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                 down_write(&sbi->gc_lock);
1242                 err = f2fs_write_checkpoint(sbi, &cpc);
1243                 up_write(&sbi->gc_lock);
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 = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1286                                         dquot->dq_dqb.dqb_bhardlimit);
1287         if (limit)
1288                 limit >>= sb->s_blocksize_bits;
1289
1290         if (limit && buf->f_blocks > limit) {
1291                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1292                 buf->f_blocks = limit;
1293                 buf->f_bfree = buf->f_bavail =
1294                         (buf->f_blocks > curblock) ?
1295                          (buf->f_blocks - curblock) : 0;
1296         }
1297
1298         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1299                                         dquot->dq_dqb.dqb_ihardlimit);
1300
1301         if (limit && buf->f_files > limit) {
1302                 buf->f_files = limit;
1303                 buf->f_ffree =
1304                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1305                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1306         }
1307
1308         spin_unlock(&dquot->dq_dqb_lock);
1309         dqput(dquot);
1310         return 0;
1311 }
1312 #endif
1313
1314 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1315 {
1316         struct super_block *sb = dentry->d_sb;
1317         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1318         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1319         block_t total_count, user_block_count, start_count;
1320         u64 avail_node_count;
1321
1322         total_count = le64_to_cpu(sbi->raw_super->block_count);
1323         user_block_count = sbi->user_block_count;
1324         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1325         buf->f_type = F2FS_SUPER_MAGIC;
1326         buf->f_bsize = sbi->blocksize;
1327
1328         buf->f_blocks = total_count - start_count;
1329         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1330                                                 sbi->current_reserved_blocks;
1331
1332         spin_lock(&sbi->stat_lock);
1333         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1334                 buf->f_bfree = 0;
1335         else
1336                 buf->f_bfree -= sbi->unusable_block_count;
1337         spin_unlock(&sbi->stat_lock);
1338
1339         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1340                 buf->f_bavail = buf->f_bfree -
1341                                 F2FS_OPTION(sbi).root_reserved_blocks;
1342         else
1343                 buf->f_bavail = 0;
1344
1345         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1346
1347         if (avail_node_count > user_block_count) {
1348                 buf->f_files = user_block_count;
1349                 buf->f_ffree = buf->f_bavail;
1350         } else {
1351                 buf->f_files = avail_node_count;
1352                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1353                                         buf->f_bavail);
1354         }
1355
1356         buf->f_namelen = F2FS_NAME_LEN;
1357         buf->f_fsid.val[0] = (u32)id;
1358         buf->f_fsid.val[1] = (u32)(id >> 32);
1359
1360 #ifdef CONFIG_QUOTA
1361         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1362                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1363                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1364         }
1365 #endif
1366         return 0;
1367 }
1368
1369 static inline void f2fs_show_quota_options(struct seq_file *seq,
1370                                            struct super_block *sb)
1371 {
1372 #ifdef CONFIG_QUOTA
1373         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1374
1375         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1376                 char *fmtname = "";
1377
1378                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1379                 case QFMT_VFS_OLD:
1380                         fmtname = "vfsold";
1381                         break;
1382                 case QFMT_VFS_V0:
1383                         fmtname = "vfsv0";
1384                         break;
1385                 case QFMT_VFS_V1:
1386                         fmtname = "vfsv1";
1387                         break;
1388                 }
1389                 seq_printf(seq, ",jqfmt=%s", fmtname);
1390         }
1391
1392         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1393                 seq_show_option(seq, "usrjquota",
1394                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1395
1396         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1397                 seq_show_option(seq, "grpjquota",
1398                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1399
1400         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1401                 seq_show_option(seq, "prjjquota",
1402                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1403 #endif
1404 }
1405
1406 static inline void f2fs_show_compress_options(struct seq_file *seq,
1407                                                         struct super_block *sb)
1408 {
1409         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1410         char *algtype = "";
1411         int i;
1412
1413         if (!f2fs_sb_has_compression(sbi))
1414                 return;
1415
1416         switch (F2FS_OPTION(sbi).compress_algorithm) {
1417         case COMPRESS_LZO:
1418                 algtype = "lzo";
1419                 break;
1420         case COMPRESS_LZ4:
1421                 algtype = "lz4";
1422                 break;
1423         }
1424         seq_printf(seq, ",compress_algorithm=%s", algtype);
1425
1426         seq_printf(seq, ",compress_log_size=%u",
1427                         F2FS_OPTION(sbi).compress_log_size);
1428
1429         for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1430                 seq_printf(seq, ",compress_extension=%s",
1431                         F2FS_OPTION(sbi).extensions[i]);
1432         }
1433 }
1434
1435 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1436 {
1437         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1438
1439         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1440                 if (test_opt(sbi, FORCE_FG_GC))
1441                         seq_printf(seq, ",background_gc=%s", "sync");
1442                 else
1443                         seq_printf(seq, ",background_gc=%s", "on");
1444         } else {
1445                 seq_printf(seq, ",background_gc=%s", "off");
1446         }
1447         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1448                 seq_puts(seq, ",disable_roll_forward");
1449         if (test_opt(sbi, DISCARD))
1450                 seq_puts(seq, ",discard");
1451         else
1452                 seq_puts(seq, ",nodiscard");
1453         if (test_opt(sbi, NOHEAP))
1454                 seq_puts(seq, ",no_heap");
1455         else
1456                 seq_puts(seq, ",heap");
1457 #ifdef CONFIG_F2FS_FS_XATTR
1458         if (test_opt(sbi, XATTR_USER))
1459                 seq_puts(seq, ",user_xattr");
1460         else
1461                 seq_puts(seq, ",nouser_xattr");
1462         if (test_opt(sbi, INLINE_XATTR))
1463                 seq_puts(seq, ",inline_xattr");
1464         else
1465                 seq_puts(seq, ",noinline_xattr");
1466         if (test_opt(sbi, INLINE_XATTR_SIZE))
1467                 seq_printf(seq, ",inline_xattr_size=%u",
1468                                         F2FS_OPTION(sbi).inline_xattr_size);
1469 #endif
1470 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1471         if (test_opt(sbi, POSIX_ACL))
1472                 seq_puts(seq, ",acl");
1473         else
1474                 seq_puts(seq, ",noacl");
1475 #endif
1476         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1477                 seq_puts(seq, ",disable_ext_identify");
1478         if (test_opt(sbi, INLINE_DATA))
1479                 seq_puts(seq, ",inline_data");
1480         else
1481                 seq_puts(seq, ",noinline_data");
1482         if (test_opt(sbi, INLINE_DENTRY))
1483                 seq_puts(seq, ",inline_dentry");
1484         else
1485                 seq_puts(seq, ",noinline_dentry");
1486         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1487                 seq_puts(seq, ",flush_merge");
1488         if (test_opt(sbi, NOBARRIER))
1489                 seq_puts(seq, ",nobarrier");
1490         if (test_opt(sbi, FASTBOOT))
1491                 seq_puts(seq, ",fastboot");
1492         if (test_opt(sbi, EXTENT_CACHE))
1493                 seq_puts(seq, ",extent_cache");
1494         else
1495                 seq_puts(seq, ",noextent_cache");
1496         if (test_opt(sbi, DATA_FLUSH))
1497                 seq_puts(seq, ",data_flush");
1498
1499         seq_puts(seq, ",mode=");
1500         if (test_opt(sbi, ADAPTIVE))
1501                 seq_puts(seq, "adaptive");
1502         else if (test_opt(sbi, LFS))
1503                 seq_puts(seq, "lfs");
1504         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1505         if (test_opt(sbi, RESERVE_ROOT))
