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
  28 #include "f2fs.h"
  29 #include "node.h"
  30 #include "segment.h"
  31 #include "xattr.h"
  32 #include "gc.h"
  33 #include "trace.h"
  34
  35 #define CREATE_TRACE_POINTS
  36 #include <trace/events/f2fs.h>
  37
  38 static struct kmem_cache *f2fs_inode_cachep;
  39
  40 #ifdef CONFIG_F2FS_FAULT_INJECTION
  41
  42 const char *f2fs_fault_name[FAULT_MAX] = {
  43         [FAULT_KMALLOC]         = "kmalloc",
  44         [FAULT_KVMALLOC]        = "kvmalloc",
  45         [FAULT_PAGE_ALLOC]      = "page alloc",
  46         [FAULT_PAGE_GET]        = "page get",
  47         [FAULT_ALLOC_BIO]       = "alloc bio",
  48         [FAULT_ALLOC_NID]       = "alloc nid",
  49         [FAULT_ORPHAN]          = "orphan",
  50         [FAULT_BLOCK]           = "no more block",
  51         [FAULT_DIR_DEPTH]       = "too big dir depth",
  52         [FAULT_EVICT_INODE]     = "evict_inode fail",
  53         [FAULT_TRUNCATE]        = "truncate fail",
  54         [FAULT_READ_IO]         = "read IO error",
  55         [FAULT_CHECKPOINT]      = "checkpoint error",
  56         [FAULT_DISCARD]         = "discard error",
  57         [FAULT_WRITE_IO]        = "write IO error",
  58 };
  59
  60 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  61                                                         unsigned int type)
  62 {
  63         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  64
  65         if (rate) {
  66                 atomic_set(&ffi->inject_ops, 0);
  67                 ffi->inject_rate = rate;
  68         }
  69
  70         if (type)
  71                 ffi->inject_type = type;
  72
  73         if (!rate && !type)
  74                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  75 }
  76 #endif
  77
  78 /* f2fs-wide shrinker description */
  79 static struct shrinker f2fs_shrinker_info = {
  80         .scan_objects = f2fs_shrink_scan,
  81         .count_objects = f2fs_shrink_count,
  82         .seeks = DEFAULT_SEEKS,
  83 };
  84
  85 enum {
  86         Opt_gc_background,
  87         Opt_disable_roll_forward,
  88         Opt_norecovery,
  89         Opt_discard,
  90         Opt_nodiscard,
  91         Opt_noheap,
  92         Opt_heap,
  93         Opt_user_xattr,
  94         Opt_nouser_xattr,
  95         Opt_acl,
  96         Opt_noacl,
  97         Opt_active_logs,
  98         Opt_disable_ext_identify,
  99         Opt_inline_xattr,
 100         Opt_noinline_xattr,
 101         Opt_inline_xattr_size,
 102         Opt_inline_data,
 103         Opt_inline_dentry,
 104         Opt_noinline_dentry,
 105         Opt_flush_merge,
 106         Opt_noflush_merge,
 107         Opt_nobarrier,
 108         Opt_fastboot,
 109         Opt_extent_cache,
 110         Opt_noextent_cache,
 111         Opt_noinline_data,
 112         Opt_data_flush,
 113         Opt_reserve_root,
 114         Opt_resgid,
 115         Opt_resuid,
 116         Opt_mode,
 117         Opt_io_size_bits,
 118         Opt_fault_injection,
 119         Opt_fault_type,
 120         Opt_lazytime,
 121         Opt_nolazytime,
 122         Opt_quota,
 123         Opt_noquota,
 124         Opt_usrquota,
 125         Opt_grpquota,
 126         Opt_prjquota,
 127         Opt_usrjquota,
 128         Opt_grpjquota,
 129         Opt_prjjquota,
 130         Opt_offusrjquota,
 131         Opt_offgrpjquota,
 132         Opt_offprjjquota,
 133         Opt_jqfmt_vfsold,
 134         Opt_jqfmt_vfsv0,
 135         Opt_jqfmt_vfsv1,
 136         Opt_whint,
 137         Opt_alloc,
 138         Opt_fsync,
 139         Opt_test_dummy_encryption,
 140         Opt_checkpoint_disable,
 141         Opt_checkpoint_disable_cap,
 142         Opt_checkpoint_disable_cap_perc,
 143         Opt_checkpoint_enable,
 144         Opt_compress_algorithm,
 145         Opt_compress_log_size,
 146         Opt_compress_extension,
 147         Opt_err,
 148 };
 149
 150 static match_table_t f2fs_tokens = {
 151         {Opt_gc_background, "background_gc=%s"},
 152         {Opt_disable_roll_forward, "disable_roll_forward"},
 153         {Opt_norecovery, "norecovery"},
 154         {Opt_discard, "discard"},
 155         {Opt_nodiscard, "nodiscard"},
 156         {Opt_noheap, "no_heap"},
 157         {Opt_heap, "heap"},
 158         {Opt_user_xattr, "user_xattr"},
 159         {Opt_nouser_xattr, "nouser_xattr"},
 160         {Opt_acl, "acl"},
 161         {Opt_noacl, "noacl"},
 162         {Opt_active_logs, "active_logs=%u"},
 163         {Opt_disable_ext_identify, "disable_ext_identify"},
 164         {Opt_inline_xattr, "inline_xattr"},
 165         {Opt_noinline_xattr, "noinline_xattr"},
 166         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 167         {Opt_inline_data, "inline_data"},
 168         {Opt_inline_dentry, "inline_dentry"},
 169         {Opt_noinline_dentry, "noinline_dentry"},
 170         {Opt_flush_merge, "flush_merge"},
 171         {Opt_noflush_merge, "noflush_merge"},
 172         {Opt_nobarrier, "nobarrier"},
 173         {Opt_fastboot, "fastboot"},
 174         {Opt_extent_cache, "extent_cache"},
 175         {Opt_noextent_cache, "noextent_cache"},
 176         {Opt_noinline_data, "noinline_data"},
 177         {Opt_data_flush, "data_flush"},
 178         {Opt_reserve_root, "reserve_root=%u"},
 179         {Opt_resgid, "resgid=%u"},
 180         {Opt_resuid, "resuid=%u"},
 181         {Opt_mode, "mode=%s"},
 182         {Opt_io_size_bits, "io_bits=%u"},
 183         {Opt_fault_injection, "fault_injection=%u"},
 184         {Opt_fault_type, "fault_type=%u"},
 185         {Opt_lazytime, "lazytime"},
 186         {Opt_nolazytime, "nolazytime"},
 187         {Opt_quota, "quota"},
 188         {Opt_noquota, "noquota"},
 189         {Opt_usrquota, "usrquota"},
 190         {Opt_grpquota, "grpquota"},
 191         {Opt_prjquota, "prjquota"},
 192         {Opt_usrjquota, "usrjquota=%s"},
 193         {Opt_grpjquota, "grpjquota=%s"},
 194         {Opt_prjjquota, "prjjquota=%s"},
 195         {Opt_offusrjquota, "usrjquota="},
 196         {Opt_offgrpjquota, "grpjquota="},
 197         {Opt_offprjjquota, "prjjquota="},
 198         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 199         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 200         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 201         {Opt_whint, "whint_mode=%s"},
 202         {Opt_alloc, "alloc_mode=%s"},
 203         {Opt_fsync, "fsync_mode=%s"},
 204         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 205         {Opt_checkpoint_disable, "checkpoint=disable"},
 206         {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 207         {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 208         {Opt_checkpoint_enable, "checkpoint=enable"},
 209         {Opt_compress_algorithm, "compress_algorithm=%s"},
 210         {Opt_compress_log_size, "compress_log_size=%u"},
 211         {Opt_compress_extension, "compress_extension=%s"},
 212         {Opt_err, NULL},
 213 };
 214
 215 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 216 {
 217         struct va_format vaf;
 218         va_list args;
 219         int level;
 220
 221         va_start(args, fmt);
 222
 223         level = printk_get_level(fmt);
 224         vaf.fmt = printk_skip_level(fmt);
 225         vaf.va = &args;
 226         printk("%c%cF2FS-fs (%s): %pV\n",
 227                KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 228
 229         va_end(args);
 230 }
 231
 232 #ifdef CONFIG_UNICODE
 233 static const struct f2fs_sb_encodings {
 234         __u16 magic;
 235         char *name;
 236         char *version;
 237 } f2fs_sb_encoding_map[] = {
 238         {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
 239 };
 240
 241 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
 242                                  const struct f2fs_sb_encodings **encoding,
 243                                  __u16 *flags)
 244 {
 245         __u16 magic = le16_to_cpu(sb->s_encoding);
 246         int i;
 247
 248         for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
 249                 if (magic == f2fs_sb_encoding_map[i].magic)
 250                         break;
 251
 252         if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
 253                 return -EINVAL;
 254
 255         *encoding = &f2fs_sb_encoding_map[i];
 256         *flags = le16_to_cpu(sb->s_encoding_flags);
 257
 258         return 0;
 259 }
 260 #endif
 261
 262 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 263 {
 264         block_t limit = min((sbi->user_block_count << 1) / 1000,
 265                         sbi->user_block_count - sbi->reserved_blocks);
 266
 267         /* limit is 0.2% */
 268         if (test_opt(sbi, RESERVE_ROOT) &&
 269                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 270                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 271                 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 272                           F2FS_OPTION(sbi).root_reserved_blocks);
 273         }
 274         if (!test_opt(sbi, RESERVE_ROOT) &&
 275                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 276                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 277                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 278                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 279                 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 280                           from_kuid_munged(&init_user_ns,
 281                                            F2FS_OPTION(sbi).s_resuid),
 282                           from_kgid_munged(&init_user_ns,
 283                                            F2FS_OPTION(sbi).s_resgid));
 284 }
 285
 286 static void init_once(void *foo)
 287 {
 288         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 289
 290         inode_init_once(&fi->vfs_inode);
 291 }
 292
 293 #ifdef CONFIG_QUOTA
 294 static const char * const quotatypes[] = INITQFNAMES;
 295 #define QTYPE2NAME(t) (quotatypes[t])
 296 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 297                                                         substring_t *args)
 298 {
 299         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 300         char *qname;
 301         int ret = -EINVAL;
 302
 303         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 304                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 305                 return -EINVAL;
 306         }
 307         if (f2fs_sb_has_quota_ino(sbi)) {
 308                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 309                 return 0;
 310         }
 311
 312         qname = match_strdup(args);
 313         if (!qname) {
 314                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 315                 return -ENOMEM;
 316         }
 317         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 318                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 319                         ret = 0;
 320                 else
 321                         f2fs_err(sbi, "%s quota file already specified",
 322                                  QTYPE2NAME(qtype));
 323                 goto errout;
 324         }
 325         if (strchr(qname, '/')) {
 326                 f2fs_err(sbi, "quotafile must be on filesystem root");
 327                 goto errout;
 328         }
 329         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 330         set_opt(sbi, QUOTA);
 331         return 0;
 332 errout:
 333         kvfree(qname);
 334         return ret;
 335 }
 336
 337 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 338 {
 339         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 340
 341         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 342                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 343                 return -EINVAL;
 344         }
 345         kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 346         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 347         return 0;
 348 }
 349
 350 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 351 {
 352         /*
 353          * We do the test below only for project quotas. 'usrquota' and
 354          * 'grpquota' mount options are allowed even without quota feature
 355          * to support legacy quotas in quota files.
 356          */
 357         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 358                 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 359                 return -1;
 360         }
 361         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 362                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 363                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 364                 if (test_opt(sbi, USRQUOTA) &&
 365                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 366                         clear_opt(sbi, USRQUOTA);
 367
 368                 if (test_opt(sbi, GRPQUOTA) &&
 369                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 370                         clear_opt(sbi, GRPQUOTA);
 371
 372                 if (test_opt(sbi, PRJQUOTA) &&
 373                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 374                         clear_opt(sbi, PRJQUOTA);
 375
 376                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 377                                 test_opt(sbi, PRJQUOTA)) {
 378                         f2fs_err(sbi, "old and new quota format mixing");
 379                         return -1;
 380                 }
 381
 382                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 383                         f2fs_err(sbi, "journaled quota format not specified");
 384                         return -1;
 385                 }
 386         }
 387
 388         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 389                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 390                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 391         }
 392         return 0;
 393 }
 394 #endif
 395
 396 static int parse_options(struct super_block *sb, char *options)
 397 {
 398         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 399         substring_t args[MAX_OPT_ARGS];
 400         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 401         char *p, *name;
 402         int arg = 0, ext_cnt;
 403         kuid_t uid;
 404         kgid_t gid;
 405 #ifdef CONFIG_QUOTA
 406         int ret;
 407 #endif
 408
 409         if (!options)
 410                 return 0;
 411
 412         while ((p = strsep(&options, ",")) != NULL) {
 413                 int token;
 414                 if (!*p)
 415                         continue;
 416                 /*
 417                  * Initialize args struct so we know whether arg was
 418                  * found; some options take optional arguments.
