                 case Opt_offusrjquota:
 641                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 642                         if (ret)
 643                                 return ret;
 644                         break;
 645                 case Opt_offgrpjquota:
 646                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 647                         if (ret)
 648                                 return ret;
 649                         break;
 650                 case Opt_offprjjquota:
 651                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 652                         if (ret)
 653                                 return ret;
 654                         break;
 655                 case Opt_jqfmt_vfsold:
 656                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 657                         break;
 658                 case Opt_jqfmt_vfsv0:
 659                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 660                         break;
 661                 case Opt_jqfmt_vfsv1:
 662                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 663                         break;
 664                 case Opt_noquota:
 665                         clear_opt(sbi, QUOTA);
 666                         clear_opt(sbi, USRQUOTA);
 667                         clear_opt(sbi, GRPQUOTA);
 668                         clear_opt(sbi, PRJQUOTA);
 669                         break;
 670 #else
 671                 case Opt_quota:
 672                 case Opt_usrquota:
 673                 case Opt_grpquota:
 674                 case Opt_prjquota:
 675                 case Opt_usrjquota:
 676                 case Opt_grpjquota:
 677                 case Opt_prjjquota:
 678                 case Opt_offusrjquota:
 679                 case Opt_offgrpjquota:
 680                 case Opt_offprjjquota:
 681                 case Opt_jqfmt_vfsold:
 682                 case Opt_jqfmt_vfsv0:
 683                 case Opt_jqfmt_vfsv1:
 684                 case Opt_noquota:
 685                         f2fs_info(sbi, "quota operations not supported");
 686                         break;
 687 #endif
 688                 case Opt_whint:
 689                         name = match_strdup(&args[0]);
 690                         if (!name)
 691                                 return -ENOMEM;
 692                         if (strlen(name) == 10 &&
 693                                         !strncmp(name, "user-based", 10)) {
 694                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 695                         } else if (strlen(name) == 3 &&
 696                                         !strncmp(name, "off", 3)) {
 697                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 698                         } else if (strlen(name) == 8 &&
 699                                         !strncmp(name, "fs-based", 8)) {
 700                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 701                         } else {
 702                                 kvfree(name);
 703                                 return -EINVAL;
 704                         }
 705                         kvfree(name);
 706                         break;
 707                 case Opt_alloc:
 708                         name = match_strdup(&args[0]);
 709                         if (!name)
 710                                 return -ENOMEM;
 711
 712                         if (strlen(name) == 7 &&
 713                                         !strncmp(name, "default", 7)) {
 714                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 715                         } else if (strlen(name) == 5 &&
 716                                         !strncmp(name, "reuse", 5)) {
 717                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 718                         } else {
 719                                 kvfree(name);
 720                                 return -EINVAL;
 721                         }
 722                         kvfree(name);
 723                         break;
 724                 case Opt_fsync:
 725                         name = match_strdup(&args[0]);
 726                         if (!name)
 727                                 return -ENOMEM;
 728                         if (strlen(name) == 5 &&
 729                                         !strncmp(name, "posix", 5)) {
 730                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 731                         } else if (strlen(name) == 6 &&
 732                                         !strncmp(name, "strict", 6)) {
 733                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 734                         } else if (strlen(name) == 9 &&
 735                                         !strncmp(name, "nobarrier", 9)) {
 736                                 F2FS_OPTION(sbi).fsync_mode =
 737                                                         FSYNC_MODE_NOBARRIER;
 738                         } else {
 739                                 kvfree(name);
 740                                 return -EINVAL;
 741                         }
 742                         kvfree(name);
 743                         break;
 744                 case Opt_test_dummy_encryption:
 745 #ifdef CONFIG_FS_ENCRYPTION
 746                         if (!f2fs_sb_has_encrypt(sbi)) {
 747                                 f2fs_err(sbi, "Encrypt feature is off");
 748                                 return -EINVAL;
 749                         }
 750
 751                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 752                         f2fs_info(sbi, "Test dummy encryption mode enabled");
 753 #else
 754                         f2fs_info(sbi, "Test dummy encryption mount option ignored");
 755 #endif
 756                         break;
 757                 case Opt_checkpoint_disable_cap_perc:
 758                         if (args->from && match_int(args, &arg))
 759                                 return -EINVAL;
 760                         if (arg < 0 || arg > 100)
 761                                 return -EINVAL;
 762                         if (arg == 100)
 763                                 F2FS_OPTION(sbi).unusable_cap =
 764                                         sbi->user_block_count;
 765                         else
 766                                 F2FS_OPTION(sbi).unusable_cap =
 767                                         (sbi->user_block_count / 100) * arg;
 768                         set_opt(sbi, DISABLE_CHECKPOINT);
 769                         break;
 770                 case Opt_checkpoint_disable_cap:
 771                         if (args->from && match_int(args, &arg))
 772                                 return -EINVAL;
 773                         F2FS_OPTION(sbi).unusable_cap = arg;
 774                         set_opt(sbi, DISABLE_CHECKPOINT);
 775                         break;
 776                 case Opt_checkpoint_disable:
 777                         set_opt(sbi, DISABLE_CHECKPOINT);
 778                         break;
 779                 case Opt_checkpoint_enable:
 780                         clear_opt(sbi, DISABLE_CHECKPOINT);
 781                         break;
 782                 default:
 783                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
 784                                  p);
 785                         return -EINVAL;
 786                 }
 787         }
 788 #ifdef CONFIG_QUOTA
 789         if (f2fs_check_quota_options(sbi))
 790                 return -EINVAL;
 791 #else
 792         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
 793                 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 794                 return -EINVAL;
 795         }
 796         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
 797                 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 798                 return -EINVAL;
 799         }
 800 #endif
 801
 802         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 803                 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
 804                          F2FS_IO_SIZE_KB(sbi));
 805                 return -EINVAL;
 806         }
 807
 808         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 809                 int min_size, max_size;
 810
 811                 if (!f2fs_sb_has_extra_attr(sbi) ||
 812                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 813                         f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
 814                         return -EINVAL;
 815                 }
 816                 if (!test_opt(sbi, INLINE_XATTR)) {
 817                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
 818                         return -EINVAL;
 819                 }
 820
 821                 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
 822                 max_size = MAX_INLINE_XATTR_SIZE;
 823
 824                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
 825                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
 826                         f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
 827                                  min_size, max_size);
 828                         return -EINVAL;
 829                 }
 830         }
 831
 832         if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
 833                 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
 834                 return -EINVAL;
 835         }
 836
 837         /* Not pass down write hints if the number of active logs is lesser
 838          * than NR_CURSEG_TYPE.
 839          */
 840         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 841                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 842         return 0;
 843 }
 844
 845 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 846 {
 847         struct f2fs_inode_info *fi;
 848
 849         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 850         if (!fi)
 851                 return NULL;
 852
 853         init_once((void *) fi);
 854
 855         /* Initialize f2fs-specific inode info */
 856         atomic_set(&fi->dirty_pages, 0);
 857         init_rwsem(&fi->i_sem);
 858         INIT_LIST_HEAD(&fi->dirty_list);
 859         INIT_LIST_HEAD(&fi->gdirty_list);
 860         INIT_LIST_HEAD(&fi->inmem_ilist);
 861         INIT_LIST_HEAD(&fi->inmem_pages);
 862         mutex_init(&fi->inmem_lock);
 863         init_rwsem(&fi->i_gc_rwsem[READ]);
 864         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 865         init_rwsem(&fi->i_mmap_sem);
 866         init_rwsem(&fi->i_xattr_sem);
 867
 868         /* Will be used by directory only */
 869         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 870
 871         return &fi->vfs_inode;
 872 }
 873
 874 static int f2fs_drop_inode(struct inode *inode)
 875 {
 876         int ret;
 877         /*
 878          * This is to avoid a deadlock condition like below.
 879          * writeback_single_inode(inode)
 880          *  - f2fs_write_data_page
 881          *    - f2fs_gc -> iput -> evict
 882          *       - inode_wait_for_writeback(inode)
 883          */
 884         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 885                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 886                         /* to avoid evict_inode call simultaneously */
 887                         atomic_inc(&inode->i_count);
 888                         spin_unlock(&inode->i_lock);
 889
 890                         /* some remained atomic pages should discarded */
 891                         if (f2fs_is_atomic_file(inode))
 892                                 f2fs_drop_inmem_pages(inode);
 893
 894                         /* should remain fi->extent_tree for writepage */
 895                         f2fs_destroy_extent_node(inode);
 896
 897                         sb_start_intwrite(inode->i_sb);
 898                         f2fs_i_size_write(inode, 0);
 899
 900                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
 901                                         inode, NULL, 0, DATA);
 902                         truncate_inode_pages_final(inode->i_mapping);
 903
 904                         if (F2FS_HAS_BLOCKS(inode))
 905                                 f2fs_truncate(inode);
 906
 907                         sb_end_intwrite(inode->i_sb);
 908
 909                         spin_lock(&inode->i_lock);
 910                         atomic_dec(&inode->i_count);
 911                 }
 912                 trace_f2fs_drop_inode(inode, 0);
 913                 return 0;
 914         }
 915         ret = generic_drop_inode(inode);
 916         trace_f2fs_drop_inode(inode, ret);
 917         return ret;
 918 }
 919
 920 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 921 {
 922         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 923         int ret = 0;
 924
 925         spin_lock(&sbi->inode_lock[DIRTY_META]);
 926         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 927                 ret = 1;
 928         } else {
 929                 set_inode_flag(inode, FI_DIRTY_INODE);
 930                 stat_inc_dirty_inode(sbi, DIRTY_META);
 931         }
 932         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 933                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 934                                 &sbi->inode_list[DIRTY_META]);
 935                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 936         }
 937         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 938         return ret;
 939 }
 940
 941 void f2fs_inode_synced(struct inode *inode)
 942 {
 943         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 944
 945         spin_lock(&sbi->inode_lock[DIRTY_META]);
 946         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 947                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 948                 return;
 949         }
 950         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 951                 list_del_init(&F2FS_I(inode)->gdirty_list);
 952                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 953         }
 954         clear_inode_flag(inode, FI_DIRTY_INODE);
 955         clear_inode_flag(inode, FI_AUTO_RECOVER);
 956         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 957         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 958 }
 959
 960 /*
 961  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 962  *
 963  * We should call set_dirty_inode to write the dirty inode through write_inode.
 964  */
 965 static void f2fs_dirty_inode(struct inode *inode, int flags)
 966 {
 967         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 968
 969         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 970                         inode->i_ino == F2FS_META_INO(sbi))
 971                 return;
 972
 973         if (flags == I_DIRTY_TIME)
 974                 return;
 975
 976         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 977                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 978
 979         f2fs_inode_dirtied(inode, false);
 980 }
 981
 982 static void f2fs_free_inode(struct inode *inode)
 983 {
 984         fscrypt_free_inode(inode);
 985         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 986 }
 987
 988 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 989 {
 990         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 991         percpu_counter_destroy(&sbi->total_valid_inode_count);
 992 }
 993
 994 static void destroy_device_list(struct f2fs_sb_info *sbi)
 995 {
 996         int i;
 997
 998         for (i = 0; i < sbi->s_ndevs; i++) {
 999                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1000 #ifdef CONFIG_BLK_DEV_ZONED
1001                 kvfree(FDEV(i).blkz_seq);
1002 #endif
1003         }
1004         kvfree(sbi->devs);
1005 }
1006
1007 static void f2fs_put_super(struct super_block *sb)
1008 {
1009         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1010         int i;
1011         bool dropped;
1012
1013         f2fs_quota_off_umount(sb);
1014
1015         /* prevent remaining shrinker jobs */
1016         mutex_lock(&sbi->umount_mutex);
1017
1018         /*
1019          * We don't need to do checkpoint when superblock is clean.
