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 #include <linux/unicode.h>
  27 #include <linux/part_stat.h>
  28
  29 #include "f2fs.h"
  30 #include "node.h"
  31 #include "segment.h"
  32 #include "xattr.h"
  33 #include "gc.h"
  34 #include "trace.h"
  35
  36 #define CREATE_TRACE_POINTS
  37 #include <trace/events/f2fs.h>
  38
  39 static struct kmem_cache *f2fs_inode_cachep;
  40
  41 #ifdef CONFIG_F2FS_FAULT_INJECTION
  42
  43 const char *f2fs_fault_name[FAULT_MAX] = {
  44         [FAULT_KMALLOC]         = "kmalloc",
  45         [FAULT_KVMALLOC]        = "kvmalloc",
  46         [FAULT_PAGE_ALLOC]      = "page alloc",
  47         [FAULT_PAGE_GET]        = "page get",
  48         [FAULT_ALLOC_BIO]       = "alloc bio",
  49         [FAULT_ALLOC_NID]       = "alloc nid",
  50         [FAULT_ORPHAN]          = "orphan",
  51         [FAULT_BLOCK]           = "no more block",
  52         [FAULT_DIR_DEPTH]       = "too big dir depth",
  53         [FAULT_EVICT_INODE]     = "evict_inode fail",
  54         [FAULT_TRUNCATE]        = "truncate fail",
  55         [FAULT_READ_IO]         = "read IO error",
  56         [FAULT_CHECKPOINT]      = "checkpoint error",
  57         [FAULT_DISCARD]         = "discard error",
  58         [FAULT_WRITE_IO]        = "write IO error",
  59 };
  60
  61 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  62                                                         unsigned int type)
  63 {
  64         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  65
  66         if (rate) {
  67                 atomic_set(&ffi->inject_ops, 0);
  68                 ffi->inject_rate = rate;
  69         }
  70
  71         if (type)
  72                 ffi->inject_type = type;
  73
  74         if (!rate && !type)
  75                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  76 }
  77 #endif
  78
  79 /* f2fs-wide shrinker description */
  80 static struct shrinker f2fs_shrinker_info = {
  81         .scan_objects = f2fs_shrink_scan,
  82         .count_objects = f2fs_shrink_count,
  83         .seeks = DEFAULT_SEEKS,
  84 };
  85
  86 enum {
  87         Opt_gc_background,
  88         Opt_disable_roll_forward,
  89         Opt_norecovery,
  90         Opt_discard,
  91         Opt_nodiscard,
  92         Opt_noheap,
  93         Opt_heap,
  94         Opt_user_xattr,
  95         Opt_nouser_xattr,
  96         Opt_acl,
  97         Opt_noacl,
  98         Opt_active_logs,
  99         Opt_disable_ext_identify,
 100         Opt_inline_xattr,
 101         Opt_noinline_xattr,
 102         Opt_inline_xattr_size,
 103         Opt_inline_data,
 104         Opt_inline_dentry,
 105         Opt_noinline_dentry,
 106         Opt_flush_merge,
 107         Opt_noflush_merge,
 108         Opt_nobarrier,
 109         Opt_fastboot,
 110         Opt_extent_cache,
 111         Opt_noextent_cache,
 112         Opt_noinline_data,
 113         Opt_data_flush,
 114         Opt_reserve_root,
 115         Opt_resgid,
 116         Opt_resuid,
 117         Opt_mode,
 118         Opt_io_size_bits,
 119         Opt_fault_injection,
 120         Opt_fault_type,
 121         Opt_lazytime,
 122         Opt_nolazytime,
 123         Opt_quota,
 124         Opt_noquota,
 125         Opt_usrquota,
 126         Opt_grpquota,
 127         Opt_prjquota,
 128         Opt_usrjquota,
 129         Opt_grpjquota,
 130         Opt_prjjquota,
 131         Opt_offusrjquota,
 132         Opt_offgrpjquota,
 133         Opt_offprjjquota,
 134         Opt_jqfmt_vfsold,
 135         Opt_jqfmt_vfsv0,
 136         Opt_jqfmt_vfsv1,
 137         Opt_whint,
 138         Opt_alloc,
 139         Opt_fsync,
 140         Opt_test_dummy_encryption,
 141         Opt_checkpoint_disable,
 142         Opt_checkpoint_disable_cap,
 143         Opt_checkpoint_disable_cap_perc,
 144         Opt_checkpoint_enable,
 145         Opt_compress_algorithm,
 146         Opt_compress_log_size,
 147         Opt_compress_extension,
 148         Opt_err,
 149 };
 150
 151 static match_table_t f2fs_tokens = {
 152         {Opt_gc_background, "background_gc=%s"},
 153         {Opt_disable_roll_forward, "disable_roll_forward"},
 154         {Opt_norecovery, "norecovery"},
 155         {Opt_discard, "discard"},
 156         {Opt_nodiscard, "nodiscard"},
 157         {Opt_noheap, "no_heap"},
 158         {Opt_heap, "heap"},
 159         {Opt_user_xattr, "user_xattr"},
 160         {Opt_nouser_xattr, "nouser_xattr"},
 161         {Opt_acl, "acl"},
 162         {Opt_noacl, "noacl"},
 163         {Opt_active_logs, "active_logs=%u"},
 164         {Opt_disable_ext_identify, "disable_ext_identify"},
 165         {Opt_inline_xattr, "inline_xattr"},
 166         {Opt_noinline_xattr, "noinline_xattr"},
 167         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 168         {Opt_inline_data, "inline_data"},
 169         {Opt_inline_dentry, "inline_dentry"},
 170         {Opt_noinline_dentry, "noinline_dentry"},
 171         {Opt_flush_merge, "flush_merge"},
 172         {Opt_noflush_merge, "noflush_merge"},
 173         {Opt_nobarrier, "nobarrier"},
 174         {Opt_fastboot, "fastboot"},
 175         {Opt_extent_cache, "extent_cache"},
 176         {Opt_noextent_cache, "noextent_cache"},
 177         {Opt_noinline_data, "noinline_data"},
 178         {Opt_data_flush, "data_flush"},
 179         {Opt_reserve_root, "reserve_root=%u"},
 180         {Opt_resgid, "resgid=%u"},
 181         {Opt_resuid, "resuid=%u"},
 182         {Opt_mode, "mode=%s"},
 183         {Opt_io_size_bits, "io_bits=%u"},
 184         {Opt_fault_injection, "fault_injection=%u"},
 185         {Opt_fault_type, "fault_type=%u"},
 186         {Opt_lazytime, "lazytime"},
 187         {Opt_nolazytime, "nolazytime"},
 188         {Opt_quota, "quota"},
 189         {Opt_noquota, "noquota"},
 190         {Opt_usrquota, "usrquota"},
 191         {Opt_grpquota, "grpquota"},
 192         {Opt_prjquota, "prjquota"},
 193         {Opt_usrjquota, "usrjquota=%s"},
 194         {Opt_grpjquota, "grpjquota=%s"},
 195         {Opt_prjjquota, "prjjquota=%s"},
 196         {Opt_offusrjquota, "usrjquota="},
 197         {Opt_offgrpjquota, "grpjquota="},
 198         {Opt_offprjjquota, "prjjquota="},
 199         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 200         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 201         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 202         {Opt_whint, "whint_mode=%s"},
 203         {Opt_alloc, "alloc_mode=%s"},
 204         {Opt_fsync, "fsync_mode=%s"},
 205         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 206         {Opt_checkpoint_disable, "checkpoint=disable"},
 207         {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 208         {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 209         {Opt_checkpoint_enable, "checkpoint=enable"},
 210         {Opt_compress_algorithm, "compress_algorithm=%s"},
 211         {Opt_compress_log_size, "compress_log_size=%u"},
 212         {Opt_compress_extension, "compress_extension=%s"},
 213         {Opt_err, NULL},
 214 };
 215
 216 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 217 {
 218         struct va_format vaf;
 219         va_list args;
 220         int level;
 221
 222         va_start(args, fmt);
 223
 224         level = printk_get_level(fmt);
 225         vaf.fmt = printk_skip_level(fmt);
 226         vaf.va = &args;
 227         printk("%c%cF2FS-fs (%s): %pV\n",
 228                KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 229
 230         va_end(args);
 231 }
 232
 233 #ifdef CONFIG_UNICODE
 234 static const struct f2fs_sb_encodings {
 235         __u16 magic;
 236         char *name;
 237         char *version;
 238 } f2fs_sb_encoding_map[] = {
 239         {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
 240 };
 241
 242 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
 243                                  const struct f2fs_sb_encodings **encoding,
 244                                  __u16 *flags)
 245 {
 246         __u16 magic = le16_to_cpu(sb->s_encoding);
 247         int i;
 248
 249         for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
 250                 if (magic == f2fs_sb_encoding_map[i].magic)
 251                         break;
 252
 253         if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
 254                 return -EINVAL;
 255
 256         *encoding = &f2fs_sb_encoding_map[i];
 257         *flags = le16_to_cpu(sb->s_encoding_flags);
 258
 259         return 0;
 260 }
 261 #endif
 262
 263 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 264 {
 265         block_t limit = min((sbi->user_block_count << 1) / 1000,
 266                         sbi->user_block_count - sbi->reserved_blocks);
 267
 268         /* limit is 0.2% */
 269         if (test_opt(sbi, RESERVE_ROOT) &&
 270                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 271                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 272                 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 273                           F2FS_OPTION(sbi).root_reserved_blocks);
 274         }
 275         if (!test_opt(sbi, RESERVE_ROOT) &&
 276                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 277                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 278                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 279                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 280                 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 281                           from_kuid_munged(&init_user_ns,
 282                                            F2FS_OPTION(sbi).s_resuid),
 283                           from_kgid_munged(&init_user_ns,
 284                                            F2FS_OPTION(sbi).s_resgid));
 285 }
 286
 287 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
 288 {
 289         if (!F2FS_OPTION(sbi).unusable_cap_perc)
 290                 return;
 291
 292         if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
 293                 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
 294         else
 295                 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
 296                                         F2FS_OPTION(sbi).unusable_cap_perc;
 297
 298         f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
 299                         F2FS_OPTION(sbi).unusable_cap,
 300                         F2FS_OPTION(sbi).unusable_cap_perc);
 301 }
 302
 303 static void init_once(void *foo)
 304 {
 305         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 306
 307         inode_init_once(&fi->vfs_inode);
 308 }
 309
 310 #ifdef CONFIG_QUOTA
 311 static const char * const quotatypes[] = INITQFNAMES;
 312 #define QTYPE2NAME(t) (quotatypes[t])
 313 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 314                                                         substring_t *args)
 315 {
 316         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 317         char *qname;
 318         int ret = -EINVAL;
 319
 320         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 321                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 322                 return -EINVAL;
 323         }
 324         if (f2fs_sb_has_quota_ino(sbi)) {
 325                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 326                 return 0;
 327         }
 328
 329         qname = match_strdup(args);
 330         if (!qname) {
 331                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 332                 return -ENOMEM;
 333         }
 334         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 335                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 336                         ret = 0;
 337                 else
 338                         f2fs_err(sbi, "%s quota file already specified",
 339                                  QTYPE2NAME(qtype));
 340                 goto errout;
 341         }
 342         if (strchr(qname, '/')) {
 343                 f2fs_err(sbi, "quotafile must be on filesystem root");
 344                 goto errout;
 345         }
 346         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 347         set_opt(sbi, QUOTA);
 348         return 0;
 349 errout:
 350         kvfree(qname);
 351         return ret;
 352 }
 353
 354 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 355 {
 356         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 357
 358         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 359                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 360                 return -EINVAL;
 361         }
 362         kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 363         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 364         return 0;
 365 }
 366
 367 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 368 {
 369         /*
 370          * We do the test below only for project quotas. 'usrquota' and
 371          * 'grpquota' mount options are allowed even without quota feature
 372          * to support legacy quotas in quota files.
 373          */
 374         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 375                 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 376                 return -1;
 377         }
 378         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 379                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 380                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 381                 if (test_opt(sbi, USRQUOTA) &&
 382                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 383                         clear_opt(sbi, USRQUOTA);
 384
 385                 if (test_opt(sbi, GRPQUOTA) &&
 386                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 387                         clear_opt(sbi, GRPQUOTA);
 388
 389                 if (test_opt(sbi, PRJQUOTA) &&
 390                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 391                         clear_opt(sbi, PRJQUOTA);
 392
 393                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 394                                 test_opt(sbi, PRJQUOTA)) {
 395                         f2fs_err(sbi, "old and new quota format mixing");
 396                         return -1;
 397                 }
 398
 399                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 400                         f2fs_err(sbi, "journaled quota format not specified");
 401                         return -1;
 402                 }
 403         }
 404
 405         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 406                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 407                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 408         }
 409         return 0;
 410 }
 411 #endif
 412
 413 static int parse_options(struct super_block *sb, char *options)
 414 {
 415         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 416         substring_t args[MAX_OPT_ARGS];
 417         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 418         char *p, *name;
 419         int arg = 0, ext_cnt;
 420         kuid_t uid;
 421         kgid_t gid;
 422 #ifdef CONFIG_QUOTA
 423         int ret;
 424 #endif
 425
 426         if (!options)
 427                 return 0;
 428
 429         while ((p = strsep(&options, ",")) != NULL) {
 430                 int token;
 431                 if (!*p)
 432                         continue;
 433                 /*
 434                  * Initialize args struct so we know whether arg was
 435                  * found; some options take optional arguments.