1506                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1507                                 F2FS_OPTION(sbi).root_reserved_blocks,
1508                                 from_kuid_munged(&init_user_ns,
1509                                         F2FS_OPTION(sbi).s_resuid),
1510                                 from_kgid_munged(&init_user_ns,
1511                                         F2FS_OPTION(sbi).s_resgid));
1512         if (F2FS_IO_SIZE_BITS(sbi))
1513                 seq_printf(seq, ",io_bits=%u",
1514                                 F2FS_OPTION(sbi).write_io_size_bits);
1515 #ifdef CONFIG_F2FS_FAULT_INJECTION
1516         if (test_opt(sbi, FAULT_INJECTION)) {
1517                 seq_printf(seq, ",fault_injection=%u",
1518                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1519                 seq_printf(seq, ",fault_type=%u",
1520                                 F2FS_OPTION(sbi).fault_info.inject_type);
1521         }
1522 #endif
1523 #ifdef CONFIG_QUOTA
1524         if (test_opt(sbi, QUOTA))
1525                 seq_puts(seq, ",quota");
1526         if (test_opt(sbi, USRQUOTA))
1527                 seq_puts(seq, ",usrquota");
1528         if (test_opt(sbi, GRPQUOTA))
1529                 seq_puts(seq, ",grpquota");
1530         if (test_opt(sbi, PRJQUOTA))
1531                 seq_puts(seq, ",prjquota");
1532 #endif
1533         f2fs_show_quota_options(seq, sbi->sb);
1534         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1535                 seq_printf(seq, ",whint_mode=%s", "user-based");
1536         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1537                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1538 #ifdef CONFIG_FS_ENCRYPTION
1539         if (F2FS_OPTION(sbi).test_dummy_encryption)
1540                 seq_puts(seq, ",test_dummy_encryption");
1541 #endif
1542
1543         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1544                 seq_printf(seq, ",alloc_mode=%s", "default");
1545         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1546                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1547
1548         if (test_opt(sbi, DISABLE_CHECKPOINT))
1549                 seq_printf(seq, ",checkpoint=disable:%u",
1550                                 F2FS_OPTION(sbi).unusable_cap);
1551         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1552                 seq_printf(seq, ",fsync_mode=%s", "posix");
1553         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1554                 seq_printf(seq, ",fsync_mode=%s", "strict");
1555         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1556                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1557
1558         f2fs_show_compress_options(seq, sbi->sb);
1559         return 0;
1560 }
1561
1562 static void default_options(struct f2fs_sb_info *sbi)
1563 {
1564         /* init some FS parameters */
1565         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1566         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1567         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1568         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1569         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1570         F2FS_OPTION(sbi).test_dummy_encryption = false;
1571         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1572         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1573         F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZO;
1574         F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1575         F2FS_OPTION(sbi).compress_ext_cnt = 0;
1576
1577         set_opt(sbi, BG_GC);
1578         set_opt(sbi, INLINE_XATTR);
1579         set_opt(sbi, INLINE_DATA);
1580         set_opt(sbi, INLINE_DENTRY);
1581         set_opt(sbi, EXTENT_CACHE);
1582         set_opt(sbi, NOHEAP);
1583         clear_opt(sbi, DISABLE_CHECKPOINT);
1584         F2FS_OPTION(sbi).unusable_cap = 0;
1585         sbi->sb->s_flags |= SB_LAZYTIME;
1586         set_opt(sbi, FLUSH_MERGE);
1587         set_opt(sbi, DISCARD);
1588         if (f2fs_sb_has_blkzoned(sbi))
1589                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1590         else
1591                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1592
1593 #ifdef CONFIG_F2FS_FS_XATTR
1594         set_opt(sbi, XATTR_USER);
1595 #endif
1596 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1597         set_opt(sbi, POSIX_ACL);
1598 #endif
1599
1600         f2fs_build_fault_attr(sbi, 0, 0);
1601 }
1602
1603 #ifdef CONFIG_QUOTA
1604 static int f2fs_enable_quotas(struct super_block *sb);
1605 #endif
1606
1607 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1608 {
1609         unsigned int s_flags = sbi->sb->s_flags;
1610         struct cp_control cpc;
1611         int err = 0;
1612         int ret;
1613         block_t unusable;
1614
1615         if (s_flags & SB_RDONLY) {
1616                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
1617                 return -EINVAL;
1618         }
1619         sbi->sb->s_flags |= SB_ACTIVE;
1620
1621         f2fs_update_time(sbi, DISABLE_TIME);
1622
1623         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1624                 down_write(&sbi->gc_lock);
1625                 err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1626                 if (err == -ENODATA) {
1627                         err = 0;
1628                         break;
1629                 }
1630                 if (err && err != -EAGAIN)
1631                         break;
1632         }
1633
1634         ret = sync_filesystem(sbi->sb);
1635         if (ret || err) {
1636                 err = ret ? ret: err;
1637                 goto restore_flag;
1638         }
1639
1640         unusable = f2fs_get_unusable_blocks(sbi);
1641         if (f2fs_disable_cp_again(sbi, unusable)) {
1642                 err = -EAGAIN;
1643                 goto restore_flag;
1644         }
1645
1646         down_write(&sbi->gc_lock);
1647         cpc.reason = CP_PAUSE;
1648         set_sbi_flag(sbi, SBI_CP_DISABLED);
1649         err = f2fs_write_checkpoint(sbi, &cpc);
1650         if (err)
1651                 goto out_unlock;
1652
1653         spin_lock(&sbi->stat_lock);
1654         sbi->unusable_block_count = unusable;
1655         spin_unlock(&sbi->stat_lock);
1656
1657 out_unlock:
1658         up_write(&sbi->gc_lock);
1659 restore_flag:
1660         sbi->sb->s_flags = s_flags;     /* Restore MS_RDONLY status */
1661         return err;
1662 }
1663
1664 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1665 {
1666         down_write(&sbi->gc_lock);
1667         f2fs_dirty_to_prefree(sbi);
1668
1669         clear_sbi_flag(sbi, SBI_CP_DISABLED);
1670         set_sbi_flag(sbi, SBI_IS_DIRTY);
1671         up_write(&sbi->gc_lock);
1672
1673         f2fs_sync_fs(sbi->sb, 1);
1674 }
1675
1676 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1677 {
1678         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1679         struct f2fs_mount_info org_mount_opt;
1680         unsigned long old_sb_flags;
1681         int err;
1682         bool need_restart_gc = false;
1683         bool need_stop_gc = false;
1684         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1685         bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1686         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1687         bool checkpoint_changed;
1688 #ifdef CONFIG_QUOTA
1689         int i, j;
1690 #endif
1691
1692         /*
1693          * Save the old mount options in case we
1694          * need to restore them.
1695          */
1696         org_mount_opt = sbi->mount_opt;
1697         old_sb_flags = sb->s_flags;
1698
1699 #ifdef CONFIG_QUOTA
1700         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1701         for (i = 0; i < MAXQUOTAS; i++) {
1702                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1703                         org_mount_opt.s_qf_names[i] =
1704                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1705                                 GFP_KERNEL);
1706                         if (!org_mount_opt.s_qf_names[i]) {
1707                                 for (j = 0; j < i; j++)
1708                                         kvfree(org_mount_opt.s_qf_names[j]);
1709                                 return -ENOMEM;
1710                         }
1711                 } else {
1712                         org_mount_opt.s_qf_names[i] = NULL;
1713                 }
1714         }
1715 #endif
1716
1717         /* recover superblocks we couldn't write due to previous RO mount */
1718         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1719                 err = f2fs_commit_super(sbi, false);
1720                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1721                           err);
1722                 if (!err)
1723                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1724         }
1725
1726         default_options(sbi);
1727
1728         /* parse mount options */
1729         err = parse_options(sb, data);
1730         if (err)
1731                 goto restore_opts;
1732         checkpoint_changed =
1733                         disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1734
1735         /*
1736          * Previous and new state of filesystem is RO,
1737          * so skip checking GC and FLUSH_MERGE conditions.