 419                  */
 420                 args[0].to = args[0].from = NULL;
 421                 token = match_token(p, f2fs_tokens, args);
 422
 423                 switch (token) {
 424                 case Opt_gc_background:
 425                         name = match_strdup(&args[0]);
 426
 427                         if (!name)
 428                                 return -ENOMEM;
 429                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 430                                 set_opt(sbi, BG_GC);
 431                                 clear_opt(sbi, FORCE_FG_GC);
 432                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 433                                 clear_opt(sbi, BG_GC);
 434                                 clear_opt(sbi, FORCE_FG_GC);
 435                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 436                                 set_opt(sbi, BG_GC);
 437                                 set_opt(sbi, FORCE_FG_GC);
 438                         } else {
 439                                 kvfree(name);
 440                                 return -EINVAL;
 441                         }
 442                         kvfree(name);
 443                         break;
 444                 case Opt_disable_roll_forward:
 445                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 446                         break;
 447                 case Opt_norecovery:
 448                         /* this option mounts f2fs with ro */
 449                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 450                         if (!f2fs_readonly(sb))
 451                                 return -EINVAL;
 452                         break;
 453                 case Opt_discard:
 454                         set_opt(sbi, DISCARD);
 455                         break;
 456                 case Opt_nodiscard:
 457                         if (f2fs_sb_has_blkzoned(sbi)) {
 458                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 459                                 return -EINVAL;
 460                         }
 461                         clear_opt(sbi, DISCARD);
 462                         break;
 463                 case Opt_noheap:
 464                         set_opt(sbi, NOHEAP);
 465                         break;
 466                 case Opt_heap:
 467                         clear_opt(sbi, NOHEAP);
 468                         break;
 469 #ifdef CONFIG_F2FS_FS_XATTR
 470                 case Opt_user_xattr:
 471                         set_opt(sbi, XATTR_USER);
 472                         break;
 473                 case Opt_nouser_xattr:
 474                         clear_opt(sbi, XATTR_USER);
 475                         break;
 476                 case Opt_inline_xattr:
 477                         set_opt(sbi, INLINE_XATTR);
 478                         break;
 479                 case Opt_noinline_xattr:
 480                         clear_opt(sbi, INLINE_XATTR);
 481                         break;
 482                 case Opt_inline_xattr_size:
 483                         if (args->from && match_int(args, &arg))
 484                                 return -EINVAL;
 485                         set_opt(sbi, INLINE_XATTR_SIZE);
 486                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 487                         break;
 488 #else
 489                 case Opt_user_xattr:
 490                         f2fs_info(sbi, "user_xattr options not supported");
 491                         break;
 492                 case Opt_nouser_xattr:
 493                         f2fs_info(sbi, "nouser_xattr options not supported");
 494                         break;
 495                 case Opt_inline_xattr:
 496                         f2fs_info(sbi, "inline_xattr options not supported");
 497                         break;
 498                 case Opt_noinline_xattr:
 499                         f2fs_info(sbi, "noinline_xattr options not supported");
 500                         break;
 501 #endif
 502 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 503                 case Opt_acl:
 504                         set_opt(sbi, POSIX_ACL);
 505                         break;
 506                 case Opt_noacl:
 507                         clear_opt(sbi, POSIX_ACL);
 508                         break;
 509 #else
 510                 case Opt_acl:
 511                         f2fs_info(sbi, "acl options not supported");
 512                         break;
 513                 case Opt_noacl:
 514                         f2fs_info(sbi, "noacl options not supported");
 515                         break;
 516 #endif
 517                 case Opt_active_logs:
 518                         if (args->from && match_int(args, &arg))
 519                                 return -EINVAL;
 520                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 521                                 return -EINVAL;
 522                         F2FS_OPTION(sbi).active_logs = arg;
 523                         break;
 524                 case Opt_disable_ext_identify:
 525                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 526                         break;
 527                 case Opt_inline_data:
 528                         set_opt(sbi, INLINE_DATA);
 529                         break;
 530                 case Opt_inline_dentry:
 531                         set_opt(sbi, INLINE_DENTRY);
 532                         break;
 533                 case Opt_noinline_dentry:
 534                         clear_opt(sbi, INLINE_DENTRY);
 535                         break;
 536                 case Opt_flush_merge:
 537                         set_opt(sbi, FLUSH_MERGE);
 538                         break;
 539                 case Opt_noflush_merge:
 540                         clear_opt(sbi, FLUSH_MERGE);
 541                         break;
 542                 case Opt_nobarrier:
 543                         set_opt(sbi, NOBARRIER);
 544                         break;
 545                 case Opt_fastboot:
 546                         set_opt(sbi, FASTBOOT);
 547                         break;
 548                 case Opt_extent_cache:
 549                         set_opt(sbi, EXTENT_CACHE);
 550                         break;
 551                 case Opt_noextent_cache:
 552                         clear_opt(sbi, EXTENT_CACHE);
 553                         break;
 554                 case Opt_noinline_data:
 555                         clear_opt(sbi, INLINE_DATA);
 556                         break;
 557                 case Opt_data_flush:
 558                         set_opt(sbi, DATA_FLUSH);
 559                         break;
 560                 case Opt_reserve_root:
 561                         if (args->from && match_int(args, &arg))
 562                                 return -EINVAL;
 563                         if (test_opt(sbi, RESERVE_ROOT)) {
 564                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 565                                           F2FS_OPTION(sbi).root_reserved_blocks);
 566                         } else {
 567                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 568                                 set_opt(sbi, RESERVE_ROOT);
 569                         }
 570                         break;
 571                 case Opt_resuid:
 572                         if (args->from && match_int(args, &arg))
 573                                 return -EINVAL;
 574                         uid = make_kuid(current_user_ns(), arg);
 575                         if (!uid_valid(uid)) {
 576                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 577                                 return -EINVAL;
 578                         }
 579                         F2FS_OPTION(sbi).s_resuid = uid;
 580                         break;
 581                 case Opt_resgid:
 582                         if (args->from && match_int(args, &arg))
 583                                 return -EINVAL;
 584                         gid = make_kgid(current_user_ns(), arg);
 585                         if (!gid_valid(gid)) {
 586                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 587                                 return -EINVAL;
 588                         }
 589                         F2FS_OPTION(sbi).s_resgid = gid;
 590                         break;
 591                 case Opt_mode:
 592                         name = match_strdup(&args[0]);
 593
 594                         if (!name)
 595                                 return -ENOMEM;
 596                         if (strlen(name) == 8 &&
 597                                         !strncmp(name, "adaptive", 8)) {
 598                                 if (f2fs_sb_has_blkzoned(sbi)) {
 599                                         f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
 600                                         kvfree(name);
 601                                         return -EINVAL;
 602                                 }
 603                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 604                         } else if (strlen(name) == 3 &&
 605                                         !strncmp(name, "lfs", 3)) {
 606                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 607                         } else {
 608                                 kvfree(name);
 609                                 return -EINVAL;
 610                         }
 611                         kvfree(name);
 612                         break;
 613                 case Opt_io_size_bits:
 614                         if (args->from && match_int(args, &arg))
 615                                 return -EINVAL;
 616                         if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
 617                                 f2fs_warn(sbi, "Not support %d, larger than %d",
 618                                           1 << arg, BIO_MAX_PAGES);
 619                                 return -EINVAL;
 620                         }
 621                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 622                         break;
 623 #ifdef CONFIG_F2FS_FAULT_INJECTION
 624                 case Opt_fault_injection:
 625                         if (args->from && match_int(args, &arg))
 626                                 return -EINVAL;
 627                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 628                         set_opt(sbi, FAULT_INJECTION);
 629                         break;
 630
 631                 case Opt_fault_type:
 632                         if (args->from && match_int(args, &arg))
 633                                 return -EINVAL;
 634                         f2fs_build_fault_attr(sbi, 0, arg);
 635                         set_opt(sbi, FAULT_INJECTION);
 636                         break;
 637 #else
 638                 case Opt_fault_injection:
 639                         f2fs_info(sbi, "fault_injection options not supported");
 640                         break;
 641
 642                 case Opt_fault_type:
 643                         f2fs_info(sbi, "fault_type options not supported");
 644                         break;
 645 #endif
 646                 case Opt_lazytime:
 647                         sb->s_flags |= SB_LAZYTIME;
 648                         break;
 649                 case Opt_nolazytime:
 650                         sb->s_flags &= ~SB_LAZYTIME;
 651                         break;
 652 #ifdef CONFIG_QUOTA
 653                 case Opt_quota:
 654                 case Opt_usrquota:
 655                         set_opt(sbi, USRQUOTA);
 656                         break;
 657                 case Opt_grpquota:
 658                         set_opt(sbi, GRPQUOTA);
 659                         break;
 660                 case Opt_prjquota:
 661                         set_opt(sbi, PRJQUOTA);
 662                         break;
 663                 case Opt_usrjquota:
 664                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 665                         if (ret)
 666                                 return ret;
 667                         break;
 668                 case Opt_grpjquota:
 669                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 670                         if (ret)
 671                                 return ret;
 672                         break;
 673                 case Opt_prjjquota:
 674                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 675                         if (ret)
 676                                 return ret;
 677                         break;
 678                 case Opt_offusrjquota:
 679                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 680                         if (ret)
 681                                 return ret;
 682                         break;
 683                 case Opt_offgrpjquota:
 684                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 685                         if (ret)
 686                                 return ret;
 687                         break;
 688                 case Opt_offprjjquota:
 689                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 690                         if (ret)
 691                                 return ret;
 692                         break;
 693                 case Opt_jqfmt_vfsold:
 694                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 695                         break;
 696                 case Opt_jqfmt_vfsv0:
 697                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 698                         break;
 699                 case Opt_jqfmt_vfsv1:
 700                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 701                         break;
 702                 case Opt_noquota:
 703                         clear_opt(sbi, QUOTA);
 704                         clear_opt(sbi, USRQUOTA);
 705                         clear_opt(sbi, GRPQUOTA);
 706                         clear_opt(sbi, PRJQUOTA);
 707                         break;
 708 #else
 709                 case Opt_quota:
 710                 case Opt_usrquota:
 711                 case Opt_grpquota:
 712                 case Opt_prjquota:
 713                 case Opt_usrjquota:
 714                 case Opt_grpjquota:
 715                 case Opt_prjjquota:
 716                 case Opt_offusrjquota:
 717                 case Opt_offgrpjquota:
 718                 case Opt_offprjjquota:
 719                 case Opt_jqfmt_vfsold:
 720                 case Opt_jqfmt_vfsv0:
 721                 case Opt_jqfmt_vfsv1:
 722                 case Opt_noquota:
 723                         f2fs_info(sbi, "quota operations not supported");
 724                         break;
 725 #endif
 726                 case Opt_whint:
 727                         name = match_strdup(&args[0]);
 728                         if (!name)
 729                                 return -ENOMEM;
 730                         if (strlen(name) == 10 &&
 731                                         !strncmp(name, "user-based", 10)) {
 732                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 733                         } else if (strlen(name) == 3 &&
 734                                         !strncmp(name, "off", 3)) {
 735                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 736                         } else if (strlen(name) == 8 &&
 737                                         !strncmp(name, "fs-based", 8)) {
 738                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 739                         } else {
 740                                 kvfree(name);
 741                                 return -EINVAL;
 742                         }
 743                         kvfree(name);
 744                         break;
 745                 case Opt_alloc:
 746                         name = match_strdup(&args[0]);
 747                         if (!name)
 748                                 return -ENOMEM;
 749
 750                         if (strlen(name) == 7 &&
 751                                         !strncmp(name, "default", 7)) {
 752                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 753                         } else if (strlen(name) == 5 &&
 754                                         !strncmp(name, "reuse", 5)) {
 755                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 756                         } else {
 757                                 kvfree(name);
 758                                 return -EINVAL;
 759                         }
 760                         kvfree(name);
 761                         break;
 762                 case Opt_fsync:
 763                         name = match_strdup(&args[0]);
 764                         if (!name)
 765                                 return -ENOMEM;
 766                         if (strlen(name) == 5 &&
 767                                         !strncmp(name, "posix", 5)) {
 768                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 769                         } else if (strlen(name) == 6 &&
 770                                         !strncmp(name, "strict", 6)) {
 771                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 772                         } else if (strlen(name) == 9 &&
 773                                         !strncmp(name, "nobarrier", 9)) {
 774                                 F2FS_OPTION(sbi).fsync_mode =
 775                                                         FSYNC_MODE_NOBARRIER;
 776                         } else {
 777                                 kvfree(name);
 778                                 return -EINVAL;
 779                         }
 780                         kvfree(name);
 781                         break;
 782                 case Opt_test_dummy_encryption:
 783 #ifdef CONFIG_FS_ENCRYPTION
 784                         if (!f2fs_sb_has_encrypt(sbi)) {
 785                                 f2fs_err(sbi, "Encrypt feature is off");
 786                                 return -EINVAL;
 787                         }
 788
 789                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 790                         f2fs_info(sbi, "Test dummy encryption mode enabled");
 791 #else
 792                         f2fs_info(sbi, "Test dummy encryption mount option ignored");
 793 #endif
 794                         break;
 795                 case Opt_checkpoint_disable_cap_perc:
 796                         if (args->from && match_int(args, &arg))
 797                                 return -EINVAL;
 798                         if (arg < 0 || arg > 100)
 799                                 return -EINVAL;
 800                         if (arg == 100)
 801                                 F2FS_OPTION(sbi).unusable_cap =
 802                                         sbi->user_block_count;
 803                         else
 804                                 F2FS_OPTION(sbi).unusable_cap =
 805                                         (sbi->user_block_count / 100) * arg;
 806                         set_opt(sbi, DISABLE_CHECKPOINT);
 807                         break;
 808                 case Opt_checkpoint_disable_cap:
 809                         if (args->from && match_int(args, &arg))
 810                                 return -EINVAL;
 811                         F2FS_OPTION(sbi).unusable_cap = arg;
 812                         set_opt(sbi, DISABLE_CHECKPOINT);
 813                         break;
 814                 case Opt_checkpoint_disable:
 815                         set_opt(sbi, DISABLE_CHECKPOINT);
 816                         break;
 817                 case Opt_checkpoint_enable:
 818                         clear_opt(sbi, DISABLE_CHECKPOINT);
 819                         break;
 820                 case Opt_compress_algorithm:
 821                         if (!f2fs_sb_has_compression(sbi)) {
 822                                 f2fs_err(sbi, "Compression feature if off");
 823                                 return -EINVAL;
 824                         }
 825                         name = match_strdup(&args[0]);
 826                         if (!name)
 827                                 return -ENOMEM;
 828                         if (strlen(name) == 3 && !strcmp(name, "lzo")) {
 829                                 F2FS_OPTION(sbi).compress_algorithm =
 830                                                                 COMPRESS_LZO;
 831                         } else if (strlen(name) == 3 &&
 832                                         !strcmp(name, "lz4")) {
 833                                 F2FS_OPTION(sbi).compress_algorithm =
 834                                                                 COMPRESS_LZ4;
 835                         } else {
 836                                 kfree(name);
 837                                 return -EINVAL;
 838                         }
 839                         kfree(name);
 840                         break;
 841                 case Opt_compress_log_size:
 842                         if (!f2fs_sb_has_compression(sbi)) {
 843                                 f2fs_err(sbi, "Compression feature is off");
 844                                 return -EINVAL;
 845                         }
 846                         if (args->from && match_int(args, &arg))
 847                                 return -EINVAL;
 848                         if (arg < MIN_COMPRESS_LOG_SIZE ||
 849                                 arg > MAX_COMPRESS_LOG_SIZE) {
 850                                 f2fs_err(sbi,
 851                                         "Compress cluster log size is out of range");
 852                                 return -EINVAL;
 853                         }
 854                         F2FS_OPTION(sbi).compress_log_size = arg;
 855                         break;
 856                 case Opt_compress_extension:
 857                         if (!f2fs_sb_has_compression(sbi)) {
 858                                 f2fs_err(sbi, "Compression feature is off");
 859                                 return -EINVAL;
 860                         }
 861                         name = match_strdup(&args[0]);
 862                         if (!name)
 863                                 return -ENOMEM;
 864
 865                         ext = F2FS_OPTION(sbi).extensions;
 866                         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
 867
 868                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
 869                                 ext_cnt >= COMPRESS_EXT_NUM) {
 870                                 f2fs_err(sbi,
 871                                         "invalid extension length/number");
 872                                 kfree(name);
 873                                 return -EINVAL;
 874                         }
 875
 876                         strcpy(ext[ext_cnt], name);
 877                         F2FS_OPTION(sbi).compress_ext_cnt++;
 878                         kfree(name);
 879                         break;
 880                 default:
 881                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
 882                                  p);
 883                         return -EINVAL;
 884                 }
 885         }
 886 #ifdef CONFIG_QUOTA
 887         if (f2fs_check_quota_options(sbi))
 888                 return -EINVAL;
 889 #else
 890         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
 891                 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 892                 return -EINVAL;
 893         }
 894         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
 895                 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 896                 return -EINVAL;
 897         }
 898 #endif
 899 #ifndef CONFIG_UNICODE
 900         if (f2fs_sb_has_casefold(sbi)) {
 901                 f2fs_err(sbi,
 902                         "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
 903                 return -EINVAL;
 904         }
 905 #endif
 906
 907         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 908                 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
 909                          F2FS_IO_SIZE_KB(sbi));
 910                 return -EINVAL;
 911         }
 912
 913         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 914                 int min_size, max_size;
 915
 916                 if (!f2fs_sb_has_extra_attr(sbi) ||
 917                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 918                         f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
 919                         return -EINVAL;
 920                 }
 921                 if (!test_opt(sbi, INLINE_XATTR)) {
 922                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
 923                         return -EINVAL;
 924                 }
 925
 926                 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
 927                 max_size = MAX_INLINE_XATTR_SIZE;
 928
 929                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
 930                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
 931                         f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
 932                                  min_size, max_size);
 933                         return -EINVAL;
 934                 }
 935         }
 936
 937         if (test_opt(sbi, DISABLE_CHECKPOINT) && test_opt(sbi, LFS)) {
 938                 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
 939                 return -EINVAL;
 940         }
 941
 942         /* Not pass down write hints if the number of active logs is lesser
 943          * than NR_CURSEG_TYPE.