 436                  */
 437                 args[0].to = args[0].from = NULL;
 438                 token = match_token(p, f2fs_tokens, args);
 439
 440                 switch (token) {
 441                 case Opt_gc_background:
 442                         name = match_strdup(&args[0]);
 443
 444                         if (!name)
 445                                 return -ENOMEM;
 446                         if (!strcmp(name, "on")) {
 447                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
 448                         } else if (!strcmp(name, "off")) {
 449                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
 450                         } else if (!strcmp(name, "sync")) {
 451                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
 452                         } else {
 453                                 kvfree(name);
 454                                 return -EINVAL;
 455                         }
 456                         kvfree(name);
 457                         break;
 458                 case Opt_disable_roll_forward:
 459                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 460                         break;
 461                 case Opt_norecovery:
 462                         /* this option mounts f2fs with ro */
 463                         set_opt(sbi, NORECOVERY);
 464                         if (!f2fs_readonly(sb))
 465                                 return -EINVAL;
 466                         break;
 467                 case Opt_discard:
 468                         set_opt(sbi, DISCARD);
 469                         break;
 470                 case Opt_nodiscard:
 471                         if (f2fs_sb_has_blkzoned(sbi)) {
 472                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 473                                 return -EINVAL;
 474                         }
 475                         clear_opt(sbi, DISCARD);
 476                         break;
 477                 case Opt_noheap:
 478                         set_opt(sbi, NOHEAP);
 479                         break;
 480                 case Opt_heap:
 481                         clear_opt(sbi, NOHEAP);
 482                         break;
 483 #ifdef CONFIG_F2FS_FS_XATTR
 484                 case Opt_user_xattr:
 485                         set_opt(sbi, XATTR_USER);
 486                         break;
 487                 case Opt_nouser_xattr:
 488                         clear_opt(sbi, XATTR_USER);
 489                         break;
 490                 case Opt_inline_xattr:
 491                         set_opt(sbi, INLINE_XATTR);
 492                         break;
 493                 case Opt_noinline_xattr:
 494                         clear_opt(sbi, INLINE_XATTR);
 495                         break;
 496                 case Opt_inline_xattr_size:
 497                         if (args->from && match_int(args, &arg))
 498                                 return -EINVAL;
 499                         set_opt(sbi, INLINE_XATTR_SIZE);
 500                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 501                         break;
 502 #else
 503                 case Opt_user_xattr:
 504                         f2fs_info(sbi, "user_xattr options not supported");
 505                         break;
 506                 case Opt_nouser_xattr:
 507                         f2fs_info(sbi, "nouser_xattr options not supported");
 508                         break;
 509                 case Opt_inline_xattr:
 510                         f2fs_info(sbi, "inline_xattr options not supported");
 511                         break;
 512                 case Opt_noinline_xattr:
 513                         f2fs_info(sbi, "noinline_xattr options not supported");
 514                         break;
 515 #endif
 516 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 517                 case Opt_acl:
 518                         set_opt(sbi, POSIX_ACL);
 519                         break;
 520                 case Opt_noacl:
 521                         clear_opt(sbi, POSIX_ACL);
 522                         break;
 523 #else
 524                 case Opt_acl:
 525                         f2fs_info(sbi, "acl options not supported");
 526                         break;
 527                 case Opt_noacl:
 528                         f2fs_info(sbi, "noacl options not supported");
 529                         break;
 530 #endif
 531                 case Opt_active_logs:
 532                         if (args->from && match_int(args, &arg))
 533                                 return -EINVAL;
 534                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 535                                 return -EINVAL;
 536                         F2FS_OPTION(sbi).active_logs = arg;
 537                         break;
 538                 case Opt_disable_ext_identify:
 539                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 540                         break;
 541                 case Opt_inline_data:
 542                         set_opt(sbi, INLINE_DATA);
 543                         break;
 544                 case Opt_inline_dentry:
 545                         set_opt(sbi, INLINE_DENTRY);
 546                         break;
 547                 case Opt_noinline_dentry:
 548                         clear_opt(sbi, INLINE_DENTRY);
 549                         break;
 550                 case Opt_flush_merge:
 551                         set_opt(sbi, FLUSH_MERGE);
 552                         break;
 553                 case Opt_noflush_merge:
 554                         clear_opt(sbi, FLUSH_MERGE);
 555                         break;
 556                 case Opt_nobarrier:
 557                         set_opt(sbi, NOBARRIER);
 558                         break;
 559                 case Opt_fastboot:
 560                         set_opt(sbi, FASTBOOT);
 561                         break;
 562                 case Opt_extent_cache:
 563                         set_opt(sbi, EXTENT_CACHE);
 564                         break;
 565                 case Opt_noextent_cache:
 566                         clear_opt(sbi, EXTENT_CACHE);
 567                         break;
 568                 case Opt_noinline_data:
 569                         clear_opt(sbi, INLINE_DATA);
 570                         break;
 571                 case Opt_data_flush:
 572                         set_opt(sbi, DATA_FLUSH);
 573                         break;
 574                 case Opt_reserve_root:
 575                         if (args->from && match_int(args, &arg))
 576                                 return -EINVAL;
 577                         if (test_opt(sbi, RESERVE_ROOT)) {
 578                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 579                                           F2FS_OPTION(sbi).root_reserved_blocks);
 580                         } else {
 581                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 582                                 set_opt(sbi, RESERVE_ROOT);
 583                         }
 584                         break;
 585                 case Opt_resuid:
 586                         if (args->from && match_int(args, &arg))
 587                                 return -EINVAL;
 588                         uid = make_kuid(current_user_ns(), arg);
 589                         if (!uid_valid(uid)) {
 590                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 591                                 return -EINVAL;
 592                         }
 593                         F2FS_OPTION(sbi).s_resuid = uid;
 594                         break;
 595                 case Opt_resgid:
 596                         if (args->from && match_int(args, &arg))
 597                                 return -EINVAL;
 598                         gid = make_kgid(current_user_ns(), arg);
 599                         if (!gid_valid(gid)) {
 600                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 601                                 return -EINVAL;
 602                         }
 603                         F2FS_OPTION(sbi).s_resgid = gid;
 604                         break;
 605                 case Opt_mode:
 606                         name = match_strdup(&args[0]);
 607
 608                         if (!name)
 609                                 return -ENOMEM;
 610                         if (!strcmp(name, "adaptive")) {
 611                                 if (f2fs_sb_has_blkzoned(sbi)) {
 612                                         f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
 613                                         kvfree(name);
 614                                         return -EINVAL;
 615                                 }
 616                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
 617                         } else if (!strcmp(name, "lfs")) {
 618                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
 619                         } else {
 620                                 kvfree(name);
 621                                 return -EINVAL;
 622                         }
 623                         kvfree(name);
 624                         break;
 625                 case Opt_io_size_bits:
 626                         if (args->from && match_int(args, &arg))
 627                                 return -EINVAL;
 628                         if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
 629                                 f2fs_warn(sbi, "Not support %d, larger than %d",
 630                                           1 << arg, BIO_MAX_PAGES);
 631                                 return -EINVAL;
 632                         }
 633                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 634                         break;
 635 #ifdef CONFIG_F2FS_FAULT_INJECTION
 636                 case Opt_fault_injection:
 637                         if (args->from && match_int(args, &arg))
 638                                 return -EINVAL;
 639                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 640                         set_opt(sbi, FAULT_INJECTION);
 641                         break;
 642
 643                 case Opt_fault_type:
 644                         if (args->from && match_int(args, &arg))
 645                                 return -EINVAL;
 646                         f2fs_build_fault_attr(sbi, 0, arg);
 647                         set_opt(sbi, FAULT_INJECTION);
 648                         break;
 649 #else
 650                 case Opt_fault_injection:
 651                         f2fs_info(sbi, "fault_injection options not supported");
 652                         break;
 653
 654                 case Opt_fault_type:
 655                         f2fs_info(sbi, "fault_type options not supported");
 656                         break;
 657 #endif
 658                 case Opt_lazytime:
 659                         sb->s_flags |= SB_LAZYTIME;
 660                         break;
 661                 case Opt_nolazytime:
 662                         sb->s_flags &= ~SB_LAZYTIME;
 663                         break;
 664 #ifdef CONFIG_QUOTA
 665                 case Opt_quota:
 666                 case Opt_usrquota:
 667                         set_opt(sbi, USRQUOTA);
 668                         break;
 669                 case Opt_grpquota:
 670                         set_opt(sbi, GRPQUOTA);
 671                         break;
 672                 case Opt_prjquota:
 673                         set_opt(sbi, PRJQUOTA);
 674                         break;
 675                 case Opt_usrjquota:
 676                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 677                         if (ret)
 678                                 return ret;
 679                         break;
 680                 case Opt_grpjquota:
 681                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 682                         if (ret)
 683                                 return ret;
 684                         break;
 685                 case Opt_prjjquota:
 686                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 687                         if (ret)
 688                                 return ret;
 689                         break;
 690                 case Opt_offusrjquota:
 691                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 692                         if (ret)
 693                                 return ret;
 694                         break;
 695                 case Opt_offgrpjquota:
 696                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 697                         if (ret)
 698                                 return ret;
 699                         break;
 700                 case Opt_offprjjquota:
 701                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 702                         if (ret)
 703                                 return ret;
 704                         break;
 705                 case Opt_jqfmt_vfsold:
 706                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 707                         break;
 708                 case Opt_jqfmt_vfsv0:
 709                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 710                         break;
 711                 case Opt_jqfmt_vfsv1:
 712                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 713                         break;
 714                 case Opt_noquota:
 715                         clear_opt(sbi, QUOTA);
 716                         clear_opt(sbi, USRQUOTA);
 717                         clear_opt(sbi, GRPQUOTA);
 718                         clear_opt(sbi, PRJQUOTA);
 719                         break;
 720 #else
 721                 case Opt_quota:
 722                 case Opt_usrquota:
 723                 case Opt_grpquota:
 724                 case Opt_prjquota:
 725                 case Opt_usrjquota:
 726                 case Opt_grpjquota:
 727                 case Opt_prjjquota:
 728                 case Opt_offusrjquota:
 729                 case Opt_offgrpjquota:
 730                 case Opt_offprjjquota:
 731                 case Opt_jqfmt_vfsold:
 732                 case Opt_jqfmt_vfsv0:
 733                 case Opt_jqfmt_vfsv1:
 734                 case Opt_noquota:
 735                         f2fs_info(sbi, "quota operations not supported");
 736                         break;
 737 #endif
 738                 case Opt_whint:
 739                         name = match_strdup(&args[0]);
 740                         if (!name)
 741                                 return -ENOMEM;
 742                         if (!strcmp(name, "user-based")) {
 743                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 744                         } else if (!strcmp(name, "off")) {
 745                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 746                         } else if (!strcmp(name, "fs-based")) {
 747                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 748                         } else {
 749                                 kvfree(name);
 750                                 return -EINVAL;
 751                         }
 752                         kvfree(name);
 753                         break;
 754                 case Opt_alloc:
 755                         name = match_strdup(&args[0]);
 756                         if (!name)
 757                                 return -ENOMEM;
 758
 759                         if (!strcmp(name, "default")) {
 760                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 761                         } else if (!strcmp(name, "reuse")) {
 762                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 763                         } else {
 764                                 kvfree(name);
 765                                 return -EINVAL;
 766                         }
 767                         kvfree(name);
 768                         break;
 769                 case Opt_fsync:
 770                         name = match_strdup(&args[0]);
 771                         if (!name)
 772                                 return -ENOMEM;
 773                         if (!strcmp(name, "posix")) {
 774                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 775                         } else if (!strcmp(name, "strict")) {
 776                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 777                         } else if (!strcmp(name, "nobarrier")) {
 778                                 F2FS_OPTION(sbi).fsync_mode =
 779                                                         FSYNC_MODE_NOBARRIER;
 780                         } else {
 781                                 kvfree(name);
 782                                 return -EINVAL;
 783                         }
 784                         kvfree(name);
 785                         break;
 786                 case Opt_test_dummy_encryption:
 787 #ifdef CONFIG_FS_ENCRYPTION
 788                         if (!f2fs_sb_has_encrypt(sbi)) {
 789                                 f2fs_err(sbi, "Encrypt feature is off");
 790                                 return -EINVAL;
 791                         }
 792
 793                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 794                         f2fs_info(sbi, "Test dummy encryption mode enabled");
 795 #else
 796                         f2fs_info(sbi, "Test dummy encryption mount option ignored");
 797 #endif
 798                         break;
 799                 case Opt_checkpoint_disable_cap_perc:
 800                         if (args->from && match_int(args, &arg))
 801                                 return -EINVAL;
 802                         if (arg < 0 || arg > 100)
 803                                 return -EINVAL;
 804                         F2FS_OPTION(sbi).unusable_cap_perc = arg;
 805                         set_opt(sbi, DISABLE_CHECKPOINT);
 806                         break;
 807                 case Opt_checkpoint_disable_cap:
 808                         if (args->from && match_int(args, &arg))
 809                                 return -EINVAL;
 810                         F2FS_OPTION(sbi).unusable_cap = arg;
 811                         set_opt(sbi, DISABLE_CHECKPOINT);
 812                         break;
 813                 case Opt_checkpoint_disable:
 814                         set_opt(sbi, DISABLE_CHECKPOINT);
 815                         break;
 816                 case Opt_checkpoint_enable:
 817                         clear_opt(sbi, DISABLE_CHECKPOINT);
 818                         break;
 819                 case Opt_compress_algorithm:
 820                         if (!f2fs_sb_has_compression(sbi)) {
 821                                 f2fs_err(sbi, "Compression feature if off");
 822                                 return -EINVAL;
 823                         }
 824                         name = match_strdup(&args[0]);
 825                         if (!name)
 826                                 return -ENOMEM;
 827                         if (!strcmp(name, "lzo")) {
 828                                 F2FS_OPTION(sbi).compress_algorithm =
 829                                                                 COMPRESS_LZO;
 830                         } else if (!strcmp(name, "lz4")) {
 831                                 F2FS_OPTION(sbi).compress_algorithm =
 832                                                                 COMPRESS_LZ4;
 833                         } else if (!strcmp(name, "zstd")) {
 834                                 F2FS_OPTION(sbi).compress_algorithm =
 835                                                                 COMPRESS_ZSTD;
 836                         } else if (!strcmp(name, "lzo-rle")) {
 837                                 F2FS_OPTION(sbi).compress_algorithm =
 838                                                                 COMPRESS_LZORLE;
 839                         } else {
 840                                 kfree(name);
 841                                 return -EINVAL;
 842                         }
 843                         kfree(name);
 844                         break;
 845                 case Opt_compress_log_size:
 846                         if (!f2fs_sb_has_compression(sbi)) {
 847                                 f2fs_err(sbi, "Compression feature is off");
 848                                 return -EINVAL;
 849                         }
 850                         if (args->from && match_int(args, &arg))
 851                                 return -EINVAL;
 852                         if (arg < MIN_COMPRESS_LOG_SIZE ||
 853                                 arg > MAX_COMPRESS_LOG_SIZE) {
 854                                 f2fs_err(sbi,
 855                                         "Compress cluster log size is out of range");
 856                                 return -EINVAL;
 857                         }
 858                         F2FS_OPTION(sbi).compress_log_size = arg;
 859                         break;
 860                 case Opt_compress_extension:
 861                         if (!f2fs_sb_has_compression(sbi)) {
 862                                 f2fs_err(sbi, "Compression feature is off");
 863                                 return -EINVAL;
 864                         }
 865                         name = match_strdup(&args[0]);
 866                         if (!name)
 867                                 return -ENOMEM;
 868
 869                         ext = F2FS_OPTION(sbi).extensions;
 870                         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
 871
 872                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
 873                                 ext_cnt >= COMPRESS_EXT_NUM) {
 874                                 f2fs_err(sbi,
 875                                         "invalid extension length/number");
 876                                 kfree(name);
 877                                 return -EINVAL;
 878                         }
 879
 880                         strcpy(ext[ext_cnt], name);
 881                         F2FS_OPTION(sbi).compress_ext_cnt++;
 882                         kfree(name);
 883                         break;
 884                 default:
 885                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
 886                                  p);
 887                         return -EINVAL;
 888                 }
 889         }
 890 #ifdef CONFIG_QUOTA
 891         if (f2fs_check_quota_options(sbi))
 892                 return -EINVAL;
 893 #else
 894         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
 895                 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 896                 return -EINVAL;
 897         }
 898         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
 899                 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 900                 return -EINVAL;
 901         }
 902 #endif
 903 #ifndef CONFIG_UNICODE
 904         if (f2fs_sb_has_casefold(sbi)) {
 905                 f2fs_err(sbi,
 906                         "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
 907                 return -EINVAL;
 908         }
 909 #endif
 910
 911         if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
 912                 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
 913                          F2FS_IO_SIZE_KB(sbi));
 914                 return -EINVAL;
 915         }
 916
 917         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 918                 int min_size, max_size;
 919
 920                 if (!f2fs_sb_has_extra_attr(sbi) ||
 921                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 922                         f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
 923                         return -EINVAL;
 924                 }
 925                 if (!test_opt(sbi, INLINE_XATTR)) {
 926                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
 927                         return -EINVAL;
 928                 }
 929
 930                 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
 931                 max_size = MAX_INLINE_XATTR_SIZE;
 932
 933                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
 934                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
 935                         f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
 936                                  min_size, max_size);
 937                         return -EINVAL;
 938                 }
 939         }
 940
 941         if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
 942                 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
 943                 return -EINVAL;
 944         }
 945
 946         /* Not pass down write hints if the number of active logs is lesser
 947          * than NR_CURSEG_TYPE.