1020          * But, the previous checkpoint was not done by umount, it needs to do
1021          * clean checkpoint again.
1022          */
1023         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1024                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1025                 struct cp_control cpc = {
1026                         .reason = CP_UMOUNT,
1027                 };
1028                 f2fs_write_checkpoint(sbi, &cpc);
1029         }
1030
1031         /* be sure to wait for any on-going discard commands */
1032         dropped = f2fs_issue_discard_timeout(sbi);
1033
1034         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1035                                         !sbi->discard_blks && !dropped) {
1036                 struct cp_control cpc = {
1037                         .reason = CP_UMOUNT | CP_TRIMMED,
1038                 };
1039                 f2fs_write_checkpoint(sbi, &cpc);
1040         }
1041
1042         /*
1043          * normally superblock is clean, so we need to release this.
1044          * In addition, EIO will skip do checkpoint, we need this as well.
1045          */
1046         f2fs_release_ino_entry(sbi, true);
1047
1048         f2fs_leave_shrinker(sbi);
1049         mutex_unlock(&sbi->umount_mutex);
1050
1051         /* our cp_error case, we can wait for any writeback page */
1052         f2fs_flush_merged_writes(sbi);
1053
1054         f2fs_wait_on_all_pages_writeback(sbi);
1055
1056         f2fs_bug_on(sbi, sbi->fsync_node_num);
1057
1058         iput(sbi->node_inode);
1059         sbi->node_inode = NULL;
1060
1061         iput(sbi->meta_inode);
1062         sbi->meta_inode = NULL;
1063
1064         /*
1065          * iput() can update stat information, if f2fs_write_checkpoint()
1066          * above failed with error.
1067          */
1068         f2fs_destroy_stats(sbi);
1069
1070         /* destroy f2fs internal modules */
1071         f2fs_destroy_node_manager(sbi);
1072         f2fs_destroy_segment_manager(sbi);
1073
1074         kvfree(sbi->ckpt);
1075
1076         f2fs_unregister_sysfs(sbi);
1077
1078         sb->s_fs_info = NULL;
1079         if (sbi->s_chksum_driver)
1080                 crypto_free_shash(sbi->s_chksum_driver);
1081         kvfree(sbi->raw_super);
1082
1083         destroy_device_list(sbi);
1084         mempool_destroy(sbi->write_io_dummy);
1085 #ifdef CONFIG_QUOTA
1086         for (i = 0; i < MAXQUOTAS; i++)
1087                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1088 #endif
1089         destroy_percpu_info(sbi);
1090         for (i = 0; i < NR_PAGE_TYPE; i++)
1091                 kvfree(sbi->write_io[i]);
1092         kvfree(sbi);
1093 }
1094
1095 int f2fs_sync_fs(struct super_block *sb, int sync)
1096 {
1097         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1098         int err = 0;
1099
1100         if (unlikely(f2fs_cp_error(sbi)))
1101                 return 0;
1102         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1103                 return 0;
1104
1105         trace_f2fs_sync_fs(sb, sync);
1106
1107         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1108                 return -EAGAIN;
1109
1110         if (sync) {
1111                 struct cp_control cpc;
1112
1113                 cpc.reason = __get_cp_reason(sbi);
1114
1115                 mutex_lock(&sbi->gc_mutex);
1116                 err = f2fs_write_checkpoint(sbi, &cpc);
1117                 mutex_unlock(&sbi->gc_mutex);
1118         }
1119         f2fs_trace_ios(NULL, 1);
1120
1121         return err;
1122 }
1123
1124 static int f2fs_freeze(struct super_block *sb)
1125 {
1126         if (f2fs_readonly(sb))
1127                 return 0;
1128
1129         /* IO error happened before */
1130         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1131                 return -EIO;
1132
1133         /* must be clean, since sync_filesystem() was already called */
1134         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1135                 return -EINVAL;
1136         return 0;
1137 }
1138
1139 static int f2fs_unfreeze(struct super_block *sb)
1140 {
1141         return 0;
1142 }
1143
1144 #ifdef CONFIG_QUOTA
1145 static int f2fs_statfs_project(struct super_block *sb,
1146                                 kprojid_t projid, struct kstatfs *buf)
1147 {
1148         struct kqid qid;
1149         struct dquot *dquot;
1150         u64 limit;
1151         u64 curblock;
1152
1153         qid = make_kqid_projid(projid);
1154         dquot = dqget(sb, qid);
1155         if (IS_ERR(dquot))
1156                 return PTR_ERR(dquot);
1157         spin_lock(&dquot->dq_dqb_lock);
1158
1159         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1160                  dquot->dq_dqb.dqb_bsoftlimit :
1161                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1162         if (limit && buf->f_blocks > limit) {
1163                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1164                 buf->f_blocks = limit;
1165                 buf->f_bfree = buf->f_bavail =
1166                         (buf->f_blocks > curblock) ?
1167                          (buf->f_blocks - curblock) : 0;
1168         }
1169
1170         limit = dquot->dq_dqb.dqb_isoftlimit ?
1171                 dquot->dq_dqb.dqb_isoftlimit :
1172                 dquot->dq_dqb.dqb_ihardlimit;
1173         if (limit && buf->f_files > limit) {
1174                 buf->f_files = limit;
1175                 buf->f_ffree =
1176                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1177                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1178         }
1179
1180         spin_unlock(&dquot->dq_dqb_lock);
1181         dqput(dquot);
1182         return 0;
1183 }
1184 #endif
1185
1186 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1187 {
1188         struct super_block *sb = dentry->d_sb;
1189         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1190         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1191         block_t total_count, user_block_count, start_count;
1192         u64 avail_node_count;
1193
1194         total_count = le64_to_cpu(sbi->raw_super->block_count);
1195         user_block_count = sbi->user_block_count;
1196         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1197         buf->f_type = F2FS_SUPER_MAGIC;
1198         buf->f_bsize = sbi->blocksize;
1199
1200         buf->f_blocks = total_count - start_count;
1201         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1202                                                 sbi->current_reserved_blocks;
1203
1204         spin_lock(&sbi->stat_lock);
1205         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1206                 buf->f_bfree = 0;
1207         else
1208                 buf->f_bfree -= sbi->unusable_block_count;
1209         spin_unlock(&sbi->stat_lock);
1210
1211         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1212                 buf->f_bavail = buf->f_bfree -
1213                                 F2FS_OPTION(sbi).root_reserved_blocks;
1214         else
1215                 buf->f_bavail = 0;
1216
1217         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1218                                                 F2FS_RESERVED_NODE_NUM;
1219
1220         if (avail_node_count > user_block_count) {
1221                 buf->f_files = user_block_count;
1222                 buf->f_ffree = buf->f_bavail;
1223         } else {
1224                 buf->f_files = avail_node_count;
1225                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1226                                         buf->f_bavail);
1227         }
1228
1229         buf->f_namelen = F2FS_NAME_LEN;
1230         buf->f_fsid.val[0] = (u32)id;
1231         buf->f_fsid.val[1] = (u32)(id >> 32);
1232
1233 #ifdef CONFIG_QUOTA
1234         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1235                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1236                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1237         }
1238 #endif
1239         return 0;
1240 }
1241
1242 static inline void f2fs_show_quota_options(struct seq_file *seq,
1243                                            struct super_block *sb)
1244 {
1245 #ifdef CONFIG_QUOTA
1246         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1247
1248         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1249                 char *fmtname = "";
1250
1251                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1252                 case QFMT_VFS_OLD:
1253                         fmtname = "vfsold";
1254                         break;
1255                 case QFMT_VFS_V0:
1256                         fmtname = "vfsv0";
1257                         break;
1258                 case QFMT_VFS_V1:
1259                         fmtname = "vfsv1";
1260                         break;
1261                 }
1262                 seq_printf(seq, ",jqfmt=%s", fmtname);
1263         }
1264
1265         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1266                 seq_show_option(seq, "usrjquota",
1267                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1268
1269         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1270                 seq_show_option(seq, "grpjquota",
1271                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1272
1273         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1274                 seq_show_option(seq, "prjjquota",
1275                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1276 #endif
1277 }
1278
1279 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1280 {
1281         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1282
1283         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1284                 if (test_opt(sbi, FORCE_FG_GC))
1285                         seq_printf(seq, ",background_gc=%s", "sync");
1286                 else
1287                         seq_printf(seq, ",background_gc=%s", "on");
1288         } else {
1289                 seq_printf(seq, ",background_gc=%s", "off");
1290         }
1291         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1292                 seq_puts(seq, ",disable_roll_forward");
1293         if (test_opt(sbi, DISCARD))
1294                 seq_puts(seq, ",discard");
1295         else
1296                 seq_puts(seq, ",nodiscard");
1297         if (test_opt(sbi, NOHEAP))
1298                 seq_puts(seq, ",no_heap");
1299         else
1300                 seq_puts(seq, ",heap");
1301 #ifdef CONFIG_F2FS_FS_XATTR
1302         if (test_opt(sbi, XATTR_USER))
1303                 seq_puts(seq, ",user_xattr");
1304         else
1305                 seq_puts(seq, ",nouser_xattr");
1306         if (test_opt(sbi, INLINE_XATTR))
1307                 seq_puts(seq, ",inline_xattr");
1308         else
1309                 seq_puts(seq, ",noinline_xattr");
1310         if (test_opt(sbi, INLINE_XATTR_SIZE))
1311                 seq_printf(seq, ",inline_xattr_size=%u",
1312                                         F2FS_OPTION(sbi).inline_xattr_size);
1313 #endif
1314 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1315         if (test_opt(sbi, POSIX_ACL))
1316                 seq_puts(seq, ",acl");
1317         else
1318                 seq_puts(seq, ",noacl");
1319 #endif
1320         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1321                 seq_puts(seq, ",disable_ext_identify");
1322         if (test_opt(sbi, INLINE_DATA))
1323                 seq_puts(seq, ",inline_data");
1324         else
1325                 seq_puts(seq, ",noinline_data");
1326         if (test_opt(sbi, INLINE_DENTRY))
1327                 seq_puts(seq, ",inline_dentry");
1328         else
1329                 seq_puts(seq, ",noinline_dentry");
1330         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1331                 seq_puts(seq, ",flush_merge");
1332         if (test_opt(sbi, NOBARRIER))
1333                 seq_puts(seq, ",nobarrier");
1334         if (test_opt(sbi, FASTBOOT))
1335                 seq_puts(seq, ",fastboot");
1336         if (test_opt(sbi, EXTENT_CACHE))
1337                 seq_puts(seq, ",extent_cache");
1338         else
1339                 seq_puts(seq, ",noextent_cache");
1340         if (test_opt(sbi, DATA_FLUSH))
1341                 seq_puts(seq, ",data_flush");
1342
1343         seq_puts(seq, ",mode=");
1344         if (test_opt(sbi, ADAPTIVE))