1738          */
1739         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1740                 goto skip;
1741
1742 #ifdef CONFIG_QUOTA
1743         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1744                 err = dquot_suspend(sb, -1);
1745                 if (err < 0)
1746                         goto restore_opts;
1747         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1748                 /* dquot_resume needs RW */
1749                 sb->s_flags &= ~SB_RDONLY;
1750                 if (sb_any_quota_suspended(sb)) {
1751                         dquot_resume(sb, -1);
1752                 } else if (f2fs_sb_has_quota_ino(sbi)) {
1753                         err = f2fs_enable_quotas(sb);
1754                         if (err)
1755                                 goto restore_opts;
1756                 }
1757         }
1758 #endif
1759         /* disallow enable/disable extent_cache dynamically */
1760         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1761                 err = -EINVAL;
1762                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
1763                 goto restore_opts;
1764         }
1765
1766         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1767                 err = -EINVAL;
1768                 f2fs_warn(sbi, "switch io_bits option is not allowed");
1769                 goto restore_opts;
1770         }
1771
1772         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1773                 err = -EINVAL;
1774                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1775                 goto restore_opts;
1776         }
1777
1778         /*
1779          * We stop the GC thread if FS is mounted as RO
1780          * or if background_gc = off is passed in mount
1781          * option. Also sync the filesystem.
1782          */
1783         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1784                 if (sbi->gc_thread) {
1785                         f2fs_stop_gc_thread(sbi);
1786                         need_restart_gc = true;
1787                 }
1788         } else if (!sbi->gc_thread) {
1789                 err = f2fs_start_gc_thread(sbi);
1790                 if (err)
1791                         goto restore_opts;
1792                 need_stop_gc = true;
1793         }
1794
1795         if (*flags & SB_RDONLY ||
1796                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1797                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1798                 sync_inodes_sb(sb);
1799
1800                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1801                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1802                 f2fs_sync_fs(sb, 1);
1803                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1804         }
1805
1806         if (checkpoint_changed) {
1807                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1808                         err = f2fs_disable_checkpoint(sbi);
1809                         if (err)
1810                                 goto restore_gc;
1811                 } else {
1812                         f2fs_enable_checkpoint(sbi);
1813                 }
1814         }
1815
1816         /*
1817          * We stop issue flush thread if FS is mounted as RO
1818          * or if flush_merge is not passed in mount option.
1819          */
1820         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1821                 clear_opt(sbi, FLUSH_MERGE);
1822                 f2fs_destroy_flush_cmd_control(sbi, false);
1823         } else {
1824                 err = f2fs_create_flush_cmd_control(sbi);
1825                 if (err)
1826                         goto restore_gc;
1827         }
1828 skip:
1829 #ifdef CONFIG_QUOTA
1830         /* Release old quota file names */
1831         for (i = 0; i < MAXQUOTAS; i++)
1832                 kvfree(org_mount_opt.s_qf_names[i]);
1833 #endif
1834         /* Update the POSIXACL Flag */
1835         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1836                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1837
1838         limit_reserve_root(sbi);
1839         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1840         return 0;
1841 restore_gc:
1842         if (need_restart_gc) {
1843                 if (f2fs_start_gc_thread(sbi))
1844                         f2fs_warn(sbi, "background gc thread has stopped");
1845         } else if (need_stop_gc) {
1846                 f2fs_stop_gc_thread(sbi);
1847         }
1848 restore_opts:
1849 #ifdef CONFIG_QUOTA
1850         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1851         for (i = 0; i < MAXQUOTAS; i++) {
1852                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1853                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1854         }
1855 #endif
1856         sbi->mount_opt = org_mount_opt;
1857         sb->s_flags = old_sb_flags;
1858         return err;
1859 }
1860
1861 #ifdef CONFIG_QUOTA
1862 /* Read data from quotafile */
1863 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1864                                size_t len, loff_t off)
1865 {
1866         struct inode *inode = sb_dqopt(sb)->files[type];
1867         struct address_space *mapping = inode->i_mapping;
1868         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1869         int offset = off & (sb->s_blocksize - 1);
1870         int tocopy;
1871         size_t toread;
1872         loff_t i_size = i_size_read(inode);
1873         struct page *page;
1874         char *kaddr;
1875
1876         if (off > i_size)
1877                 return 0;
1878
1879         if (off + len > i_size)
1880                 len = i_size - off;
1881         toread = len;
1882         while (toread > 0) {
1883                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1884 repeat:
1885                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1886                 if (IS_ERR(page)) {
1887                         if (PTR_ERR(page) == -ENOMEM) {
1888                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1889                                 goto repeat;
1890                         }
1891                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1892                         return PTR_ERR(page);
1893                 }
1894
1895                 lock_page(page);
1896
1897                 if (unlikely(page->mapping != mapping)) {
1898                         f2fs_put_page(page, 1);
1899                         goto repeat;
1900                 }
1901                 if (unlikely(!PageUptodate(page))) {
1902                         f2fs_put_page(page, 1);
1903                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1904                         return -EIO;
1905                 }
1906
1907                 kaddr = kmap_atomic(page);
1908                 memcpy(data, kaddr + offset, tocopy);
1909                 kunmap_atomic(kaddr);
1910                 f2fs_put_page(page, 1);
1911
1912                 offset = 0;
1913                 toread -= tocopy;
1914                 data += tocopy;
1915                 blkidx++;
1916         }
1917         return len;
1918 }
1919
1920 /* Write to quotafile */
1921 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1922                                 const char *data, size_t len, loff_t off)
1923 {
1924         struct inode *inode = sb_dqopt(sb)->files[type];
1925         struct address_space *mapping = inode->i_mapping;
1926         const struct address_space_operations *a_ops = mapping->a_ops;
1927         int offset = off & (sb->s_blocksize - 1);
1928         size_t towrite = len;
1929         struct page *page;
1930         char *kaddr;
1931         int err = 0;
1932         int tocopy;
1933
1934         while (towrite > 0) {
1935                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1936                                                                 towrite);
1937 retry:
1938                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1939                                                         &page, NULL);
1940                 if (unlikely(err)) {
1941                         if (err == -ENOMEM) {
1942                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1943                                 goto retry;
1944                         }
1945                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1946                         break;
1947                 }
1948
1949                 kaddr = kmap_atomic(page);
1950                 memcpy(kaddr + offset, data, tocopy);
1951                 kunmap_atomic(kaddr);
1952                 flush_dcache_page(page);
1953
1954                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1955                                                 page, NULL);
1956                 offset = 0;
1957                 towrite -= tocopy;
1958                 off += tocopy;
1959                 data += tocopy;
1960                 cond_resched();
1961         }
1962
1963         if (len == towrite)
1964                 return err;
1965         inode->i_mtime = inode->i_ctime = current_time(inode);
1966         f2fs_mark_inode_dirty_sync(inode, false);
1967         return len - towrite;
1968 }
1969
1970 static struct dquot **f2fs_get_dquots(struct inode *inode)
1971 {
1972         return F2FS_I(inode)->i_dquot;
1973 }
1974
1975 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1976 {
1977         return &F2FS_I(inode)->i_reserved_quota;
1978 }
1979
1980 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1981 {
1982         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1983                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1984                 return 0;
1985         }
1986
1987         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1988                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1989 }
1990
1991 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1992 {
1993         int enabled = 0;
1994         int i, err;
1995
1996         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
1997                 err = f2fs_enable_quotas(sbi->sb);
1998                 if (err) {
1999                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2000                         return 0;
2001                 }
2002                 return 1;
2003         }
2004
2005         for (i = 0; i < MAXQUOTAS; i++) {
2006                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2007                         err = f2fs_quota_on_mount(sbi, i);
2008                         if (!err) {
2009                                 enabled = 1;
2010                                 continue;
2011                         }
2012                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2013                                  err, i);
2014                 }
2015         }
2016         return enabled;
2017 }
2018
2019 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2020                              unsigned int flags)
2021 {
2022         struct inode *qf_inode;
2023         unsigned long qf_inum;
2024         int err;
2025
2026         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2027
2028         qf_inum = f2fs_qf_ino(sb, type);
2029         if (!qf_inum)
2030                 return -EPERM;
2031
2032         qf_inode = f2fs_iget(sb, qf_inum);
2033         if (IS_ERR(qf_inode)) {
2034                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2035                 return PTR_ERR(qf_inode);
2036         }
2037
2038         /* Don't account quota for quota files to avoid recursion */
2039         qf_inode->i_flags |= S_NOQUOTA;
2040         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2041         iput(qf_inode);
2042         return err;
2043 }
2044
2045 static int f2fs_enable_quotas(struct super_block *sb)
2046 {
2047         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2048         int type, err = 0;
2049         unsigned long qf_inum;
2050         bool quota_mopt[MAXQUOTAS] = {
2051                 test_opt(sbi, USRQUOTA),
2052                 test_opt(sbi, GRPQUOTA),
2053                 test_opt(sbi, PRJQUOTA),
2054         };
2055
2056         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2057                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2058                 return 0;
2059         }
2060
2061         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2062
2063         for (type = 0; type < MAXQUOTAS; type++) {
2064                 qf_inum = f2fs_qf_ino(sb, type);
2065                 if (qf_inum) {
2066                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2067                                 DQUOT_USAGE_ENABLED |
2068                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2069                         if (err) {
2070                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2071                                          type, err);
2072                                 for (type--; type >= 0; type--)
2073                                         dquot_quota_off(sb, type);
2074                                 set_sbi_flag(F2FS_SB(sb),
2075                                                 SBI_QUOTA_NEED_REPAIR);
2076                                 return err;
2077                         }
2078                 }
2079         }
2080         return 0;
2081 }
2082
2083 int f2fs_quota_sync(struct super_block *sb, int type)
2084 {
2085         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2086         struct quota_info *dqopt = sb_dqopt(sb);
2087         int cnt;
2088         int ret;
2089
2090         /*
2091          * do_quotactl
2092          *  f2fs_quota_sync
2093          *  down_read(quota_sem)
2094          *  dquot_writeback_dquots()
2095          *  f2fs_dquot_commit
2096          *                            block_operation
2097          *                            down_read(quota_sem)
2098          */
2099         f2fs_lock_op(sbi);
2100
2101         down_read(&sbi->quota_sem);
2102         ret = dquot_writeback_dquots(sb, type);
2103         if (ret)
2104                 goto out;
2105
2106         /*
2107          * Now when everything is written we can discard the pagecache so
2108          * that userspace sees the changes.