 948          */
 949         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 950                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 951         return 0;
 952 }
 953
 954 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 955 {
 956         struct f2fs_inode_info *fi;
 957
 958         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 959         if (!fi)
 960                 return NULL;
 961
 962         init_once((void *) fi);
 963
 964         /* Initialize f2fs-specific inode info */
 965         atomic_set(&fi->dirty_pages, 0);
 966         init_rwsem(&fi->i_sem);
 967         spin_lock_init(&fi->i_size_lock);
 968         INIT_LIST_HEAD(&fi->dirty_list);
 969         INIT_LIST_HEAD(&fi->gdirty_list);
 970         INIT_LIST_HEAD(&fi->inmem_ilist);
 971         INIT_LIST_HEAD(&fi->inmem_pages);
 972         mutex_init(&fi->inmem_lock);
 973         init_rwsem(&fi->i_gc_rwsem[READ]);
 974         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 975         init_rwsem(&fi->i_mmap_sem);
 976         init_rwsem(&fi->i_xattr_sem);
 977
 978         /* Will be used by directory only */
 979         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 980
 981         return &fi->vfs_inode;
 982 }
 983
 984 static int f2fs_drop_inode(struct inode *inode)
 985 {
 986         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 987         int ret;
 988
 989         /*
 990          * during filesystem shutdown, if checkpoint is disabled,
 991          * drop useless meta/node dirty pages.
 992          */
 993         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
 994                 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 995                         inode->i_ino == F2FS_META_INO(sbi)) {
 996                         trace_f2fs_drop_inode(inode, 1);
 997                         return 1;
 998                 }
 999         }
1000
1001         /*
1002          * This is to avoid a deadlock condition like below.
1003          * writeback_single_inode(inode)
1004          *  - f2fs_write_data_page
1005          *    - f2fs_gc -> iput -> evict
1006          *       - inode_wait_for_writeback(inode)
1007          */
1008         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1009                 if (!inode->i_nlink && !is_bad_inode(inode)) {
1010                         /* to avoid evict_inode call simultaneously */
1011                         atomic_inc(&inode->i_count);
1012                         spin_unlock(&inode->i_lock);
1013
1014                         /* some remained atomic pages should discarded */
1015                         if (f2fs_is_atomic_file(inode))
1016                                 f2fs_drop_inmem_pages(inode);
1017
1018                         /* should remain fi->extent_tree for writepage */
1019                         f2fs_destroy_extent_node(inode);
1020
1021                         sb_start_intwrite(inode->i_sb);
1022                         f2fs_i_size_write(inode, 0);
1023
1024                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1025                                         inode, NULL, 0, DATA);
1026                         truncate_inode_pages_final(inode->i_mapping);
1027
1028                         if (F2FS_HAS_BLOCKS(inode))
1029                                 f2fs_truncate(inode);
1030
1031                         sb_end_intwrite(inode->i_sb);
1032
1033                         spin_lock(&inode->i_lock);
1034                         atomic_dec(&inode->i_count);
1035                 }
1036                 trace_f2fs_drop_inode(inode, 0);
1037                 return 0;
1038         }
1039         ret = generic_drop_inode(inode);
1040         if (!ret)
1041                 ret = fscrypt_drop_inode(inode);
1042         trace_f2fs_drop_inode(inode, ret);
1043         return ret;
1044 }
1045
1046 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1047 {
1048         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1049         int ret = 0;
1050
1051         spin_lock(&sbi->inode_lock[DIRTY_META]);
1052         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1053                 ret = 1;
1054         } else {
1055                 set_inode_flag(inode, FI_DIRTY_INODE);
1056                 stat_inc_dirty_inode(sbi, DIRTY_META);
1057         }
1058         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1059                 list_add_tail(&F2FS_I(inode)->gdirty_list,
1060                                 &sbi->inode_list[DIRTY_META]);
1061                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1062         }
1063         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1064         return ret;
1065 }
1066
1067 void f2fs_inode_synced(struct inode *inode)
1068 {
1069         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1070
1071         spin_lock(&sbi->inode_lock[DIRTY_META]);
1072         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1073                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1074                 return;
1075         }
1076         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1077                 list_del_init(&F2FS_I(inode)->gdirty_list);
1078                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1079         }
1080         clear_inode_flag(inode, FI_DIRTY_INODE);
1081         clear_inode_flag(inode, FI_AUTO_RECOVER);
1082         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1083         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1084 }
1085
1086 /*
1087  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1088  *
1089  * We should call set_dirty_inode to write the dirty inode through write_inode.
1090  */
1091 static void f2fs_dirty_inode(struct inode *inode, int flags)
1092 {
1093         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1094
1095         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1096                         inode->i_ino == F2FS_META_INO(sbi))
1097                 return;
1098
1099         if (flags == I_DIRTY_TIME)
1100                 return;
1101
1102         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1103                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1104
1105         f2fs_inode_dirtied(inode, false);
1106 }
1107
1108 static void f2fs_free_inode(struct inode *inode)
1109 {
1110         fscrypt_free_inode(inode);
1111         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1112 }
1113
1114 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1115 {
1116         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1117         percpu_counter_destroy(&sbi->total_valid_inode_count);
1118 }
1119
1120 static void destroy_device_list(struct f2fs_sb_info *sbi)
1121 {
1122         int i;
1123
1124         for (i = 0; i < sbi->s_ndevs; i++) {
1125                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1126 #ifdef CONFIG_BLK_DEV_ZONED
1127                 kvfree(FDEV(i).blkz_seq);
1128 #endif
1129         }
1130         kvfree(sbi->devs);
1131 }
1132
1133 static void f2fs_put_super(struct super_block *sb)
1134 {
1135         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1136         int i;
1137         bool dropped;
1138
1139         f2fs_quota_off_umount(sb);
1140
1141         /* prevent remaining shrinker jobs */
1142         mutex_lock(&sbi->umount_mutex);
1143
1144         /*
1145          * We don't need to do checkpoint when superblock is clean.
1146          * But, the previous checkpoint was not done by umount, it needs to do
1147          * clean checkpoint again.
1148          */
1149         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1150                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1151                 struct cp_control cpc = {
1152                         .reason = CP_UMOUNT,
1153                 };
1154                 f2fs_write_checkpoint(sbi, &cpc);
1155         }
1156
1157         /* be sure to wait for any on-going discard commands */
1158         dropped = f2fs_issue_discard_timeout(sbi);
1159
1160         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1161                                         !sbi->discard_blks && !dropped) {
1162                 struct cp_control cpc = {
1163                         .reason = CP_UMOUNT | CP_TRIMMED,
1164                 };
1165                 f2fs_write_checkpoint(sbi, &cpc);
1166         }
1167
1168         /*
1169          * normally superblock is clean, so we need to release this.
1170          * In addition, EIO will skip do checkpoint, we need this as well.
1171          */
1172         f2fs_release_ino_entry(sbi, true);
1173
1174         f2fs_leave_shrinker(sbi);
1175         mutex_unlock(&sbi->umount_mutex);
1176
1177         /* our cp_error case, we can wait for any writeback page */
1178         f2fs_flush_merged_writes(sbi);
1179
1180         f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1181
1182         f2fs_bug_on(sbi, sbi->fsync_node_num);
1183
1184         iput(sbi->node_inode);
1185         sbi->node_inode = NULL;
1186
1187         iput(sbi->meta_inode);
1188         sbi->meta_inode = NULL;
1189
1190         /*
1191          * iput() can update stat information, if f2fs_write_checkpoint()
1192          * above failed with error.
1193          */
1194         f2fs_destroy_stats(sbi);
1195
1196         /* destroy f2fs internal modules */
1197         f2fs_destroy_node_manager(sbi);
1198         f2fs_destroy_segment_manager(sbi);
1199
1200         f2fs_destroy_post_read_wq(sbi);
1201
1202         kvfree(sbi->ckpt);
1203
1204         f2fs_unregister_sysfs(sbi);
1205
1206         sb->s_fs_info = NULL;
1207         if (sbi->s_chksum_driver)
1208                 crypto_free_shash(sbi->s_chksum_driver);
1209         kvfree(sbi->raw_super);
1210
1211         destroy_device_list(sbi);
1212         f2fs_destroy_xattr_caches(sbi);
1213         mempool_destroy(sbi->write_io_dummy);
1214 #ifdef CONFIG_QUOTA
1215         for (i = 0; i < MAXQUOTAS; i++)
1216                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1217 #endif
1218         destroy_percpu_info(sbi);
1219         for (i = 0; i < NR_PAGE_TYPE; i++)
1220                 kvfree(sbi->write_io[i]);
1221 #ifdef CONFIG_UNICODE
1222         utf8_unload(sbi->s_encoding);
1223 #endif
1224         kvfree(sbi);
1225 }
1226
1227 int f2fs_sync_fs(struct super_block *sb, int sync)
1228 {
1229         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1230         int err = 0;
1231
1232         if (unlikely(f2fs_cp_error(sbi)))
1233                 return 0;
1234         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1235                 return 0;
1236
1237         trace_f2fs_sync_fs(sb, sync);
1238
1239         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1240                 return -EAGAIN;
1241
1242         if (sync) {
1243                 struct cp_control cpc;
1244
1245                 cpc.reason = __get_cp_reason(sbi);
1246
1247                 down_write(&sbi->gc_lock);
1248                 err = f2fs_write_checkpoint(sbi, &cpc);
1249                 up_write(&sbi->gc_lock);
1250         }
1251         f2fs_trace_ios(NULL, 1);
1252
1253         return err;
1254 }
1255
1256 static int f2fs_freeze(struct super_block *sb)
1257 {
1258         if (f2fs_readonly(sb))
1259                 return 0;
1260
1261         /* IO error happened before */
1262         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1263                 return -EIO;
1264
1265         /* must be clean, since sync_filesystem() was already called */
1266         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1267                 return -EINVAL;
1268         return 0;
1269 }
1270
1271 static int f2fs_unfreeze(struct super_block *sb)
1272 {
1273         return 0;
1274 }
1275
1276 #ifdef CONFIG_QUOTA
1277 static int f2fs_statfs_project(struct super_block *sb,
1278                                 kprojid_t projid, struct kstatfs *buf)
1279 {
1280         struct kqid qid;
1281         struct dquot *dquot;
1282         u64 limit;
1283         u64 curblock;
1284
1285         qid = make_kqid_projid(projid);
1286         dquot = dqget(sb, qid);
1287         if (IS_ERR(dquot))
1288                 return PTR_ERR(dquot);
1289         spin_lock(&dquot->dq_dqb_lock);
1290
1291         limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1292                                         dquot->dq_dqb.dqb_bhardlimit);
1293         if (limit)
1294                 limit >>= sb->s_blocksize_bits;
1295
1296         if (limit && buf->f_blocks > limit) {
1297                 curblock = (dquot->dq_dqb.dqb_curspace +
1298                             dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1299                 buf->f_blocks = limit;
1300                 buf->f_bfree = buf->f_bavail =
1301                         (buf->f_blocks > curblock) ?