1345                 seq_puts(seq, "adaptive");
1346         else if (test_opt(sbi, LFS))
1347                 seq_puts(seq, "lfs");
1348         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1349         if (test_opt(sbi, RESERVE_ROOT))
1350                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1351                                 F2FS_OPTION(sbi).root_reserved_blocks,
1352                                 from_kuid_munged(&init_user_ns,
1353                                         F2FS_OPTION(sbi).s_resuid),
1354                                 from_kgid_munged(&init_user_ns,
1355                                         F2FS_OPTION(sbi).s_resgid));
1356         if (F2FS_IO_SIZE_BITS(sbi))
1357                 seq_printf(seq, ",io_bits=%u",
1358                                 F2FS_OPTION(sbi).write_io_size_bits);
1359 #ifdef CONFIG_F2FS_FAULT_INJECTION
1360         if (test_opt(sbi, FAULT_INJECTION)) {
1361                 seq_printf(seq, ",fault_injection=%u",
1362                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1363                 seq_printf(seq, ",fault_type=%u",
1364                                 F2FS_OPTION(sbi).fault_info.inject_type);
1365         }
1366 #endif
1367 #ifdef CONFIG_QUOTA
1368         if (test_opt(sbi, QUOTA))
1369                 seq_puts(seq, ",quota");
1370         if (test_opt(sbi, USRQUOTA))
1371                 seq_puts(seq, ",usrquota");
1372         if (test_opt(sbi, GRPQUOTA))
1373                 seq_puts(seq, ",grpquota");
1374         if (test_opt(sbi, PRJQUOTA))
1375                 seq_puts(seq, ",prjquota");
1376 #endif
1377         f2fs_show_quota_options(seq, sbi->sb);
1378         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1379                 seq_printf(seq, ",whint_mode=%s", "user-based");
1380         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1381                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1382 #ifdef CONFIG_FS_ENCRYPTION
1383         if (F2FS_OPTION(sbi).test_dummy_encryption)
1384                 seq_puts(seq, ",test_dummy_encryption");
1385 #endif
1386
1387         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1388                 seq_printf(seq, ",alloc_mode=%s", "default");
1389         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1390                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1391
1392         if (test_opt(sbi, DISABLE_CHECKPOINT))
1393                 seq_printf(seq, ",checkpoint=disable:%u",
1394                                 F2FS_OPTION(sbi).unusable_cap);
1395         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1396                 seq_printf(seq, ",fsync_mode=%s", "posix");
1397         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1398                 seq_printf(seq, ",fsync_mode=%s", "strict");
1399         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1400                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1401         return 0;
1402 }
1403
1404 static void default_options(struct f2fs_sb_info *sbi)
1405 {
1406         /* init some FS parameters */
1407         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1408         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1409         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1410         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1411         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1412         F2FS_OPTION(sbi).test_dummy_encryption = false;
1413         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1414         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1415
1416         set_opt(sbi, BG_GC);
1417         set_opt(sbi, INLINE_XATTR);
1418         set_opt(sbi, INLINE_DATA);
1419         set_opt(sbi, INLINE_DENTRY);
1420         set_opt(sbi, EXTENT_CACHE);
1421         set_opt(sbi, NOHEAP);
1422         clear_opt(sbi, DISABLE_CHECKPOINT);
1423         F2FS_OPTION(sbi).unusable_cap = 0;
1424         sbi->sb->s_flags |= SB_LAZYTIME;
1425         set_opt(sbi, FLUSH_MERGE);
1426         set_opt(sbi, DISCARD);
1427         if (f2fs_sb_has_blkzoned(sbi))
1428                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1429         else
1430                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1431
1432 #ifdef CONFIG_F2FS_FS_XATTR
1433         set_opt(sbi, XATTR_USER);
1434 #endif
1435 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1436         set_opt(sbi, POSIX_ACL);
1437 #endif
1438
1439         f2fs_build_fault_attr(sbi, 0, 0);
1440 }
1441
1442 #ifdef CONFIG_QUOTA
1443 static int f2fs_enable_quotas(struct super_block *sb);
1444 #endif
1445
1446 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1447 {
1448         unsigned int s_flags = sbi->sb->s_flags;
1449         struct cp_control cpc;
1450         int err = 0;
1451         int ret;
1452         block_t unusable;
1453
1454         if (s_flags & SB_RDONLY) {
1455                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1456                 return -EINVAL;
1457         }
1458         sbi->sb->s_flags |= SB_ACTIVE;
1459
1460         f2fs_update_time(sbi, DISABLE_TIME);
1461
1462         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1463                 mutex_lock(&sbi->gc_mutex);
1464                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1465                 if (err == -ENODATA) {
1466                         err = 0;
1467                         break;
1468                 }
1469                 if (err && err != -EAGAIN)
1470                         break;
1471         }
1472
1473         ret = sync_filesystem(sbi->sb);
1474         if (ret || err) {
1475                 err = ret ? ret: err;
1476                 goto restore_flag;
1477         }
1478
1479         unusable = f2fs_get_unusable_blocks(sbi);
1480         if (f2fs_disable_cp_again(sbi, unusable)) {
1481                 err = -EAGAIN;
1482                 goto restore_flag;
1483         }
1484
1485         mutex_lock(&sbi->gc_mutex);
1486         cpc.reason = CP_PAUSE;
1487         set_sbi_flag(sbi, SBI_CP_DISABLED);
1488         err = f2fs_write_checkpoint(sbi, &cpc);
1489         if (err)
1490                 goto out_unlock;
1491
1492         spin_lock(&sbi->stat_lock);
1493         sbi->unusable_block_count = unusable;
1494         spin_unlock(&sbi->stat_lock);
1495
1496 out_unlock:
1497         mutex_unlock(&sbi->gc_mutex);
1498 restore_flag:
1499         sbi->sb->s_flags = s_flags;     /* Restore MS_RDONLY status */
1500         return err;
1501 }
1502
1503 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1504 {
1505         mutex_lock(&sbi->gc_mutex);
1506         f2fs_dirty_to_prefree(sbi);
1507
1508         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1509         set_sbi_flag(sbi, SBI_IS_DIRTY);
1510         mutex_unlock(&sbi->gc_mutex);
1511
1512         f2fs_sync_fs(sbi->sb, 1);
1513 }
1514
1515 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1516 {
1517         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1518         struct f2fs_mount_info org_mount_opt;
1519         unsigned long old_sb_flags;
1520         int err;
1521         bool need_restart_gc = false;
1522         bool need_stop_gc = false;
1523         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1524         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1525         bool checkpoint_changed;
1526 #ifdef CONFIG_QUOTA
1527         int i, j;
1528 #endif
1529
1530         /*
1531          * Save the old mount options in case we
1532          * need to restore them.
1533          */
1534         org_mount_opt = sbi->mount_opt;
1535         old_sb_flags = sb->s_flags;
1536
1537 #ifdef CONFIG_QUOTA
1538         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1539         for (i = 0; i < MAXQUOTAS; i++) {
1540                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1541                         org_mount_opt.s_qf_names[i] =
1542                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1543                                 GFP_KERNEL);
1544                         if (!org_mount_opt.s_qf_names[i]) {
1545                                 for (j = 0; j < i; j++)
1546                                         kvfree(org_mount_opt.s_qf_names[j]);
1547                                 return -ENOMEM;
1548                         }
1549                 } else {
1550                         org_mount_opt.s_qf_names[i] = NULL;
1551                 }
1552         }
1553 #endif
1554
1555         /* recover superblocks we couldn't write due to previous RO mount */
1556         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1557                 err = f2fs_commit_super(sbi, false);
1558                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1559                           err);
1560                 if (!err)
1561                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1562         }
1563
1564         default_options(sbi);
1565
1566         /* parse mount options */
1567         err = parse_options(sb, data);
1568         if (err)
1569                 goto restore_opts;
1570         checkpoint_changed =
1571                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1572
1573         /*
1574          * Previous and new state of filesystem is RO,
1575          * so skip checking GC and FLUSH_MERGE conditions.
1576          */
1577         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1578                 goto skip;
1579
1580 #ifdef CONFIG_QUOTA
1581         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1582                 err = dquot_suspend(sb, -1);
1583                 if (err < 0)
1584                         goto restore_opts;
1585         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1586                 /* dquot_resume needs RW */
1587                 sb->s_flags &= ~SB_RDONLY;
1588                 if (sb_any_quota_suspended(sb)) {
1589                         dquot_resume(sb, -1);
1590                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1591                         err = f2fs_enable_quotas(sb);
1592                         if (err)
1593                                 goto restore_opts;
1594                 }
1595         }
1596 #endif
1597         /* disallow enable/disable extent_cache dynamically */
1598         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1599                 err = -EINVAL;
1600                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1601                 goto restore_opts;
1602         }
1603
1604         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1605                 err = -EINVAL;
1606                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1607                 goto restore_opts;
1608         }
1609
1610         /*
1611          * We stop the GC thread if FS is mounted as RO
1612          * or if background_gc = off is passed in mount
1613          * option. Also sync the filesystem.
1614          */
1615         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1616                 if (sbi->gc_thread) {
1617                         f2fs_stop_gc_thread(sbi);
1618                         need_restart_gc = true;
1619                 }
1620         } else if (!sbi->gc_thread) {
1621                 err = f2fs_start_gc_thread(sbi);
1622                 if (err)
1623                         goto restore_opts;
1624                 need_stop_gc = true;
1625         }
1626
1627         if (*flags & SB_RDONLY ||
1628                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1629                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1630                 sync_inodes_sb(sb);
1631
1632                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1633                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1634                 f2fs_sync_fs(sb, 1);
1635                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1636         }
1637
1638         if (checkpoint_changed) {
1639                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1640                         err = f2fs_disable_checkpoint(sbi);
1641                         if (err)
1642                                 goto restore_gc;
1643                 } else {
1644                         f2fs_enable_checkpoint(sbi);
1645                 }
1646         }
1647
1648         /*
1649          * We stop issue flush thread if FS is mounted as RO
1650          * or if flush_merge is not passed in mount option.