2109          */
2110         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2111                 struct address_space *mapping;
2112
2113                 if (type != -1 && cnt != type)
2114                         continue;
2115                 if (!sb_has_quota_active(sb, cnt))
2116                         continue;
2117
2118                 mapping = dqopt->files[cnt]->i_mapping;
2119
2120                 ret = filemap_fdatawrite(mapping);
2121                 if (ret)
2122                         goto out;
2123
2124                 /* if we are using journalled quota */
2125                 if (is_journalled_quota(sbi))
2126                         continue;
2127
2128                 ret = filemap_fdatawait(mapping);
2129                 if (ret)
2130                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2131
2132                 inode_lock(dqopt->files[cnt]);
2133                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2134                 inode_unlock(dqopt->files[cnt]);
2135         }
2136 out:
2137         if (ret)
2138                 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2139         up_read(&sbi->quota_sem);
2140         f2fs_unlock_op(sbi);
2141         return ret;
2142 }
2143
2144 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2145                                                         const struct path *path)
2146 {
2147         struct inode *inode;
2148         int err;
2149
2150         /* if quota sysfile exists, deny enabling quota with specific file */
2151         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2152                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2153                 return -EBUSY;
2154         }
2155
2156         err = f2fs_quota_sync(sb, type);
2157         if (err)
2158                 return err;
2159
2160         err = dquot_quota_on(sb, type, format_id, path);
2161         if (err)
2162                 return err;
2163
2164         inode = d_inode(path->dentry);
2165
2166         inode_lock(inode);
2167         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2168         f2fs_set_inode_flags(inode);
2169         inode_unlock(inode);
2170         f2fs_mark_inode_dirty_sync(inode, false);
2171
2172         return 0;
2173 }
2174
2175 static int __f2fs_quota_off(struct super_block *sb, int type)
2176 {
2177         struct inode *inode = sb_dqopt(sb)->files[type];
2178         int err;
2179
2180         if (!inode || !igrab(inode))
2181                 return dquot_quota_off(sb, type);
2182
2183         err = f2fs_quota_sync(sb, type);
2184         if (err)
2185                 goto out_put;
2186
2187         err = dquot_quota_off(sb, type);
2188         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2189                 goto out_put;
2190
2191         inode_lock(inode);
2192         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2193         f2fs_set_inode_flags(inode);
2194         inode_unlock(inode);
2195         f2fs_mark_inode_dirty_sync(inode, false);
2196 out_put:
2197         iput(inode);
2198         return err;
2199 }
2200
2201 static int f2fs_quota_off(struct super_block *sb, int type)
2202 {
2203         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2204         int err;
2205
2206         err = __f2fs_quota_off(sb, type);
2207
2208         /*
2209          * quotactl can shutdown journalled quota, result in inconsistence
2210          * between quota record and fs data by following updates, tag the
2211          * flag to let fsck be aware of it.
2212          */
2213         if (is_journalled_quota(sbi))
2214                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2215         return err;
2216 }
2217
2218 void f2fs_quota_off_umount(struct super_block *sb)
2219 {
2220         int type;
2221         int err;
2222
2223         for (type = 0; type < MAXQUOTAS; type++) {
2224                 err = __f2fs_quota_off(sb, type);
2225                 if (err) {
2226                         int ret = dquot_quota_off(sb, type);
2227
2228                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2229                                  type, err, ret);
2230                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2231                 }
2232         }
2233         /*
2234          * In case of checkpoint=disable, we must flush quota blocks.
2235          * This can cause NULL exception for node_inode in end_io, since
2236          * put_super already dropped it.