1302                          (buf->f_blocks - curblock) : 0;
1303         }
1304
1305         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1306                                         dquot->dq_dqb.dqb_ihardlimit);
1307
1308         if (limit && buf->f_files > limit) {
1309                 buf->f_files = limit;
1310                 buf->f_ffree =
1311                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1312                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1313         }
1314
1315         spin_unlock(&dquot->dq_dqb_lock);
1316         dqput(dquot);
1317         return 0;
1318 }
1319 #endif
1320
1321 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1322 {
1323         struct super_block *sb = dentry->d_sb;
1324         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1325         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1326         block_t total_count, user_block_count, start_count;
1327         u64 avail_node_count;
1328
1329         total_count = le64_to_cpu(sbi->raw_super->block_count);
1330         user_block_count = sbi->user_block_count;
1331         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1332         buf->f_type = F2FS_SUPER_MAGIC;
1333         buf->f_bsize = sbi->blocksize;
1334
1335         buf->f_blocks = total_count - start_count;
1336         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1337                                                 sbi->current_reserved_blocks;
1338
1339         spin_lock(&sbi->stat_lock);
1340         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1341                 buf->f_bfree = 0;
1342         else
1343                 buf->f_bfree -= sbi->unusable_block_count;
1344         spin_unlock(&sbi->stat_lock);
1345
1346         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1347                 buf->f_bavail = buf->f_bfree -
1348                                 F2FS_OPTION(sbi).root_reserved_blocks;
1349         else
1350                 buf->f_bavail = 0;
1351
1352         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1353
1354         if (avail_node_count > user_block_count) {
1355                 buf->f_files = user_block_count;
1356                 buf->f_ffree = buf->f_bavail;
1357         } else {
1358                 buf->f_files = avail_node_count;
1359                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1360                                         buf->f_bavail);
1361         }
1362
1363         buf->f_namelen = F2FS_NAME_LEN;
1364         buf->f_fsid.val[0] = (u32)id;
1365         buf->f_fsid.val[1] = (u32)(id >> 32);
1366
1367 #ifdef CONFIG_QUOTA
1368         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1369                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1370                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1371         }
1372 #endif
1373         return 0;
1374 }
1375
1376 static inline void f2fs_show_quota_options(struct seq_file *seq,
1377                                            struct super_block *sb)
1378 {
1379 #ifdef CONFIG_QUOTA
1380         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1381
1382         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1383                 char *fmtname = "";
1384
1385                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1386                 case QFMT_VFS_OLD:
1387                         fmtname = "vfsold";
1388                         break;
1389                 case QFMT_VFS_V0:
1390                         fmtname = "vfsv0";
1391                         break;
1392                 case QFMT_VFS_V1:
1393                         fmtname = "vfsv1";
1394                         break;
1395                 }
1396                 seq_printf(seq, ",jqfmt=%s", fmtname);
1397         }
1398
1399         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1400                 seq_show_option(seq, "usrjquota",
1401                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1402
1403         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1404                 seq_show_option(seq, "grpjquota",
1405                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1406
1407         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1408                 seq_show_option(seq, "prjjquota",
1409                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1410 #endif
1411 }
1412
1413 static inline void f2fs_show_compress_options(struct seq_file *seq,
1414                                                         struct super_block *sb)
1415 {
1416         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1417         char *algtype = "";
1418         int i;
1419
1420         if (!f2fs_sb_has_compression(sbi))
1421                 return;
1422
1423         switch (F2FS_OPTION(sbi).compress_algorithm) {
1424         case COMPRESS_LZO:
1425                 algtype = "lzo";
1426                 break;
1427         case COMPRESS_LZ4:
1428                 algtype = "lz4";
1429                 break;
1430         case COMPRESS_ZSTD:
1431                 algtype = "zstd";
1432                 break;
1433         case COMPRESS_LZORLE:
1434                 algtype = "lzo-rle";
1435                 break;
1436         }
1437         seq_printf(seq, ",compress_algorithm=%s", algtype);
1438
1439         seq_printf(seq, ",compress_log_size=%u",
1440                         F2FS_OPTION(sbi).compress_log_size);
1441
1442         for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1443                 seq_printf(seq, ",compress_extension=%s",
1444                         F2FS_OPTION(sbi).extensions[i]);
1445         }
1446 }
1447
1448 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1449 {
1450         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1451
1452         if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1453                 seq_printf(seq, ",background_gc=%s", "sync");
1454         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1455                 seq_printf(seq, ",background_gc=%s", "on");
1456         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1457                 seq_printf(seq, ",background_gc=%s", "off");
1458
1459         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1460                 seq_puts(seq, ",disable_roll_forward");
1461         if (test_opt(sbi, NORECOVERY))
1462                 seq_puts(seq, ",norecovery");
1463         if (test_opt(sbi, DISCARD))
1464                 seq_puts(seq, ",discard");
1465         else
1466                 seq_puts(seq, ",nodiscard");
1467         if (test_opt(sbi, NOHEAP))
1468                 seq_puts(seq, ",no_heap");
1469         else
1470                 seq_puts(seq, ",heap");
1471 #ifdef CONFIG_F2FS_FS_XATTR
1472         if (test_opt(sbi, XATTR_USER))
1473                 seq_puts(seq, ",user_xattr");
1474         else
1475                 seq_puts(seq, ",nouser_xattr");
1476         if (test_opt(sbi, INLINE_XATTR))
1477                 seq_puts(seq, ",inline_xattr");
1478         else
1479                 seq_puts(seq, ",noinline_xattr");
1480         if (test_opt(sbi, INLINE_XATTR_SIZE))
1481                 seq_printf(seq, ",inline_xattr_size=%u",
1482                                         F2FS_OPTION(sbi).inline_xattr_size);
1483 #endif
1484 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1485         if (test_opt(sbi, POSIX_ACL))
1486                 seq_puts(seq, ",acl");
1487         else
1488                 seq_puts(seq, ",noacl");
1489 #endif
1490         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1491                 seq_puts(seq, ",disable_ext_identify");
1492         if (test_opt(sbi, INLINE_DATA))
1493                 seq_puts(seq, ",inline_data");
1494         else
1495                 seq_puts(seq, ",noinline_data");
1496         if (test_opt(sbi, INLINE_DENTRY))
1497                 seq_puts(seq, ",inline_dentry");
1498         else
1499                 seq_puts(seq, ",noinline_dentry");
1500         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1501                 seq_puts(seq, ",flush_merge");
1502         if (test_opt(sbi, NOBARRIER))
1503                 seq_puts(seq, ",nobarrier");
1504         if (test_opt(sbi, FASTBOOT))
1505                 seq_puts(seq, ",fastboot");
1506         if (test_opt(sbi, EXTENT_CACHE))
1507                 seq_puts(seq, ",extent_cache");
1508         else
1509                 seq_puts(seq, ",noextent_cache");
1510         if (test_opt(sbi, DATA_FLUSH))
1511                 seq_puts(seq, ",data_flush");
1512
1513         seq_puts(seq, ",mode=");
1514         if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1515                 seq_puts(seq, "adaptive");
1516         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1517                 seq_puts(seq, "lfs");
1518         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1519         if (test_opt(sbi, RESERVE_ROOT))
1520                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1521                                 F2FS_OPTION(sbi).root_reserved_blocks,
1522                                 from_kuid_munged(&init_user_ns,
1523                                         F2FS_OPTION(sbi).s_resuid),
1524                                 from_kgid_munged(&init_user_ns,
1525                                         F2FS_OPTION(sbi).s_resgid));
1526         if (F2FS_IO_SIZE_BITS(sbi))
1527                 seq_printf(seq, ",io_bits=%u",
1528                                 F2FS_OPTION(sbi).write_io_size_bits);
1529 #ifdef CONFIG_F2FS_FAULT_INJECTION
1530         if (test_opt(sbi, FAULT_INJECTION)) {
1531                 seq_printf(seq, ",fault_injection=%u",
1532                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1533                 seq_printf(seq, ",fault_type=%u",
1534                                 F2FS_OPTION(sbi).fault_info.inject_type);
1535         }
1536 #endif
1537 #ifdef CONFIG_QUOTA
1538         if (test_opt(sbi, QUOTA))
1539                 seq_puts(seq, ",quota");
1540         if (test_opt(sbi, USRQUOTA))
1541                 seq_puts(seq, ",usrquota");
1542         if (test_opt(sbi, GRPQUOTA))
1543                 seq_puts(seq, ",grpquota");
1544         if (test_opt(sbi, PRJQUOTA))
1545                 seq_puts(seq, ",prjquota");
1546 #endif
1547         f2fs_show_quota_options(seq, sbi->sb);
1548         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1549                 seq_printf(seq, ",whint_mode=%s", "user-based");
1550         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1551                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1552 #ifdef CONFIG_FS_ENCRYPTION
1553         if (F2FS_OPTION(sbi).test_dummy_encryption)
1554                 seq_puts(seq, ",test_dummy_encryption");
1555 #endif
1556
1557         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1558                 seq_printf(seq, ",alloc_mode=%s", "default");
1559         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1560                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1561
1562         if (test_opt(sbi, DISABLE_CHECKPOINT))
1563                 seq_printf(seq, ",checkpoint=disable:%u",
1564                                 F2FS_OPTION(sbi).unusable_cap);
1565         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1566                 seq_printf(seq, ",fsync_mode=%s", "posix");
1567         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1568                 seq_printf(seq, ",fsync_mode=%s", "strict");
1569         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1570                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1571
1572         f2fs_show_compress_options(seq, sbi->sb);
1573         return 0;
1574 }
1575
1576 static void default_options(struct f2fs_sb_info *sbi)
1577 {
1578         /* init some FS parameters */
1579         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1580         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1581         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1582         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1583         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1584         F2FS_OPTION(sbi).test_dummy_encryption = false;
1585         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1586         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1587         F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1588         F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1589         F2FS_OPTION(sbi).compress_ext_cnt = 0;
1590         F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1591
1592         set_opt(sbi, INLINE_XATTR);
1593         set_opt(sbi, INLINE_DATA);
1594         set_opt(sbi, INLINE_DENTRY);
1595         set_opt(sbi, EXTENT_CACHE);
1596         set_opt(sbi, NOHEAP);
1597         clear_opt(sbi, DISABLE_CHECKPOINT);
1598         F2FS_OPTION(sbi).unusable_cap = 0;
1599         sbi->sb->s_flags |= SB_LAZYTIME;
1600         set_opt(sbi, FLUSH_MERGE);
1601         set_opt(sbi, DISCARD);
1602         if (f2fs_sb_has_blkzoned(sbi))
1603                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1604         else
1605                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1606
1607 #ifdef CONFIG_F2FS_FS_XATTR
1608         set_opt(sbi, XATTR_USER);
1609 #endif
1610 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1611         set_opt(sbi, POSIX_ACL);
1612 #endif
1613
1614         f2fs_build_fault_attr(sbi, 0, 0);
1615 }
1616
1617 #ifdef CONFIG_QUOTA
1618 static int f2fs_enable_quotas(struct super_block *sb);
1619 #endif
1620
1621 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1622 {
1623         unsigned int s_flags = sbi->sb->s_flags;
1624         struct cp_control cpc;
1625         int err = 0;
1626         int ret;
1627         block_t unusable;
1628
1629         if (s_flags & SB_RDONLY) {
1630                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1631                 return -EINVAL;
1632         }
1633         sbi->sb->s_flags |= SB_ACTIVE;
1634
1635         f2fs_update_time(sbi, DISABLE_TIME);
1636
1637         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1638                 down_write(&sbi->gc_lock);
1639                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1640                 if (err == -ENODATA) {
1641                         err = 0;
1642                         break;
1643                 }
1644                 if (err && err != -EAGAIN)
1645                         break;
1646         }
1647
1648         ret = sync_filesystem(sbi->sb);
1649         if (ret || err) {
1650                 err = ret ? ret: err;
1651                 goto restore_flag;
1652         }
1653
1654         unusable = f2fs_get_unusable_blocks(sbi);
1655         if (f2fs_disable_cp_again(sbi, unusable)) {
1656                 err = -EAGAIN;
1657                 goto restore_flag;
1658         }
1659
1660         down_write(&sbi->gc_lock);
1661         cpc.reason = CP_PAUSE;
1662         set_sbi_flag(sbi, SBI_CP_DISABLED);
1663         err = f2fs_write_checkpoint(sbi, &cpc);
1664         if (err)
1665                 goto out_unlock;
1666
1667         spin_lock(&sbi->stat_lock);
1668         sbi->unusable_block_count = unusable;
1669         spin_unlock(&sbi->stat_lock);
1670
1671 out_unlock:
1672         up_write(&sbi->gc_lock);
1673 restore_flag:
1674         sbi->sb->s_flags = s_flags;     /* Restore SB_RDONLY status */
1675         return err;
1676 }
1677
1678 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1679 {
1680         down_write(&sbi->gc_lock);
1681         f2fs_dirty_to_prefree(sbi);
1682
1683         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1684         set_sbi_flag(sbi, SBI_IS_DIRTY);
1685         up_write(&sbi->gc_lock);
1686
1687         f2fs_sync_fs(sbi->sb, 1);
1688 }
1689
1690 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1691 {
1692         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1693         struct f2fs_mount_info org_mount_opt;
1694         unsigned long old_sb_flags;
1695         int err;
1696         bool need_restart_gc = false;
1697         bool need_stop_gc = false;
1698         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1699         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1700         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1701         bool checkpoint_changed;
1702 #ifdef CONFIG_QUOTA
1703         int i, j;
1704 #endif
1705
1706         /*
1707          * Save the old mount options in case we
1708          * need to restore them.
1709          */
1710         org_mount_opt = sbi->mount_opt;
1711         old_sb_flags = sb->s_flags;
1712
1713 #ifdef CONFIG_QUOTA
1714         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1715         for (i = 0; i < MAXQUOTAS; i++) {
1716                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1717                         org_mount_opt.s_qf_names[i] =
1718                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1719                                 GFP_KERNEL);
1720                         if (!org_mount_opt.s_qf_names[i]) {
1721                                 for (j = 0; j < i; j++)
1722                                         kvfree(org_mount_opt.s_qf_names[j]);
1723                                 return -ENOMEM;
1724                         }
1725                 } else {
1726                         org_mount_opt.s_qf_names[i] = NULL;
1727                 }
1728         }
1729 #endif
1730
1731         /* recover superblocks we couldn't write due to previous RO mount */
1732         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1733                 err = f2fs_commit_super(sbi, false);
1734                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1735                           err);
1736                 if (!err)
1737                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1738         }
1739
1740         default_options(sbi);
1741
1742         /* parse mount options */
1743         err = parse_options(sb, data);
1744         if (err)
1745                 goto restore_opts;
1746         checkpoint_changed =
1747                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1748
1749         /*
1750          * Previous and new state of filesystem is RO,
1751          * so skip checking GC and FLUSH_MERGE conditions.
1752          */
1753         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1754                 goto skip;
1755
1756 #ifdef CONFIG_QUOTA
1757         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1758                 err = dquot_suspend(sb, -1);
1759                 if (err < 0)
1760                         goto restore_opts;
1761         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1762                 /* dquot_resume needs RW */
1763                 sb->s_flags &= ~SB_RDONLY;
1764                 if (sb_any_quota_suspended(sb)) {
1765                         dquot_resume(sb, -1);
1766                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1767                         err = f2fs_enable_quotas(sb);
1768                         if (err)
1769                                 goto restore_opts;
1770                 }
1771         }
1772 #endif
1773         /* disallow enable/disable extent_cache dynamically */
1774         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1775                 err = -EINVAL;
1776                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1777                 goto restore_opts;
1778         }
1779
1780         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1781                 err = -EINVAL;
1782                 f2fs_warn(sbi, "switch io_bits option is not allowed");
1783                 goto restore_opts;
1784         }
1785
1786         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1787                 err = -EINVAL;
1788                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1789                 goto restore_opts;
1790         }
1791
1792         /*
1793          * We stop the GC thread if FS is mounted as RO
1794          * or if background_gc = off is passed in mount
1795          * option. Also sync the filesystem.
1796          */
1797         if ((*flags & SB_RDONLY) ||
1798                         F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) {
1799                 if (sbi->gc_thread) {
1800                         f2fs_stop_gc_thread(sbi);
1801                         need_restart_gc = true;
1802                 }
1803         } else if (!sbi->gc_thread) {
1804                 err = f2fs_start_gc_thread(sbi);
1805                 if (err)
1806                         goto restore_opts;
1807                 need_stop_gc = true;
1808         }
1809
1810         if (*flags & SB_RDONLY ||
1811                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1812                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1813                 sync_inodes_sb(sb);
1814
1815                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1816                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1817                 f2fs_sync_fs(sb, 1);
1818                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1819         }
1820
1821         if (checkpoint_changed) {
1822                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1823                         err = f2fs_disable_checkpoint(sbi);
1824                         if (err)
1825                                 goto restore_gc;
1826                 } else {
1827                         f2fs_enable_checkpoint(sbi);
1828                 }
1829         }
1830
1831         /*
1832          * We stop issue flush thread if FS is mounted as RO
1833          * or if flush_merge is not passed in mount option.