1651          */
1652         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1653                 clear_opt(sbi, FLUSH_MERGE);
1654                 f2fs_destroy_flush_cmd_control(sbi, false);
1655         } else {
1656                 err = f2fs_create_flush_cmd_control(sbi);
1657                 if (err)
1658                         goto restore_gc;
1659         }
1660 skip:
1661 #ifdef CONFIG_QUOTA
1662         /* Release old quota file names */
1663         for (i = 0; i < MAXQUOTAS; i++)
1664                 kvfree(org_mount_opt.s_qf_names[i]);
1665 #endif
1666         /* Update the POSIXACL Flag */
1667         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1668                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1669
1670         limit_reserve_root(sbi);
1671         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1672         return 0;
1673 restore_gc:
1674         if (need_restart_gc) {
1675                 if (f2fs_start_gc_thread(sbi))
1676                         f2fs_warn(sbi, "background gc thread has stopped");
1677         } else if (need_stop_gc) {
1678                 f2fs_stop_gc_thread(sbi);
1679         }
1680 restore_opts:
1681 #ifdef CONFIG_QUOTA
1682         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1683         for (i = 0; i < MAXQUOTAS; i++) {
1684                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1685                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1686         }
1687 #endif
1688         sbi->mount_opt = org_mount_opt;
1689         sb->s_flags = old_sb_flags;
1690         return err;
1691 }
1692
1693 #ifdef CONFIG_QUOTA
1694 /* Read data from quotafile */
1695 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1696                                size_t len, loff_t off)
1697 {
1698         struct inode *inode = sb_dqopt(sb)->files[type];
1699         struct address_space *mapping = inode->i_mapping;
1700         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1701         int offset = off & (sb->s_blocksize - 1);
1702         int tocopy;
1703         size_t toread;
1704         loff_t i_size = i_size_read(inode);
1705         struct page *page;
1706         char *kaddr;
1707
1708         if (off > i_size)
1709                 return 0;
1710
1711         if (off + len > i_size)
1712                 len = i_size - off;
1713         toread = len;
1714         while (toread > 0) {
1715                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1716 repeat:
1717                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1718                 if (IS_ERR(page)) {
1719                         if (PTR_ERR(page) == -ENOMEM) {
1720                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1721                                 goto repeat;
1722                         }
1723                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1724                         return PTR_ERR(page);
1725                 }
1726
1727                 lock_page(page);
1728
1729                 if (unlikely(page->mapping != mapping)) {
1730                         f2fs_put_page(page, 1);
1731                         goto repeat;
1732                 }
1733                 if (unlikely(!PageUptodate(page))) {
1734                         f2fs_put_page(page, 1);
1735                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1736                         return -EIO;
1737                 }
1738
1739                 kaddr = kmap_atomic(page);
1740                 memcpy(data, kaddr + offset, tocopy);
1741                 kunmap_atomic(kaddr);
1742                 f2fs_put_page(page, 1);
1743
1744                 offset = 0;
1745                 toread -= tocopy;
1746                 data += tocopy;
1747                 blkidx++;
1748         }
1749         return len;
1750 }
1751
1752 /* Write to quotafile */
1753 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1754                                 const char *data, size_t len, loff_t off)
1755 {
1756         struct inode *inode = sb_dqopt(sb)->files[type];
1757         struct address_space *mapping = inode->i_mapping;
1758         const struct address_space_operations *a_ops = mapping->a_ops;
1759         int offset = off & (sb->s_blocksize - 1);
1760         size_t towrite = len;
1761         struct page *page;
1762         char *kaddr;
1763         int err = 0;
1764         int tocopy;
1765
1766         while (towrite > 0) {
1767                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1768                                                                 towrite);
1769 retry:
1770                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1771                                                         &page, NULL);
1772                 if (unlikely(err)) {
1773                         if (err == -ENOMEM) {
1774                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1775                                 goto retry;
1776                         }
1777                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1778                         break;
1779                 }
1780
1781                 kaddr = kmap_atomic(page);
1782                 memcpy(kaddr + offset, data, tocopy);
1783                 kunmap_atomic(kaddr);
1784                 flush_dcache_page(page);
1785
1786                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1787                                                 page, NULL);
1788                 offset = 0;
1789                 towrite -= tocopy;
1790                 off += tocopy;
1791                 data += tocopy;
1792                 cond_resched();
1793         }
1794
1795         if (len == towrite)
1796                 return err;
1797         inode->i_mtime = inode->i_ctime = current_time(inode);
1798         f2fs_mark_inode_dirty_sync(inode, false);
1799         return len - towrite;
1800 }
1801
1802 static struct dquot **f2fs_get_dquots(struct inode *inode)
1803 {
1804         return F2FS_I(inode)->i_dquot;
1805 }
1806
1807 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1808 {
1809         return &F2FS_I(inode)->i_reserved_quota;
1810 }
1811
1812 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1813 {
1814         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1815                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1816                 return 0;
1817         }
1818
1819         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1820                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1821 }
1822
1823 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1824 {
1825         int enabled = 0;
1826         int i, err;
1827
1828         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1829                 err = f2fs_enable_quotas(sbi->sb);
1830                 if (err) {
1831                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
1832                         return 0;
1833                 }
1834                 return 1;
1835         }
1836
1837         for (i = 0; i < MAXQUOTAS; i++) {
1838                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1839                         err = f2fs_quota_on_mount(sbi, i);
1840                         if (!err) {
1841                                 enabled = 1;
1842                                 continue;
1843                         }
1844                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
1845                                  err, i);
1846                 }
1847         }
1848         return enabled;
1849 }
1850
1851 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1852                              unsigned int flags)
1853 {
1854         struct inode *qf_inode;
1855         unsigned long qf_inum;
1856         int err;
1857
1858         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
1859
1860         qf_inum = f2fs_qf_ino(sb, type);
1861         if (!qf_inum)
1862                 return -EPERM;
1863
1864         qf_inode = f2fs_iget(sb, qf_inum);
1865         if (IS_ERR(qf_inode)) {
1866                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
1867                 return PTR_ERR(qf_inode);
1868         }
1869
1870         /* Don't account quota for quota files to avoid recursion */
1871         qf_inode->i_flags |= S_NOQUOTA;
1872         err = dquot_enable(qf_inode, type, format_id, flags);
1873         iput(qf_inode);
1874         return err;
1875 }
1876
1877 static int f2fs_enable_quotas(struct super_block *sb)
1878 {
1879         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1880         int type, err = 0;
1881         unsigned long qf_inum;
1882         bool quota_mopt[MAXQUOTAS] = {
1883                 test_opt(sbi, USRQUOTA),
1884                 test_opt(sbi, GRPQUOTA),
1885                 test_opt(sbi, PRJQUOTA),
1886         };
1887
1888         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
1889                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
1890                 return 0;
1891         }
1892
1893         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1894
1895         for (type = 0; type < MAXQUOTAS; type++) {
1896                 qf_inum = f2fs_qf_ino(sb, type);
1897                 if (qf_inum) {
1898                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1899                                 DQUOT_USAGE_ENABLED |
1900                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1901                         if (err) {
1902                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
1903                                          type, err);
1904                                 for (type--; type >= 0; type--)
1905                                         dquot_quota_off(sb, type);
1906                                 set_sbi_flag(F2FS_SB(sb),
1907                                                 SBI_QUOTA_NEED_REPAIR);
1908                                 return err;
1909                         }
1910                 }
1911         }
1912         return 0;
1913 }
1914
1915 int f2fs_quota_sync(struct super_block *sb, int type)
1916 {
1917         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1918         struct quota_info *dqopt = sb_dqopt(sb);
1919         int cnt;
1920         int ret;
1921
1922         /*
1923          * do_quotactl
1924          *  f2fs_quota_sync
1925          *  down_read(quota_sem)
1926          *  dquot_writeback_dquots()
1927          *  f2fs_dquot_commit
1928          *                            block_operation
1929          *                            down_read(quota_sem)
1930          */
1931         f2fs_lock_op(sbi);
1932
1933         down_read(&sbi->quota_sem);
1934         ret = dquot_writeback_dquots(sb, type);
1935         if (ret)
1936                 goto out;
1937
1938         /*
1939          * Now when everything is written we can discard the pagecache so
1940          * that userspace sees the changes.
1941          */
1942         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1943                 struct address_space *mapping;
1944
1945                 if (type != -1 && cnt != type)
1946                         continue;
1947                 if (!sb_has_quota_active(sb, cnt))
1948                         continue;
1949
1950                 mapping = dqopt->files[cnt]->i_mapping;
1951
1952                 ret = filemap_fdatawrite(mapping);
1953                 if (ret)
1954                         goto out;
1955
1956                 /* if we are using journalled quota */
1957                 if (is_journalled_quota(sbi))
1958                         continue;
1959
1960                 ret = filemap_fdatawait(mapping);
1961                 if (ret)
1962                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1963
1964                 inode_lock(dqopt->files[cnt]);
1965                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1966                 inode_unlock(dqopt->files[cnt]);
1967         }
1968 out:
1969         if (ret)
1970                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1971         up_read(&sbi->quota_sem);
1972         f2fs_unlock_op(sbi);
1973         return ret;
1974 }
1975
1976 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1977                                                         const struct path *path)
1978 {
1979         struct inode *inode;
1980         int err;
1981
1982         err = f2fs_quota_sync(sb, type);
1983         if (err)
1984                 return err;
1985
1986         err = dquot_quota_on(sb, type, format_id, path);
1987         if (err)
1988                 return err;
1989
1990         inode = d_inode(path->dentry);
1991
1992         inode_lock(inode);
1993         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1994         f2fs_set_inode_flags(inode);
1995         inode_unlock(inode);
1996         f2fs_mark_inode_dirty_sync(inode, false);
1997
1998         return 0;
1999 }
2000
2001 static int f2fs_quota_off(struct super_block *sb, int type)
2002 {
2003         struct inode *inode = sb_dqopt(sb)->files[type];
2004         int err;
2005
2006         if (!inode || !igrab(inode))
2007                 return dquot_quota_off(sb, type);
2008
2009         err = f2fs_quota_sync(sb, type);
2010         if (err)
2011                 goto out_put;
2012
2013         err = dquot_quota_off(sb, type);
2014         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2015                 goto out_put;
2016
2017         inode_lock(inode);
2018         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2019         f2fs_set_inode_flags(inode);
2020         inode_unlock(inode);
2021         f2fs_mark_inode_dirty_sync(inode, false);
2022 out_put:
2023         iput(inode);
2024         return err;
2025 }
2026
2027 void f2fs_quota_off_umount(struct super_block *sb)
2028 {
2029         int type;
2030         int err;
2031
2032         for (type = 0; type < MAXQUOTAS; type++) {
2033                 err = f2fs_quota_off(sb, type);
2034                 if (err) {
2035                         int ret = dquot_quota_off(sb, type);
2036
2037                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2038                                  type, err, ret);
2039                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2040                 }
2041         }
2042         /*
2043          * In case of checkpoint=disable, we must flush quota blocks.
2044          * This can cause NULL exception for node_inode in end_io, since
2045          * put_super already dropped it.