 944          */
 945         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 946                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 947         return 0;
 948 }
 949
 950 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 951 {
 952         struct f2fs_inode_info *fi;
 953
 954         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 955         if (!fi)
 956                 return NULL;
 957
 958         init_once((void *) fi);
 959
 960         /* Initialize f2fs-specific inode info */
 961         atomic_set(&fi->dirty_pages, 0);
 962         init_rwsem(&fi->i_sem);
 963         INIT_LIST_HEAD(&fi->dirty_list);
 964         INIT_LIST_HEAD(&fi->gdirty_list);
 965         INIT_LIST_HEAD(&fi->inmem_ilist);
 966         INIT_LIST_HEAD(&fi->inmem_pages);
 967         mutex_init(&fi->inmem_lock);
 968         init_rwsem(&fi->i_gc_rwsem[READ]);
 969         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 970         init_rwsem(&fi->i_mmap_sem);
 971         init_rwsem(&fi->i_xattr_sem);
 972
 973         /* Will be used by directory only */
 974         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 975
 976         return &fi->vfs_inode;
 977 }
 978
 979 static int f2fs_drop_inode(struct inode *inode)
 980 {
 981         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 982         int ret;
 983
 984         /*
 985          * during filesystem shutdown, if checkpoint is disabled,
 986          * drop useless meta/node dirty pages.
 987          */
 988         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
 989                 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 990                         inode->i_ino == F2FS_META_INO(sbi)) {
 991                         trace_f2fs_drop_inode(inode, 1);
 992                         return 1;
 993                 }
 994         }
 995
 996         /*
 997          * This is to avoid a deadlock condition like below.
 998          * writeback_single_inode(inode)
 999          *  - f2fs_write_data_page
1000          *    - f2fs_gc -> iput -> evict
1001          *       - inode_wait_for_writeback(inode)
1002          */
1003         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1004                 if (!inode->i_nlink && !is_bad_inode(inode)) {
1005                         /* to avoid evict_inode call simultaneously */
1006                         atomic_inc(&inode->i_count);
1007                         spin_unlock(&inode->i_lock);
1008
1009                         /* some remained atomic pages should discarded */
1010                         if (f2fs_is_atomic_file(inode))
1011                                 f2fs_drop_inmem_pages(inode);
1012
1013                         /* should remain fi->extent_tree for writepage */
1014                         f2fs_destroy_extent_node(inode);
1015
1016                         sb_start_intwrite(inode->i_sb);
1017                         f2fs_i_size_write(inode, 0);
1018
1019                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1020                                         inode, NULL, 0, DATA);
1021                         truncate_inode_pages_final(inode->i_mapping);
1022
1023                         if (F2FS_HAS_BLOCKS(inode))
1024                                 f2fs_truncate(inode);
1025
1026                         sb_end_intwrite(inode->i_sb);
1027
1028                         spin_lock(&inode->i_lock);
1029                         atomic_dec(&inode->i_count);
1030                 }
1031                 trace_f2fs_drop_inode(inode, 0);
1032                 return 0;
1033         }
1034         ret = generic_drop_inode(inode);
1035         if (!ret)
1036                 ret = fscrypt_drop_inode(inode);
1037         trace_f2fs_drop_inode(inode, ret);
1038         return ret;
1039 }
1040
1041 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1042 {
1043         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1044         int ret = 0;
1045
1046         spin_lock(&sbi->inode_lock[DIRTY_META]);
1047         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1048                 ret = 1;
1049         } else {
1050                 set_inode_flag(inode, FI_DIRTY_INODE);
1051                 stat_inc_dirty_inode(sbi, DIRTY_META);
1052         }
1053         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1054                 list_add_tail(&F2FS_I(inode)->gdirty_list,
1055                                 &sbi->inode_list[DIRTY_META]);
1056                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1057         }
1058         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1059         return ret;
1060 }
1061
1062 void f2fs_inode_synced(struct inode *inode)
1063 {
1064         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1065
1066         spin_lock(&sbi->inode_lock[DIRTY_META]);
1067         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1068                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1069                 return;
1070         }
1071         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1072                 list_del_init(&F2FS_I(inode)->gdirty_list);
1073                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1074         }
1075         clear_inode_flag(inode, FI_DIRTY_INODE);
1076         clear_inode_flag(inode, FI_AUTO_RECOVER);
1077         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1078         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1079 }
1080
1081 /*
1082  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1083  *
1084  * We should call set_dirty_inode to write the dirty inode through write_inode.
1085  */
1086 static void f2fs_dirty_inode(struct inode *inode, int flags)
1087 {
1088         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1089
1090         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1091                         inode->i_ino == F2FS_META_INO(sbi))
1092                 return;
1093
1094         if (flags == I_DIRTY_TIME)
1095                 return;
1096
1097         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1098                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1099
1100         f2fs_inode_dirtied(inode, false);
1101 }
1102
1103 static void f2fs_free_inode(struct inode *inode)
1104 {
1105         fscrypt_free_inode(inode);
1106         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1107 }
1108
1109 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1110 {
1111         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1112         percpu_counter_destroy(&sbi->total_valid_inode_count);
1113 }
1114
1115 static void destroy_device_list(struct f2fs_sb_info *sbi)
1116 {
1117         int i;
1118
1119         for (i = 0; i < sbi->s_ndevs; i++) {
1120                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1121 #ifdef CONFIG_BLK_DEV_ZONED
1122                 kvfree(FDEV(i).blkz_seq);
1123 #endif
1124         }
1125         kvfree(sbi->devs);
1126 }
1127
1128 static void f2fs_put_super(struct super_block *sb)
1129 {
1130         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1131         int i;
1132         bool dropped;
1133
1134         f2fs_quota_off_umount(sb);
1135
1136         /* prevent remaining shrinker jobs */
1137         mutex_lock(&sbi->umount_mutex);
1138
1139         /*
1140          * We don't need to do checkpoint when superblock is clean.
1141          * But, the previous checkpoint was not done by umount, it needs to do
1142          * clean checkpoint again.
1143          */
1144         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1145                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1146                 struct cp_control cpc = {
1147                         .reason = CP_UMOUNT,
1148                 };
1149                 f2fs_write_checkpoint(sbi, &cpc);
1150         }
1151
1152         /* be sure to wait for any on-going discard commands */
1153         dropped = f2fs_issue_discard_timeout(sbi);
1154
1155         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1156                                         !sbi->discard_blks && !dropped) {
1157                 struct cp_control cpc = {
1158                         .reason = CP_UMOUNT | CP_TRIMMED,
1159                 };
1160                 f2fs_write_checkpoint(sbi, &cpc);
1161         }
1162
1163         /*
1164          * normally superblock is clean, so we need to release this.
1165          * In addition, EIO will skip do checkpoint, we need this as well.
1166          */
1167         f2fs_release_ino_entry(sbi, true);
1168
1169         f2fs_leave_shrinker(sbi);
1170         mutex_unlock(&sbi->umount_mutex);
1171
1172         /* our cp_error case, we can wait for any writeback page */
1173         f2fs_flush_merged_writes(sbi);
1174
1175         f2fs_wait_on_all_pages_writeback(sbi);
1176
1177         f2fs_bug_on(sbi, sbi->fsync_node_num);
1178
1179         iput(sbi->node_inode);
1180         sbi->node_inode = NULL;
1181
1182         iput(sbi->meta_inode);
1183         sbi->meta_inode = NULL;
1184
1185         /*
1186          * iput() can update stat information, if f2fs_write_checkpoint()
1187          * above failed with error.
1188          */
1189         f2fs_destroy_stats(sbi);
1190
1191         /* destroy f2fs internal modules */
1192         f2fs_destroy_node_manager(sbi);
1193         f2fs_destroy_segment_manager(sbi);
1194
1195         f2fs_destroy_post_read_wq(sbi);
1196
1197         kvfree(sbi->ckpt);
1198
1199         f2fs_unregister_sysfs(sbi);
1200
1201         sb->s_fs_info = NULL;
1202         if (sbi->s_chksum_driver)
1203                 crypto_free_shash(sbi->s_chksum_driver);
1204         kvfree(sbi->raw_super);
1205
1206         destroy_device_list(sbi);
1207         mempool_destroy(sbi->write_io_dummy);
1208 #ifdef CONFIG_QUOTA
1209         for (i = 0; i < MAXQUOTAS; i++)
1210                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1211 #endif
1212         destroy_percpu_info(sbi);
1213         for (i = 0; i < NR_PAGE_TYPE; i++)
1214                 kvfree(sbi->write_io[i]);
1215 #ifdef CONFIG_UNICODE
1216         utf8_unload(sbi->s_encoding);
1217 #endif
1218         kvfree(sbi);
1219 }
1220
1221 int f2fs_sync_fs(struct super_block *sb, int sync)
1222 {
1223         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1224         int err = 0;
1225
1226         if (unlikely(f2fs_cp_error(sbi)))
1227                 return 0;
1228         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1229                 return 0;
1230
1231         trace_f2fs_sync_fs(sb, sync);
1232
1233         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1234                 return -EAGAIN;
1235
1236         if (sync) {
1237                 struct cp_control cpc;
1238
1239                 cpc.reason = __get_cp_reason(sbi);
1240
1241                 mutex_lock(&sbi->gc_mutex);
1242                 err = f2fs_write_checkpoint(sbi, &cpc);
1243                 mutex_unlock(&sbi->gc_mutex);
1244         }
1245         f2fs_trace_ios(NULL, 1);
1246
1247         return err;
1248 }
1249
1250 static int f2fs_freeze(struct super_block *sb)
1251 {
1252         if (f2fs_readonly(sb))
1253                 return 0;
1254
1255         /* IO error happened before */
1256         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1257                 return -EIO;
1258
1259         /* must be clean, since sync_filesystem() was already called */
1260         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1261                 return -EINVAL;
1262         return 0;
1263 }
1264
1265 static int f2fs_unfreeze(struct super_block *sb)
1266 {
1267         return 0;
1268 }
1269
1270 #ifdef CONFIG_QUOTA
1271 static int f2fs_statfs_project(struct super_block *sb,
1272                                 kprojid_t projid, struct kstatfs *buf)
1273 {
1274         struct kqid qid;
1275         struct dquot *dquot;
1276         u64 limit;
1277         u64 curblock;
1278
1279         qid = make_kqid_projid(projid);
1280         dquot = dqget(sb, qid);
1281         if (IS_ERR(dquot))
1282                 return PTR_ERR(dquot);
1283         spin_lock(&dquot->dq_dqb_lock);
1284
1285         limit = 0;
1286         if (dquot->dq_dqb.dqb_bsoftlimit)
1287                 limit = dquot->dq_dqb.dqb_bsoftlimit;
1288         if (dquot->dq_dqb.dqb_bhardlimit &&
1289                         (!limit || dquot->dq_dqb.dqb_bhardlimit < limit))
1290                 limit = dquot->dq_dqb.dqb_bhardlimit;
1291
1292         if (limit && buf->f_blocks > limit) {
1293                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1294                 buf->f_blocks = limit;
1295                 buf->f_bfree = buf->f_bavail =
1296                         (buf->f_blocks > curblock) ?