2237          */
2238         sync_filesystem(sb);
2239 }
2240
2241 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2242 {
2243         struct quota_info *dqopt = sb_dqopt(sb);
2244         int type;
2245
2246         for (type = 0; type < MAXQUOTAS; type++) {
2247                 if (!dqopt->files[type])
2248                         continue;
2249                 f2fs_inode_synced(dqopt->files[type]);
2250         }
2251 }
2252
2253 static int f2fs_dquot_commit(struct dquot *dquot)
2254 {
2255         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2256         int ret;
2257
2258         down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2259         ret = dquot_commit(dquot);
2260         if (ret < 0)
2261                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2262         up_read(&sbi->quota_sem);
2263         return ret;
2264 }
2265
2266 static int f2fs_dquot_acquire(struct dquot *dquot)
2267 {
2268         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2269         int ret;
2270
2271         down_read(&sbi->quota_sem);
2272         ret = dquot_acquire(dquot);
2273         if (ret < 0)
2274                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2275         up_read(&sbi->quota_sem);
2276         return ret;
2277 }
2278
2279 static int f2fs_dquot_release(struct dquot *dquot)
2280 {
2281         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2282         int ret = dquot_release(dquot);
2283
2284         if (ret < 0)
2285                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2286         return ret;
2287 }
2288
2289 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2290 {
2291         struct super_block *sb = dquot->dq_sb;
2292         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2293         int ret = dquot_mark_dquot_dirty(dquot);
2294
2295         /* if we are using journalled quota */
2296         if (is_journalled_quota(sbi))
2297                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2298
2299         return ret;
2300 }
2301
2302 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2303 {
2304         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2305         int ret = dquot_commit_info(sb, type);
2306
2307         if (ret < 0)
2308                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2309         return ret;
2310 }
2311
2312 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2313 {
2314         *projid = F2FS_I(inode)->i_projid;
2315         return 0;
2316 }
2317
2318 static const struct dquot_operations f2fs_quota_operations = {
2319         .get_reserved_space = f2fs_get_reserved_space,
2320         .write_dquot    = f2fs_dquot_commit,
2321         .acquire_dquot  = f2fs_dquot_acquire,
2322         .release_dquot  = f2fs_dquot_release,
2323         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2324         .write_info     = f2fs_dquot_commit_info,
2325         .alloc_dquot    = dquot_alloc,
2326         .destroy_dquot  = dquot_destroy,
2327         .get_projid     = f2fs_get_projid,
2328         .get_next_id    = dquot_get_next_id,
2329 };
2330
2331 static const struct quotactl_ops f2fs_quotactl_ops = {
2332         .quota_on       = f2fs_quota_on,
2333         .quota_off      = f2fs_quota_off,
2334         .quota_sync     = f2fs_quota_sync,
2335         .get_state      = dquot_get_state,
2336         .set_info       = dquot_set_dqinfo,
2337         .get_dqblk      = dquot_get_dqblk,
2338         .set_dqblk      = dquot_set_dqblk,
2339         .get_nextdqblk  = dquot_get_next_dqblk,
2340 };
2341 #else
2342 int f2fs_quota_sync(struct super_block *sb, int type)
2343 {
2344         return 0;
2345 }
2346
2347 void f2fs_quota_off_umount(struct super_block *sb)
2348 {
2349 }
2350 #endif
2351
2352 static const struct super_operations f2fs_sops = {
2353         .alloc_inode    = f2fs_alloc_inode,
2354         .free_inode     = f2fs_free_inode,
2355         .drop_inode     = f2fs_drop_inode,
2356         .write_inode    = f2fs_write_inode,
2357         .dirty_inode    = f2fs_dirty_inode,
2358         .show_options   = f2fs_show_options,
2359 #ifdef CONFIG_QUOTA
2360         .quota_read     = f2fs_quota_read,
2361         .quota_write    = f2fs_quota_write,
2362         .get_dquots     = f2fs_get_dquots,
2363 #endif
2364         .evict_inode    = f2fs_evict_inode,
2365         .put_super      = f2fs_put_super,
2366         .sync_fs        = f2fs_sync_fs,
2367         .freeze_fs      = f2fs_freeze,
2368         .unfreeze_fs    = f2fs_unfreeze,
2369         .statfs         = f2fs_statfs,
2370         .remount_fs     = f2fs_remount,
2371 };
2372
2373 #ifdef CONFIG_FS_ENCRYPTION
2374 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2375 {
2376         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2377                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2378                                 ctx, len, NULL);
2379 }
2380
2381 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2382                                                         void *fs_data)
2383 {
2384         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2385
2386         /*
2387          * Encrypting the root directory is not allowed because fsck
2388          * expects lost+found directory to exist and remain unencrypted
2389          * if LOST_FOUND feature is enabled.
2390          *
2391          */
2392         if (f2fs_sb_has_lost_found(sbi) &&
2393                         inode->i_ino == F2FS_ROOT_INO(sbi))
2394                 return -EPERM;
2395
2396         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2397                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2398                                 ctx, len, fs_data, XATTR_CREATE);
2399 }
2400
2401 static bool f2fs_dummy_context(struct inode *inode)
2402 {
2403         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2404 }
2405
2406 static bool f2fs_has_stable_inodes(struct super_block *sb)
2407 {
2408         return true;
2409 }
2410
2411 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2412                                        int *ino_bits_ret, int *lblk_bits_ret)
2413 {
2414         *ino_bits_ret = 8 * sizeof(nid_t);
2415         *lblk_bits_ret = 8 * sizeof(block_t);
2416 }
2417
2418 static const struct fscrypt_operations f2fs_cryptops = {
2419         .key_prefix             = "f2fs:",
2420         .get_context            = f2fs_get_context,
2421         .set_context            = f2fs_set_context,
2422         .dummy_context          = f2fs_dummy_context,
2423         .empty_dir              = f2fs_empty_dir,
2424         .max_namelen            = F2FS_NAME_LEN,
2425         .has_stable_inodes      = f2fs_has_stable_inodes,
2426         .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
2427 };
2428 #endif
2429
2430 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2431                 u64 ino, u32 generation)
2432 {
2433         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2434         struct inode *inode;
2435
2436         if (f2fs_check_nid_range(sbi, ino))
2437                 return ERR_PTR(-ESTALE);
2438
2439         /*
2440          * f2fs_iget isn't quite right if the inode is currently unallocated!
2441          * However f2fs_iget currently does appropriate checks to handle stale
2442          * inodes so everything is OK.
2443          */
2444         inode = f2fs_iget(sb, ino);
2445         if (IS_ERR(inode))
2446                 return ERR_CAST(inode);
2447         if (unlikely(generation && inode->i_generation != generation)) {
2448                 /* we didn't find the right inode.. */
2449                 iput(inode);
2450                 return ERR_PTR(-ESTALE);
2451         }
2452         return inode;
2453 }
2454
2455 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2456                 int fh_len, int fh_type)
2457 {
2458         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2459                                     f2fs_nfs_get_inode);
2460 }
2461
2462 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2463                 int fh_len, int fh_type)
2464 {
2465         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2466                                     f2fs_nfs_get_inode);
2467 }
2468
2469 static const struct export_operations f2fs_export_ops = {
2470         .fh_to_dentry = f2fs_fh_to_dentry,
2471         .fh_to_parent = f2fs_fh_to_parent,
2472         .get_parent = f2fs_get_parent,
2473 };
2474
2475 static loff_t max_file_blocks(void)
2476 {
2477         loff_t result = 0;
2478         loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2479
2480         /*
2481          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2482          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2483          * space in inode.i_addr, it will be more safe to reassign
2484          * result as zero.
2485          */
2486
2487         /* two direct node blocks */
2488         result += (leaf_count * 2);
2489
2490         /* two indirect node blocks */
2491         leaf_count *= NIDS_PER_BLOCK;
2492         result += (leaf_count * 2);
2493
2494         /* one double indirect node block */
2495         leaf_count *= NIDS_PER_BLOCK;
2496         result += leaf_count;
2497
2498         return result;
2499 }
2500
2501 static int __f2fs_commit_super(struct buffer_head *bh,
2502                         struct f2fs_super_block *super)
2503 {
2504         lock_buffer(bh);
2505         if (super)
2506                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2507         set_buffer_dirty(bh);
2508         unlock_buffer(bh);
2509
2510         /* it's rare case, we can do fua all the time */
2511         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2512 }
2513
2514 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2515                                         struct buffer_head *bh)
2516 {
2517         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2518                                         (bh->b_data + F2FS_SUPER_OFFSET);
2519         struct super_block *sb = sbi->sb;
2520         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2521         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2522         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2523         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2524         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2525         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2526         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2527         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2528         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2529         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2530         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2531         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2532         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2533         u64 main_end_blkaddr = main_blkaddr +
2534                                 (segment_count_main << log_blocks_per_seg);
2535         u64 seg_end_blkaddr = segment0_blkaddr +
2536                                 (segment_count << log_blocks_per_seg);
2537
2538         if (segment0_blkaddr != cp_blkaddr) {
2539                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2540                           segment0_blkaddr, cp_blkaddr);
2541                 return true;
2542         }
2543
2544         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2545                                                         sit_blkaddr) {
2546                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2547                           cp_blkaddr, sit_blkaddr,
2548                           segment_count_ckpt << log_blocks_per_seg);
2549                 return true;
2550         }
2551
2552         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2553                                                         nat_blkaddr) {
2554                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2555                           sit_blkaddr, nat_blkaddr,
2556                           segment_count_sit << log_blocks_per_seg);
2557                 return true;
2558         }
2559
2560         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2561                                                         ssa_blkaddr) {
2562                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2563                           nat_blkaddr, ssa_blkaddr,
2564                           segment_count_nat << log_blocks_per_seg);
2565                 return true;
2566         }
2567
2568         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2569                                                         main_blkaddr) {