1834          */
1835         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1836                 clear_opt(sbi, FLUSH_MERGE);
1837                 f2fs_destroy_flush_cmd_control(sbi, false);
1838         } else {
1839                 err = f2fs_create_flush_cmd_control(sbi);
1840                 if (err)
1841                         goto restore_gc;
1842         }
1843 skip:
1844 #ifdef CONFIG_QUOTA
1845         /* Release old quota file names */
1846         for (i = 0; i < MAXQUOTAS; i++)
1847                 kvfree(org_mount_opt.s_qf_names[i]);
1848 #endif
1849         /* Update the POSIXACL Flag */
1850         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1851                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1852
1853         limit_reserve_root(sbi);
1854         adjust_unusable_cap_perc(sbi);
1855         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1856         return 0;
1857 restore_gc:
1858         if (need_restart_gc) {
1859                 if (f2fs_start_gc_thread(sbi))
1860                         f2fs_warn(sbi, "background gc thread has stopped");
1861         } else if (need_stop_gc) {
1862                 f2fs_stop_gc_thread(sbi);
1863         }
1864 restore_opts:
1865 #ifdef CONFIG_QUOTA
1866         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1867         for (i = 0; i < MAXQUOTAS; i++) {
1868                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1869                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1870         }
1871 #endif
1872         sbi->mount_opt = org_mount_opt;
1873         sb->s_flags = old_sb_flags;
1874         return err;
1875 }
1876
1877 #ifdef CONFIG_QUOTA
1878 /* Read data from quotafile */
1879 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1880                                size_t len, loff_t off)
1881 {
1882         struct inode *inode = sb_dqopt(sb)->files[type];
1883         struct address_space *mapping = inode->i_mapping;
1884         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1885         int offset = off & (sb->s_blocksize - 1);
1886         int tocopy;
1887         size_t toread;
1888         loff_t i_size = i_size_read(inode);
1889         struct page *page;
1890         char *kaddr;
1891
1892         if (off > i_size)
1893                 return 0;
1894
1895         if (off + len > i_size)
1896                 len = i_size - off;
1897         toread = len;
1898         while (toread > 0) {
1899                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1900 repeat:
1901                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1902                 if (IS_ERR(page)) {
1903                         if (PTR_ERR(page) == -ENOMEM) {
1904                                 congestion_wait(BLK_RW_ASYNC,
1905                                                 DEFAULT_IO_TIMEOUT);
1906                                 goto repeat;
1907                         }
1908                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1909                         return PTR_ERR(page);
1910                 }
1911
1912                 lock_page(page);
1913
1914                 if (unlikely(page->mapping != mapping)) {
1915                         f2fs_put_page(page, 1);
1916                         goto repeat;
1917                 }
1918                 if (unlikely(!PageUptodate(page))) {
1919                         f2fs_put_page(page, 1);
1920                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1921                         return -EIO;
1922                 }
1923
1924                 kaddr = kmap_atomic(page);
1925                 memcpy(data, kaddr + offset, tocopy);
1926                 kunmap_atomic(kaddr);
1927                 f2fs_put_page(page, 1);
1928
1929                 offset = 0;
1930                 toread -= tocopy;
1931                 data += tocopy;
1932                 blkidx++;
1933         }
1934         return len;
1935 }
1936
1937 /* Write to quotafile */
1938 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1939                                 const char *data, size_t len, loff_t off)
1940 {
1941         struct inode *inode = sb_dqopt(sb)->files[type];
1942         struct address_space *mapping = inode->i_mapping;
1943         const struct address_space_operations *a_ops = mapping->a_ops;
1944         int offset = off & (sb->s_blocksize - 1);
1945         size_t towrite = len;
1946         struct page *page;
1947         void *fsdata = NULL;
1948         char *kaddr;
1949         int err = 0;
1950         int tocopy;
1951
1952         while (towrite > 0) {
1953                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1954                                                                 towrite);
1955 retry:
1956                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1957                                                         &page, &fsdata);
1958                 if (unlikely(err)) {
1959                         if (err == -ENOMEM) {
1960                                 congestion_wait(BLK_RW_ASYNC,
1961                                                 DEFAULT_IO_TIMEOUT);
1962                                 goto retry;
1963                         }
1964                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1965                         break;
1966                 }
1967
1968                 kaddr = kmap_atomic(page);
1969                 memcpy(kaddr + offset, data, tocopy);
1970                 kunmap_atomic(kaddr);
1971                 flush_dcache_page(page);
1972
1973                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1974                                                 page, fsdata);
1975                 offset = 0;
1976                 towrite -= tocopy;
1977                 off += tocopy;
1978                 data += tocopy;
1979                 cond_resched();
1980         }
1981
1982         if (len == towrite)
1983                 return err;
1984         inode->i_mtime = inode->i_ctime = current_time(inode);
1985         f2fs_mark_inode_dirty_sync(inode, false);
1986         return len - towrite;
1987 }
1988
1989 static struct dquot **f2fs_get_dquots(struct inode *inode)
1990 {
1991         return F2FS_I(inode)->i_dquot;
1992 }
1993
1994 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1995 {
1996         return &F2FS_I(inode)->i_reserved_quota;
1997 }
1998
1999 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2000 {
2001         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2002                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2003                 return 0;
2004         }
2005
2006         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2007                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
2008 }
2009
2010 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2011 {
2012         int enabled = 0;
2013         int i, err;
2014
2015         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2016                 err = f2fs_enable_quotas(sbi->sb);
2017                 if (err) {
2018                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2019                         return 0;
2020                 }
2021                 return 1;
2022         }
2023
2024         for (i = 0; i < MAXQUOTAS; i++) {
2025                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2026                         err = f2fs_quota_on_mount(sbi, i);
2027                         if (!err) {
2028                                 enabled = 1;
2029                                 continue;
2030                         }
2031                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2032                                  err, i);
2033                 }
2034         }
2035         return enabled;
2036 }
2037
2038 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2039                              unsigned int flags)
2040 {
2041         struct inode *qf_inode;
2042         unsigned long qf_inum;
2043         int err;
2044
2045         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2046
2047         qf_inum = f2fs_qf_ino(sb, type);
2048         if (!qf_inum)
2049                 return -EPERM;
2050
2051         qf_inode = f2fs_iget(sb, qf_inum);
2052         if (IS_ERR(qf_inode)) {
2053                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2054                 return PTR_ERR(qf_inode);
2055         }
2056
2057         /* Don't account quota for quota files to avoid recursion */
2058         qf_inode->i_flags |= S_NOQUOTA;
2059         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2060         iput(qf_inode);
2061         return err;
2062 }
2063
2064 static int f2fs_enable_quotas(struct super_block *sb)
2065 {
2066         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2067         int type, err = 0;
2068         unsigned long qf_inum;
2069         bool quota_mopt[MAXQUOTAS] = {
2070                 test_opt(sbi, USRQUOTA),
2071                 test_opt(sbi, GRPQUOTA),
2072                 test_opt(sbi, PRJQUOTA),
2073         };
2074
2075         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2076                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2077                 return 0;
2078         }
2079
2080         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2081
2082         for (type = 0; type < MAXQUOTAS; type++) {
2083                 qf_inum = f2fs_qf_ino(sb, type);
2084                 if (qf_inum) {
2085                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2086                                 DQUOT_USAGE_ENABLED |
2087                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2088                         if (err) {
2089                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2090                                          type, err);
2091                                 for (type--; type >= 0; type--)
2092                                         dquot_quota_off(sb, type);
2093                                 set_sbi_flag(F2FS_SB(sb),
2094                                                 SBI_QUOTA_NEED_REPAIR);
2095                                 return err;
2096                         }
2097                 }
2098         }
2099         return 0;
2100 }
2101
2102 int f2fs_quota_sync(struct super_block *sb, int type)
2103 {
2104         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2105         struct quota_info *dqopt = sb_dqopt(sb);
2106         int cnt;
2107         int ret;
2108
2109         /*
2110          * do_quotactl
2111          *  f2fs_quota_sync
2112          *  down_read(quota_sem)
2113          *  dquot_writeback_dquots()
2114          *  f2fs_dquot_commit
2115          *                            block_operation
2116          *                            down_read(quota_sem)
2117          */
2118         f2fs_lock_op(sbi);
2119
2120         down_read(&sbi->quota_sem);
2121         ret = dquot_writeback_dquots(sb, type);
2122         if (ret)
2123                 goto out;
2124
2125         /*
2126          * Now when everything is written we can discard the pagecache so
2127          * that userspace sees the changes.
2128          */
2129         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2130                 struct address_space *mapping;
2131
2132                 if (type != -1 && cnt != type)
2133                         continue;
2134                 if (!sb_has_quota_active(sb, cnt))
2135                         continue;
2136
2137                 mapping = dqopt->files[cnt]->i_mapping;
2138
2139                 ret = filemap_fdatawrite(mapping);
2140                 if (ret)
2141                         goto out;
2142
2143                 /* if we are using journalled quota */
2144                 if (is_journalled_quota(sbi))
2145                         continue;
2146
2147                 ret = filemap_fdatawait(mapping);
2148                 if (ret)
2149                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2150
2151                 inode_lock(dqopt->files[cnt]);
2152                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2153                 inode_unlock(dqopt->files[cnt]);
2154         }
2155 out:
2156         if (ret)
2157                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2158         up_read(&sbi->quota_sem);
2159         f2fs_unlock_op(sbi);
2160         return ret;
2161 }
2162
2163 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2164                                                         const struct path *path)
2165 {
2166         struct inode *inode;
2167         int err;
2168
2169         /* if quota sysfile exists, deny enabling quota with specific file */
2170         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2171                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2172                 return -EBUSY;
2173         }
2174
2175         err = f2fs_quota_sync(sb, type);
2176         if (err)
2177                 return err;
2178
2179         err = dquot_quota_on(sb, type, format_id, path);
2180         if (err)
2181                 return err;
2182
2183         inode = d_inode(path->dentry);
2184
2185         inode_lock(inode);
2186         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2187         f2fs_set_inode_flags(inode);
2188         inode_unlock(inode);
2189         f2fs_mark_inode_dirty_sync(inode, false);
2190
2191         return 0;
2192 }
2193
2194 static int __f2fs_quota_off(struct super_block *sb, int type)
2195 {
2196         struct inode *inode = sb_dqopt(sb)->files[type];
2197         int err;
2198
2199         if (!inode || !igrab(inode))
2200                 return dquot_quota_off(sb, type);
2201
2202         err = f2fs_quota_sync(sb, type);
2203         if (err)
2204                 goto out_put;
2205
2206         err = dquot_quota_off(sb, type);
2207         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2208                 goto out_put;
2209
2210         inode_lock(inode);
2211         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2212         f2fs_set_inode_flags(inode);
2213         inode_unlock(inode);
2214         f2fs_mark_inode_dirty_sync(inode, false);
2215 out_put:
2216         iput(inode);
2217         return err;
2218 }
2219
2220 static int f2fs_quota_off(struct super_block *sb, int type)
2221 {
2222         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2223         int err;
2224
2225         err = __f2fs_quota_off(sb, type);
2226
2227         /*
2228          * quotactl can shutdown journalled quota, result in inconsistence
2229          * between quota record and fs data by following updates, tag the
2230          * flag to let fsck be aware of it.
2231          */
2232         if (is_journalled_quota(sbi))
2233                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2234         return err;
2235 }
2236
2237 void f2fs_quota_off_umount(struct super_block *sb)
2238 {
2239         int type;
2240         int err;
2241
2242         for (type = 0; type < MAXQUOTAS; type++) {
2243                 err = __f2fs_quota_off(sb, type);
2244                 if (err) {
2245                         int ret = dquot_quota_off(sb, type);
2246
2247                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2248                                  type, err, ret);
2249                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2250                 }
2251         }
2252         /*
2253          * In case of checkpoint=disable, we must flush quota blocks.
2254          * This can cause NULL exception for node_inode in end_io, since
2255          * put_super already dropped it.
2256          */
2257         sync_filesystem(sb);
2258 }
2259
2260 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2261 {
2262         struct quota_info *dqopt = sb_dqopt(sb);
2263         int type;
2264
2265         for (type = 0; type < MAXQUOTAS; type++) {
2266                 if (!dqopt->files[type])
2267                         continue;
2268                 f2fs_inode_synced(dqopt->files[type]);
2269         }
2270 }
2271
2272 static int f2fs_dquot_commit(struct dquot *dquot)
2273 {
2274         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2275         int ret;
2276
2277         down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2278         ret = dquot_commit(dquot);
2279         if (ret < 0)
2280                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2281         up_read(&sbi->quota_sem);
2282         return ret;
2283 }
2284
2285 static int f2fs_dquot_acquire(struct dquot *dquot)
2286 {
2287         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2288         int ret;
2289
2290         down_read(&sbi->quota_sem);
2291         ret = dquot_acquire(dquot);
2292         if (ret < 0)
2293                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2294         up_read(&sbi->quota_sem);
2295         return ret;
2296 }
2297
2298 static int f2fs_dquot_release(struct dquot *dquot)
2299 {
2300         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2301         int ret = dquot_release(dquot);
2302
2303         if (ret < 0)
2304                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2305         return ret;
2306 }
2307
2308 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2309 {
2310         struct super_block *sb = dquot->dq_sb;
2311         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2312         int ret = dquot_mark_dquot_dirty(dquot);
2313
2314         /* if we are using journalled quota */
2315         if (is_journalled_quota(sbi))
2316                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2317
2318         return ret;
2319 }
2320
2321 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2322 {
2323         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2324         int ret = dquot_commit_info(sb, type);
2325
2326         if (ret < 0)
2327                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2328         return ret;
2329 }
2330
2331 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2332 {
2333         *projid = F2FS_I(inode)->i_projid;
2334         return 0;
2335 }
2336
2337 static const struct dquot_operations f2fs_quota_operations = {
2338         .get_reserved_space = f2fs_get_reserved_space,
2339         .write_dquot    = f2fs_dquot_commit,
2340         .acquire_dquot  = f2fs_dquot_acquire,
2341         .release_dquot  = f2fs_dquot_release,
2342         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2343         .write_info     = f2fs_dquot_commit_info,
2344         .alloc_dquot    = dquot_alloc,
2345         .destroy_dquot  = dquot_destroy,
2346         .get_projid     = f2fs_get_projid,
2347         .get_next_id    = dquot_get_next_id,
2348 };
2349
2350 static const struct quotactl_ops f2fs_quotactl_ops = {
2351         .quota_on       = f2fs_quota_on,
2352         .quota_off      = f2fs_quota_off,
2353         .quota_sync     = f2fs_quota_sync,
2354         .get_state      = dquot_get_state,
2355         .set_info       = dquot_set_dqinfo,
2356         .get_dqblk      = dquot_get_dqblk,
2357         .set_dqblk      = dquot_set_dqblk,
2358         .get_nextdqblk  = dquot_get_next_dqblk,
2359 };
2360 #else
2361 int f2fs_quota_sync(struct super_block *sb, int type)
2362 {
2363         return 0;
2364 }
2365
2366 void f2fs_quota_off_umount(struct super_block *sb)
2367 {
2368 }
2369 #endif
2370
2371 static const struct super_operations f2fs_sops = {
2372         .alloc_inode    = f2fs_alloc_inode,
2373         .free_inode     = f2fs_free_inode,
2374         .drop_inode     = f2fs_drop_inode,
2375         .write_inode    = f2fs_write_inode,
2376         .dirty_inode    = f2fs_dirty_inode,
2377         .show_options   = f2fs_show_options,
2378 #ifdef CONFIG_QUOTA
2379         .quota_read     = f2fs_quota_read,
2380         .quota_write    = f2fs_quota_write,
2381         .get_dquots     = f2fs_get_dquots,
2382 #endif
2383         .evict_inode    = f2fs_evict_inode,
2384         .put_super      = f2fs_put_super,
2385         .sync_fs        = f2fs_sync_fs,
2386         .freeze_fs      = f2fs_freeze,
2387         .unfreeze_fs    = f2fs_unfreeze,
2388         .statfs         = f2fs_statfs,
2389         .remount_fs     = f2fs_remount,
2390 };
2391
2392 #ifdef CONFIG_FS_ENCRYPTION
2393 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2394 {
2395         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2396                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2397                                 ctx, len, NULL);
2398 }
2399
2400 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2401                                                         void *fs_data)
2402 {
2403         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2404
2405         /*
2406          * Encrypting the root directory is not allowed because fsck
2407          * expects lost+found directory to exist and remain unencrypted
2408          * if LOST_FOUND feature is enabled.