1297                          (buf->f_blocks - curblock) : 0;
1298         }
1299
1300         limit = 0;
1301         if (dquot->dq_dqb.dqb_isoftlimit)
1302                 limit = dquot->dq_dqb.dqb_isoftlimit;
1303         if (dquot->dq_dqb.dqb_ihardlimit &&
1304                         (!limit || dquot->dq_dqb.dqb_ihardlimit < limit))
1305                 limit = dquot->dq_dqb.dqb_ihardlimit;
1306
1307         if (limit && buf->f_files > limit) {
1308                 buf->f_files = limit;
1309                 buf->f_ffree =
1310                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1311                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1312         }
1313
1314         spin_unlock(&dquot->dq_dqb_lock);
1315         dqput(dquot);
1316         return 0;
1317 }
1318 #endif
1319
1320 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1321 {
1322         struct super_block *sb = dentry->d_sb;
1323         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1324         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1325         block_t total_count, user_block_count, start_count;
1326         u64 avail_node_count;
1327
1328         total_count = le64_to_cpu(sbi->raw_super->block_count);
1329         user_block_count = sbi->user_block_count;
1330         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1331         buf->f_type = F2FS_SUPER_MAGIC;
1332         buf->f_bsize = sbi->blocksize;
1333
1334         buf->f_blocks = total_count - start_count;
1335         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1336                                                 sbi->current_reserved_blocks;
1337
1338         spin_lock(&sbi->stat_lock);
1339         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1340                 buf->f_bfree = 0;
1341         else
1342                 buf->f_bfree -= sbi->unusable_block_count;
1343         spin_unlock(&sbi->stat_lock);
1344
1345         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1346                 buf->f_bavail = buf->f_bfree -
1347                                 F2FS_OPTION(sbi).root_reserved_blocks;
1348         else
1349                 buf->f_bavail = 0;
1350
1351         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1352
1353         if (avail_node_count > user_block_count) {
1354                 buf->f_files = user_block_count;
1355                 buf->f_ffree = buf->f_bavail;
1356         } else {
1357                 buf->f_files = avail_node_count;
1358                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1359                                         buf->f_bavail);
1360         }
1361
1362         buf->f_namelen = F2FS_NAME_LEN;
1363         buf->f_fsid.val[0] = (u32)id;
1364         buf->f_fsid.val[1] = (u32)(id >> 32);
1365
1366 #ifdef CONFIG_QUOTA
1367         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1368                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1369                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1370         }
1371 #endif
1372         return 0;
1373 }
1374
1375 static inline void f2fs_show_quota_options(struct seq_file *seq,
1376                                            struct super_block *sb)
1377 {
1378 #ifdef CONFIG_QUOTA
1379         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1380
1381         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1382                 char *fmtname = "";
1383
1384                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1385                 case QFMT_VFS_OLD:
1386                         fmtname = "vfsold";
1387                         break;
1388                 case QFMT_VFS_V0:
1389                         fmtname = "vfsv0";
1390                         break;
1391                 case QFMT_VFS_V1:
1392                         fmtname = "vfsv1";
1393                         break;
1394                 }
1395                 seq_printf(seq, ",jqfmt=%s", fmtname);
1396         }
1397
1398         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1399                 seq_show_option(seq, "usrjquota",
1400                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1401
1402         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1403                 seq_show_option(seq, "grpjquota",
1404                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1405
1406         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1407                 seq_show_option(seq, "prjjquota",
1408                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1409 #endif
1410 }
1411
1412 static inline void f2fs_show_compress_options(struct seq_file *seq,
1413                                                         struct super_block *sb)
1414 {
1415         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1416         char *algtype = "";
1417         int i;
1418
1419         if (!f2fs_sb_has_compression(sbi))
1420                 return;
1421
1422         switch (F2FS_OPTION(sbi).compress_algorithm) {
1423         case COMPRESS_LZO:
1424                 algtype = "lzo";
1425                 break;
1426         case COMPRESS_LZ4:
1427                 algtype = "lz4";
1428                 break;
1429         }
1430         seq_printf(seq, ",compress_algorithm=%s", algtype);
1431
1432         seq_printf(seq, ",compress_log_size=%u",
1433                         F2FS_OPTION(sbi).compress_log_size);
1434
1435         for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1436                 seq_printf(seq, ",compress_extension=%s",
1437                         F2FS_OPTION(sbi).extensions[i]);
1438         }
1439 }
1440
1441 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1442 {
1443         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1444
1445         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1446                 if (test_opt(sbi, FORCE_FG_GC))
1447                         seq_printf(seq, ",background_gc=%s", "sync");
1448                 else
1449                         seq_printf(seq, ",background_gc=%s", "on");
1450         } else {
1451                 seq_printf(seq, ",background_gc=%s", "off");
1452         }
1453         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1454                 seq_puts(seq, ",disable_roll_forward");
1455         if (test_opt(sbi, DISCARD))
1456                 seq_puts(seq, ",discard");
1457         else
1458                 seq_puts(seq, ",nodiscard");
1459         if (test_opt(sbi, NOHEAP))
1460                 seq_puts(seq, ",no_heap");
1461         else
1462                 seq_puts(seq, ",heap");
1463 #ifdef CONFIG_F2FS_FS_XATTR
1464         if (test_opt(sbi, XATTR_USER))
1465                 seq_puts(seq, ",user_xattr");
1466         else
1467                 seq_puts(seq, ",nouser_xattr");
1468         if (test_opt(sbi, INLINE_XATTR))
1469                 seq_puts(seq, ",inline_xattr");
1470         else
1471                 seq_puts(seq, ",noinline_xattr");
1472         if (test_opt(sbi, INLINE_XATTR_SIZE))
1473                 seq_printf(seq, ",inline_xattr_size=%u",
1474                                         F2FS_OPTION(sbi).inline_xattr_size);
1475 #endif
1476 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1477         if (test_opt(sbi, POSIX_ACL))
1478                 seq_puts(seq, ",acl");
1479         else
1480                 seq_puts(seq, ",noacl");
1481 #endif
1482         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1483                 seq_puts(seq, ",disable_ext_identify");
1484         if (test_opt(sbi, INLINE_DATA))
1485                 seq_puts(seq, ",inline_data");
1486         else
1487                 seq_puts(seq, ",noinline_data");
1488         if (test_opt(sbi, INLINE_DENTRY))
1489                 seq_puts(seq, ",inline_dentry");
1490         else
1491                 seq_puts(seq, ",noinline_dentry");
1492         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1493                 seq_puts(seq, ",flush_merge");
1494         if (test_opt(sbi, NOBARRIER))
1495                 seq_puts(seq, ",nobarrier");
1496         if (test_opt(sbi, FASTBOOT))
1497                 seq_puts(seq, ",fastboot");
1498         if (test_opt(sbi, EXTENT_CACHE))
1499                 seq_puts(seq, ",extent_cache");
1500         else
1501                 seq_puts(seq, ",noextent_cache");
1502         if (test_opt(sbi, DATA_FLUSH))
1503                 seq_puts(seq, ",data_flush");
1504
1505         seq_puts(seq, ",mode=");
1506         if (test_opt(sbi, ADAPTIVE))
1507                 seq_puts(seq, "adaptive");
1508         else if (test_opt(sbi, LFS))
1509                 seq_puts(seq, "lfs");
1510         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1511         if (test_opt(sbi, RESERVE_ROOT))
1512                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1513                                 F2FS_OPTION(sbi).root_reserved_blocks,
1514                                 from_kuid_munged(&init_user_ns,
1515                                         F2FS_OPTION(sbi).s_resuid),
1516                                 from_kgid_munged(&init_user_ns,
1517                                         F2FS_OPTION(sbi).s_resgid));
1518         if (F2FS_IO_SIZE_BITS(sbi))
1519                 seq_printf(seq, ",io_bits=%u",
1520                                 F2FS_OPTION(sbi).write_io_size_bits);
1521 #ifdef CONFIG_F2FS_FAULT_INJECTION
1522         if (test_opt(sbi, FAULT_INJECTION)) {
1523                 seq_printf(seq, ",fault_injection=%u",
1524                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1525                 seq_printf(seq, ",fault_type=%u",
1526                                 F2FS_OPTION(sbi).fault_info.inject_type);
1527         }
1528 #endif
1529 #ifdef CONFIG_QUOTA
1530         if (test_opt(sbi, QUOTA))
1531                 seq_puts(seq, ",quota");
1532         if (test_opt(sbi, USRQUOTA))
1533                 seq_puts(seq, ",usrquota");
1534         if (test_opt(sbi, GRPQUOTA))
1535                 seq_puts(seq, ",grpquota");
1536         if (test_opt(sbi, PRJQUOTA))
1537                 seq_puts(seq, ",prjquota");
1538 #endif
1539         f2fs_show_quota_options(seq, sbi->sb);
1540         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1541                 seq_printf(seq, ",whint_mode=%s", "user-based");
1542         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1543                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1544 #ifdef CONFIG_FS_ENCRYPTION
1545         if (F2FS_OPTION(sbi).test_dummy_encryption)
1546                 seq_puts(seq, ",test_dummy_encryption");
1547 #endif
1548
1549         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1550                 seq_printf(seq, ",alloc_mode=%s", "default");
1551         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1552                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1553
1554         if (test_opt(sbi, DISABLE_CHECKPOINT))
1555                 seq_printf(seq, ",checkpoint=disable:%u",
1556                                 F2FS_OPTION(sbi).unusable_cap);
1557         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1558                 seq_printf(seq, ",fsync_mode=%s", "posix");
1559         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1560                 seq_printf(seq, ",fsync_mode=%s", "strict");
1561         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1562                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1563
1564         f2fs_show_compress_options(seq, sbi->sb);
1565         return 0;
1566 }
1567
1568 static void default_options(struct f2fs_sb_info *sbi)
1569 {
1570         /* init some FS parameters */
1571         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1572         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1573         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1574         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1575         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1576         F2FS_OPTION(sbi).test_dummy_encryption = false;
1577         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1578         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1579         F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZO;
1580         F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1581         F2FS_OPTION(sbi).compress_ext_cnt = 0;
1582
1583         set_opt(sbi, BG_GC);
1584         set_opt(sbi, INLINE_XATTR);
1585         set_opt(sbi, INLINE_DATA);
1586         set_opt(sbi, INLINE_DENTRY);
1587         set_opt(sbi, EXTENT_CACHE);
1588         set_opt(sbi, NOHEAP);
1589         clear_opt(sbi, DISABLE_CHECKPOINT);
1590         F2FS_OPTION(sbi).unusable_cap = 0;
1591         sbi->sb->s_flags |= SB_LAZYTIME;
1592         set_opt(sbi, FLUSH_MERGE);
1593         set_opt(sbi, DISCARD);
1594         if (f2fs_sb_has_blkzoned(sbi))
1595                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1596         else
1597                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1598
1599 #ifdef CONFIG_F2FS_FS_XATTR
1600         set_opt(sbi, XATTR_USER);
1601 #endif
1602 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1603         set_opt(sbi, POSIX_ACL);
1604 #endif
1605
1606         f2fs_build_fault_attr(sbi, 0, 0);
1607 }
1608
1609 #ifdef CONFIG_QUOTA
1610 static int f2fs_enable_quotas(struct super_block *sb);
1611 #endif
1612
1613 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1614 {
1615         unsigned int s_flags = sbi->sb->s_flags;
1616         struct cp_control cpc;
1617         int err = 0;
1618         int ret;
1619         block_t unusable;
1620
1621         if (s_flags & SB_RDONLY) {
1622                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1623                 return -EINVAL;
1624         }
1625         sbi->sb->s_flags |= SB_ACTIVE;
1626
1627         f2fs_update_time(sbi, DISABLE_TIME);
1628
1629         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1630                 mutex_lock(&sbi->gc_mutex);
1631                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1632                 if (err == -ENODATA) {
1633                         err = 0;
1634                         break;
1635                 }
1636                 if (err && err != -EAGAIN)
1637                         break;
1638         }
1639
1640         ret = sync_filesystem(sbi->sb);
1641         if (ret || err) {
1642                 err = ret ? ret: err;
1643                 goto restore_flag;
1644         }
1645
1646         unusable = f2fs_get_unusable_blocks(sbi);
1647         if (f2fs_disable_cp_again(sbi, unusable)) {
1648                 err = -EAGAIN;
1649                 goto restore_flag;
1650         }
1651
1652         mutex_lock(&sbi->gc_mutex);
1653         cpc.reason = CP_PAUSE;
1654         set_sbi_flag(sbi, SBI_CP_DISABLED);
1655         err = f2fs_write_checkpoint(sbi, &cpc);
1656         if (err)
1657                 goto out_unlock;
1658
1659         spin_lock(&sbi->stat_lock);
1660         sbi->unusable_block_count = unusable;
1661         spin_unlock(&sbi->stat_lock);
1662
1663 out_unlock:
1664         mutex_unlock(&sbi->gc_mutex);
1665 restore_flag:
1666         sbi->sb->s_flags = s_flags;     /* Restore MS_RDONLY status */
1667         return err;
1668 }
1669
1670 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1671 {
1672         mutex_lock(&sbi->gc_mutex);
1673         f2fs_dirty_to_prefree(sbi);
1674
1675         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1676         set_sbi_flag(sbi, SBI_IS_DIRTY);
1677         mutex_unlock(&sbi->gc_mutex);
1678
1679         f2fs_sync_fs(sbi->sb, 1);
1680 }
1681
1682 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1683 {
1684         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1685         struct f2fs_mount_info org_mount_opt;
1686         unsigned long old_sb_flags;
1687         int err;
1688         bool need_restart_gc = false;
1689         bool need_stop_gc = false;
1690         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1691         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1692         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1693         bool checkpoint_changed;
1694 #ifdef CONFIG_QUOTA
1695         int i, j;
1696 #endif
1697
1698         /*
1699          * Save the old mount options in case we
1700          * need to restore them.
1701          */
1702         org_mount_opt = sbi->mount_opt;
1703         old_sb_flags = sb->s_flags;
1704
1705 #ifdef CONFIG_QUOTA
1706         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1707         for (i = 0; i < MAXQUOTAS; i++) {
1708                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1709                         org_mount_opt.s_qf_names[i] =
1710                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1711                                 GFP_KERNEL);
1712                         if (!org_mount_opt.s_qf_names[i]) {
1713                                 for (j = 0; j < i; j++)
1714                                         kvfree(org_mount_opt.s_qf_names[j]);
1715                                 return -ENOMEM;
1716                         }
1717                 } else {
1718                         org_mount_opt.s_qf_names[i] = NULL;
1719                 }
1720         }
1721 #endif
1722
1723         /* recover superblocks we couldn't write due to previous RO mount */
1724         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1725                 err = f2fs_commit_super(sbi, false);
1726                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1727                           err);
1728                 if (!err)
1729                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1730         }
1731
1732         default_options(sbi);
1733
1734         /* parse mount options */
1735         err = parse_options(sb, data);
1736         if (err)
1737                 goto restore_opts;
1738         checkpoint_changed =
1739                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1740
1741         /*
1742          * Previous and new state of filesystem is RO,
1743          * so skip checking GC and FLUSH_MERGE conditions.
1744          */
1745         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1746                 goto skip;
1747
1748 #ifdef CONFIG_QUOTA
1749         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1750                 err = dquot_suspend(sb, -1);
1751                 if (err < 0)
1752                         goto restore_opts;
1753         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1754                 /* dquot_resume needs RW */
1755                 sb->s_flags &= ~SB_RDONLY;
1756                 if (sb_any_quota_suspended(sb)) {
1757                         dquot_resume(sb, -1);
1758                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1759                         err = f2fs_enable_quotas(sb);
1760                         if (err)
1761                                 goto restore_opts;
1762                 }
1763         }
1764 #endif
1765         /* disallow enable/disable extent_cache dynamically */
1766         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1767                 err = -EINVAL;
1768                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1769                 goto restore_opts;
1770         }
1771
1772         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1773                 err = -EINVAL;
1774                 f2fs_warn(sbi, "switch io_bits option is not allowed");
1775                 goto restore_opts;
1776         }
1777
1778         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1779                 err = -EINVAL;
1780                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1781                 goto restore_opts;
1782         }
1783
1784         /*
1785          * We stop the GC thread if FS is mounted as RO
1786          * or if background_gc = off is passed in mount
1787          * option. Also sync the filesystem.
1788          */
1789         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1790                 if (sbi->gc_thread) {
1791                         f2fs_stop_gc_thread(sbi);
1792                         need_restart_gc = true;
1793                 }
1794         } else if (!sbi->gc_thread) {
1795                 err = f2fs_start_gc_thread(sbi);
1796                 if (err)
1797                         goto restore_opts;
1798                 need_stop_gc = true;
1799         }
1800
1801         if (*flags & SB_RDONLY ||
1802                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1803                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1804                 sync_inodes_sb(sb);
1805
1806                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1807                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1808                 f2fs_sync_fs(sb, 1);
1809                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1810         }
1811
1812         if (checkpoint_changed) {
1813                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1814                         err = f2fs_disable_checkpoint(sbi);
1815                         if (err)
1816                                 goto restore_gc;
1817                 } else {
1818                         f2fs_enable_checkpoint(sbi);
1819                 }
1820         }
1821
1822         /*
1823          * We stop issue flush thread if FS is mounted as RO
1824          * or if flush_merge is not passed in mount option.