2570                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2571                           ssa_blkaddr, main_blkaddr,
2572                           segment_count_ssa << log_blocks_per_seg);
2573                 return true;
2574         }
2575
2576         if (main_end_blkaddr > seg_end_blkaddr) {
2577                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2578                           main_blkaddr,
2579                           segment0_blkaddr +
2580                           (segment_count << log_blocks_per_seg),
2581                           segment_count_main << log_blocks_per_seg);
2582                 return true;
2583         } else if (main_end_blkaddr < seg_end_blkaddr) {
2584                 int err = 0;
2585                 char *res;
2586
2587                 /* fix in-memory information all the time */
2588                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2589                                 segment0_blkaddr) >> log_blocks_per_seg);
2590
2591                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2592                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2593                         res = "internally";
2594                 } else {
2595                         err = __f2fs_commit_super(bh, NULL);
2596                         res = err ? "failed" : "done";
2597                 }
2598                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2599                           res, main_blkaddr,
2600                           segment0_blkaddr +
2601                           (segment_count << log_blocks_per_seg),
2602                           segment_count_main << log_blocks_per_seg);
2603                 if (err)
2604                         return true;
2605         }
2606         return false;
2607 }
2608
2609 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2610                                 struct buffer_head *bh)
2611 {
2612         block_t segment_count, segs_per_sec, secs_per_zone;
2613         block_t total_sections, blocks_per_seg;
2614         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2615                                         (bh->b_data + F2FS_SUPER_OFFSET);
2616         unsigned int blocksize;
2617         size_t crc_offset = 0;
2618         __u32 crc = 0;
2619
2620         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2621                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2622                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2623                 return -EINVAL;
2624         }
2625
2626         /* Check checksum_offset and crc in superblock */
2627         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2628                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
2629                 if (crc_offset !=
2630                         offsetof(struct f2fs_super_block, crc)) {
2631                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2632                                   crc_offset);
2633                         return -EFSCORRUPTED;
2634                 }
2635                 crc = le32_to_cpu(raw_super->crc);
2636                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2637                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2638                         return -EFSCORRUPTED;
2639                 }
2640         }
2641
2642         /* Currently, support only 4KB page cache size */
2643         if (F2FS_BLKSIZE != PAGE_SIZE) {
2644                 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2645                           PAGE_SIZE);
2646                 return -EFSCORRUPTED;
2647         }
2648
2649         /* Currently, support only 4KB block size */
2650         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2651         if (blocksize != F2FS_BLKSIZE) {
2652                 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2653                           blocksize);
2654                 return -EFSCORRUPTED;
2655         }
2656
2657         /* check log blocks per segment */
2658         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2659                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2660                           le32_to_cpu(raw_super->log_blocks_per_seg));
2661                 return -EFSCORRUPTED;
2662         }
2663
2664         /* Currently, support 512/1024/2048/4096 bytes sector size */
2665         if (le32_to_cpu(raw_super->log_sectorsize) >
2666                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2667                 le32_to_cpu(raw_super->log_sectorsize) <
2668                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2669                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
2670                           le32_to_cpu(raw_super->log_sectorsize));
2671                 return -EFSCORRUPTED;
2672         }
2673         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2674                 le32_to_cpu(raw_super->log_sectorsize) !=
2675                         F2FS_MAX_LOG_SECTOR_SIZE) {
2676                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2677                           le32_to_cpu(raw_super->log_sectors_per_block),
2678                           le32_to_cpu(raw_super->log_sectorsize));
2679                 return -EFSCORRUPTED;
2680         }
2681
2682         segment_count = le32_to_cpu(raw_super->segment_count);
2683         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2684         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2685         total_sections = le32_to_cpu(raw_super->section_count);
2686
2687         /* blocks_per_seg should be 512, given the above check */
2688         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2689
2690         if (segment_count > F2FS_MAX_SEGMENT ||
2691                                 segment_count < F2FS_MIN_SEGMENTS) {
2692                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2693                 return -EFSCORRUPTED;
2694         }
2695
2696         if (total_sections > segment_count ||
2697                         total_sections < F2FS_MIN_SEGMENTS ||
2698                         segs_per_sec > segment_count || !segs_per_sec) {
2699                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2700                           segment_count, total_sections, segs_per_sec);
2701                 return -EFSCORRUPTED;
2702         }
2703
2704         if ((segment_count / segs_per_sec) < total_sections) {
2705                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2706                           segment_count, segs_per_sec, total_sections);
2707                 return -EFSCORRUPTED;
2708         }
2709
2710         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2711                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2712                           segment_count, le64_to_cpu(raw_super->block_count));
2713                 return -EFSCORRUPTED;
2714         }
2715
2716         if (RDEV(0).path[0]) {
2717                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2718                 int i = 1;
2719
2720                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
2721                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2722                         i++;
2723                 }
2724                 if (segment_count != dev_seg_count) {
2725                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2726                                         segment_count, dev_seg_count);
2727                         return -EFSCORRUPTED;
2728                 }
2729         }
2730
2731         if (secs_per_zone > total_sections || !secs_per_zone) {
2732                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2733                           secs_per_zone, total_sections);
2734                 return -EFSCORRUPTED;
2735         }
2736         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2737                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2738                         (le32_to_cpu(raw_super->extension_count) +
2739                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2740                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2741                           le32_to_cpu(raw_super->extension_count),
2742                           raw_super->hot_ext_count,
2743                           F2FS_MAX_EXTENSION);
2744                 return -EFSCORRUPTED;
2745         }
2746
2747         if (le32_to_cpu(raw_super->cp_payload) >
2748                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2749                 f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2750                           le32_to_cpu(raw_super->cp_payload),
2751                           blocks_per_seg - F2FS_CP_PACKS);
2752                 return -EFSCORRUPTED;
2753         }
2754
2755         /* check reserved ino info */
2756         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2757                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2758                 le32_to_cpu(raw_super->root_ino) != 3) {
2759                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2760                           le32_to_cpu(raw_super->node_ino),
2761                           le32_to_cpu(raw_super->meta_ino),
2762                           le32_to_cpu(raw_super->root_ino));
2763                 return -EFSCORRUPTED;
2764         }
2765
2766         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2767         if (sanity_check_area_boundary(sbi, bh))
2768                 return -EFSCORRUPTED;
2769
2770         return 0;
2771 }
2772
2773 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2774 {
2775         unsigned int total, fsmeta;
2776         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2777         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2778         unsigned int ovp_segments, reserved_segments;
2779         unsigned int main_segs, blocks_per_seg;
2780         unsigned int sit_segs, nat_segs;
2781         unsigned int sit_bitmap_size, nat_bitmap_size;
2782         unsigned int log_blocks_per_seg;
2783         unsigned int segment_count_main;
2784         unsigned int cp_pack_start_sum, cp_payload;
2785         block_t user_block_count, valid_user_blocks;
2786         block_t avail_node_count, valid_node_count;
2787         int i, j;
2788
2789         total = le32_to_cpu(raw_super->segment_count);
2790         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2791         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2792         fsmeta += sit_segs;
2793         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2794         fsmeta += nat_segs;
2795         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2796         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2797
2798         if (unlikely(fsmeta >= total))
2799                 return 1;
2800
2801         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2802         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2803
2804         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2805                         ovp_segments == 0 || reserved_segments == 0)) {
2806                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2807                 return 1;
2808         }
2809
2810         user_block_count = le64_to_cpu(ckpt->user_block_count);
2811         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2812         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2813         if (!user_block_count || user_block_count >=
2814                         segment_count_main << log_blocks_per_seg) {
2815                 f2fs_err(sbi, "Wrong user_block_count: %u",
2816                          user_block_count);
2817                 return 1;
2818         }
2819
2820         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2821         if (valid_user_blocks > user_block_count) {
2822                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2823                          valid_user_blocks, user_block_count);
2824                 return 1;
2825         }
2826
2827         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2828         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2829         if (valid_node_count > avail_node_count) {
2830                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2831                          valid_node_count, avail_node_count);
2832                 return 1;
2833         }
2834
2835         main_segs = le32_to_cpu(raw_super->segment_count_main);
2836         blocks_per_seg = sbi->blocks_per_seg;
2837
2838         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2839                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2840                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2841                         return 1;