2409          *
2410          */
2411         if (f2fs_sb_has_lost_found(sbi) &&
2412                         inode->i_ino == F2FS_ROOT_INO(sbi))
2413                 return -EPERM;
2414
2415         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2416                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2417                                 ctx, len, fs_data, XATTR_CREATE);
2418 }
2419
2420 static bool f2fs_dummy_context(struct inode *inode)
2421 {
2422         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2423 }
2424
2425 static bool f2fs_has_stable_inodes(struct super_block *sb)
2426 {
2427         return true;
2428 }
2429
2430 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2431                                        int *ino_bits_ret, int *lblk_bits_ret)
2432 {
2433         *ino_bits_ret = 8 * sizeof(nid_t);
2434         *lblk_bits_ret = 8 * sizeof(block_t);
2435 }
2436
2437 static const struct fscrypt_operations f2fs_cryptops = {
2438         .key_prefix             = "f2fs:",
2439         .get_context            = f2fs_get_context,
2440         .set_context            = f2fs_set_context,
2441         .dummy_context          = f2fs_dummy_context,
2442         .empty_dir              = f2fs_empty_dir,
2443         .max_namelen            = F2FS_NAME_LEN,
2444         .has_stable_inodes      = f2fs_has_stable_inodes,
2445         .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
2446 };
2447 #endif
2448
2449 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2450                 u64 ino, u32 generation)
2451 {
2452         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2453         struct inode *inode;
2454
2455         if (f2fs_check_nid_range(sbi, ino))
2456                 return ERR_PTR(-ESTALE);
2457
2458         /*
2459          * f2fs_iget isn't quite right if the inode is currently unallocated!
2460          * However f2fs_iget currently does appropriate checks to handle stale
2461          * inodes so everything is OK.
2462          */
2463         inode = f2fs_iget(sb, ino);
2464         if (IS_ERR(inode))
2465                 return ERR_CAST(inode);
2466         if (unlikely(generation && inode->i_generation != generation)) {
2467                 /* we didn't find the right inode.. */
2468                 iput(inode);
2469                 return ERR_PTR(-ESTALE);
2470         }
2471         return inode;
2472 }
2473
2474 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2475                 int fh_len, int fh_type)
2476 {
2477         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2478                                     f2fs_nfs_get_inode);
2479 }
2480
2481 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2482                 int fh_len, int fh_type)
2483 {
2484         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2485                                     f2fs_nfs_get_inode);
2486 }
2487
2488 static const struct export_operations f2fs_export_ops = {
2489         .fh_to_dentry = f2fs_fh_to_dentry,
2490         .fh_to_parent = f2fs_fh_to_parent,
2491         .get_parent = f2fs_get_parent,
2492 };
2493
2494 static loff_t max_file_blocks(void)
2495 {
2496         loff_t result = 0;
2497         loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2498
2499         /*
2500          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2501          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2502          * space in inode.i_addr, it will be more safe to reassign
2503          * result as zero.
2504          */
2505
2506         /* two direct node blocks */
2507         result += (leaf_count * 2);
2508
2509         /* two indirect node blocks */
2510         leaf_count *= NIDS_PER_BLOCK;
2511         result += (leaf_count * 2);
2512
2513         /* one double indirect node block */
2514         leaf_count *= NIDS_PER_BLOCK;
2515         result += leaf_count;
2516
2517         return result;
2518 }
2519
2520 static int __f2fs_commit_super(struct buffer_head *bh,
2521                         struct f2fs_super_block *super)
2522 {
2523         lock_buffer(bh);
2524         if (super)
2525                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2526         set_buffer_dirty(bh);
2527         unlock_buffer(bh);
2528
2529         /* it's rare case, we can do fua all the time */
2530         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2531 }
2532
2533 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2534                                         struct buffer_head *bh)
2535 {
2536         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2537                                         (bh->b_data + F2FS_SUPER_OFFSET);
2538         struct super_block *sb = sbi->sb;
2539         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2540         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2541         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2542         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2543         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2544         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2545         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2546         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2547         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2548         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2549         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2550         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2551         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2552         u64 main_end_blkaddr = main_blkaddr +
2553                                 (segment_count_main << log_blocks_per_seg);
2554         u64 seg_end_blkaddr = segment0_blkaddr +
2555                                 (segment_count << log_blocks_per_seg);
2556
2557         if (segment0_blkaddr != cp_blkaddr) {
2558                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2559                           segment0_blkaddr, cp_blkaddr);
2560                 return true;
2561         }
2562
2563         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2564                                                         sit_blkaddr) {
2565                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2566                           cp_blkaddr, sit_blkaddr,
2567                           segment_count_ckpt << log_blocks_per_seg);
2568                 return true;
2569         }
2570
2571         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2572                                                         nat_blkaddr) {
2573                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2574                           sit_blkaddr, nat_blkaddr,
2575                           segment_count_sit << log_blocks_per_seg);
2576                 return true;
2577         }
2578
2579         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2580                                                         ssa_blkaddr) {
2581                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2582                           nat_blkaddr, ssa_blkaddr,
2583                           segment_count_nat << log_blocks_per_seg);
2584                 return true;
2585         }
2586
2587         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2588                                                         main_blkaddr) {
2589                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2590                           ssa_blkaddr, main_blkaddr,
2591                           segment_count_ssa << log_blocks_per_seg);
2592                 return true;
2593         }
2594
2595         if (main_end_blkaddr > seg_end_blkaddr) {
2596                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2597                           main_blkaddr,
2598                           segment0_blkaddr +
2599                           (segment_count << log_blocks_per_seg),
2600                           segment_count_main << log_blocks_per_seg);
2601                 return true;
2602         } else if (main_end_blkaddr < seg_end_blkaddr) {
2603                 int err = 0;
2604                 char *res;
2605
2606                 /* fix in-memory information all the time */
2607                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2608                                 segment0_blkaddr) >> log_blocks_per_seg);
2609
2610                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2611                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2612                         res = "internally";
2613                 } else {
2614                         err = __f2fs_commit_super(bh, NULL);
2615                         res = err ? "failed" : "done";
2616                 }
2617                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2618                           res, main_blkaddr,
2619                           segment0_blkaddr +
2620                           (segment_count << log_blocks_per_seg),
2621                           segment_count_main << log_blocks_per_seg);
2622                 if (err)
2623                         return true;
2624         }
2625         return false;
2626 }
2627
2628 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2629                                 struct buffer_head *bh)
2630 {
2631         block_t segment_count, segs_per_sec, secs_per_zone;
2632         block_t total_sections, blocks_per_seg;
2633         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2634                                         (bh->b_data + F2FS_SUPER_OFFSET);
2635         unsigned int blocksize;
2636         size_t crc_offset = 0;
2637         __u32 crc = 0;
2638
2639         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2640                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2641                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2642                 return -EINVAL;
2643         }
2644
2645         /* Check checksum_offset and crc in superblock */
2646         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2647                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2648                 if (crc_offset !=
2649                         offsetof(struct f2fs_super_block, crc)) {
2650                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2651                                   crc_offset);
2652                         return -EFSCORRUPTED;
2653                 }
2654                 crc = le32_to_cpu(raw_super->crc);
2655                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2656                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2657                         return -EFSCORRUPTED;
2658                 }
2659         }
2660
2661         /* Currently, support only 4KB page cache size */
2662         if (F2FS_BLKSIZE != PAGE_SIZE) {
2663                 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2664                           PAGE_SIZE);
2665                 return -EFSCORRUPTED;
2666         }
2667
2668         /* Currently, support only 4KB block size */
2669         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2670         if (blocksize != F2FS_BLKSIZE) {
2671                 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2672                           blocksize);
2673                 return -EFSCORRUPTED;
2674         }
2675
2676         /* check log blocks per segment */
2677         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2678                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2679                           le32_to_cpu(raw_super->log_blocks_per_seg));
2680                 return -EFSCORRUPTED;
2681         }
2682
2683         /* Currently, support 512/1024/2048/4096 bytes sector size */
2684         if (le32_to_cpu(raw_super->log_sectorsize) >
2685                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2686                 le32_to_cpu(raw_super->log_sectorsize) <
2687                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2688                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2689                           le32_to_cpu(raw_super->log_sectorsize));
2690                 return -EFSCORRUPTED;
2691         }
2692         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2693                 le32_to_cpu(raw_super->log_sectorsize) !=
2694                         F2FS_MAX_LOG_SECTOR_SIZE) {
2695                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2696                           le32_to_cpu(raw_super->log_sectors_per_block),
2697                           le32_to_cpu(raw_super->log_sectorsize));
2698                 return -EFSCORRUPTED;
2699         }
2700
2701         segment_count = le32_to_cpu(raw_super->segment_count);
2702         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2703         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2704         total_sections = le32_to_cpu(raw_super->section_count);
2705
2706         /* blocks_per_seg should be 512, given the above check */
2707         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2708
2709         if (segment_count > F2FS_MAX_SEGMENT ||
2710                                 segment_count < F2FS_MIN_SEGMENTS) {
2711                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2712                 return -EFSCORRUPTED;
2713         }
2714
2715         if (total_sections > segment_count ||
2716                         total_sections < F2FS_MIN_SEGMENTS ||
2717                         segs_per_sec > segment_count || !segs_per_sec) {
2718                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2719                           segment_count, total_sections, segs_per_sec);
2720                 return -EFSCORRUPTED;
2721         }
2722
2723         if ((segment_count / segs_per_sec) < total_sections) {
2724                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2725                           segment_count, segs_per_sec, total_sections);
2726                 return -EFSCORRUPTED;
2727         }
2728
2729         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2730                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2731                           segment_count, le64_to_cpu(raw_super->block_count));
2732                 return -EFSCORRUPTED;
2733         }
2734
2735         if (RDEV(0).path[0]) {
2736                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2737                 int i = 1;
2738
2739                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2740                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2741                         i++;
2742                 }
2743                 if (segment_count != dev_seg_count) {
2744                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2745                                         segment_count, dev_seg_count);
2746                         return -EFSCORRUPTED;
2747                 }
2748         }
2749
2750         if (secs_per_zone > total_sections || !secs_per_zone) {
2751                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2752                           secs_per_zone, total_sections);
2753                 return -EFSCORRUPTED;
2754         }
2755         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2756                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2757                         (le32_to_cpu(raw_super->extension_count) +
2758                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2759                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2760                           le32_to_cpu(raw_super->extension_count),
2761                           raw_super->hot_ext_count,
2762                           F2FS_MAX_EXTENSION);
2763                 return -EFSCORRUPTED;
2764         }
2765
2766         if (le32_to_cpu(raw_super->cp_payload) >
2767                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2768                 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2769                           le32_to_cpu(raw_super->cp_payload),
2770                           blocks_per_seg - F2FS_CP_PACKS);
2771                 return -EFSCORRUPTED;
2772         }
2773
2774         /* check reserved ino info */
2775         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2776                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2777                 le32_to_cpu(raw_super->root_ino) != 3) {
2778                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2779                           le32_to_cpu(raw_super->node_ino),
2780                           le32_to_cpu(raw_super->meta_ino),
2781                           le32_to_cpu(raw_super->root_ino));
2782                 return -EFSCORRUPTED;
2783         }
2784
2785         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2786         if (sanity_check_area_boundary(sbi, bh))
2787                 return -EFSCORRUPTED;
2788
2789         return 0;
2790 }
2791
2792 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2793 {
2794         unsigned int total, fsmeta;
2795         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2796         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2797         unsigned int ovp_segments, reserved_segments;
2798         unsigned int main_segs, blocks_per_seg;
2799         unsigned int sit_segs, nat_segs;
2800         unsigned int sit_bitmap_size, nat_bitmap_size;
2801         unsigned int log_blocks_per_seg;
2802         unsigned int segment_count_main;
2803         unsigned int cp_pack_start_sum, cp_payload;
2804         block_t user_block_count, valid_user_blocks;
2805         block_t avail_node_count, valid_node_count;
2806         int i, j;
2807
2808         total = le32_to_cpu(raw_super->segment_count);
2809         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2810         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2811         fsmeta += sit_segs;
2812         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2813         fsmeta += nat_segs;
2814         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2815         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2816
2817         if (unlikely(fsmeta >= total))
2818                 return 1;
2819
2820         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2821         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2822
2823         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2824                         ovp_segments == 0 || reserved_segments == 0)) {
2825                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2826                 return 1;
2827         }
2828
2829         user_block_count = le64_to_cpu(ckpt->user_block_count);
2830         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2831         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2832         if (!user_block_count || user_block_count >=
2833                         segment_count_main << log_blocks_per_seg) {
2834                 f2fs_err(sbi, "Wrong user_block_count: %u",
2835                          user_block_count);
2836                 return 1;
2837         }
2838
2839         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2840         if (valid_user_blocks > user_block_count) {
2841                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2842                          valid_user_blocks, user_block_count);
2843                 return 1;
2844         }
2845
2846         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2847         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2848         if (valid_node_count > avail_node_count) {
2849                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2850                          valid_node_count, avail_node_count);
2851                 return 1;
2852         }
2853
2854         main_segs = le32_to_cpu(raw_super->segment_count_main);
2855         blocks_per_seg = sbi->blocks_per_seg;
2856
2857         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2858                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2859                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2860                         return 1;
2861                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2862                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2863                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2864                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2865                                          i, j,
2866                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2867                                 return 1;
2868                         }
2869                 }
2870         }
2871         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2872                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2873                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2874                         return 1;
2875                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2876                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2877                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2878                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2879                                          i, j,
2880                                          le32_to_cpu(ckpt->cur_data_segno[i]));
2881                                 return 1;
2882                         }
2883                 }
2884         }
2885         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2886                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2887                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2888                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2889                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2890                                          i, j,
2891                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2892                                 return 1;
2893                         }
2894                 }
2895         }
2896
2897         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2898         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2899
2900         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2901                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2902                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2903                          sit_bitmap_size, nat_bitmap_size);
2904                 return 1;
2905         }
2906
2907         cp_pack_start_sum = __start_sum_addr(sbi);
2908         cp_payload = __cp_payload(sbi);
2909         if (cp_pack_start_sum < cp_payload + 1 ||
2910                 cp_pack_start_sum > blocks_per_seg - 1 -
2911                         NR_CURSEG_TYPE) {
2912                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2913                          cp_pack_start_sum);
2914                 return 1;
2915         }
2916
2917         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2918                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2919                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2920                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2921                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2922                           le32_to_cpu(ckpt->checksum_offset));
2923                 return 1;
2924         }
2925
2926         if (unlikely(f2fs_cp_error(sbi))) {
2927                 f2fs_err(sbi, "A bug case: need to run fsck");
2928                 return 1;
2929         }
2930         return 0;
2931 }
2932
2933 static void init_sb_info(struct f2fs_sb_info *sbi)
2934 {
2935         struct f2fs_super_block *raw_super = sbi->raw_super;
2936         int i;
2937
2938         sbi->log_sectors_per_block =
2939                 le32_to_cpu(raw_super->log_sectors_per_block);
2940         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2941         sbi->blocksize = 1 << sbi->log_blocksize;
2942         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2943         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2944         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2945         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2946         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2947         sbi->total_node_count =
2948                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2949                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2950         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2951         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2952         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2953         sbi->cur_victim_sec = NULL_SECNO;
2954         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2955         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2956         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2957         sbi->migration_granularity = sbi->segs_per_sec;
2958
2959         sbi->dir_level = DEF_DIR_LEVEL;
2960         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2961         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2962         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2963         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2964         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2965         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2966                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2967         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2968
2969         for (i = 0; i < NR_COUNT_TYPE; i++)
2970                 atomic_set(&sbi->nr_pages[i], 0);
2971
2972         for (i = 0; i < META; i++)
2973                 atomic_set(&sbi->wb_sync_req[i], 0);
2974
2975         INIT_LIST_HEAD(&sbi->s_list);
2976         mutex_init(&sbi->umount_mutex);
2977         init_rwsem(&sbi->io_order_lock);
2978         spin_lock_init(&sbi->cp_lock);
2979
2980         sbi->dirty_device = 0;
2981         spin_lock_init(&sbi->dev_lock);
2982
2983         init_rwsem(&sbi->sb_lock);
2984         init_rwsem(&sbi->pin_sem);
2985 }
2986
2987 static int init_percpu_info(struct f2fs_sb_info *sbi)
2988 {
2989         int err;
2990
2991         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2992         if (err)
2993                 return err;
2994
2995         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2996                                                                 GFP_KERNEL);
2997         if (err)
2998                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2999
3000         return err;
3001 }
3002
3003 #ifdef CONFIG_BLK_DEV_ZONED
3004 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3005                                void *data)
3006 {
3007         struct f2fs_dev_info *dev = data;
3008
3009         if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
3010                 set_bit(idx, dev->blkz_seq);
3011         return 0;
3012 }
3013
3014 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3015 {
3016         struct block_device *bdev = FDEV(devi).bdev;
3017         sector_t nr_sectors = bdev->bd_part->nr_sects;
3018         int ret;
3019
3020         if (!f2fs_sb_has_blkzoned(sbi))
3021                 return 0;
3022
3023         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3024                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3025                 return -EINVAL;
3026         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3027         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3028                                 __ilog2_u32(sbi->blocks_per_blkz))
3029                 return -EINVAL;
3030         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3031         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3032                                         sbi->log_blocks_per_blkz;
3033         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3034                 FDEV(devi).nr_blkz++;
3035
3036         FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
3037                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
3038                                         * sizeof(unsigned long),
3039                                         GFP_KERNEL);
3040         if (!FDEV(devi).blkz_seq)
3041                 return -ENOMEM;
3042
3043         /* Get block zones type */
3044         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3045                                   &FDEV(devi));
3046         if (ret < 0)
3047                 return ret;
3048
3049         return 0;
3050 }
3051 #endif
3052
3053 /*
3054  * Read f2fs raw super block.