1825          */
1826         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1827                 clear_opt(sbi, FLUSH_MERGE);
1828                 f2fs_destroy_flush_cmd_control(sbi, false);
1829         } else {
1830                 err = f2fs_create_flush_cmd_control(sbi);
1831                 if (err)
1832                         goto restore_gc;
1833         }
1834 skip:
1835 #ifdef CONFIG_QUOTA
1836         /* Release old quota file names */
1837         for (i = 0; i < MAXQUOTAS; i++)
1838                 kvfree(org_mount_opt.s_qf_names[i]);
1839 #endif
1840         /* Update the POSIXACL Flag */
1841         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1842                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1843
1844         limit_reserve_root(sbi);
1845         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1846         return 0;
1847 restore_gc:
1848         if (need_restart_gc) {
1849                 if (f2fs_start_gc_thread(sbi))
1850                         f2fs_warn(sbi, "background gc thread has stopped");
1851         } else if (need_stop_gc) {
1852                 f2fs_stop_gc_thread(sbi);
1853         }
1854 restore_opts:
1855 #ifdef CONFIG_QUOTA
1856         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1857         for (i = 0; i < MAXQUOTAS; i++) {
1858                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1859                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1860         }
1861 #endif
1862         sbi->mount_opt = org_mount_opt;
1863         sb->s_flags = old_sb_flags;
1864         return err;
1865 }
1866
1867 #ifdef CONFIG_QUOTA
1868 /* Read data from quotafile */
1869 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1870                                size_t len, loff_t off)
1871 {
1872         struct inode *inode = sb_dqopt(sb)->files[type];
1873         struct address_space *mapping = inode->i_mapping;
1874         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1875         int offset = off & (sb->s_blocksize - 1);
1876         int tocopy;
1877         size_t toread;
1878         loff_t i_size = i_size_read(inode);
1879         struct page *page;
1880         char *kaddr;
1881
1882         if (off > i_size)
1883                 return 0;
1884
1885         if (off + len > i_size)
1886                 len = i_size - off;
1887         toread = len;
1888         while (toread > 0) {
1889                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1890 repeat:
1891                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1892                 if (IS_ERR(page)) {
1893                         if (PTR_ERR(page) == -ENOMEM) {
1894                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1895                                 goto repeat;
1896                         }
1897                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1898                         return PTR_ERR(page);
1899                 }
1900
1901                 lock_page(page);
1902
1903                 if (unlikely(page->mapping != mapping)) {
1904                         f2fs_put_page(page, 1);
1905                         goto repeat;
1906                 }
1907                 if (unlikely(!PageUptodate(page))) {
1908                         f2fs_put_page(page, 1);
1909                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1910                         return -EIO;
1911                 }
1912
1913                 kaddr = kmap_atomic(page);
1914                 memcpy(data, kaddr + offset, tocopy);
1915                 kunmap_atomic(kaddr);
1916                 f2fs_put_page(page, 1);
1917
1918                 offset = 0;
1919                 toread -= tocopy;
1920                 data += tocopy;
1921                 blkidx++;
1922         }
1923         return len;
1924 }
1925
1926 /* Write to quotafile */
1927 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1928                                 const char *data, size_t len, loff_t off)
1929 {
1930         struct inode *inode = sb_dqopt(sb)->files[type];
1931         struct address_space *mapping = inode->i_mapping;
1932         const struct address_space_operations *a_ops = mapping->a_ops;
1933         int offset = off & (sb->s_blocksize - 1);
1934         size_t towrite = len;
1935         struct page *page;
1936         char *kaddr;
1937         int err = 0;
1938         int tocopy;
1939
1940         while (towrite > 0) {
1941                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1942                                                                 towrite);
1943 retry:
1944                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1945                                                         &page, NULL);
1946                 if (unlikely(err)) {
1947                         if (err == -ENOMEM) {
1948                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1949                                 goto retry;
1950                         }
1951                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1952                         break;
1953                 }
1954
1955                 kaddr = kmap_atomic(page);
1956                 memcpy(kaddr + offset, data, tocopy);
1957                 kunmap_atomic(kaddr);
1958                 flush_dcache_page(page);
1959
1960                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1961                                                 page, NULL);
1962                 offset = 0;
1963                 towrite -= tocopy;
1964                 off += tocopy;
1965                 data += tocopy;
1966                 cond_resched();
1967         }
1968
1969         if (len == towrite)
1970                 return err;
1971         inode->i_mtime = inode->i_ctime = current_time(inode);
1972         f2fs_mark_inode_dirty_sync(inode, false);
1973         return len - towrite;
1974 }
1975
1976 static struct dquot **f2fs_get_dquots(struct inode *inode)
1977 {
1978         return F2FS_I(inode)->i_dquot;
1979 }
1980
1981 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1982 {
1983         return &F2FS_I(inode)->i_reserved_quota;
1984 }
1985
1986 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1987 {
1988         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1989                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1990                 return 0;
1991         }
1992
1993         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1994                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1995 }
1996
1997 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1998 {
1999         int enabled = 0;
2000         int i, err;
2001
2002         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2003                 err = f2fs_enable_quotas(sbi->sb);
2004                 if (err) {
2005                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2006                         return 0;
2007                 }
2008                 return 1;
2009         }
2010
2011         for (i = 0; i < MAXQUOTAS; i++) {
2012                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2013                         err = f2fs_quota_on_mount(sbi, i);
2014                         if (!err) {
2015                                 enabled = 1;
2016                                 continue;
2017                         }
2018                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2019                                  err, i);
2020                 }
2021         }
2022         return enabled;
2023 }
2024
2025 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2026                              unsigned int flags)
2027 {
2028         struct inode *qf_inode;
2029         unsigned long qf_inum;
2030         int err;
2031
2032         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2033
2034         qf_inum = f2fs_qf_ino(sb, type);
2035         if (!qf_inum)
2036                 return -EPERM;
2037
2038         qf_inode = f2fs_iget(sb, qf_inum);
2039         if (IS_ERR(qf_inode)) {
2040                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2041                 return PTR_ERR(qf_inode);
2042         }
2043
2044         /* Don't account quota for quota files to avoid recursion */
2045         qf_inode->i_flags |= S_NOQUOTA;
2046         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2047         iput(qf_inode);
2048         return err;
2049 }
2050
2051 static int f2fs_enable_quotas(struct super_block *sb)
2052 {
2053         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2054         int type, err = 0;
2055         unsigned long qf_inum;
2056         bool quota_mopt[MAXQUOTAS] = {
2057                 test_opt(sbi, USRQUOTA),
2058                 test_opt(sbi, GRPQUOTA),
2059                 test_opt(sbi, PRJQUOTA),
2060         };
2061
2062         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2063                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2064                 return 0;
2065         }
2066
2067         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2068
2069         for (type = 0; type < MAXQUOTAS; type++) {
2070                 qf_inum = f2fs_qf_ino(sb, type);
2071                 if (qf_inum) {
2072                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2073                                 DQUOT_USAGE_ENABLED |
2074                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2075                         if (err) {
2076                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2077                                          type, err);
2078                                 for (type--; type >= 0; type--)
2079                                         dquot_quota_off(sb, type);
2080                                 set_sbi_flag(F2FS_SB(sb),
2081                                                 SBI_QUOTA_NEED_REPAIR);
2082                                 return err;
2083                         }
2084                 }
2085         }
2086         return 0;
2087 }
2088
2089 int f2fs_quota_sync(struct super_block *sb, int type)
2090 {
2091         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2092         struct quota_info *dqopt = sb_dqopt(sb);
2093         int cnt;
2094         int ret;
2095
2096         /*
2097          * do_quotactl
2098          *  f2fs_quota_sync
2099          *  down_read(quota_sem)
2100          *  dquot_writeback_dquots()
2101          *  f2fs_dquot_commit
2102          *                            block_operation
2103          *                            down_read(quota_sem)
2104          */
2105         f2fs_lock_op(sbi);
2106
2107         down_read(&sbi->quota_sem);
2108         ret = dquot_writeback_dquots(sb, type);
2109         if (ret)
2110                 goto out;
2111
2112         /*
2113          * Now when everything is written we can discard the pagecache so
2114          * that userspace sees the changes.
2115          */
2116         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2117                 struct address_space *mapping;
2118
2119                 if (type != -1 && cnt != type)
2120                         continue;
2121                 if (!sb_has_quota_active(sb, cnt))
2122                         continue;
2123
2124                 mapping = dqopt->files[cnt]->i_mapping;
2125
2126                 ret = filemap_fdatawrite(mapping);
2127                 if (ret)
2128                         goto out;
2129
2130                 /* if we are using journalled quota */
2131                 if (is_journalled_quota(sbi))
2132                         continue;
2133
2134                 ret = filemap_fdatawait(mapping);
2135                 if (ret)
2136                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2137
2138                 inode_lock(dqopt->files[cnt]);
2139                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2140                 inode_unlock(dqopt->files[cnt]);
2141         }
2142 out:
2143         if (ret)
2144                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2145         up_read(&sbi->quota_sem);
2146         f2fs_unlock_op(sbi);
2147         return ret;
2148 }
2149
2150 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2151                                                         const struct path *path)
2152 {
2153         struct inode *inode;
2154         int err;
2155
2156         /* if quota sysfile exists, deny enabling quota with specific file */
2157         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2158                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2159                 return -EBUSY;
2160         }
2161
2162         err = f2fs_quota_sync(sb, type);
2163         if (err)
2164                 return err;
2165
2166         err = dquot_quota_on(sb, type, format_id, path);
2167         if (err)
2168                 return err;
2169
2170         inode = d_inode(path->dentry);
2171
2172         inode_lock(inode);
2173         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2174         f2fs_set_inode_flags(inode);
2175         inode_unlock(inode);
2176         f2fs_mark_inode_dirty_sync(inode, false);
2177
2178         return 0;
2179 }
2180
2181 static int __f2fs_quota_off(struct super_block *sb, int type)
2182 {
2183         struct inode *inode = sb_dqopt(sb)->files[type];
2184         int err;
2185
2186         if (!inode || !igrab(inode))
2187                 return dquot_quota_off(sb, type);
2188
2189         err = f2fs_quota_sync(sb, type);
2190         if (err)
2191                 goto out_put;
2192
2193         err = dquot_quota_off(sb, type);
2194         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2195                 goto out_put;
2196
2197         inode_lock(inode);
2198         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2199         f2fs_set_inode_flags(inode);
2200         inode_unlock(inode);
2201         f2fs_mark_inode_dirty_sync(inode, false);
2202 out_put:
2203         iput(inode);
2204         return err;
2205 }
2206
2207 static int f2fs_quota_off(struct super_block *sb, int type)
2208 {
2209         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2210         int err;
2211
2212         err = __f2fs_quota_off(sb, type);
2213
2214         /*
2215          * quotactl can shutdown journalled quota, result in inconsistence
2216          * between quota record and fs data by following updates, tag the
2217          * flag to let fsck be aware of it.
2218          */
2219         if (is_journalled_quota(sbi))
2220                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2221         return err;
2222 }
2223
2224 void f2fs_quota_off_umount(struct super_block *sb)
2225 {
2226         int type;
2227         int err;
2228
2229         for (type = 0; type < MAXQUOTAS; type++) {
2230                 err = __f2fs_quota_off(sb, type);
2231                 if (err) {
2232                         int ret = dquot_quota_off(sb, type);
2233
2234                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2235                                  type, err, ret);
2236                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2237                 }
2238         }
2239         /*
2240          * In case of checkpoint=disable, we must flush quota blocks.
2241          * This can cause NULL exception for node_inode in end_io, since
2242          * put_super already dropped it.
2243          */
2244         sync_filesystem(sb);
2245 }
2246
2247 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2248 {
2249         struct quota_info *dqopt = sb_dqopt(sb);
2250         int type;
2251
2252         for (type = 0; type < MAXQUOTAS; type++) {
2253                 if (!dqopt->files[type])
2254                         continue;
2255                 f2fs_inode_synced(dqopt->files[type]);
2256         }
2257 }
2258
2259 static int f2fs_dquot_commit(struct dquot *dquot)
2260 {
2261         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2262         int ret;
2263
2264         down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2265         ret = dquot_commit(dquot);
2266         if (ret < 0)
2267                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2268         up_read(&sbi->quota_sem);
2269         return ret;
2270 }
2271
2272 static int f2fs_dquot_acquire(struct dquot *dquot)
2273 {
2274         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2275         int ret;
2276
2277         down_read(&sbi->quota_sem);
2278         ret = dquot_acquire(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_release(struct dquot *dquot)
2286 {
2287         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2288         int ret = dquot_release(dquot);
2289
2290         if (ret < 0)
2291                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2292         return ret;
2293 }
2294
2295 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2296 {
2297         struct super_block *sb = dquot->dq_sb;
2298         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2299         int ret = dquot_mark_dquot_dirty(dquot);
2300
2301         /* if we are using journalled quota */
2302         if (is_journalled_quota(sbi))
2303                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2304
2305         return ret;
2306 }
2307
2308 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2309 {
2310         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2311         int ret = dquot_commit_info(sb, type);
2312
2313         if (ret < 0)
2314                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2315         return ret;
2316 }
2317
2318 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2319 {
2320         *projid = F2FS_I(inode)->i_projid;
2321         return 0;
2322 }
2323
2324 static const struct dquot_operations f2fs_quota_operations = {
2325         .get_reserved_space = f2fs_get_reserved_space,
2326         .write_dquot    = f2fs_dquot_commit,
2327         .acquire_dquot  = f2fs_dquot_acquire,
2328         .release_dquot  = f2fs_dquot_release,
2329         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2330         .write_info     = f2fs_dquot_commit_info,
2331         .alloc_dquot    = dquot_alloc,
2332         .destroy_dquot  = dquot_destroy,
2333         .get_projid     = f2fs_get_projid,
2334         .get_next_id    = dquot_get_next_id,
2335 };
2336
2337 static const struct quotactl_ops f2fs_quotactl_ops = {
2338         .quota_on       = f2fs_quota_on,
2339         .quota_off      = f2fs_quota_off,
2340         .quota_sync     = f2fs_quota_sync,
2341         .get_state      = dquot_get_state,
2342         .set_info       = dquot_set_dqinfo,
2343         .get_dqblk      = dquot_get_dqblk,
2344         .set_dqblk      = dquot_set_dqblk,
2345         .get_nextdqblk  = dquot_get_next_dqblk,
2346 };
2347 #else
2348 int f2fs_quota_sync(struct super_block *sb, int type)
2349 {
2350         return 0;
2351 }
2352
2353 void f2fs_quota_off_umount(struct super_block *sb)
2354 {
2355 }
2356 #endif
2357
2358 static const struct super_operations f2fs_sops = {
2359         .alloc_inode    = f2fs_alloc_inode,
2360         .free_inode     = f2fs_free_inode,
2361         .drop_inode     = f2fs_drop_inode,
2362         .write_inode    = f2fs_write_inode,
2363         .dirty_inode    = f2fs_dirty_inode,
2364         .show_options   = f2fs_show_options,
2365 #ifdef CONFIG_QUOTA
2366         .quota_read     = f2fs_quota_read,
2367         .quota_write    = f2fs_quota_write,
2368         .get_dquots     = f2fs_get_dquots,
2369 #endif
2370         .evict_inode    = f2fs_evict_inode,
2371         .put_super      = f2fs_put_super,
2372         .sync_fs        = f2fs_sync_fs,
2373         .freeze_fs      = f2fs_freeze,
2374         .unfreeze_fs    = f2fs_unfreeze,
2375         .statfs         = f2fs_statfs,
2376         .remount_fs     = f2fs_remount,
2377 };
2378
2379 #ifdef CONFIG_FS_ENCRYPTION
2380 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2381 {
2382         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2383                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2384                                 ctx, len, NULL);
2385 }
2386
2387 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2388                                                         void *fs_data)
2389 {
2390         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2391
2392         /*
2393          * Encrypting the root directory is not allowed because fsck
2394          * expects lost+found directory to exist and remain unencrypted
2395          * if LOST_FOUND feature is enabled.