2842                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2843                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2844                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2845                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2846                                          i, j,
2847                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2848                                 return 1;
2849                         }
2850                 }
2851         }
2852         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2853                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2854                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2855                         return 1;
2856                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2857                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2858                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2859                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2860                                          i, j,
2861                                          le32_to_cpu(ckpt->cur_data_segno[i]));
2862                                 return 1;
2863                         }
2864                 }
2865         }
2866         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2867                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2868                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2869                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2870                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2871                                          i, j,
2872                                          le32_to_cpu(ckpt->cur_node_segno[i]));
2873                                 return 1;
2874                         }
2875                 }
2876         }
2877
2878         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2879         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2880
2881         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2882                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2883                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2884                          sit_bitmap_size, nat_bitmap_size);
2885                 return 1;
2886         }
2887
2888         cp_pack_start_sum = __start_sum_addr(sbi);
2889         cp_payload = __cp_payload(sbi);
2890         if (cp_pack_start_sum < cp_payload + 1 ||
2891                 cp_pack_start_sum > blocks_per_seg - 1 -
2892                         NR_CURSEG_TYPE) {
2893                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2894                          cp_pack_start_sum);
2895                 return 1;
2896         }
2897
2898         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2899                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2900                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2901                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2902                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2903                           le32_to_cpu(ckpt->checksum_offset));
2904                 return 1;
2905         }
2906
2907         if (unlikely(f2fs_cp_error(sbi))) {
2908                 f2fs_err(sbi, "A bug case: need to run fsck");
2909                 return 1;
2910         }
2911         return 0;
2912 }
2913
2914 static void init_sb_info(struct f2fs_sb_info *sbi)
2915 {
2916         struct f2fs_super_block *raw_super = sbi->raw_super;
2917         int i;
2918
2919         sbi->log_sectors_per_block =
2920                 le32_to_cpu(raw_super->log_sectors_per_block);
2921         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2922         sbi->blocksize = 1 << sbi->log_blocksize;
2923         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2924         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2925         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2926         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2927         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2928         sbi->total_node_count =
2929                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2930                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2931         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2932         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2933         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2934         sbi->cur_victim_sec = NULL_SECNO;
2935         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2936         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2937         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2938         sbi->migration_granularity = sbi->segs_per_sec;
2939
2940         sbi->dir_level = DEF_DIR_LEVEL;
2941         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2942         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2943         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2944         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2945         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2946         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2947                                 DEF_UMOUNT_DISCARD_TIMEOUT;
2948         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2949
2950         for (i = 0; i < NR_COUNT_TYPE; i++)
2951                 atomic_set(&sbi->nr_pages[i], 0);
2952
2953         for (i = 0; i < META; i++)
2954                 atomic_set(&sbi->wb_sync_req[i], 0);
2955
2956         INIT_LIST_HEAD(&sbi->s_list);
2957         mutex_init(&sbi->umount_mutex);
2958         init_rwsem(&sbi->io_order_lock);
2959         spin_lock_init(&sbi->cp_lock);
2960
2961         sbi->dirty_device = 0;
2962         spin_lock_init(&sbi->dev_lock);
2963
2964         init_rwsem(&sbi->sb_lock);
2965         init_rwsem(&sbi->pin_sem);
2966 }
2967
2968 static int init_percpu_info(struct f2fs_sb_info *sbi)
2969 {
2970         int err;
2971
2972         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2973         if (err)
2974                 return err;
2975
2976         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2977                                                                 GFP_KERNEL);
2978         if (err)
2979                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2980
2981         return err;
2982 }
2983
2984 #ifdef CONFIG_BLK_DEV_ZONED
2985 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
2986                                void *data)
2987 {
2988         struct f2fs_dev_info *dev = data;
2989
2990         if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
2991                 set_bit(idx, dev->blkz_seq);
2992         return 0;
2993 }
2994
2995 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2996 {
2997         struct block_device *bdev = FDEV(devi).bdev;
2998         sector_t nr_sectors = bdev->bd_part->nr_sects;
2999         int ret;
3000
3001         if (!f2fs_sb_has_blkzoned(sbi))
3002                 return 0;
3003
3004         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3005                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3006                 return -EINVAL;
3007         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3008         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3009                                 __ilog2_u32(sbi->blocks_per_blkz))
3010                 return -EINVAL;
3011         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3012         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3013                                         sbi->log_blocks_per_blkz;
3014         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3015                 FDEV(devi).nr_blkz++;
3016
3017         FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
3018                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
3019                                         * sizeof(unsigned long),
3020                                         GFP_KERNEL);
3021         if (!FDEV(devi).blkz_seq)
3022                 return -ENOMEM;
3023
3024         /* Get block zones type */
3025         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3026                                   &FDEV(devi));
3027         if (ret < 0)
3028                 return ret;
3029
3030         return 0;
3031 }
3032 #endif
3033
3034 /*
3035  * Read f2fs raw super block.
3036  * Because we have two copies of super block, so read both of them
3037  * to get the first valid one. If any one of them is broken, we pass
3038  * them recovery flag back to the caller.
3039  */
3040 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3041                         struct f2fs_super_block **raw_super,
3042                         int *valid_super_block, int *recovery)
3043 {
3044         struct super_block *sb = sbi->sb;
3045         int block;
3046         struct buffer_head *bh;
3047         struct f2fs_super_block *super;
3048         int err = 0;
3049
3050         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3051         if (!super)
3052                 return -ENOMEM;
3053
3054         for (block = 0; block < 2; block++) {
3055                 bh = sb_bread(sb, block);
3056                 if (!bh) {
3057                         f2fs_err(sbi, "Unable to read %dth superblock",
3058                                  block + 1);
3059                         err = -EIO;
3060                         *recovery = 1;
3061                         continue;
3062                 }
3063
3064                 /* sanity checking of raw super */
3065                 err = sanity_check_raw_super(sbi, bh);
3066                 if (err) {
3067                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3068                                  block + 1);
3069                         brelse(bh);
3070                         *recovery = 1;
3071                         continue;
3072                 }
3073
3074                 if (!*raw_super) {
3075                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3076                                                         sizeof(*super));
3077                         *valid_super_block = block;
3078                         *raw_super = super;
3079                 }
3080                 brelse(bh);
3081         }
3082
3083         /* No valid superblock */
3084         if (!*raw_super)
3085                 kvfree(super);
3086         else
3087                 err = 0;
3088
3089         return err;
3090 }
3091
3092 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3093 {
3094         struct buffer_head *bh;
3095         __u32 crc = 0;
3096         int err;
3097
3098         if ((recover && f2fs_readonly(sbi->sb)) ||
3099                                 bdev_read_only(sbi->sb->s_bdev)) {
3100                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3101                 return -EROFS;
3102         }
3103
3104         /* we should update superblock crc here */
3105         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3106                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3107                                 offsetof(struct f2fs_super_block, crc));
3108                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3109         }
3110
3111         /* write back-up superblock first */
3112         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3113         if (!bh)
3114                 return -EIO;
3115         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3116         brelse(bh);
3117
3118         /* if we are in recovery path, skip writing valid superblock */
3119         if (recover || err)
3120                 return err;
3121
3122         /* write current valid superblock */
3123         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3124         if (!bh)
3125                 return -EIO;
3126         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3127         brelse(bh);
3128         return err;
3129 }
3130
3131 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3132 {
3133         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3134         unsigned int max_devices = MAX_DEVICES;
3135         int i;
3136
3137         /* Initialize single device information */
3138         if (!RDEV(0).path[0]) {
3139                 if (!bdev_is_zoned(sbi->sb->s_bdev))
3140                         return 0;
3141                 max_devices = 1;
3142         }
3143
3144         /*
3145          * Initialize multiple devices information, or single
3146          * zoned block device information.