3055  * Because we have two copies of super block, so read both of them
3056  * to get the first valid one. If any one of them is broken, we pass
3057  * them recovery flag back to the caller.
3058  */
3059 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3060                         struct f2fs_super_block **raw_super,
3061                         int *valid_super_block, int *recovery)
3062 {
3063         struct super_block *sb = sbi->sb;
3064         int block;
3065         struct buffer_head *bh;
3066         struct f2fs_super_block *super;
3067         int err = 0;
3068
3069         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3070         if (!super)
3071                 return -ENOMEM;
3072
3073         for (block = 0; block < 2; block++) {
3074                 bh = sb_bread(sb, block);
3075                 if (!bh) {
3076                         f2fs_err(sbi, "Unable to read %dth superblock",
3077                                  block + 1);
3078                         err = -EIO;
3079                         *recovery = 1;
3080                         continue;
3081                 }
3082
3083                 /* sanity checking of raw super */
3084                 err = sanity_check_raw_super(sbi, bh);
3085                 if (err) {
3086                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3087                                  block + 1);
3088                         brelse(bh);
3089                         *recovery = 1;
3090                         continue;
3091                 }
3092
3093                 if (!*raw_super) {
3094                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3095                                                         sizeof(*super));
3096                         *valid_super_block = block;
3097                         *raw_super = super;
3098                 }
3099                 brelse(bh);
3100         }
3101
3102         /* No valid superblock */
3103         if (!*raw_super)
3104                 kvfree(super);
3105         else
3106                 err = 0;
3107
3108         return err;
3109 }
3110
3111 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3112 {
3113         struct buffer_head *bh;
3114         __u32 crc = 0;
3115         int err;
3116
3117         if ((recover && f2fs_readonly(sbi->sb)) ||
3118                                 bdev_read_only(sbi->sb->s_bdev)) {
3119                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3120                 return -EROFS;
3121         }
3122
3123         /* we should update superblock crc here */
3124         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3125                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3126                                 offsetof(struct f2fs_super_block, crc));
3127                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3128         }
3129
3130         /* write back-up superblock first */
3131         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3132         if (!bh)
3133                 return -EIO;
3134         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3135         brelse(bh);
3136
3137         /* if we are in recovery path, skip writing valid superblock */
3138         if (recover || err)
3139                 return err;
3140
3141         /* write current valid superblock */
3142         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3143         if (!bh)
3144                 return -EIO;
3145         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3146         brelse(bh);
3147         return err;
3148 }
3149
3150 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3151 {
3152         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3153         unsigned int max_devices = MAX_DEVICES;
3154         int i;
3155
3156         /* Initialize single device information */
3157         if (!RDEV(0).path[0]) {
3158                 if (!bdev_is_zoned(sbi->sb->s_bdev))
3159                         return 0;
3160                 max_devices = 1;
3161         }
3162
3163         /*
3164          * Initialize multiple devices information, or single
3165          * zoned block device information.
3166          */
3167         sbi->devs = f2fs_kzalloc(sbi,
3168                                  array_size(max_devices,
3169                                             sizeof(struct f2fs_dev_info)),
3170                                  GFP_KERNEL);
3171         if (!sbi->devs)
3172                 return -ENOMEM;
3173
3174         for (i = 0; i < max_devices; i++) {
3175
3176                 if (i > 0 && !RDEV(i).path[0])
3177                         break;
3178
3179                 if (max_devices == 1) {
3180                         /* Single zoned block device mount */
3181                         FDEV(0).bdev =
3182                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3183                                         sbi->sb->s_mode, sbi->sb->s_type);
3184                 } else {
3185                         /* Multi-device mount */
3186                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3187                         FDEV(i).total_segments =
3188                                 le32_to_cpu(RDEV(i).total_segments);
3189                         if (i == 0) {
3190                                 FDEV(i).start_blk = 0;
3191                                 FDEV(i).end_blk = FDEV(i).start_blk +
3192                                     (FDEV(i).total_segments <<
3193                                     sbi->log_blocks_per_seg) - 1 +
3194                                     le32_to_cpu(raw_super->segment0_blkaddr);
3195                         } else {
3196                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3197                                 FDEV(i).end_blk = FDEV(i).start_blk +
3198                                         (FDEV(i).total_segments <<
3199                                         sbi->log_blocks_per_seg) - 1;
3200                         }
3201                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3202                                         sbi->sb->s_mode, sbi->sb->s_type);
3203                 }
3204                 if (IS_ERR(FDEV(i).bdev))
3205                         return PTR_ERR(FDEV(i).bdev);
3206
3207                 /* to release errored devices */
3208                 sbi->s_ndevs = i + 1;
3209
3210 #ifdef CONFIG_BLK_DEV_ZONED
3211                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3212                                 !f2fs_sb_has_blkzoned(sbi)) {
3213                         f2fs_err(sbi, "Zoned block device feature not enabled\n");
3214                         return -EINVAL;
3215                 }
3216                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3217                         if (init_blkz_info(sbi, i)) {
3218                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3219                                 return -EINVAL;
3220                         }
3221                         if (max_devices == 1)
3222                                 break;
3223                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3224                                   i, FDEV(i).path,
3225                                   FDEV(i).total_segments,
3226                                   FDEV(i).start_blk, FDEV(i).end_blk,
3227                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3228                                   "Host-aware" : "Host-managed");
3229                         continue;
3230                 }
3231 #endif
3232                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3233                           i, FDEV(i).path,
3234                           FDEV(i).total_segments,
3235                           FDEV(i).start_blk, FDEV(i).end_blk);
3236         }
3237         f2fs_info(sbi,
3238                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3239         return 0;
3240 }
3241
3242 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3243 {
3244 #ifdef CONFIG_UNICODE
3245         if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3246                 const struct f2fs_sb_encodings *encoding_info;
3247                 struct unicode_map *encoding;
3248                 __u16 encoding_flags;
3249
3250                 if (f2fs_sb_has_encrypt(sbi)) {
3251                         f2fs_err(sbi,
3252                                 "Can't mount with encoding and encryption");
3253                         return -EINVAL;
3254                 }
3255
3256                 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3257                                           &encoding_flags)) {
3258                         f2fs_err(sbi,
3259                                  "Encoding requested by superblock is unknown");
3260                         return -EINVAL;
3261                 }
3262
3263                 encoding = utf8_load(encoding_info->version);
3264                 if (IS_ERR(encoding)) {
3265                         f2fs_err(sbi,
3266                                  "can't mount with superblock charset: %s-%s "
3267                                  "not supported by the kernel. flags: 0x%x.",
3268                                  encoding_info->name, encoding_info->version,
3269                                  encoding_flags);
3270                         return PTR_ERR(encoding);
3271                 }
3272                 f2fs_info(sbi, "Using encoding defined by superblock: "
3273                          "%s-%s with flags 0x%hx", encoding_info->name,
3274                          encoding_info->version?:"\b", encoding_flags);
3275
3276                 sbi->s_encoding = encoding;
3277                 sbi->s_encoding_flags = encoding_flags;
3278                 sbi->sb->s_d_op = &f2fs_dentry_ops;
3279         }
3280 #else
3281         if (f2fs_sb_has_casefold(sbi)) {
3282                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3283                 return -EINVAL;
3284         }
3285 #endif
3286         return 0;
3287 }
3288
3289 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3290 {
3291         struct f2fs_sm_info *sm_i = SM_I(sbi);
3292
3293         /* adjust parameters according to the volume size */
3294         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3295                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3296                 sm_i->dcc_info->discard_granularity = 1;
3297                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3298         }
3299
3300         sbi->readdir_ra = 1;
3301 }
3302
3303 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3304 {
3305         struct f2fs_sb_info *sbi;
3306         struct f2fs_super_block *raw_super;
3307         struct inode *root;
3308         int err;
3309         bool skip_recovery = false, need_fsck = false;
3310         char *options = NULL;
3311         int recovery, i, valid_super_block;
3312         struct curseg_info *seg_i;
3313         int retry_cnt = 1;
3314
3315 try_onemore:
3316         err = -EINVAL;
3317         raw_super = NULL;
3318         valid_super_block = -1;
3319         recovery = 0;
3320
3321         /* allocate memory for f2fs-specific super block info */
3322         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3323         if (!sbi)
3324                 return -ENOMEM;
3325
3326         sbi->sb = sb;
3327
3328         /* Load the checksum driver */
3329         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3330         if (IS_ERR(sbi->s_chksum_driver)) {
3331                 f2fs_err(sbi, "Cannot load crc32 driver.");
3332                 err = PTR_ERR(sbi->s_chksum_driver);
3333                 sbi->s_chksum_driver = NULL;
3334                 goto free_sbi;
3335         }
3336
3337         /* set a block size */
3338         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3339                 f2fs_err(sbi, "unable to set blocksize");
3340                 goto free_sbi;
3341         }
3342
3343         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3344                                                                 &recovery);
3345         if (err)
3346                 goto free_sbi;
3347
3348         sb->s_fs_info = sbi;
3349         sbi->raw_super = raw_super;
3350
3351         /* precompute checksum seed for metadata */
3352         if (f2fs_sb_has_inode_chksum(sbi))
3353                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3354                                                 sizeof(raw_super->uuid));
3355
3356         /*
3357          * The BLKZONED feature indicates that the drive was formatted with
3358          * zone alignment optimization. This is optional for host-aware
3359          * devices, but mandatory for host-managed zoned block devices.