2396          *
2397          */
2398         if (f2fs_sb_has_lost_found(sbi) &&
2399                         inode->i_ino == F2FS_ROOT_INO(sbi))
2400                 return -EPERM;
2401
2402         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2403                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2404                                 ctx, len, fs_data, XATTR_CREATE);
2405 }
2406
2407 static bool f2fs_dummy_context(struct inode *inode)
2408 {
2409         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2410 }
2411
2412 static bool f2fs_has_stable_inodes(struct super_block *sb)
2413 {
2414         return true;
2415 }
2416
2417 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2418                                        int *ino_bits_ret, int *lblk_bits_ret)
2419 {
2420         *ino_bits_ret = 8 * sizeof(nid_t);
2421         *lblk_bits_ret = 8 * sizeof(block_t);
2422 }
2423
2424 static const struct fscrypt_operations f2fs_cryptops = {
2425         .key_prefix             = "f2fs:",
2426         .get_context            = f2fs_get_context,
2427         .set_context            = f2fs_set_context,
2428         .dummy_context          = f2fs_dummy_context,
2429         .empty_dir              = f2fs_empty_dir,
2430         .max_namelen            = F2FS_NAME_LEN,
2431         .has_stable_inodes      = f2fs_has_stable_inodes,
2432         .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
2433 };
2434 #endif
2435
2436 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2437                 u64 ino, u32 generation)
2438 {
2439         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2440         struct inode *inode;
2441
2442         if (f2fs_check_nid_range(sbi, ino))
2443                 return ERR_PTR(-ESTALE);
2444
2445         /*
2446          * f2fs_iget isn't quite right if the inode is currently unallocated!
2447          * However f2fs_iget currently does appropriate checks to handle stale
2448          * inodes so everything is OK.
2449          */
2450         inode = f2fs_iget(sb, ino);
2451         if (IS_ERR(inode))
2452                 return ERR_CAST(inode);
2453         if (unlikely(generation && inode->i_generation != generation)) {
2454                 /* we didn't find the right inode.. */
2455                 iput(inode);
2456                 return ERR_PTR(-ESTALE);
2457         }
2458         return inode;
2459 }
2460
2461 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2462                 int fh_len, int fh_type)
2463 {
2464         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2465                                     f2fs_nfs_get_inode);
2466 }
2467
2468 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2469                 int fh_len, int fh_type)
2470 {
2471         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2472                                     f2fs_nfs_get_inode);
2473 }
2474
2475 static const struct export_operations f2fs_export_ops = {
2476         .fh_to_dentry = f2fs_fh_to_dentry,
2477         .fh_to_parent = f2fs_fh_to_parent,
2478         .get_parent = f2fs_get_parent,
2479 };
2480
2481 static loff_t max_file_blocks(void)
2482 {
2483         loff_t result = 0;
2484         loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2485
2486         /*
2487          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2488          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2489          * space in inode.i_addr, it will be more safe to reassign
2490          * result as zero.
2491          */
2492
2493         /* two direct node blocks */
2494         result += (leaf_count * 2);
2495
2496         /* two indirect node blocks */
2497         leaf_count *= NIDS_PER_BLOCK;
2498         result += (leaf_count * 2);
2499
2500         /* one double indirect node block */
2501         leaf_count *= NIDS_PER_BLOCK;
2502         result += leaf_count;
2503
2504         return result;
2505 }
2506
2507 static int __f2fs_commit_super(struct buffer_head *bh,
2508                         struct f2fs_super_block *super)
2509 {
2510         lock_buffer(bh);
2511         if (super)
2512                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2513         set_buffer_dirty(bh);
2514         unlock_buffer(bh);
2515
2516         /* it's rare case, we can do fua all the time */
2517         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2518 }
2519
2520 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2521                                         struct buffer_head *bh)
2522 {
2523         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2524                                         (bh->b_data + F2FS_SUPER_OFFSET);
2525         struct super_block *sb = sbi->sb;
2526         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2527         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2528         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2529         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2530         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2531         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2532         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2533         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2534         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2535         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2536         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2537         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2538         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2539         u64 main_end_blkaddr = main_blkaddr +
2540                                 (segment_count_main << log_blocks_per_seg);
2541         u64 seg_end_blkaddr = segment0_blkaddr +
2542                                 (segment_count << log_blocks_per_seg);
2543
2544         if (segment0_blkaddr != cp_blkaddr) {
2545                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2546                           segment0_blkaddr, cp_blkaddr);
2547                 return true;
2548         }
2549
2550         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2551                                                         sit_blkaddr) {
2552                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2553                           cp_blkaddr, sit_blkaddr,
2554                           segment_count_ckpt << log_blocks_per_seg);
2555                 return true;
2556         }
2557
2558         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2559                                                         nat_blkaddr) {
2560                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2561                           sit_blkaddr, nat_blkaddr,
2562                           segment_count_sit << log_blocks_per_seg);
2563                 return true;
2564         }
2565
2566         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2567                                                         ssa_blkaddr) {
2568                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2569                           nat_blkaddr, ssa_blkaddr,
2570                           segment_count_nat << log_blocks_per_seg);
2571                 return true;
2572         }
2573
2574         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2575                                                         main_blkaddr) {
2576                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2577                           ssa_blkaddr, main_blkaddr,
2578                           segment_count_ssa << log_blocks_per_seg);
2579                 return true;
2580         }
2581
2582         if (main_end_blkaddr > seg_end_blkaddr) {
2583                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2584                           main_blkaddr,
2585                           segment0_blkaddr +
2586                           (segment_count << log_blocks_per_seg),
2587                           segment_count_main << log_blocks_per_seg);
2588                 return true;
2589         } else if (main_end_blkaddr < seg_end_blkaddr) {
2590                 int err = 0;
2591                 char *res;
2592
2593                 /* fix in-memory information all the time */
2594                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2595                                 segment0_blkaddr) >> log_blocks_per_seg);
2596
2597                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2598                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2599                         res = "internally";
2600                 } else {
2601                         err = __f2fs_commit_super(bh, NULL);
2602                         res = err ? "failed" : "done";
2603                 }
2604                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2605                           res, main_blkaddr,
2606                           segment0_blkaddr +
2607                           (segment_count << log_blocks_per_seg),
2608                           segment_count_main << log_blocks_per_seg);
2609                 if (err)
2610                         return true;
2611         }
2612         return false;
2613 }
2614
2615 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2616                                 struct buffer_head *bh)
2617 {
2618         block_t segment_count, segs_per_sec, secs_per_zone;
2619         block_t total_sections, blocks_per_seg;
2620         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2621                                         (bh->b_data + F2FS_SUPER_OFFSET);
2622         unsigned int blocksize;
2623         size_t crc_offset = 0;
2624         __u32 crc = 0;
2625
2626         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2627                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2628                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2629                 return -EINVAL;
2630         }
2631
2632         /* Check checksum_offset and crc in superblock */
2633         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2634                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2635                 if (crc_offset !=
2636                         offsetof(struct f2fs_super_block, crc)) {
2637                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2638                                   crc_offset);
2639                         return -EFSCORRUPTED;
2640                 }
2641                 crc = le32_to_cpu(raw_super->crc);
2642                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2643                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2644                         return -EFSCORRUPTED;
2645                 }
2646         }
2647
2648         /* Currently, support only 4KB page cache size */
2649         if (F2FS_BLKSIZE != PAGE_SIZE) {
2650                 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2651                           PAGE_SIZE);
2652                 return -EFSCORRUPTED;
2653         }
2654
2655         /* Currently, support only 4KB block size */
2656         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2657         if (blocksize != F2FS_BLKSIZE) {
2658                 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2659                           blocksize);
2660                 return -EFSCORRUPTED;
2661         }
2662
2663         /* check log blocks per segment */
2664         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2665                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2666                           le32_to_cpu(raw_super->log_blocks_per_seg));
2667                 return -EFSCORRUPTED;
2668         }
2669
2670         /* Currently, support 512/1024/2048/4096 bytes sector size */
2671         if (le32_to_cpu(raw_super->log_sectorsize) >
2672                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2673                 le32_to_cpu(raw_super->log_sectorsize) <
2674                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2675                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2676                           le32_to_cpu(raw_super->log_sectorsize));
2677                 return -EFSCORRUPTED;
2678         }
2679         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2680                 le32_to_cpu(raw_super->log_sectorsize) !=
2681                         F2FS_MAX_LOG_SECTOR_SIZE) {
2682                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2683                           le32_to_cpu(raw_super->log_sectors_per_block),
2684                           le32_to_cpu(raw_super->log_sectorsize));
2685                 return -EFSCORRUPTED;
2686         }
2687
2688         segment_count = le32_to_cpu(raw_super->segment_count);
2689         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2690         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2691         total_sections = le32_to_cpu(raw_super->section_count);
2692
2693         /* blocks_per_seg should be 512, given the above check */
2694         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2695
2696         if (segment_count > F2FS_MAX_SEGMENT ||
2697                                 segment_count < F2FS_MIN_SEGMENTS) {
2698                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2699                 return -EFSCORRUPTED;
2700         }
2701
2702         if (total_sections > segment_count ||
2703                         total_sections < F2FS_MIN_SEGMENTS ||
2704                         segs_per_sec > segment_count || !segs_per_sec) {
2705                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2706                           segment_count, total_sections, segs_per_sec);
2707                 return -EFSCORRUPTED;
2708         }
2709
2710         if ((segment_count / segs_per_sec) < total_sections) {
2711                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2712                           segment_count, segs_per_sec, total_sections);
2713                 return -EFSCORRUPTED;
2714         }
2715
2716         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2717                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2718                           segment_count, le64_to_cpu(raw_super->block_count));
2719                 return -EFSCORRUPTED;
2720         }
2721
2722         if (RDEV(0).path[0]) {
2723                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2724                 int i = 1;
2725
2726                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2727                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2728                         i++;
2729                 }
2730                 if (segment_count != dev_seg_count) {
2731                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2732                                         segment_count, dev_seg_count);
2733                         return -EFSCORRUPTED;
2734                 }
2735         }
2736
2737         if (secs_per_zone > total_sections || !secs_per_zone) {
2738                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2739                           secs_per_zone, total_sections);
2740                 return -EFSCORRUPTED;
2741         }
2742         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2743                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2744                         (le32_to_cpu(raw_super->extension_count) +
2745                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2746                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2747                           le32_to_cpu(raw_super->extension_count),
2748                           raw_super->hot_ext_count,
2749                           F2FS_MAX_EXTENSION);
2750                 return -EFSCORRUPTED;
2751         }
2752
2753         if (le32_to_cpu(raw_super->cp_payload) >
2754                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2755                 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2756                           le32_to_cpu(raw_super->cp_payload),
2757                           blocks_per_seg - F2FS_CP_PACKS);
2758                 return -EFSCORRUPTED;
2759         }
2760
2761         /* check reserved ino info */
2762         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2763                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2764                 le32_to_cpu(raw_super->root_ino) != 3) {
2765                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2766                           le32_to_cpu(raw_super->node_ino),
2767                           le32_to_cpu(raw_super->meta_ino),
2768                           le32_to_cpu(raw_super->root_ino));
2769                 return -EFSCORRUPTED;
2770         }
2771
2772         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2773         if (sanity_check_area_boundary(sbi, bh))
2774                 return -EFSCORRUPTED;
2775
2776         return 0;
2777 }
2778
2779 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2780 {
2781         unsigned int total, fsmeta;
2782         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2783         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2784         unsigned int ovp_segments, reserved_segments;
2785         unsigned int main_segs, blocks_per_seg;
2786         unsigned int sit_segs, nat_segs;
2787         unsigned int sit_bitmap_size, nat_bitmap_size;
2788         unsigned int log_blocks_per_seg;
2789         unsigned int segment_count_main;
2790         unsigned int cp_pack_start_sum, cp_payload;
2791         block_t user_block_count, valid_user_blocks;
2792         block_t avail_node_count, valid_node_count;
2793         int i, j;
2794
2795         total = le32_to_cpu(raw_super->segment_count);
2796         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2797         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2798         fsmeta += sit_segs;
2799         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2800         fsmeta += nat_segs;
2801         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2802         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2803
2804         if (unlikely(fsmeta >= total))
2805                 return 1;
2806
2807         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2808         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2809
2810         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2811                         ovp_segments == 0 || reserved_segments == 0)) {
2812                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2813                 return 1;
2814         }
2815
2816         user_block_count = le64_to_cpu(ckpt->user_block_count);
2817         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2818         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2819         if (!user_block_count || user_block_count >=
2820                         segment_count_main << log_blocks_per_seg) {
2821                 f2fs_err(sbi, "Wrong user_block_count: %u",
2822                          user_block_count);
2823                 return 1;
2824         }
2825
2826         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2827         if (valid_user_blocks > user_block_count) {
2828                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2829                          valid_user_blocks, user_block_count);
2830                 return 1;
2831         }
2832
2833         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2834         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2835         if (valid_node_count > avail_node_count) {
2836                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2837                          valid_node_count, avail_node_count);
2838                 return 1;
2839         }
2840
2841         main_segs = le32_to_cpu(raw_super->segment_count_main);
2842         blocks_per_seg = sbi->blocks_per_seg;
2843
2844         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2845                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2846                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2847                         return 1;
2848                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2849                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2850                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2851                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2852                                          i, j,
2853                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2854                                 return 1;
2855                         }
2856                 }
2857         }
2858         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2859                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2860                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2861                         return 1;
2862                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2863                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2864                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2865                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2866                                          i, j,
2867                                          le32_to_cpu(ckpt->cur_data_segno[i]));
2868                                 return 1;
2869                         }
2870                 }
2871         }
2872         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2873                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2874                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2875                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2876                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2877                                          i, j,
2878                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2879                                 return 1;
2880                         }
2881                 }
2882         }
2883
2884         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2885         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2886
2887         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2888                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2889                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2890                          sit_bitmap_size, nat_bitmap_size);
2891                 return 1;
2892         }
2893
2894         cp_pack_start_sum = __start_sum_addr(sbi);
2895         cp_payload = __cp_payload(sbi);
2896         if (cp_pack_start_sum < cp_payload + 1 ||
2897                 cp_pack_start_sum > blocks_per_seg - 1 -
2898                         NR_CURSEG_TYPE) {
2899                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2900                          cp_pack_start_sum);
2901                 return 1;
2902         }
2903
2904         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2905                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2906                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2907                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2908                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2909                           le32_to_cpu(ckpt->checksum_offset));
2910                 return 1;
2911         }
2912
2913         if (unlikely(f2fs_cp_error(sbi))) {
2914                 f2fs_err(sbi, "A bug case: need to run fsck");
2915                 return 1;
2916         }
2917         return 0;
2918 }
2919
2920 static void init_sb_info(struct f2fs_sb_info *sbi)
2921 {
2922         struct f2fs_super_block *raw_super = sbi->raw_super;
2923         int i;
2924
2925         sbi->log_sectors_per_block =
2926                 le32_to_cpu(raw_super->log_sectors_per_block);
2927         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2928         sbi->blocksize = 1 << sbi->log_blocksize;
2929         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2930         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2931         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2932         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2933         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2934         sbi->total_node_count =
2935                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2936                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2937         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2938         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2939         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2940         sbi->cur_victim_sec = NULL_SECNO;
2941         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2942         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2943         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2944         sbi->migration_granularity = sbi->segs_per_sec;
2945
2946         sbi->dir_level = DEF_DIR_LEVEL;
2947         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2948         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2949         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2950         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2951         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2952         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2953                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2954         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2955
2956         for (i = 0; i < NR_COUNT_TYPE; i++)
2957                 atomic_set(&sbi->nr_pages[i], 0);
2958
2959         for (i = 0; i < META; i++)
2960                 atomic_set(&sbi->wb_sync_req[i], 0);
2961
2962         INIT_LIST_HEAD(&sbi->s_list);
2963         mutex_init(&sbi->umount_mutex);
2964         init_rwsem(&sbi->io_order_lock);
2965         spin_lock_init(&sbi->cp_lock);
2966
2967         sbi->dirty_device = 0;
2968         spin_lock_init(&sbi->dev_lock);
2969
2970         init_rwsem(&sbi->sb_lock);
2971         init_rwsem(&sbi->pin_sem);
2972 }
2973
2974 static int init_percpu_info(struct f2fs_sb_info *sbi)
2975 {
2976         int err;
2977
2978         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2979         if (err)
2980                 return err;
2981
2982         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2983                                                                 GFP_KERNEL);
2984         if (err)
2985                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2986
2987         return err;
2988 }
2989
2990 #ifdef CONFIG_BLK_DEV_ZONED
2991 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
2992                                void *data)
2993 {
2994         struct f2fs_dev_info *dev = data;
2995
2996         if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
2997                 set_bit(idx, dev->blkz_seq);
2998         return 0;
2999 }
3000
3001 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3002 {
3003         struct block_device *bdev = FDEV(devi).bdev;
3004         sector_t nr_sectors = bdev->bd_part->nr_sects;
3005         int ret;
3006
3007         if (!f2fs_sb_has_blkzoned(sbi))
3008                 return 0;
3009
3010         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3011                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3012                 return -EINVAL;
3013         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3014         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3015                                 __ilog2_u32(sbi->blocks_per_blkz))
3016                 return -EINVAL;
3017         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3018         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3019                                         sbi->log_blocks_per_blkz;
3020         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3021                 FDEV(devi).nr_blkz++;
3022
3023         FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
3024                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
3025                                         * sizeof(unsigned long),
3026                                         GFP_KERNEL);
3027         if (!FDEV(devi).blkz_seq)
3028                 return -ENOMEM;
3029
3030         /* Get block zones type */
3031         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3032                                   &FDEV(devi));
3033         if (ret < 0)
3034                 return ret;
3035
3036         return 0;
3037 }
3038 #endif
3039
3040 /*
3041  * Read f2fs raw super block.