2046          */
2047         sync_filesystem(sb);
2048 }
2049
2050 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2051 {
2052         struct quota_info *dqopt = sb_dqopt(sb);
2053         int type;
2054
2055         for (type = 0; type < MAXQUOTAS; type++) {
2056                 if (!dqopt->files[type])
2057                         continue;
2058                 f2fs_inode_synced(dqopt->files[type]);
2059         }
2060 }
2061
2062 static int f2fs_dquot_commit(struct dquot *dquot)
2063 {
2064         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2065         int ret;
2066
2067         down_read(&sbi->quota_sem);
2068         ret = dquot_commit(dquot);
2069         if (ret < 0)
2070                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2071         up_read(&sbi->quota_sem);
2072         return ret;
2073 }
2074
2075 static int f2fs_dquot_acquire(struct dquot *dquot)
2076 {
2077         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2078         int ret;
2079
2080         down_read(&sbi->quota_sem);
2081         ret = dquot_acquire(dquot);
2082         if (ret < 0)
2083                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2084         up_read(&sbi->quota_sem);
2085         return ret;
2086 }
2087
2088 static int f2fs_dquot_release(struct dquot *dquot)
2089 {
2090         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2091         int ret;
2092
2093         down_read(&sbi->quota_sem);
2094         ret = dquot_release(dquot);
2095         if (ret < 0)
2096                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2097         up_read(&sbi->quota_sem);
2098         return ret;
2099 }
2100
2101 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2102 {
2103         struct super_block *sb = dquot->dq_sb;
2104         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2105         int ret;
2106
2107         down_read(&sbi->quota_sem);
2108         ret = dquot_mark_dquot_dirty(dquot);
2109
2110         /* if we are using journalled quota */
2111         if (is_journalled_quota(sbi))
2112                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2113
2114         up_read(&sbi->quota_sem);
2115         return ret;
2116 }
2117
2118 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2119 {
2120         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2121         int ret;
2122
2123         down_read(&sbi->quota_sem);
2124         ret = dquot_commit_info(sb, type);
2125         if (ret < 0)
2126                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2127         up_read(&sbi->quota_sem);
2128         return ret;
2129 }
2130
2131 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2132 {
2133         *projid = F2FS_I(inode)->i_projid;
2134         return 0;
2135 }
2136
2137 static const struct dquot_operations f2fs_quota_operations = {
2138         .get_reserved_space = f2fs_get_reserved_space,
2139         .write_dquot    = f2fs_dquot_commit,
2140         .acquire_dquot  = f2fs_dquot_acquire,
2141         .release_dquot  = f2fs_dquot_release,
2142         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2143         .write_info     = f2fs_dquot_commit_info,
2144         .alloc_dquot    = dquot_alloc,
2145         .destroy_dquot  = dquot_destroy,
2146         .get_projid     = f2fs_get_projid,
2147         .get_next_id    = dquot_get_next_id,
2148 };
2149
2150 static const struct quotactl_ops f2fs_quotactl_ops = {
2151         .quota_on       = f2fs_quota_on,
2152         .quota_off      = f2fs_quota_off,
2153         .quota_sync     = f2fs_quota_sync,
2154         .get_state      = dquot_get_state,
2155         .set_info       = dquot_set_dqinfo,
2156         .get_dqblk      = dquot_get_dqblk,
2157         .set_dqblk      = dquot_set_dqblk,
2158         .get_nextdqblk  = dquot_get_next_dqblk,
2159 };
2160 #else
2161 int f2fs_quota_sync(struct super_block *sb, int type)
2162 {
2163         return 0;
2164 }
2165
2166 void f2fs_quota_off_umount(struct super_block *sb)
2167 {
2168 }
2169 #endif
2170
2171 static const struct super_operations f2fs_sops = {
2172         .alloc_inode    = f2fs_alloc_inode,
2173         .free_inode     = f2fs_free_inode,
2174         .drop_inode     = f2fs_drop_inode,
2175         .write_inode    = f2fs_write_inode,
2176         .dirty_inode    = f2fs_dirty_inode,
2177         .show_options   = f2fs_show_options,
2178 #ifdef CONFIG_QUOTA
2179         .quota_read     = f2fs_quota_read,
2180         .quota_write    = f2fs_quota_write,
2181         .get_dquots     = f2fs_get_dquots,
2182 #endif
2183         .evict_inode    = f2fs_evict_inode,
2184         .put_super      = f2fs_put_super,
2185         .sync_fs        = f2fs_sync_fs,
2186         .freeze_fs      = f2fs_freeze,
2187         .unfreeze_fs    = f2fs_unfreeze,
2188         .statfs         = f2fs_statfs,
2189         .remount_fs     = f2fs_remount,
2190 };
2191
2192 #ifdef CONFIG_FS_ENCRYPTION
2193 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2194 {
2195         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2196                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2197                                 ctx, len, NULL);
2198 }
2199
2200 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2201                                                         void *fs_data)
2202 {
2203         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2204
2205         /*
2206          * Encrypting the root directory is not allowed because fsck
2207          * expects lost+found directory to exist and remain unencrypted
2208          * if LOST_FOUND feature is enabled.
2209          *
2210          */
2211         if (f2fs_sb_has_lost_found(sbi) &&
2212                         inode->i_ino == F2FS_ROOT_INO(sbi))
2213                 return -EPERM;
2214
2215         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2216                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2217                                 ctx, len, fs_data, XATTR_CREATE);
2218 }
2219
2220 static bool f2fs_dummy_context(struct inode *inode)
2221 {
2222         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2223 }
2224
2225 static const struct fscrypt_operations f2fs_cryptops = {
2226         .key_prefix     = "f2fs:",
2227         .get_context    = f2fs_get_context,
2228         .set_context    = f2fs_set_context,
2229         .dummy_context  = f2fs_dummy_context,
2230         .empty_dir      = f2fs_empty_dir,
2231         .max_namelen    = F2FS_NAME_LEN,
2232 };
2233 #endif
2234
2235 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2236                 u64 ino, u32 generation)
2237 {
2238         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2239         struct inode *inode;
2240
2241         if (f2fs_check_nid_range(sbi, ino))
2242                 return ERR_PTR(-ESTALE);
2243
2244         /*
2245          * f2fs_iget isn't quite right if the inode is currently unallocated!
2246          * However f2fs_iget currently does appropriate checks to handle stale
2247          * inodes so everything is OK.
2248          */
2249         inode = f2fs_iget(sb, ino);
2250         if (IS_ERR(inode))
2251                 return ERR_CAST(inode);
2252         if (unlikely(generation && inode->i_generation != generation)) {
2253                 /* we didn't find the right inode.. */
2254                 iput(inode);
2255                 return ERR_PTR(-ESTALE);
2256         }
2257         return inode;
2258 }
2259
2260 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2261                 int fh_len, int fh_type)
2262 {
2263         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2264                                     f2fs_nfs_get_inode);
2265 }
2266
2267 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2268                 int fh_len, int fh_type)
2269 {
2270         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2271                                     f2fs_nfs_get_inode);
2272 }
2273
2274 static const struct export_operations f2fs_export_ops = {
2275         .fh_to_dentry = f2fs_fh_to_dentry,
2276         .fh_to_parent = f2fs_fh_to_parent,
2277         .get_parent = f2fs_get_parent,
2278 };
2279
2280 static loff_t max_file_blocks(void)
2281 {
2282         loff_t result = 0;
2283         loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2284
2285         /*
2286          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2287          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2288          * space in inode.i_addr, it will be more safe to reassign
2289          * result as zero.
2290          */
2291
2292         /* two direct node blocks */
2293         result += (leaf_count * 2);
2294
2295         /* two indirect node blocks */
2296         leaf_count *= NIDS_PER_BLOCK;
2297         result += (leaf_count * 2);
2298
2299         /* one double indirect node block */
2300         leaf_count *= NIDS_PER_BLOCK;
2301         result += leaf_count;
2302
2303         return result;
2304 }
2305
2306 static int __f2fs_commit_super(struct buffer_head *bh,
2307                         struct f2fs_super_block *super)
2308 {
2309         lock_buffer(bh);
2310         if (super)
2311                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2312         set_buffer_dirty(bh);
2313         unlock_buffer(bh);
2314
2315         /* it's rare case, we can do fua all the time */
2316         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2317 }
2318
2319 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2320                                         struct buffer_head *bh)
2321 {
2322         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2323                                         (bh->b_data + F2FS_SUPER_OFFSET);
2324         struct super_block *sb = sbi->sb;
2325         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2326         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2327         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2328         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2329         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2330         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2331         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2332         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2333         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2334         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2335         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2336         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2337         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2338         u64 main_end_blkaddr = main_blkaddr +
2339                                 (segment_count_main << log_blocks_per_seg);
2340         u64 seg_end_blkaddr = segment0_blkaddr +
2341                                 (segment_count << log_blocks_per_seg);
2342
2343         if (segment0_blkaddr != cp_blkaddr) {
2344                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2345                           segment0_blkaddr, cp_blkaddr);
2346                 return true;
2347         }
2348
2349         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2350                                                         sit_blkaddr) {
2351                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2352                           cp_blkaddr, sit_blkaddr,
2353                           segment_count_ckpt << log_blocks_per_seg);
2354                 return true;
2355         }
2356
2357         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2358                                                         nat_blkaddr) {
2359                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2360                           sit_blkaddr, nat_blkaddr,
2361                           segment_count_sit << log_blocks_per_seg);
2362                 return true;
2363         }
2364
2365         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2366                                                         ssa_blkaddr) {
2367                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2368                           nat_blkaddr, ssa_blkaddr,
2369                           segment_count_nat << log_blocks_per_seg);
2370                 return true;
2371         }
2372
2373         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2374                                                         main_blkaddr) {
2375                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2376                           ssa_blkaddr, main_blkaddr,
2377                           segment_count_ssa << log_blocks_per_seg);
2378                 return true;
2379         }
2380
2381         if (main_end_blkaddr > seg_end_blkaddr) {
2382                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2383                           main_blkaddr,
2384                           segment0_blkaddr +
2385                           (segment_count << log_blocks_per_seg),
2386                           segment_count_main << log_blocks_per_seg);
2387                 return true;
2388         } else if (main_end_blkaddr < seg_end_blkaddr) {
2389                 int err = 0;
2390                 char *res;
2391
2392                 /* fix in-memory information all the time */
2393                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2394                                 segment0_blkaddr) >> log_blocks_per_seg);
2395
2396                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2397                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2398                         res = "internally";
2399                 } else {
2400                         err = __f2fs_commit_super(bh, NULL);
2401                         res = err ? "failed" : "done";
2402                 }
2403                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2404                           res, main_blkaddr,
2405                           segment0_blkaddr +
2406                           (segment_count << log_blocks_per_seg),
2407                           segment_count_main << log_blocks_per_seg);
2408                 if (err)
2409                         return true;
2410         }
2411         return false;
2412 }
2413
2414 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2415                                 struct buffer_head *bh)
2416 {
2417         block_t segment_count, segs_per_sec, secs_per_zone;
2418         block_t total_sections, blocks_per_seg;
2419         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2420                                         (bh->b_data + F2FS_SUPER_OFFSET);
2421         unsigned int blocksize;
2422         size_t crc_offset = 0;
2423         __u32 crc = 0;
2424
2425         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2426                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2427                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2428                 return -EINVAL;
2429         }
2430
2431         /* Check checksum_offset and crc in superblock */
2432         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2433                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2434                 if (crc_offset !=
2435                         offsetof(struct f2fs_super_block, crc)) {
2436                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2437                                   crc_offset);
2438                         return -EFSCORRUPTED;
2439                 }
2440                 crc = le32_to_cpu(raw_super->crc);
2441                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2442                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2443                         return -EFSCORRUPTED;
2444                 }
2445         }
2446
2447         /* Currently, support only 4KB page cache size */
2448         if (F2FS_BLKSIZE != PAGE_SIZE) {
2449                 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2450                           PAGE_SIZE);
2451                 return -EFSCORRUPTED;
2452         }
2453
2454         /* Currently, support only 4KB block size */
2455         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2456         if (blocksize != F2FS_BLKSIZE) {
2457                 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2458                           blocksize);
2459                 return -EFSCORRUPTED;
2460         }
2461
2462         /* check log blocks per segment */
2463         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2464                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2465                           le32_to_cpu(raw_super->log_blocks_per_seg));
2466                 return -EFSCORRUPTED;
2467         }
2468
2469         /* Currently, support 512/1024/2048/4096 bytes sector size */
2470         if (le32_to_cpu(raw_super->log_sectorsize) >
2471                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2472                 le32_to_cpu(raw_super->log_sectorsize) <
2473                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2474                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2475                           le32_to_cpu(raw_super->log_sectorsize));
2476                 return -EFSCORRUPTED;
2477         }
2478         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2479                 le32_to_cpu(raw_super->log_sectorsize) !=
2480                         F2FS_MAX_LOG_SECTOR_SIZE) {
2481                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2482                           le32_to_cpu(raw_super->log_sectors_per_block),
2483                           le32_to_cpu(raw_super->log_sectorsize));
2484                 return -EFSCORRUPTED;
2485         }
2486
2487         segment_count = le32_to_cpu(raw_super->segment_count);
2488         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2489         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2490         total_sections = le32_to_cpu(raw_super->section_count);
2491
2492         /* blocks_per_seg should be 512, given the above check */
2493         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2494
2495         if (segment_count > F2FS_MAX_SEGMENT ||
2496                                 segment_count < F2FS_MIN_SEGMENTS) {
2497                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2498                 return -EFSCORRUPTED;
2499         }
2500
2501         if (total_sections > segment_count ||
2502                         total_sections < F2FS_MIN_SEGMENTS ||
2503                         segs_per_sec > segment_count || !segs_per_sec) {
2504                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2505                           segment_count, total_sections, segs_per_sec);
2506                 return -EFSCORRUPTED;
2507         }
2508
2509         if ((segment_count / segs_per_sec) < total_sections) {
2510                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2511                           segment_count, segs_per_sec, total_sections);
2512                 return -EFSCORRUPTED;
2513         }