3360          */
3361 #ifndef CONFIG_BLK_DEV_ZONED
3362         if (f2fs_sb_has_blkzoned(sbi)) {
3363                 f2fs_err(sbi, "Zoned block device support is not enabled");
3364                 err = -EOPNOTSUPP;
3365                 goto free_sb_buf;
3366         }
3367 #endif
3368         default_options(sbi);
3369         /* parse mount options */
3370         options = kstrdup((const char *)data, GFP_KERNEL);
3371         if (data && !options) {
3372                 err = -ENOMEM;
3373                 goto free_sb_buf;
3374         }
3375
3376         err = parse_options(sb, options);
3377         if (err)
3378                 goto free_options;
3379
3380         sbi->max_file_blocks = max_file_blocks();
3381         sb->s_maxbytes = sbi->max_file_blocks <<
3382                                 le32_to_cpu(raw_super->log_blocksize);
3383         sb->s_max_links = F2FS_LINK_MAX;
3384
3385         err = f2fs_setup_casefold(sbi);
3386         if (err)
3387                 goto free_options;
3388
3389 #ifdef CONFIG_QUOTA
3390         sb->dq_op = &f2fs_quota_operations;
3391         sb->s_qcop = &f2fs_quotactl_ops;
3392         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3393
3394         if (f2fs_sb_has_quota_ino(sbi)) {
3395                 for (i = 0; i < MAXQUOTAS; i++) {
3396                         if (f2fs_qf_ino(sbi->sb, i))
3397                                 sbi->nquota_files++;
3398                 }
3399         }
3400 #endif
3401
3402         sb->s_op = &f2fs_sops;
3403 #ifdef CONFIG_FS_ENCRYPTION
3404         sb->s_cop = &f2fs_cryptops;
3405 #endif
3406 #ifdef CONFIG_FS_VERITY
3407         sb->s_vop = &f2fs_verityops;
3408 #endif
3409         sb->s_xattr = f2fs_xattr_handlers;
3410         sb->s_export_op = &f2fs_export_ops;
3411         sb->s_magic = F2FS_SUPER_MAGIC;
3412         sb->s_time_gran = 1;
3413         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3414                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3415         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3416         sb->s_iflags |= SB_I_CGROUPWB;
3417
3418         /* init f2fs-specific super block info */
3419         sbi->valid_super_block = valid_super_block;
3420         init_rwsem(&sbi->gc_lock);
3421         mutex_init(&sbi->writepages);
3422         mutex_init(&sbi->cp_mutex);
3423         init_rwsem(&sbi->node_write);
3424         init_rwsem(&sbi->node_change);
3425
3426         /* disallow all the data/node/meta page writes */
3427         set_sbi_flag(sbi, SBI_POR_DOING);
3428         spin_lock_init(&sbi->stat_lock);
3429
3430         /* init iostat info */
3431         spin_lock_init(&sbi->iostat_lock);
3432         sbi->iostat_enable = false;
3433         sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3434
3435         for (i = 0; i < NR_PAGE_TYPE; i++) {
3436                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3437                 int j;
3438
3439                 sbi->write_io[i] =
3440                         f2fs_kmalloc(sbi,
3441                                      array_size(n,
3442                                                 sizeof(struct f2fs_bio_info)),
3443                                      GFP_KERNEL);
3444                 if (!sbi->write_io[i]) {
3445                         err = -ENOMEM;
3446                         goto free_bio_info;
3447                 }
3448
3449                 for (j = HOT; j < n; j++) {
3450                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3451                         sbi->write_io[i][j].sbi = sbi;
3452                         sbi->write_io[i][j].bio = NULL;
3453                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3454                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3455                         INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3456                         init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3457                 }
3458         }
3459
3460         init_rwsem(&sbi->cp_rwsem);
3461         init_rwsem(&sbi->quota_sem);
3462         init_waitqueue_head(&sbi->cp_wait);
3463         init_sb_info(sbi);
3464
3465         err = init_percpu_info(sbi);
3466         if (err)
3467                 goto free_bio_info;
3468
3469         if (F2FS_IO_ALIGNED(sbi)) {
3470                 sbi->write_io_dummy =
3471                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3472                 if (!sbi->write_io_dummy) {
3473                         err = -ENOMEM;
3474                         goto free_percpu;
3475                 }
3476         }
3477
3478         /* init per sbi slab cache */
3479         err = f2fs_init_xattr_caches(sbi);
3480         if (err)
3481                 goto free_io_dummy;
3482
3483         /* get an inode for meta space */
3484         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3485         if (IS_ERR(sbi->meta_inode)) {
3486                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3487                 err = PTR_ERR(sbi->meta_inode);
3488                 goto free_xattr_cache;
3489         }
3490
3491         err = f2fs_get_valid_checkpoint(sbi);
3492         if (err) {
3493                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3494                 goto free_meta_inode;
3495         }
3496
3497         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3498                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3499         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3500                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3501                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3502         }
3503
3504         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3505                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3506
3507         /* Initialize device list */
3508         err = f2fs_scan_devices(sbi);
3509         if (err) {
3510                 f2fs_err(sbi, "Failed to find devices");
3511                 goto free_devices;
3512         }
3513
3514         err = f2fs_init_post_read_wq(sbi);
3515         if (err) {
3516                 f2fs_err(sbi, "Failed to initialize post read workqueue");
3517                 goto free_devices;
3518         }
3519
3520         sbi->total_valid_node_count =
3521                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3522         percpu_counter_set(&sbi->total_valid_inode_count,
3523                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3524         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3525         sbi->total_valid_block_count =
3526                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3527         sbi->last_valid_block_count = sbi->total_valid_block_count;
3528         sbi->reserved_blocks = 0;
3529         sbi->current_reserved_blocks = 0;
3530         limit_reserve_root(sbi);
3531         adjust_unusable_cap_perc(sbi);
3532
3533         for (i = 0; i < NR_INODE_TYPE; i++) {
3534                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3535                 spin_lock_init(&sbi->inode_lock[i]);
3536         }
3537         mutex_init(&sbi->flush_lock);
3538
3539         f2fs_init_extent_cache_info(sbi);
3540
3541         f2fs_init_ino_entry_info(sbi);
3542
3543         f2fs_init_fsync_node_info(sbi);
3544
3545         /* setup f2fs internal modules */
3546         err = f2fs_build_segment_manager(sbi);
3547         if (err) {
3548                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3549                          err);
3550                 goto free_sm;
3551         }
3552         err = f2fs_build_node_manager(sbi);
3553         if (err) {
3554                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3555                          err);
3556                 goto free_nm;
3557         }
3558
3559         /* For write statistics */
3560         if (sb->s_bdev->bd_part)
3561                 sbi->sectors_written_start =
3562                         (u64)part_stat_read(sb->s_bdev->bd_part,
3563                                             sectors[STAT_WRITE]);
3564
3565         /* Read accumulated write IO statistics if exists */
3566         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3567         if (__exist_node_summaries(sbi))
3568                 sbi->kbytes_written =
3569                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3570
3571         f2fs_build_gc_manager(sbi);
3572
3573         err = f2fs_build_stats(sbi);
3574         if (err)
3575                 goto free_nm;
3576
3577         /* get an inode for node space */
3578         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3579         if (IS_ERR(sbi->node_inode)) {
3580                 f2fs_err(sbi, "Failed to read node inode");
3581                 err = PTR_ERR(sbi->node_inode);
3582                 goto free_stats;
3583         }
3584
3585         /* read root inode and dentry */
3586         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3587         if (IS_ERR(root)) {
3588                 f2fs_err(sbi, "Failed to read root inode");
3589                 err = PTR_ERR(root);
3590                 goto free_node_inode;
3591         }
3592         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3593                         !root->i_size || !root->i_nlink) {
3594                 iput(root);
3595                 err = -EINVAL;
3596                 goto free_node_inode;
3597         }
3598
3599         sb->s_root = d_make_root(root); /* allocate root dentry */
3600         if (!sb->s_root) {
3601                 err = -ENOMEM;
3602                 goto free_node_inode;
3603         }
3604
3605         err = f2fs_register_sysfs(sbi);
3606         if (err)
3607                 goto free_root_inode;
3608
3609 #ifdef CONFIG_QUOTA
3610         /* Enable quota usage during mount */
3611         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3612                 err = f2fs_enable_quotas(sb);
3613                 if (err)
3614                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3615         }
3616 #endif
3617         /* if there are any orphan inodes, free them */
3618         err = f2fs_recover_orphan_inodes(sbi);
3619         if (err)
3620                 goto free_meta;
3621
3622         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3623                 goto reset_checkpoint;
3624
3625         /* recover fsynced data */
3626         if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3627                         !test_opt(sbi, NORECOVERY)) {
3628                 /*
3629                  * mount should be failed, when device has readonly mode, and
3630                  * previous checkpoint was not done by clean system shutdown.
3631                  */
3632                 if (f2fs_hw_is_readonly(sbi)) {
3633                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3634                                 err = -EROFS;
3635                                 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3636                                 goto free_meta;
3637                         }
3638                         f2fs_info(sbi, "write access unavailable, skipping recovery");
3639                         goto reset_checkpoint;
3640                 }
3641
3642                 if (need_fsck)
3643                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3644
3645                 if (skip_recovery)
3646                         goto reset_checkpoint;
3647
3648                 err = f2fs_recover_fsync_data(sbi, false);
3649                 if (err < 0) {
3650                         if (err != -ENOMEM)
3651                                 skip_recovery = true;
3652                         need_fsck = true;
3653                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3654                                  err);
3655                         goto free_meta;
3656                 }
3657         } else {
3658                 err = f2fs_recover_fsync_data(sbi, true);
3659
3660                 if (!f2fs_readonly(sb) && err > 0) {
3661                         err = -EINVAL;
3662                         f2fs_err(sbi, "Need to recover fsync data");
3663                         goto free_meta;
3664                 }
3665         }
3666
3667         /*
3668          * If the f2fs is not readonly and fsync data recovery succeeds,
3669          * check zoned block devices' write pointer consistency.
3670          */
3671         if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3672                 err = f2fs_check_write_pointer(sbi);
3673                 if (err)
3674                         goto free_meta;
3675         }
3676
3677 reset_checkpoint:
3678         /* f2fs_recover_fsync_data() cleared this already */
3679         clear_sbi_flag(sbi, SBI_POR_DOING);
3680
3681         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3682                 err = f2fs_disable_checkpoint(sbi);
3683                 if (err)
3684                         goto sync_free_meta;
3685         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3686                 f2fs_enable_checkpoint(sbi);
3687         }
3688
3689         /*
3690          * If filesystem is not mounted as read-only then
3691          * do start the gc_thread.
3692          */
3693         if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) {
3694                 /* After POR, we can run background GC thread.*/
3695                 err = f2fs_start_gc_thread(sbi);
3696                 if (err)
3697                         goto sync_free_meta;
3698         }
3699         kvfree(options);
3700
3701         /* recover broken superblock */
3702         if (recovery) {
3703                 err = f2fs_commit_super(sbi, true);
3704                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3705                           sbi->valid_super_block ? 1 : 2, err);
3706         }
3707
3708         f2fs_join_shrinker(sbi);
3709
3710         f2fs_tuning_parameters(sbi);
3711
3712         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3713                     cur_cp_version(F2FS_CKPT(sbi)));
3714         f2fs_update_time(sbi, CP_TIME);
3715         f2fs_update_time(sbi, REQ_TIME);
3716         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3717         return 0;
3718
3719 sync_free_meta:
3720         /* safe to flush all the data */
3721         sync_filesystem(sbi->sb);
3722         retry_cnt = 0;
3723
3724 free_meta:
3725 #ifdef CONFIG_QUOTA
3726         f2fs_truncate_quota_inode_pages(sb);
3727         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3728                 f2fs_quota_off_umount(sbi->sb);
3729 #endif
3730         /*
3731          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3732          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3733          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3734          * falls into an infinite loop in f2fs_sync_meta_pages().
3735          */
3736         truncate_inode_pages_final(META_MAPPING(sbi));
3737         /* evict some inodes being cached by GC */
3738         evict_inodes(sb);
3739         f2fs_unregister_sysfs(sbi);
3740 free_root_inode:
3741         dput(sb->s_root);
3742         sb->s_root = NULL;
3743 free_node_inode:
3744         f2fs_release_ino_entry(sbi, true);
3745         truncate_inode_pages_final(NODE_MAPPING(sbi));
3746         iput(sbi->node_inode);
3747         sbi->node_inode = NULL;
3748 free_stats:
3749         f2fs_destroy_stats(sbi);
3750 free_nm:
3751         f2fs_destroy_node_manager(sbi);
3752 free_sm:
3753         f2fs_destroy_segment_manager(sbi);
3754         f2fs_destroy_post_read_wq(sbi);
3755 free_devices:
3756         destroy_device_list(sbi);
3757         kvfree(sbi->ckpt);
3758 free_meta_inode:
3759         make_bad_inode(sbi->meta_inode);
3760         iput(sbi->meta_inode);
3761         sbi->meta_inode = NULL;
3762 free_xattr_cache:
3763         f2fs_destroy_xattr_caches(sbi);
3764 free_io_dummy:
3765         mempool_destroy(sbi->write_io_dummy);
3766 free_percpu:
3767         destroy_percpu_info(sbi);
3768 free_bio_info:
3769         for (i = 0; i < NR_PAGE_TYPE; i++)
3770                 kvfree(sbi->write_io[i]);
3771
3772 #ifdef CONFIG_UNICODE
3773         utf8_unload(sbi->s_encoding);
3774 #endif
3775 free_options:
3776 #ifdef CONFIG_QUOTA
3777         for (i = 0; i < MAXQUOTAS; i++)
3778                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3779 #endif
3780         kvfree(options);
3781 free_sb_buf:
3782         kvfree(raw_super);
3783 free_sbi:
3784         if (sbi->s_chksum_driver)
3785                 crypto_free_shash(sbi->s_chksum_driver);
3786         kvfree(sbi);
3787
3788         /* give only one another chance */
3789         if (retry_cnt > 0 && skip_recovery) {
3790                 retry_cnt--;
3791                 shrink_dcache_sb(sb);
3792                 goto try_onemore;
3793         }
3794         return err;
3795 }
3796
3797 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3798                         const char *dev_name, void *data)
3799 {
3800         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3801 }
3802
3803 static void kill_f2fs_super(struct super_block *sb)
3804 {
3805         if (sb->s_root) {
3806                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3807
3808                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3809                 f2fs_stop_gc_thread(sbi);
3810                 f2fs_stop_discard_thread(sbi);
3811
3812                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3813                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3814                         struct cp_control cpc = {
3815                                 .reason = CP_UMOUNT,
3816                         };
3817                         f2fs_write_checkpoint(sbi, &cpc);
3818                 }
3819
3820                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3821                         sb->s_flags &= ~SB_RDONLY;
3822         }
3823         kill_block_super(sb);
3824 }
3825
3826 static struct file_system_type f2fs_fs_type = {
3827         .owner          = THIS_MODULE,
3828         .name           = "f2fs",
3829         .mount          = f2fs_mount,
3830         .kill_sb        = kill_f2fs_super,
3831         .fs_flags       = FS_REQUIRES_DEV,
3832 };
3833 MODULE_ALIAS_FS("f2fs");
3834
3835 static int __init init_inodecache(void)
3836 {
3837         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3838                         sizeof(struct f2fs_inode_info), 0,
3839                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3840         if (!f2fs_inode_cachep)
3841                 return -ENOMEM;
3842         return 0;
3843 }
3844
3845 static void destroy_inodecache(void)
3846 {
3847         /*
3848          * Make sure all delayed rcu free inodes are flushed before we
3849          * destroy cache.
3850          */
3851         rcu_barrier();
3852         kmem_cache_destroy(f2fs_inode_cachep);
3853 }
3854
3855 static int __init init_f2fs_fs(void)
3856 {
3857         int err;
3858
3859         if (PAGE_SIZE != F2FS_BLKSIZE) {
3860                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3861                                 PAGE_SIZE, F2FS_BLKSIZE);
3862                 return -EINVAL;
3863         }
3864
3865         f2fs_build_trace_ios();
3866
3867         err = init_inodecache();
3868         if (err)
3869                 goto fail;
3870         err = f2fs_create_node_manager_caches();
3871         if (err)
3872                 goto free_inodecache;
3873         err = f2fs_create_segment_manager_caches();
3874         if (err)
3875                 goto free_node_manager_caches;
3876         err = f2fs_create_checkpoint_caches();
3877         if (err)
3878                 goto free_segment_manager_caches;
3879         err = f2fs_create_extent_cache();
3880         if (err)
3881                 goto free_checkpoint_caches;
3882         err = f2fs_init_sysfs();
3883         if (err)
3884                 goto free_extent_cache;
3885         err = register_shrinker(&f2fs_shrinker_info);
3886         if (err)
3887                 goto free_sysfs;
3888         err = register_filesystem(&f2fs_fs_type);
3889         if (err)
3890                 goto free_shrinker;
3891         f2fs_create_root_stats();
3892         err = f2fs_init_post_read_processing();
3893         if (err)
3894                 goto free_root_stats;
3895         err = f2fs_init_bio_entry_cache();
3896         if (err)
3897                 goto free_post_read;
3898         err = f2fs_init_bioset();
3899         if (err)
3900                 goto free_bio_enrty_cache;
3901         err = f2fs_init_compress_mempool();
3902         if (err)
3903                 goto free_bioset;
3904         return 0;
3905 free_bioset:
3906         f2fs_destroy_bioset();
3907 free_bio_enrty_cache:
3908         f2fs_destroy_bio_entry_cache();
3909 free_post_read:
3910         f2fs_destroy_post_read_processing();
3911 free_root_stats:
3912         f2fs_destroy_root_stats();
3913         unregister_filesystem(&f2fs_fs_type);
3914 free_shrinker:
3915         unregister_shrinker(&f2fs_shrinker_info);
3916 free_sysfs:
3917         f2fs_exit_sysfs();
3918 free_extent_cache:
3919         f2fs_destroy_extent_cache();
3920 free_checkpoint_caches:
3921         f2fs_destroy_checkpoint_caches();
3922 free_segment_manager_caches:
3923         f2fs_destroy_segment_manager_caches();
3924 free_node_manager_caches:
3925         f2fs_destroy_node_manager_caches();
3926 free_inodecache:
3927         destroy_inodecache();
3928 fail:
3929         return err;
3930 }
3931
3932 static void __exit exit_f2fs_fs(void)
3933 {
3934         f2fs_destroy_compress_mempool();
3935         f2fs_destroy_bioset();
3936         f2fs_destroy_bio_entry_cache();
3937         f2fs_destroy_post_read_processing();
3938         f2fs_destroy_root_stats();
3939         unregister_filesystem(&f2fs_fs_type);
3940         unregister_shrinker(&f2fs_shrinker_info);
3941         f2fs_exit_sysfs();
3942         f2fs_destroy_extent_cache();
3943         f2fs_destroy_checkpoint_caches();
3944         f2fs_destroy_segment_manager_caches();
3945         f2fs_destroy_node_manager_caches();
3946         destroy_inodecache();
3947         f2fs_destroy_trace_ios();
3948 }
3949
3950 module_init(init_f2fs_fs)
3951 module_exit(exit_f2fs_fs)
3952
3953 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3954 MODULE_DESCRIPTION("Flash Friendly File System");
3955 MODULE_LICENSE("GPL");
3956