3042  * Because we have two copies of super block, so read both of them
3043  * to get the first valid one. If any one of them is broken, we pass
3044  * them recovery flag back to the caller.
3045  */
3046 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3047                         struct f2fs_super_block **raw_super,
3048                         int *valid_super_block, int *recovery)
3049 {
3050         struct super_block *sb = sbi->sb;
3051         int block;
3052         struct buffer_head *bh;
3053         struct f2fs_super_block *super;
3054         int err = 0;
3055
3056         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3057         if (!super)
3058                 return -ENOMEM;
3059
3060         for (block = 0; block < 2; block++) {
3061                 bh = sb_bread(sb, block);
3062                 if (!bh) {
3063                         f2fs_err(sbi, "Unable to read %dth superblock",
3064                                  block + 1);
3065                         err = -EIO;
3066                         *recovery = 1;
3067                         continue;
3068                 }
3069
3070                 /* sanity checking of raw super */
3071                 err = sanity_check_raw_super(sbi, bh);
3072                 if (err) {
3073                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3074                                  block + 1);
3075                         brelse(bh);
3076                         *recovery = 1;
3077                         continue;
3078                 }
3079
3080                 if (!*raw_super) {
3081                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3082                                                         sizeof(*super));
3083                         *valid_super_block = block;
3084                         *raw_super = super;
3085                 }
3086                 brelse(bh);
3087         }
3088
3089         /* No valid superblock */
3090         if (!*raw_super)
3091                 kvfree(super);
3092         else
3093                 err = 0;
3094
3095         return err;
3096 }
3097
3098 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3099 {
3100         struct buffer_head *bh;
3101         __u32 crc = 0;
3102         int err;
3103
3104         if ((recover && f2fs_readonly(sbi->sb)) ||
3105                                 bdev_read_only(sbi->sb->s_bdev)) {
3106                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3107                 return -EROFS;
3108         }
3109
3110         /* we should update superblock crc here */
3111         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3112                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3113                                 offsetof(struct f2fs_super_block, crc));
3114                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3115         }
3116
3117         /* write back-up superblock first */
3118         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3119         if (!bh)
3120                 return -EIO;
3121         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3122         brelse(bh);
3123
3124         /* if we are in recovery path, skip writing valid superblock */
3125         if (recover || err)
3126                 return err;
3127
3128         /* write current valid superblock */
3129         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3130         if (!bh)
3131                 return -EIO;
3132         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3133         brelse(bh);
3134         return err;
3135 }
3136
3137 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3138 {
3139         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3140         unsigned int max_devices = MAX_DEVICES;
3141         int i;
3142
3143         /* Initialize single device information */
3144         if (!RDEV(0).path[0]) {
3145                 if (!bdev_is_zoned(sbi->sb->s_bdev))
3146                         return 0;
3147                 max_devices = 1;
3148         }
3149
3150         /*
3151          * Initialize multiple devices information, or single
3152          * zoned block device information.
3153          */
3154         sbi->devs = f2fs_kzalloc(sbi,
3155                                  array_size(max_devices,
3156                                             sizeof(struct f2fs_dev_info)),
3157                                  GFP_KERNEL);
3158         if (!sbi->devs)
3159                 return -ENOMEM;
3160
3161         for (i = 0; i < max_devices; i++) {
3162
3163                 if (i > 0 && !RDEV(i).path[0])
3164                         break;
3165
3166                 if (max_devices == 1) {
3167                         /* Single zoned block device mount */
3168                         FDEV(0).bdev =
3169                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3170                                         sbi->sb->s_mode, sbi->sb->s_type);
3171                 } else {
3172                         /* Multi-device mount */
3173                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3174                         FDEV(i).total_segments =
3175                                 le32_to_cpu(RDEV(i).total_segments);
3176                         if (i == 0) {
3177                                 FDEV(i).start_blk = 0;
3178                                 FDEV(i).end_blk = FDEV(i).start_blk +
3179                                     (FDEV(i).total_segments <<
3180                                     sbi->log_blocks_per_seg) - 1 +
3181                                     le32_to_cpu(raw_super->segment0_blkaddr);
3182                         } else {
3183                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3184                                 FDEV(i).end_blk = FDEV(i).start_blk +
3185                                         (FDEV(i).total_segments <<
3186                                         sbi->log_blocks_per_seg) - 1;
3187                         }
3188                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3189                                         sbi->sb->s_mode, sbi->sb->s_type);
3190                 }
3191                 if (IS_ERR(FDEV(i).bdev))
3192                         return PTR_ERR(FDEV(i).bdev);
3193
3194                 /* to release errored devices */
3195                 sbi->s_ndevs = i + 1;
3196
3197 #ifdef CONFIG_BLK_DEV_ZONED
3198                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3199                                 !f2fs_sb_has_blkzoned(sbi)) {
3200                         f2fs_err(sbi, "Zoned block device feature not enabled\n");
3201                         return -EINVAL;
3202                 }
3203                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3204                         if (init_blkz_info(sbi, i)) {
3205                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3206                                 return -EINVAL;
3207                         }
3208                         if (max_devices == 1)
3209                                 break;
3210                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3211                                   i, FDEV(i).path,
3212                                   FDEV(i).total_segments,
3213                                   FDEV(i).start_blk, FDEV(i).end_blk,
3214                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3215                                   "Host-aware" : "Host-managed");
3216                         continue;
3217                 }
3218 #endif
3219                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3220                           i, FDEV(i).path,
3221                           FDEV(i).total_segments,
3222                           FDEV(i).start_blk, FDEV(i).end_blk);
3223         }
3224         f2fs_info(sbi,
3225                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3226         return 0;
3227 }
3228
3229 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3230 {
3231 #ifdef CONFIG_UNICODE
3232         if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3233                 const struct f2fs_sb_encodings *encoding_info;
3234                 struct unicode_map *encoding;
3235                 __u16 encoding_flags;
3236
3237                 if (f2fs_sb_has_encrypt(sbi)) {
3238                         f2fs_err(sbi,
3239                                 "Can't mount with encoding and encryption");
3240                         return -EINVAL;
3241                 }
3242
3243                 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3244                                           &encoding_flags)) {
3245                         f2fs_err(sbi,
3246                                  "Encoding requested by superblock is unknown");
3247                         return -EINVAL;
3248                 }
3249
3250                 encoding = utf8_load(encoding_info->version);
3251                 if (IS_ERR(encoding)) {
3252                         f2fs_err(sbi,
3253                                  "can't mount with superblock charset: %s-%s "
3254                                  "not supported by the kernel. flags: 0x%x.",
3255                                  encoding_info->name, encoding_info->version,
3256                                  encoding_flags);
3257                         return PTR_ERR(encoding);
3258                 }
3259                 f2fs_info(sbi, "Using encoding defined by superblock: "
3260                          "%s-%s with flags 0x%hx", encoding_info->name,
3261                          encoding_info->version?:"\b", encoding_flags);
3262
3263                 sbi->s_encoding = encoding;
3264                 sbi->s_encoding_flags = encoding_flags;
3265                 sbi->sb->s_d_op = &f2fs_dentry_ops;
3266         }
3267 #else
3268         if (f2fs_sb_has_casefold(sbi)) {
3269                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3270                 return -EINVAL;
3271         }
3272 #endif
3273         return 0;
3274 }
3275
3276 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3277 {
3278         struct f2fs_sm_info *sm_i = SM_I(sbi);
3279
3280         /* adjust parameters according to the volume size */
3281         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3282                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3283                 sm_i->dcc_info->discard_granularity = 1;
3284                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3285         }
3286
3287         sbi->readdir_ra = 1;
3288 }
3289
3290 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3291 {
3292         struct f2fs_sb_info *sbi;
3293         struct f2fs_super_block *raw_super;
3294         struct inode *root;
3295         int err;
3296         bool skip_recovery = false, need_fsck = false;
3297         char *options = NULL;
3298         int recovery, i, valid_super_block;
3299         struct curseg_info *seg_i;
3300         int retry_cnt = 1;
3301
3302 try_onemore:
3303         err = -EINVAL;
3304         raw_super = NULL;
3305         valid_super_block = -1;
3306         recovery = 0;
3307
3308         /* allocate memory for f2fs-specific super block info */
3309         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3310         if (!sbi)
3311                 return -ENOMEM;
3312
3313         sbi->sb = sb;
3314
3315         /* Load the checksum driver */
3316         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3317         if (IS_ERR(sbi->s_chksum_driver)) {
3318                 f2fs_err(sbi, "Cannot load crc32 driver.");
3319                 err = PTR_ERR(sbi->s_chksum_driver);
3320                 sbi->s_chksum_driver = NULL;
3321                 goto free_sbi;
3322         }
3323
3324         /* set a block size */
3325         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3326                 f2fs_err(sbi, "unable to set blocksize");
3327                 goto free_sbi;
3328         }
3329
3330         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3331                                                                 &recovery);
3332         if (err)
3333                 goto free_sbi;
3334
3335         sb->s_fs_info = sbi;
3336         sbi->raw_super = raw_super;
3337
3338         /* precompute checksum seed for metadata */
3339         if (f2fs_sb_has_inode_chksum(sbi))
3340                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3341                                                 sizeof(raw_super->uuid));
3342
3343         /*
3344          * The BLKZONED feature indicates that the drive was formatted with
3345          * zone alignment optimization. This is optional for host-aware
3346          * devices, but mandatory for host-managed zoned block devices.