2514
2515         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2516                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2517                           segment_count, le64_to_cpu(raw_super->block_count));
2518                 return -EFSCORRUPTED;
2519         }
2520
2521         if (secs_per_zone > total_sections || !secs_per_zone) {
2522                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2523                           secs_per_zone, total_sections);
2524                 return -EFSCORRUPTED;
2525         }
2526         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2527                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2528                         (le32_to_cpu(raw_super->extension_count) +
2529                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2530                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2531                           le32_to_cpu(raw_super->extension_count),
2532                           raw_super->hot_ext_count,
2533                           F2FS_MAX_EXTENSION);
2534                 return -EFSCORRUPTED;
2535         }
2536
2537         if (le32_to_cpu(raw_super->cp_payload) >
2538                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2539                 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2540                           le32_to_cpu(raw_super->cp_payload),
2541                           blocks_per_seg - F2FS_CP_PACKS);
2542                 return -EFSCORRUPTED;
2543         }
2544
2545         /* check reserved ino info */
2546         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2547                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2548                 le32_to_cpu(raw_super->root_ino) != 3) {
2549                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2550                           le32_to_cpu(raw_super->node_ino),
2551                           le32_to_cpu(raw_super->meta_ino),
2552                           le32_to_cpu(raw_super->root_ino));
2553                 return -EFSCORRUPTED;
2554         }
2555
2556         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2557         if (sanity_check_area_boundary(sbi, bh))
2558                 return -EFSCORRUPTED;
2559
2560         return 0;
2561 }
2562
2563 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2564 {
2565         unsigned int total, fsmeta;
2566         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2567         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2568         unsigned int ovp_segments, reserved_segments;
2569         unsigned int main_segs, blocks_per_seg;
2570         unsigned int sit_segs, nat_segs;
2571         unsigned int sit_bitmap_size, nat_bitmap_size;
2572         unsigned int log_blocks_per_seg;
2573         unsigned int segment_count_main;
2574         unsigned int cp_pack_start_sum, cp_payload;
2575         block_t user_block_count, valid_user_blocks;
2576         block_t avail_node_count, valid_node_count;
2577         int i, j;
2578
2579         total = le32_to_cpu(raw_super->segment_count);
2580         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2581         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2582         fsmeta += sit_segs;
2583         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2584         fsmeta += nat_segs;
2585         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2586         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2587
2588         if (unlikely(fsmeta >= total))
2589                 return 1;
2590
2591         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2592         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2593
2594         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2595                         ovp_segments == 0 || reserved_segments == 0)) {
2596                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2597                 return 1;
2598         }
2599
2600         user_block_count = le64_to_cpu(ckpt->user_block_count);
2601         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2602         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2603         if (!user_block_count || user_block_count >=
2604                         segment_count_main << log_blocks_per_seg) {
2605                 f2fs_err(sbi, "Wrong user_block_count: %u",
2606                          user_block_count);
2607                 return 1;
2608         }
2609
2610         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2611         if (valid_user_blocks > user_block_count) {
2612                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2613                          valid_user_blocks, user_block_count);
2614                 return 1;
2615         }
2616
2617         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2618         avail_node_count = sbi->total_node_count - sbi->nquota_files -
2619                                                 F2FS_RESERVED_NODE_NUM;
2620         if (valid_node_count > avail_node_count) {
2621                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2622                          valid_node_count, avail_node_count);
2623                 return 1;
2624         }
2625
2626         main_segs = le32_to_cpu(raw_super->segment_count_main);
2627         blocks_per_seg = sbi->blocks_per_seg;
2628
2629         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2630                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2631                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2632                         return 1;
2633                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2634                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2635                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2636                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2637                                          i, j,
2638                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2639                                 return 1;
2640                         }
2641                 }
2642         }
2643         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2644                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2645                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2646                         return 1;
2647                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2648                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2649                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2650                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2651                                          i, j,
2652                                          le32_to_cpu(ckpt->cur_data_segno[i]));
2653                                 return 1;
2654                         }
2655                 }
2656         }
2657         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2658                 for (j = i; j < NR_CURSEG_DATA_TYPE; j++) {
2659                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2660                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2661                                 f2fs_err(sbi, "Data segment (%u) and Data segment (%u) has the same segno: %u",
2662                                          i, j,
2663                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2664                                 return 1;
2665                         }
2666                 }
2667         }
2668
2669         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2670         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2671
2672         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2673                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2674                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2675                          sit_bitmap_size, nat_bitmap_size);
2676                 return 1;
2677         }
2678
2679         cp_pack_start_sum = __start_sum_addr(sbi);
2680         cp_payload = __cp_payload(sbi);
2681         if (cp_pack_start_sum < cp_payload + 1 ||
2682                 cp_pack_start_sum > blocks_per_seg - 1 -
2683                         NR_CURSEG_TYPE) {
2684                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2685                          cp_pack_start_sum);
2686                 return 1;
2687         }
2688
2689         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2690                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2691                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2692                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2693                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2694                           le32_to_cpu(ckpt->checksum_offset));
2695                 return 1;
2696         }
2697
2698         if (unlikely(f2fs_cp_error(sbi))) {
2699                 f2fs_err(sbi, "A bug case: need to run fsck");
2700                 return 1;
2701         }
2702         return 0;
2703 }
2704
2705 static void init_sb_info(struct f2fs_sb_info *sbi)
2706 {
2707         struct f2fs_super_block *raw_super = sbi->raw_super;
2708         int i;
2709
2710         sbi->log_sectors_per_block =
2711                 le32_to_cpu(raw_super->log_sectors_per_block);
2712         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2713         sbi->blocksize = 1 << sbi->log_blocksize;
2714         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2715         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2716         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2717         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2718         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2719         sbi->total_node_count =
2720                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2721                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2722         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2723         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2724         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2725         sbi->cur_victim_sec = NULL_SECNO;
2726         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2727         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2728         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2729         sbi->migration_granularity = sbi->segs_per_sec;
2730
2731         sbi->dir_level = DEF_DIR_LEVEL;
2732         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2733         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2734         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2735         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2736         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2737         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2738                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2739         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2740
2741         for (i = 0; i < NR_COUNT_TYPE; i++)
2742                 atomic_set(&sbi->nr_pages[i], 0);
2743
2744         for (i = 0; i < META; i++)
2745                 atomic_set(&sbi->wb_sync_req[i], 0);
2746
2747         INIT_LIST_HEAD(&sbi->s_list);
2748         mutex_init(&sbi->umount_mutex);
2749         init_rwsem(&sbi->io_order_lock);
2750         spin_lock_init(&sbi->cp_lock);
2751
2752         sbi->dirty_device = 0;
2753         spin_lock_init(&sbi->dev_lock);
2754
2755         init_rwsem(&sbi->sb_lock);
2756 }
2757
2758 static int init_percpu_info(struct f2fs_sb_info *sbi)
2759 {
2760         int err;
2761
2762         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2763         if (err)
2764                 return err;
2765
2766         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2767                                                                 GFP_KERNEL);
2768         if (err)
2769                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2770
2771         return err;
2772 }
2773
2774 #ifdef CONFIG_BLK_DEV_ZONED
2775 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2776 {
2777         struct block_device *bdev = FDEV(devi).bdev;
2778         sector_t nr_sectors = bdev->bd_part->nr_sects;
2779         sector_t sector = 0;
2780         struct blk_zone *zones;
2781         unsigned int i, nr_zones;
2782         unsigned int n = 0;
2783         int err = -EIO;
2784
2785         if (!f2fs_sb_has_blkzoned(sbi))
2786                 return 0;
2787
2788         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2789                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2790                 return -EINVAL;
2791         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2792         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2793                                 __ilog2_u32(sbi->blocks_per_blkz))
2794                 return -EINVAL;
2795         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2796         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2797                                         sbi->log_blocks_per_blkz;
2798         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2799                 FDEV(devi).nr_blkz++;
2800
2801         FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
2802                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
2803                                         * sizeof(unsigned long),
2804                                         GFP_KERNEL);
2805         if (!FDEV(devi).blkz_seq)
2806                 return -ENOMEM;
2807
2808 #define F2FS_REPORT_NR_ZONES   4096
2809
2810         zones = f2fs_kzalloc(sbi,
2811                              array_size(F2FS_REPORT_NR_ZONES,
2812                                         sizeof(struct blk_zone)),
2813                              GFP_KERNEL);
2814         if (!zones)
2815                 return -ENOMEM;
2816
2817         /* Get block zones type */
2818         while (zones && sector < nr_sectors) {
2819
2820                 nr_zones = F2FS_REPORT_NR_ZONES;
2821                 err = blkdev_report_zones(bdev, sector, zones, &nr_zones);
2822                 if (err)
2823                         break;
2824                 if (!nr_zones) {
2825                         err = -EIO;
2826                         break;
2827                 }
2828
2829                 for (i = 0; i < nr_zones; i++) {
2830                         if (zones[i].type != BLK_ZONE_TYPE_CONVENTIONAL)
2831                                 set_bit(n, FDEV(devi).blkz_seq);
2832                         sector += zones[i].len;
2833                         n++;
2834                 }
2835         }
2836
2837         kvfree(zones);
2838
2839         return err;
2840 }
2841 #endif
2842
2843 /*
2844  * Read f2fs raw super block.
2845  * Because we have two copies of super block, so read both of them
2846  * to get the first valid one. If any one of them is broken, we pass
2847  * them recovery flag back to the caller.
2848  */
2849 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2850                         struct f2fs_super_block **raw_super,
2851                         int *valid_super_block, int *recovery)
2852 {
2853         struct super_block *sb = sbi->sb;
2854         int block;
2855         struct buffer_head *bh;
2856         struct f2fs_super_block *super;
2857         int err = 0;
2858
2859         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2860         if (!super)
2861                 return -ENOMEM;
2862
2863         for (block = 0; block < 2; block++) {
2864                 bh = sb_bread(sb, block);
2865                 if (!bh) {
2866                         f2fs_err(sbi, "Unable to read %dth superblock",
2867                                  block + 1);
2868                         err = -EIO;
2869                         continue;
2870                 }
2871
2872                 /* sanity checking of raw super */
2873                 err = sanity_check_raw_super(sbi, bh);
2874                 if (err) {
2875                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
2876                                  block + 1);
2877                         brelse(bh);
2878                         continue;
2879                 }
2880
2881                 if (!*raw_super) {
2882                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2883                                                         sizeof(*super));
2884                         *valid_super_block = block;
2885                         *raw_super = super;
2886                 }
2887                 brelse(bh);
2888         }
2889
2890         /* Fail to read any one of the superblocks*/
2891         if (err < 0)
2892                 *recovery = 1;
2893
2894         /* No valid superblock */
2895         if (!*raw_super)
2896                 kvfree(super);
2897         else
2898                 err = 0;
2899
2900         return err;
2901 }
2902
2903 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2904 {
2905         struct buffer_head *bh;
2906         __u32 crc = 0;
2907         int err;
2908
2909         if ((recover && f2fs_readonly(sbi->sb)) ||
2910                                 bdev_read_only(sbi->sb->s_bdev)) {
2911                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2912                 return -EROFS;
2913         }
2914
2915         /* we should update superblock crc here */
2916         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
2917                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
2918                                 offsetof(struct f2fs_super_block, crc));
2919                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
2920         }
2921
2922         /* write back-up superblock first */
2923         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2924         if (!bh)
2925                 return -EIO;
2926         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2927         brelse(bh);
2928
2929         /* if we are in recovery path, skip writing valid superblock */
2930         if (recover || err)
2931                 return err;
2932
2933         /* write current valid superblock */
2934         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2935         if (!bh)
2936                 return -EIO;
2937         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2938         brelse(bh);
2939         return err;
2940 }
2941
2942 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2943 {
2944         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2945         unsigned int max_devices = MAX_DEVICES;
2946         int i;
2947
2948         /* Initialize single device information */
2949         if (!RDEV(0).path[0]) {
2950                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2951                         return 0;
2952                 max_devices = 1;
2953         }
2954
2955         /*
2956          * Initialize multiple devices information, or single
2957          * zoned block device information.