3347          */
3348 #ifndef CONFIG_BLK_DEV_ZONED
3349         if (f2fs_sb_has_blkzoned(sbi)) {
3350                 f2fs_err(sbi, "Zoned block device support is not enabled");
3351                 err = -EOPNOTSUPP;
3352                 goto free_sb_buf;
3353         }
3354 #endif
3355         default_options(sbi);
3356         /* parse mount options */
3357         options = kstrdup((const char *)data, GFP_KERNEL);
3358         if (data && !options) {
3359                 err = -ENOMEM;
3360                 goto free_sb_buf;
3361         }
3362
3363         err = parse_options(sb, options);
3364         if (err)
3365                 goto free_options;
3366
3367         sbi->max_file_blocks = max_file_blocks();
3368         sb->s_maxbytes = sbi->max_file_blocks <<
3369                                 le32_to_cpu(raw_super->log_blocksize);
3370         sb->s_max_links = F2FS_LINK_MAX;
3371
3372         err = f2fs_setup_casefold(sbi);
3373         if (err)
3374                 goto free_options;
3375
3376 #ifdef CONFIG_QUOTA
3377         sb->dq_op = &f2fs_quota_operations;
3378         sb->s_qcop = &f2fs_quotactl_ops;
3379         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3380
3381         if (f2fs_sb_has_quota_ino(sbi)) {
3382                 for (i = 0; i < MAXQUOTAS; i++) {
3383                         if (f2fs_qf_ino(sbi->sb, i))
3384                                 sbi->nquota_files++;
3385                 }
3386         }
3387 #endif
3388
3389         sb->s_op = &f2fs_sops;
3390 #ifdef CONFIG_FS_ENCRYPTION
3391         sb->s_cop = &f2fs_cryptops;
3392 #endif
3393 #ifdef CONFIG_FS_VERITY
3394         sb->s_vop = &f2fs_verityops;
3395 #endif
3396         sb->s_xattr = f2fs_xattr_handlers;
3397         sb->s_export_op = &f2fs_export_ops;
3398         sb->s_magic = F2FS_SUPER_MAGIC;
3399         sb->s_time_gran = 1;
3400         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3401                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3402         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3403         sb->s_iflags |= SB_I_CGROUPWB;
3404
3405         /* init f2fs-specific super block info */
3406         sbi->valid_super_block = valid_super_block;
3407         mutex_init(&sbi->gc_mutex);
3408         mutex_init(&sbi->writepages);
3409         mutex_init(&sbi->cp_mutex);
3410         mutex_init(&sbi->resize_mutex);
3411         init_rwsem(&sbi->node_write);
3412         init_rwsem(&sbi->node_change);
3413
3414         /* disallow all the data/node/meta page writes */
3415         set_sbi_flag(sbi, SBI_POR_DOING);
3416         spin_lock_init(&sbi->stat_lock);
3417
3418         /* init iostat info */
3419         spin_lock_init(&sbi->iostat_lock);
3420         sbi->iostat_enable = false;
3421
3422         for (i = 0; i < NR_PAGE_TYPE; i++) {
3423                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3424                 int j;
3425
3426                 sbi->write_io[i] =
3427                         f2fs_kmalloc(sbi,
3428                                      array_size(n,
3429                                                 sizeof(struct f2fs_bio_info)),
3430                                      GFP_KERNEL);
3431                 if (!sbi->write_io[i]) {
3432                         err = -ENOMEM;
3433                         goto free_bio_info;
3434                 }
3435
3436                 for (j = HOT; j < n; j++) {
3437                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3438                         sbi->write_io[i][j].sbi = sbi;
3439                         sbi->write_io[i][j].bio = NULL;
3440                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3441                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3442                         INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3443                         init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3444                 }
3445         }
3446
3447         init_rwsem(&sbi->cp_rwsem);
3448         init_rwsem(&sbi->quota_sem);
3449         init_waitqueue_head(&sbi->cp_wait);
3450         init_sb_info(sbi);
3451
3452         err = init_percpu_info(sbi);
3453         if (err)
3454                 goto free_bio_info;
3455
3456         if (F2FS_IO_ALIGNED(sbi)) {
3457                 sbi->write_io_dummy =
3458                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3459                 if (!sbi->write_io_dummy) {
3460                         err = -ENOMEM;
3461                         goto free_percpu;
3462                 }
3463         }
3464
3465         /* get an inode for meta space */
3466         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3467         if (IS_ERR(sbi->meta_inode)) {
3468                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3469                 err = PTR_ERR(sbi->meta_inode);
3470                 goto free_io_dummy;
3471         }
3472
3473         err = f2fs_get_valid_checkpoint(sbi);
3474         if (err) {
3475                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3476                 goto free_meta_inode;
3477         }
3478
3479         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3480                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3481         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3482                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3483                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3484         }
3485
3486         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3487                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3488
3489         /* Initialize device list */
3490         err = f2fs_scan_devices(sbi);
3491         if (err) {
3492                 f2fs_err(sbi, "Failed to find devices");
3493                 goto free_devices;
3494         }
3495
3496         err = f2fs_init_post_read_wq(sbi);
3497         if (err) {
3498                 f2fs_err(sbi, "Failed to initialize post read workqueue");
3499                 goto free_devices;
3500         }
3501
3502         sbi->total_valid_node_count =
3503                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3504         percpu_counter_set(&sbi->total_valid_inode_count,
3505                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3506         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3507         sbi->total_valid_block_count =
3508                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3509         sbi->last_valid_block_count = sbi->total_valid_block_count;
3510         sbi->reserved_blocks = 0;
3511         sbi->current_reserved_blocks = 0;
3512         limit_reserve_root(sbi);
3513
3514         for (i = 0; i < NR_INODE_TYPE; i++) {
3515                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3516                 spin_lock_init(&sbi->inode_lock[i]);
3517         }
3518         mutex_init(&sbi->flush_lock);
3519
3520         f2fs_init_extent_cache_info(sbi);
3521
3522         f2fs_init_ino_entry_info(sbi);
3523
3524         f2fs_init_fsync_node_info(sbi);
3525
3526         /* setup f2fs internal modules */
3527         err = f2fs_build_segment_manager(sbi);
3528         if (err) {
3529                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3530                          err);
3531                 goto free_sm;
3532         }
3533         err = f2fs_build_node_manager(sbi);
3534         if (err) {
3535                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3536                          err);
3537                 goto free_nm;
3538         }
3539
3540         /* For write statistics */
3541         if (sb->s_bdev->bd_part)
3542                 sbi->sectors_written_start =
3543                         (u64)part_stat_read(sb->s_bdev->bd_part,
3544                                             sectors[STAT_WRITE]);
3545
3546         /* Read accumulated write IO statistics if exists */
3547         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3548         if (__exist_node_summaries(sbi))
3549                 sbi->kbytes_written =
3550                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3551
3552         f2fs_build_gc_manager(sbi);
3553
3554         err = f2fs_build_stats(sbi);
3555         if (err)
3556                 goto free_nm;
3557
3558         /* get an inode for node space */
3559         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3560         if (IS_ERR(sbi->node_inode)) {
3561                 f2fs_err(sbi, "Failed to read node inode");
3562                 err = PTR_ERR(sbi->node_inode);
3563                 goto free_stats;
3564         }
3565
3566         /* read root inode and dentry */
3567         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3568         if (IS_ERR(root)) {
3569                 f2fs_err(sbi, "Failed to read root inode");
3570                 err = PTR_ERR(root);
3571                 goto free_node_inode;
3572         }
3573         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3574                         !root->i_size || !root->i_nlink) {
3575                 iput(root);
3576                 err = -EINVAL;
3577                 goto free_node_inode;
3578         }
3579
3580         sb->s_root = d_make_root(root); /* allocate root dentry */
3581         if (!sb->s_root) {
3582                 err = -ENOMEM;
3583                 goto free_node_inode;
3584         }
3585
3586         err = f2fs_register_sysfs(sbi);
3587         if (err)
3588                 goto free_root_inode;
3589
3590 #ifdef CONFIG_QUOTA
3591         /* Enable quota usage during mount */
3592         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3593                 err = f2fs_enable_quotas(sb);
3594                 if (err)
3595                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3596         }
3597 #endif
3598         /* if there are nt orphan nodes free them */
3599         err = f2fs_recover_orphan_inodes(sbi);
3600         if (err)
3601                 goto free_meta;
3602
3603         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3604                 goto reset_checkpoint;
3605
3606         /* recover fsynced data */
3607         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3608                 /*
3609                  * mount should be failed, when device has readonly mode, and
3610                  * previous checkpoint was not done by clean system shutdown.
3611                  */
3612                 if (f2fs_hw_is_readonly(sbi)) {
3613                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3614                                 err = -EROFS;
3615                                 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3616                                 goto free_meta;
3617                         }
3618                         f2fs_info(sbi, "write access unavailable, skipping recovery");
3619                         goto reset_checkpoint;
3620                 }
3621
3622                 if (need_fsck)
3623                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3624
3625                 if (skip_recovery)
3626                         goto reset_checkpoint;
3627
3628                 err = f2fs_recover_fsync_data(sbi, false);
3629                 if (err < 0) {
3630                         if (err != -ENOMEM)
3631                                 skip_recovery = true;
3632                         need_fsck = true;
3633                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3634                                  err);
3635                         goto free_meta;
3636                 }
3637         } else {
3638                 err = f2fs_recover_fsync_data(sbi, true);
3639
3640                 if (!f2fs_readonly(sb) && err > 0) {
3641                         err = -EINVAL;
3642                         f2fs_err(sbi, "Need to recover fsync data");
3643                         goto free_meta;
3644                 }
3645         }
3646
3647         /*
3648          * If the f2fs is not readonly and fsync data recovery succeeds,
3649          * check zoned block devices' write pointer consistency.
3650          */
3651         if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3652                 err = f2fs_check_write_pointer(sbi);
3653                 if (err)
3654                         goto free_meta;
3655         }
3656
3657 reset_checkpoint:
3658         /* f2fs_recover_fsync_data() cleared this already */
3659         clear_sbi_flag(sbi, SBI_POR_DOING);
3660
3661         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3662                 err = f2fs_disable_checkpoint(sbi);
3663                 if (err)
3664                         goto sync_free_meta;
3665         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3666                 f2fs_enable_checkpoint(sbi);
3667         }
3668
3669         /*
3670          * If filesystem is not mounted as read-only then
3671          * do start the gc_thread.
3672          */
3673         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3674                 /* After POR, we can run background GC thread.*/
3675                 err = f2fs_start_gc_thread(sbi);
3676                 if (err)
3677                         goto sync_free_meta;
3678         }
3679         kvfree(options);
3680
3681         /* recover broken superblock */
3682         if (recovery) {
3683                 err = f2fs_commit_super(sbi, true);
3684                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3685                           sbi->valid_super_block ? 1 : 2, err);
3686         }
3687
3688         f2fs_join_shrinker(sbi);
3689
3690         f2fs_tuning_parameters(sbi);
3691
3692         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3693                     cur_cp_version(F2FS_CKPT(sbi)));
3694         f2fs_update_time(sbi, CP_TIME);
3695         f2fs_update_time(sbi, REQ_TIME);
3696         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3697         return 0;
3698
3699 sync_free_meta:
3700         /* safe to flush all the data */
3701         sync_filesystem(sbi->sb);
3702         retry_cnt = 0;
3703
3704 free_meta:
3705 #ifdef CONFIG_QUOTA
3706         f2fs_truncate_quota_inode_pages(sb);
3707         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3708                 f2fs_quota_off_umount(sbi->sb);
3709 #endif
3710         /*
3711          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3712          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3713          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3714          * falls into an infinite loop in f2fs_sync_meta_pages().
3715          */
3716         truncate_inode_pages_final(META_MAPPING(sbi));
3717         /* evict some inodes being cached by GC */
3718         evict_inodes(sb);
3719         f2fs_unregister_sysfs(sbi);
3720 free_root_inode:
3721         dput(sb->s_root);
3722         sb->s_root = NULL;
3723 free_node_inode:
3724         f2fs_release_ino_entry(sbi, true);
3725         truncate_inode_pages_final(NODE_MAPPING(sbi));
3726         iput(sbi->node_inode);
3727         sbi->node_inode = NULL;
3728 free_stats:
3729         f2fs_destroy_stats(sbi);
3730 free_nm:
3731         f2fs_destroy_node_manager(sbi);
3732 free_sm:
3733         f2fs_destroy_segment_manager(sbi);
3734         f2fs_destroy_post_read_wq(sbi);
3735 free_devices:
3736         destroy_device_list(sbi);
3737         kvfree(sbi->ckpt);
3738 free_meta_inode:
3739         make_bad_inode(sbi->meta_inode);
3740         iput(sbi->meta_inode);
3741         sbi->meta_inode = NULL;
3742 free_io_dummy:
3743         mempool_destroy(sbi->write_io_dummy);
3744 free_percpu:
3745         destroy_percpu_info(sbi);
3746 free_bio_info:
3747         for (i = 0; i < NR_PAGE_TYPE; i++)
3748                 kvfree(sbi->write_io[i]);
3749
3750 #ifdef CONFIG_UNICODE
3751         utf8_unload(sbi->s_encoding);
3752 #endif
3753 free_options:
3754 #ifdef CONFIG_QUOTA
3755         for (i = 0; i < MAXQUOTAS; i++)
3756                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3757 #endif
3758         kvfree(options);
3759 free_sb_buf:
3760         kvfree(raw_super);
3761 free_sbi:
3762         if (sbi->s_chksum_driver)
3763                 crypto_free_shash(sbi->s_chksum_driver);
3764         kvfree(sbi);
3765
3766         /* give only one another chance */
3767         if (retry_cnt > 0 && skip_recovery) {
3768                 retry_cnt--;
3769                 shrink_dcache_sb(sb);
3770                 goto try_onemore;
3771         }
3772         return err;
3773 }
3774
3775 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3776                         const char *dev_name, void *data)
3777 {
3778         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3779 }
3780
3781 static void kill_f2fs_super(struct super_block *sb)
3782 {
3783         if (sb->s_root) {
3784                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3785
3786                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3787                 f2fs_stop_gc_thread(sbi);
3788                 f2fs_stop_discard_thread(sbi);
3789
3790                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3791                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3792                         struct cp_control cpc = {
3793                                 .reason = CP_UMOUNT,
3794                         };
3795                         f2fs_write_checkpoint(sbi, &cpc);
3796                 }
3797
3798                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3799                         sb->s_flags &= ~SB_RDONLY;
3800         }
3801         kill_block_super(sb);
3802 }
3803
3804 static struct file_system_type f2fs_fs_type = {
3805         .owner          = THIS_MODULE,
3806         .name           = "f2fs",
3807         .mount          = f2fs_mount,
3808         .kill_sb        = kill_f2fs_super,
3809         .fs_flags       = FS_REQUIRES_DEV,
3810 };
3811 MODULE_ALIAS_FS("f2fs");
3812
3813 static int __init init_inodecache(void)
3814 {
3815         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3816                         sizeof(struct f2fs_inode_info), 0,
3817                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3818         if (!f2fs_inode_cachep)
3819                 return -ENOMEM;
3820         return 0;
3821 }
3822
3823 static void destroy_inodecache(void)
3824 {
3825         /*
3826          * Make sure all delayed rcu free inodes are flushed before we
3827          * destroy cache.
3828          */
3829         rcu_barrier();
3830         kmem_cache_destroy(f2fs_inode_cachep);
3831 }
3832
3833 static int __init init_f2fs_fs(void)
3834 {
3835         int err;
3836
3837         if (PAGE_SIZE != F2FS_BLKSIZE) {
3838                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3839                                 PAGE_SIZE, F2FS_BLKSIZE);
3840                 return -EINVAL;
3841         }
3842
3843         f2fs_build_trace_ios();
3844
3845         err = init_inodecache();
3846         if (err)
3847                 goto fail;
3848         err = f2fs_create_node_manager_caches();
3849         if (err)
3850                 goto free_inodecache;
3851         err = f2fs_create_segment_manager_caches();
3852         if (err)
3853                 goto free_node_manager_caches;
3854         err = f2fs_create_checkpoint_caches();
3855         if (err)
3856                 goto free_segment_manager_caches;
3857         err = f2fs_create_extent_cache();
3858         if (err)
3859                 goto free_checkpoint_caches;
3860         err = f2fs_init_sysfs();
3861         if (err)
3862                 goto free_extent_cache;
3863         err = register_shrinker(&f2fs_shrinker_info);
3864         if (err)
3865                 goto free_sysfs;
3866         err = register_filesystem(&f2fs_fs_type);
3867         if (err)
3868                 goto free_shrinker;
3869         f2fs_create_root_stats();
3870         err = f2fs_init_post_read_processing();
3871         if (err)
3872                 goto free_root_stats;
3873         err = f2fs_init_bio_entry_cache();
3874         if (err)
3875                 goto free_post_read;
3876         err = f2fs_init_bioset();
3877         if (err)
3878                 goto free_bio_enrty_cache;
3879         return 0;
3880 free_bio_enrty_cache:
3881         f2fs_destroy_bio_entry_cache();
3882 free_post_read:
3883         f2fs_destroy_post_read_processing();
3884 free_root_stats:
3885         f2fs_destroy_root_stats();
3886         unregister_filesystem(&f2fs_fs_type);
3887 free_shrinker:
3888         unregister_shrinker(&f2fs_shrinker_info);
3889 free_sysfs:
3890         f2fs_exit_sysfs();
3891 free_extent_cache:
3892         f2fs_destroy_extent_cache();
3893 free_checkpoint_caches:
3894         f2fs_destroy_checkpoint_caches();
3895 free_segment_manager_caches:
3896         f2fs_destroy_segment_manager_caches();
3897 free_node_manager_caches:
3898         f2fs_destroy_node_manager_caches();
3899 free_inodecache:
3900         destroy_inodecache();
3901 fail:
3902         return err;
3903 }
3904
3905 static void __exit exit_f2fs_fs(void)
3906 {
3907         f2fs_destroy_bioset();
3908         f2fs_destroy_bio_entry_cache();
3909         f2fs_destroy_post_read_processing();
3910         f2fs_destroy_root_stats();
3911         unregister_filesystem(&f2fs_fs_type);
3912         unregister_shrinker(&f2fs_shrinker_info);
3913         f2fs_exit_sysfs();
3914         f2fs_destroy_extent_cache();
3915         f2fs_destroy_checkpoint_caches();
3916         f2fs_destroy_segment_manager_caches();
3917         f2fs_destroy_node_manager_caches();
3918         destroy_inodecache();
3919         f2fs_destroy_trace_ios();
3920 }
3921
3922 module_init(init_f2fs_fs)
3923 module_exit(exit_f2fs_fs)
3924
3925 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3926 MODULE_DESCRIPTION("Flash Friendly File System");
3927 MODULE_LICENSE("GPL");
3928