2958          */
2959         sbi->devs = f2fs_kzalloc(sbi,
2960                                  array_size(max_devices,
2961                                             sizeof(struct f2fs_dev_info)),
2962                                  GFP_KERNEL);
2963         if (!sbi->devs)
2964                 return -ENOMEM;
2965
2966         for (i = 0; i < max_devices; i++) {
2967
2968                 if (i > 0 && !RDEV(i).path[0])
2969                         break;
2970
2971                 if (max_devices == 1) {
2972                         /* Single zoned block device mount */
2973                         FDEV(0).bdev =
2974                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2975                                         sbi->sb->s_mode, sbi->sb->s_type);
2976                 } else {
2977                         /* Multi-device mount */
2978                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2979                         FDEV(i).total_segments =
2980                                 le32_to_cpu(RDEV(i).total_segments);
2981                         if (i == 0) {
2982                                 FDEV(i).start_blk = 0;
2983                                 FDEV(i).end_blk = FDEV(i).start_blk +
2984                                     (FDEV(i).total_segments <<
2985                                     sbi->log_blocks_per_seg) - 1 +
2986                                     le32_to_cpu(raw_super->segment0_blkaddr);
2987                         } else {
2988                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2989                                 FDEV(i).end_blk = FDEV(i).start_blk +
2990                                         (FDEV(i).total_segments <<
2991                                         sbi->log_blocks_per_seg) - 1;
2992                         }
2993                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2994                                         sbi->sb->s_mode, sbi->sb->s_type);
2995                 }
2996                 if (IS_ERR(FDEV(i).bdev))
2997                         return PTR_ERR(FDEV(i).bdev);
2998
2999                 /* to release errored devices */
3000                 sbi->s_ndevs = i + 1;
3001
3002 #ifdef CONFIG_BLK_DEV_ZONED
3003                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3004                                 !f2fs_sb_has_blkzoned(sbi)) {
3005                         f2fs_err(sbi, "Zoned block device feature not enabled\n");
3006                         return -EINVAL;
3007                 }
3008                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3009                         if (init_blkz_info(sbi, i)) {
3010                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3011                                 return -EINVAL;
3012                         }
3013                         if (max_devices == 1)
3014                                 break;
3015                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3016                                   i, FDEV(i).path,
3017                                   FDEV(i).total_segments,
3018                                   FDEV(i).start_blk, FDEV(i).end_blk,
3019                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3020                                   "Host-aware" : "Host-managed");
3021                         continue;
3022                 }
3023 #endif
3024                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3025                           i, FDEV(i).path,
3026                           FDEV(i).total_segments,
3027                           FDEV(i).start_blk, FDEV(i).end_blk);
3028         }
3029         f2fs_info(sbi,
3030                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3031         return 0;
3032 }
3033
3034 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3035 {
3036         struct f2fs_sm_info *sm_i = SM_I(sbi);
3037
3038         /* adjust parameters according to the volume size */
3039         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3040                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3041                 sm_i->dcc_info->discard_granularity = 1;
3042                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3043         }
3044
3045         sbi->readdir_ra = 1;
3046 }
3047
3048 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3049 {
3050         struct f2fs_sb_info *sbi;
3051         struct f2fs_super_block *raw_super;
3052         struct inode *root;
3053         int err;
3054         bool skip_recovery = false, need_fsck = false;
3055         char *options = NULL;
3056         int recovery, i, valid_super_block;
3057         struct curseg_info *seg_i;
3058         int retry_cnt = 1;
3059
3060 try_onemore:
3061         err = -EINVAL;
3062         raw_super = NULL;
3063         valid_super_block = -1;
3064         recovery = 0;
3065
3066         /* allocate memory for f2fs-specific super block info */
3067         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3068         if (!sbi)
3069                 return -ENOMEM;
3070
3071         sbi->sb = sb;
3072
3073         /* Load the checksum driver */
3074         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3075         if (IS_ERR(sbi->s_chksum_driver)) {
3076                 f2fs_err(sbi, "Cannot load crc32 driver.");
3077                 err = PTR_ERR(sbi->s_chksum_driver);
3078                 sbi->s_chksum_driver = NULL;
3079                 goto free_sbi;
3080         }
3081
3082         /* set a block size */
3083         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3084                 f2fs_err(sbi, "unable to set blocksize");
3085                 goto free_sbi;
3086         }
3087
3088         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3089                                                                 &recovery);
3090         if (err)
3091                 goto free_sbi;
3092
3093         sb->s_fs_info = sbi;
3094         sbi->raw_super = raw_super;
3095
3096         /* precompute checksum seed for metadata */
3097         if (f2fs_sb_has_inode_chksum(sbi))
3098                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3099                                                 sizeof(raw_super->uuid));
3100
3101         /*
3102          * The BLKZONED feature indicates that the drive was formatted with
3103          * zone alignment optimization. This is optional for host-aware
3104          * devices, but mandatory for host-managed zoned block devices.
3105          */
3106 #ifndef CONFIG_BLK_DEV_ZONED
3107         if (f2fs_sb_has_blkzoned(sbi)) {
3108                 f2fs_err(sbi, "Zoned block device support is not enabled");
3109                 err = -EOPNOTSUPP;
3110                 goto free_sb_buf;
3111         }
3112 #endif
3113         default_options(sbi);
3114         /* parse mount options */
3115         options = kstrdup((const char *)data, GFP_KERNEL);
3116         if (data && !options) {
3117                 err = -ENOMEM;
3118                 goto free_sb_buf;
3119         }
3120
3121         err = parse_options(sb, options);
3122         if (err)
3123                 goto free_options;
3124
3125         sbi->max_file_blocks = max_file_blocks();
3126         sb->s_maxbytes = sbi->max_file_blocks <<
3127                                 le32_to_cpu(raw_super->log_blocksize);
3128         sb->s_max_links = F2FS_LINK_MAX;
3129
3130 #ifdef CONFIG_QUOTA
3131         sb->dq_op = &f2fs_quota_operations;
3132         sb->s_qcop = &f2fs_quotactl_ops;
3133         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3134
3135         if (f2fs_sb_has_quota_ino(sbi)) {
3136                 for (i = 0; i < MAXQUOTAS; i++) {
3137                         if (f2fs_qf_ino(sbi->sb, i))
3138                                 sbi->nquota_files++;
3139                 }
3140         }
3141 #endif
3142
3143         sb->s_op = &f2fs_sops;
3144 #ifdef CONFIG_FS_ENCRYPTION
3145         sb->s_cop = &f2fs_cryptops;
3146 #endif
3147         sb->s_xattr = f2fs_xattr_handlers;
3148         sb->s_export_op = &f2fs_export_ops;
3149         sb->s_magic = F2FS_SUPER_MAGIC;
3150         sb->s_time_gran = 1;
3151         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3152                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3153         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3154         sb->s_iflags |= SB_I_CGROUPWB;
3155
3156         /* init f2fs-specific super block info */
3157         sbi->valid_super_block = valid_super_block;
3158         mutex_init(&sbi->gc_mutex);
3159         mutex_init(&sbi->writepages);
3160         mutex_init(&sbi->cp_mutex);
3161         mutex_init(&sbi->resize_mutex);
3162         init_rwsem(&sbi->node_write);
3163         init_rwsem(&sbi->node_change);
3164
3165         /* disallow all the data/node/meta page writes */
3166         set_sbi_flag(sbi, SBI_POR_DOING);
3167         spin_lock_init(&sbi->stat_lock);
3168
3169         /* init iostat info */
3170         spin_lock_init(&sbi->iostat_lock);
3171         sbi->iostat_enable = false;
3172
3173         for (i = 0; i < NR_PAGE_TYPE; i++) {
3174                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3175                 int j;
3176
3177                 sbi->write_io[i] =
3178                         f2fs_kmalloc(sbi,
3179                                      array_size(n,
3180                                                 sizeof(struct f2fs_bio_info)),
3181                                      GFP_KERNEL);
3182                 if (!sbi->write_io[i]) {
3183                         err = -ENOMEM;
3184                         goto free_bio_info;
3185                 }
3186
3187                 for (j = HOT; j < n; j++) {
3188                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3189                         sbi->write_io[i][j].sbi = sbi;
3190                         sbi->write_io[i][j].bio = NULL;
3191                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3192                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3193                 }
3194         }
3195
3196         init_rwsem(&sbi->cp_rwsem);
3197         init_rwsem(&sbi->quota_sem);
3198         init_waitqueue_head(&sbi->cp_wait);
3199         init_sb_info(sbi);
3200
3201         err = init_percpu_info(sbi);
3202         if (err)
3203                 goto free_bio_info;
3204
3205         if (F2FS_IO_SIZE(sbi) > 1) {
3206                 sbi->write_io_dummy =
3207                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3208                 if (!sbi->write_io_dummy) {
3209                         err = -ENOMEM;
3210                         goto free_percpu;
3211                 }
3212         }
3213
3214         /* get an inode for meta space */
3215         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3216         if (IS_ERR(sbi->meta_inode)) {
3217                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3218                 err = PTR_ERR(sbi->meta_inode);
3219                 goto free_io_dummy;
3220         }
3221
3222         err = f2fs_get_valid_checkpoint(sbi);
3223         if (err) {
3224                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3225                 goto free_meta_inode;
3226         }
3227
3228         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3229                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3230         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3231                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3232                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3233         }
3234
3235         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3236                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3237
3238         /* Initialize device list */
3239         err = f2fs_scan_devices(sbi);
3240         if (err) {
3241                 f2fs_err(sbi, "Failed to find devices");
3242                 goto free_devices;
3243         }
3244
3245         sbi->total_valid_node_count =
3246                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3247         percpu_counter_set(&sbi->total_valid_inode_count,
3248                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3249         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3250         sbi->total_valid_block_count =
3251                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3252         sbi->last_valid_block_count = sbi->total_valid_block_count;
3253         sbi->reserved_blocks = 0;
3254         sbi->current_reserved_blocks = 0;
3255         limit_reserve_root(sbi);
3256
3257         for (i = 0; i < NR_INODE_TYPE; i++) {
3258                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3259                 spin_lock_init(&sbi->inode_lock[i]);
3260         }
3261         mutex_init(&sbi->flush_lock);
3262
3263         f2fs_init_extent_cache_info(sbi);
3264
3265         f2fs_init_ino_entry_info(sbi);
3266
3267         f2fs_init_fsync_node_info(sbi);
3268
3269         /* setup f2fs internal modules */
3270         err = f2fs_build_segment_manager(sbi);
3271         if (err) {
3272                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3273                          err);
3274                 goto free_sm;
3275         }
3276         err = f2fs_build_node_manager(sbi);
3277         if (err) {
3278                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3279                          err);
3280                 goto free_nm;
3281         }
3282
3283         /* For write statistics */
3284         if (sb->s_bdev->bd_part)
3285                 sbi->sectors_written_start =
3286                         (u64)part_stat_read(sb->s_bdev->bd_part,
3287                                             sectors[STAT_WRITE]);
3288
3289         /* Read accumulated write IO statistics if exists */
3290         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3291         if (__exist_node_summaries(sbi))
3292                 sbi->kbytes_written =
3293                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3294
3295         f2fs_build_gc_manager(sbi);
3296
3297         err = f2fs_build_stats(sbi);
3298         if (err)
3299                 goto free_nm;
3300
3301         /* get an inode for node space */
3302         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3303         if (IS_ERR(sbi->node_inode)) {
3304                 f2fs_err(sbi, "Failed to read node inode");
3305                 err = PTR_ERR(sbi->node_inode);
3306                 goto free_stats;
3307         }
3308
3309         /* read root inode and dentry */
3310         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3311         if (IS_ERR(root)) {
3312                 f2fs_err(sbi, "Failed to read root inode");
3313                 err = PTR_ERR(root);
3314                 goto free_node_inode;
3315         }
3316         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3317                         !root->i_size || !root->i_nlink) {
3318                 iput(root);
3319                 err = -EINVAL;
3320                 goto free_node_inode;
3321         }
3322
3323         sb->s_root = d_make_root(root); /* allocate root dentry */
3324         if (!sb->s_root) {
3325                 err = -ENOMEM;
3326                 goto free_node_inode;
3327         }
3328
3329         err = f2fs_register_sysfs(sbi);
3330         if (err)
3331                 goto free_root_inode;
3332
3333 #ifdef CONFIG_QUOTA
3334         /* Enable quota usage during mount */
3335         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3336                 err = f2fs_enable_quotas(sb);
3337                 if (err)
3338                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3339         }
3340 #endif
3341         /* if there are nt orphan nodes free them */
3342         err = f2fs_recover_orphan_inodes(sbi);
3343         if (err)
3344                 goto free_meta;
3345
3346         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3347                 goto reset_checkpoint;
3348
3349         /* recover fsynced data */
3350         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3351                 /*
3352                  * mount should be failed, when device has readonly mode, and
3353                  * previous checkpoint was not done by clean system shutdown.
3354                  */
3355                 if (f2fs_hw_is_readonly(sbi)) {
3356                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3357                                 err = -EROFS;
3358                                 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3359                                 goto free_meta;
3360                         }
3361                         f2fs_info(sbi, "write access unavailable, skipping recovery");
3362                         goto reset_checkpoint;
3363                 }
3364
3365                 if (need_fsck)
3366                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3367
3368                 if (skip_recovery)
3369                         goto reset_checkpoint;
3370
3371                 err = f2fs_recover_fsync_data(sbi, false);
3372                 if (err < 0) {
3373                         if (err != -ENOMEM)
3374                                 skip_recovery = true;
3375                         need_fsck = true;
3376                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3377                                  err);
3378                         goto free_meta;
3379                 }
3380         } else {
3381                 err = f2fs_recover_fsync_data(sbi, true);
3382
3383                 if (!f2fs_readonly(sb) && err > 0) {
3384                         err = -EINVAL;
3385                         f2fs_err(sbi, "Need to recover fsync data");
3386                         goto free_meta;
3387                 }
3388         }
3389 reset_checkpoint:
3390         /* f2fs_recover_fsync_data() cleared this already */
3391         clear_sbi_flag(sbi, SBI_POR_DOING);
3392
3393         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3394                 err = f2fs_disable_checkpoint(sbi);
3395                 if (err)
3396                         goto sync_free_meta;
3397         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3398                 f2fs_enable_checkpoint(sbi);
3399         }
3400
3401         /*
3402          * If filesystem is not mounted as read-only then
3403          * do start the gc_thread.
3404          */
3405         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3406                 /* After POR, we can run background GC thread.*/
3407                 err = f2fs_start_gc_thread(sbi);
3408                 if (err)
3409                         goto sync_free_meta;
3410         }
3411         kvfree(options);
3412
3413         /* recover broken superblock */
3414         if (recovery) {
3415                 err = f2fs_commit_super(sbi, true);
3416                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3417                           sbi->valid_super_block ? 1 : 2, err);
3418         }
3419
3420         f2fs_join_shrinker(sbi);
3421
3422         f2fs_tuning_parameters(sbi);
3423
3424         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3425                     cur_cp_version(F2FS_CKPT(sbi)));
3426         f2fs_update_time(sbi, CP_TIME);
3427         f2fs_update_time(sbi, REQ_TIME);
3428         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3429         return 0;
3430
3431 sync_free_meta:
3432         /* safe to flush all the data */
3433         sync_filesystem(sbi->sb);
3434         retry_cnt = 0;
3435
3436 free_meta:
3437 #ifdef CONFIG_QUOTA
3438         f2fs_truncate_quota_inode_pages(sb);
3439         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3440                 f2fs_quota_off_umount(sbi->sb);
3441 #endif
3442         /*
3443          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3444          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3445          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3446          * falls into an infinite loop in f2fs_sync_meta_pages().
3447          */
3448         truncate_inode_pages_final(META_MAPPING(sbi));
3449         /* evict some inodes being cached by GC */
3450         evict_inodes(sb);
3451         f2fs_unregister_sysfs(sbi);
3452 free_root_inode:
3453         dput(sb->s_root);
3454         sb->s_root = NULL;
3455 free_node_inode:
3456         f2fs_release_ino_entry(sbi, true);
3457         truncate_inode_pages_final(NODE_MAPPING(sbi));
3458         iput(sbi->node_inode);
3459         sbi->node_inode = NULL;
3460 free_stats:
3461         f2fs_destroy_stats(sbi);
3462 free_nm:
3463         f2fs_destroy_node_manager(sbi);
3464 free_sm:
3465         f2fs_destroy_segment_manager(sbi);
3466 free_devices:
3467         destroy_device_list(sbi);
3468         kvfree(sbi->ckpt);
3469 free_meta_inode:
3470         make_bad_inode(sbi->meta_inode);
3471         iput(sbi->meta_inode);
3472         sbi->meta_inode = NULL;
3473 free_io_dummy:
3474         mempool_destroy(sbi->write_io_dummy);
3475 free_percpu:
3476         destroy_percpu_info(sbi);
3477 free_bio_info:
3478         for (i = 0; i < NR_PAGE_TYPE; i++)
3479                 kvfree(sbi->write_io[i]);
3480 free_options:
3481 #ifdef CONFIG_QUOTA
3482         for (i = 0; i < MAXQUOTAS; i++)
3483                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3484 #endif
3485         kvfree(options);
3486 free_sb_buf:
3487         kvfree(raw_super);
3488 free_sbi:
3489         if (sbi->s_chksum_driver)
3490                 crypto_free_shash(sbi->s_chksum_driver);
3491         kvfree(sbi);
3492
3493         /* give only one another chance */
3494         if (retry_cnt > 0 && skip_recovery) {
3495                 retry_cnt--;
3496                 shrink_dcache_sb(sb);
3497                 goto try_onemore;
3498         }
3499         return err;
3500 }
3501
3502 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3503                         const char *dev_name, void *data)
3504 {
3505         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3506 }
3507
3508 static void kill_f2fs_super(struct super_block *sb)
3509 {
3510         if (sb->s_root) {
3511                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3512
3513                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3514                 f2fs_stop_gc_thread(sbi);
3515                 f2fs_stop_discard_thread(sbi);
3516
3517                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3518                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3519                         struct cp_control cpc = {
3520                                 .reason = CP_UMOUNT,
3521                         };
3522                         f2fs_write_checkpoint(sbi, &cpc);
3523                 }
3524
3525                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3526                         sb->s_flags &= ~SB_RDONLY;
3527         }
3528         kill_block_super(sb);
3529 }
3530
3531 static struct file_system_type f2fs_fs_type = {
3532         .owner          = THIS_MODULE,
3533         .name           = "f2fs",
3534         .mount          = f2fs_mount,
3535         .kill_sb        = kill_f2fs_super,
3536         .fs_flags       = FS_REQUIRES_DEV,
3537 };
3538 MODULE_ALIAS_FS("f2fs");
3539
3540 static int __init init_inodecache(void)
3541 {
3542         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3543                         sizeof(struct f2fs_inode_info), 0,
3544                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3545         if (!f2fs_inode_cachep)
3546                 return -ENOMEM;
3547         return 0;
3548 }
3549
3550 static void destroy_inodecache(void)
3551 {
3552         /*
3553          * Make sure all delayed rcu free inodes are flushed before we
3554          * destroy cache.
3555          */
3556         rcu_barrier();
3557         kmem_cache_destroy(f2fs_inode_cachep);
3558 }
3559
3560 static int __init init_f2fs_fs(void)
3561 {
3562         int err;
3563
3564         if (PAGE_SIZE != F2FS_BLKSIZE) {
3565                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3566                                 PAGE_SIZE, F2FS_BLKSIZE);
3567                 return -EINVAL;
3568         }
3569
3570         f2fs_build_trace_ios();
3571
3572         err = init_inodecache();
3573         if (err)
3574                 goto fail;
3575         err = f2fs_create_node_manager_caches();
3576         if (err)
3577                 goto free_inodecache;
3578         err = f2fs_create_segment_manager_caches();
3579         if (err)
3580                 goto free_node_manager_caches;
3581         err = f2fs_create_checkpoint_caches();
3582         if (err)
3583                 goto free_segment_manager_caches;
3584         err = f2fs_create_extent_cache();
3585         if (err)
3586                 goto free_checkpoint_caches;
3587         err = f2fs_init_sysfs();
3588         if (err)
3589                 goto free_extent_cache;
3590         err = register_shrinker(&f2fs_shrinker_info);
3591         if (err)
3592                 goto free_sysfs;
3593         err = register_filesystem(&f2fs_fs_type);
3594         if (err)
3595                 goto free_shrinker;
3596         f2fs_create_root_stats();
3597         err = f2fs_init_post_read_processing();
3598         if (err)
3599                 goto free_root_stats;
3600         return 0;
3601
3602 free_root_stats:
3603         f2fs_destroy_root_stats();
3604         unregister_filesystem(&f2fs_fs_type);
3605 free_shrinker:
3606         unregister_shrinker(&f2fs_shrinker_info);
3607 free_sysfs:
3608         f2fs_exit_sysfs();
3609 free_extent_cache:
3610         f2fs_destroy_extent_cache();
3611 free_checkpoint_caches:
3612         f2fs_destroy_checkpoint_caches();
3613 free_segment_manager_caches:
3614         f2fs_destroy_segment_manager_caches();
3615 free_node_manager_caches:
3616         f2fs_destroy_node_manager_caches();
3617 free_inodecache:
3618         destroy_inodecache();
3619 fail:
3620         return err;
3621 }
3622
3623 static void __exit exit_f2fs_fs(void)
3624 {
3625         f2fs_destroy_post_read_processing();
3626         f2fs_destroy_root_stats();
3627         unregister_filesystem(&f2fs_fs_type);
3628         unregister_shrinker(&f2fs_shrinker_info);
3629         f2fs_exit_sysfs();
3630         f2fs_destroy_extent_cache();
3631         f2fs_destroy_checkpoint_caches();
3632         f2fs_destroy_segment_manager_caches();
3633         f2fs_destroy_node_manager_caches();
3634         destroy_inodecache();
3635         f2fs_destroy_trace_ios();
3636 }
3637
3638 module_init(init_f2fs_fs)
3639 module_exit(exit_f2fs_fs)
3640
3641 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3642 MODULE_DESCRIPTION("Flash Friendly File System");
3643 MODULE_LICENSE("GPL");
3644