3147          */
3148         sbi->devs = f2fs_kzalloc(sbi,
3149                                  array_size(max_devices,
3150                                             sizeof(struct f2fs_dev_info)),
3151                                  GFP_KERNEL);
3152         if (!sbi->devs)
3153                 return -ENOMEM;
3154
3155         for (i = 0; i < max_devices; i++) {
3156
3157                 if (i > 0 && !RDEV(i).path[0])
3158                         break;
3159
3160                 if (max_devices == 1) {
3161                         /* Single zoned block device mount */
3162                         FDEV(0).bdev =
3163                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3164                                         sbi->sb->s_mode, sbi->sb->s_type);
3165                 } else {
3166                         /* Multi-device mount */
3167                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3168                         FDEV(i).total_segments =
3169                                 le32_to_cpu(RDEV(i).total_segments);
3170                         if (i == 0) {
3171                                 FDEV(i).start_blk = 0;
3172                                 FDEV(i).end_blk = FDEV(i).start_blk +
3173                                     (FDEV(i).total_segments <<
3174                                     sbi->log_blocks_per_seg) - 1 +
3175                                     le32_to_cpu(raw_super->segment0_blkaddr);
3176                         } else {
3177                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3178                                 FDEV(i).end_blk = FDEV(i).start_blk +
3179                                         (FDEV(i).total_segments <<
3180                                         sbi->log_blocks_per_seg) - 1;
3181                         }
3182                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3183                                         sbi->sb->s_mode, sbi->sb->s_type);
3184                 }
3185                 if (IS_ERR(FDEV(i).bdev))
3186                         return PTR_ERR(FDEV(i).bdev);
3187
3188                 /* to release errored devices */
3189                 sbi->s_ndevs = i + 1;
3190
3191 #ifdef CONFIG_BLK_DEV_ZONED
3192                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3193                                 !f2fs_sb_has_blkzoned(sbi)) {
3194                         f2fs_err(sbi, "Zoned block device feature not enabled\n");
3195                         return -EINVAL;
3196                 }
3197                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3198                         if (init_blkz_info(sbi, i)) {
3199                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3200                                 return -EINVAL;
3201                         }
3202                         if (max_devices == 1)
3203                                 break;
3204                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3205                                   i, FDEV(i).path,
3206                                   FDEV(i).total_segments,
3207                                   FDEV(i).start_blk, FDEV(i).end_blk,
3208                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3209                                   "Host-aware" : "Host-managed");
3210                         continue;
3211                 }
3212 #endif
3213                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3214                           i, FDEV(i).path,
3215                           FDEV(i).total_segments,
3216                           FDEV(i).start_blk, FDEV(i).end_blk);
3217         }
3218         f2fs_info(sbi,
3219                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3220         return 0;
3221 }
3222
3223 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3224 {
3225 #ifdef CONFIG_UNICODE
3226         if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3227                 const struct f2fs_sb_encodings *encoding_info;
3228                 struct unicode_map *encoding;
3229                 __u16 encoding_flags;
3230
3231                 if (f2fs_sb_has_encrypt(sbi)) {
3232                         f2fs_err(sbi,
3233                                 "Can't mount with encoding and encryption");
3234                         return -EINVAL;
3235                 }
3236
3237                 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3238                                           &encoding_flags)) {
3239                         f2fs_err(sbi,
3240                                  "Encoding requested by superblock is unknown");
3241                         return -EINVAL;
3242                 }
3243
3244                 encoding = utf8_load(encoding_info->version);
3245                 if (IS_ERR(encoding)) {
3246                         f2fs_err(sbi,
3247                                  "can't mount with superblock charset: %s-%s "
3248                                  "not supported by the kernel. flags: 0x%x.",
3249                                  encoding_info->name, encoding_info->version,
3250                                  encoding_flags);
3251                         return PTR_ERR(encoding);
3252                 }
3253                 f2fs_info(sbi, "Using encoding defined by superblock: "
3254                          "%s-%s with flags 0x%hx", encoding_info->name,
3255                          encoding_info->version?:"\b", encoding_flags);
3256
3257                 sbi->s_encoding = encoding;
3258                 sbi->s_encoding_flags = encoding_flags;
3259                 sbi->sb->s_d_op = &f2fs_dentry_ops;
3260         }
3261 #else
3262         if (f2fs_sb_has_casefold(sbi)) {
3263                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3264                 return -EINVAL;
3265         }
3266 #endif
3267         return 0;
3268 }
3269
3270 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3271 {
3272         struct f2fs_sm_info *sm_i = SM_I(sbi);
3273
3274         /* adjust parameters according to the volume size */
3275         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3276                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3277                 sm_i->dcc_info->discard_granularity = 1;
3278                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3279         }
3280
3281         sbi->readdir_ra = 1;
3282 }
3283
3284 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3285 {
3286         struct f2fs_sb_info *sbi;
3287         struct f2fs_super_block *raw_super;
3288         struct inode *root;
3289         int err;
3290         bool skip_recovery = false, need_fsck = false;
3291         char *options = NULL;
3292         int recovery, i, valid_super_block;
3293         struct curseg_info *seg_i;
3294         int retry_cnt = 1;
3295
3296 try_onemore:
3297         err = -EINVAL;
3298         raw_super = NULL;
3299         valid_super_block = -1;
3300         recovery = 0;
3301
3302         /* allocate memory for f2fs-specific super block info */
3303         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3304         if (!sbi)
3305                 return -ENOMEM;
3306
3307         sbi->sb = sb;
3308
3309         /* Load the checksum driver */
3310         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3311         if (IS_ERR(sbi->s_chksum_driver)) {
3312                 f2fs_err(sbi, "Cannot load crc32 driver.");
3313                 err = PTR_ERR(sbi->s_chksum_driver);
3314                 sbi->s_chksum_driver = NULL;
3315                 goto free_sbi;
3316         }
3317
3318         /* set a block size */
3319         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3320                 f2fs_err(sbi, "unable to set blocksize");
3321                 goto free_sbi;
3322         }
3323
3324         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3325                                                                 &recovery);
3326         if (err)
3327                 goto free_sbi;
3328
3329         sb->s_fs_info = sbi;
3330         sbi->raw_super = raw_super;
3331
3332         /* precompute checksum seed for metadata */
3333         if (f2fs_sb_has_inode_chksum(sbi))
3334                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3335                                                 sizeof(raw_super->uuid));
3336
3337         /*
3338          * The BLKZONED feature indicates that the drive was formatted with
3339          * zone alignment optimization. This is optional for host-aware
3340          * devices, but mandatory for host-managed zoned block devices.
3341          */
3342 #ifndef CONFIG_BLK_DEV_ZONED
3343         if (f2fs_sb_has_blkzoned(sbi)) {
3344                 f2fs_err(sbi, "Zoned block device support is not enabled");
3345                 err = -EOPNOTSUPP;
3346                 goto free_sb_buf;
3347         }
3348 #endif
3349         default_options(sbi);
3350         /* parse mount options */
3351         options = kstrdup((const char *)data, GFP_KERNEL);
3352         if (data && !options) {
3353                 err = -ENOMEM;
3354                 goto free_sb_buf;
3355         }
3356
3357         err = parse_options(sb, options);
3358         if (err)
3359                 goto free_options;
3360
3361         sbi->max_file_blocks = max_file_blocks();
3362         sb->s_maxbytes = sbi->max_file_blocks <<
3363                                 le32_to_cpu(raw_super->log_blocksize);
3364         sb->s_max_links = F2FS_LINK_MAX;
3365
3366         err = f2fs_setup_casefold(sbi);
3367         if (err)
3368                 goto free_options;
3369
3370 #ifdef CONFIG_QUOTA
3371         sb->dq_op = &f2fs_quota_operations;
3372         sb->s_qcop = &f2fs_quotactl_ops;
3373         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3374
3375         if (f2fs_sb_has_quota_ino(sbi)) {
3376                 for (i = 0; i < MAXQUOTAS; i++) {
3377                         if (f2fs_qf_ino(sbi->sb, i))
3378                                 sbi->nquota_files++;
3379                 }
3380         }
3381 #endif
3382
3383         sb->s_op = &f2fs_sops;
3384 #ifdef CONFIG_FS_ENCRYPTION
3385         sb->s_cop = &f2fs_cryptops;
3386 #endif
3387 #ifdef CONFIG_FS_VERITY
3388         sb->s_vop = &f2fs_verityops;
3389 #endif
3390         sb->s_xattr = f2fs_xattr_handlers;
3391         sb->s_export_op = &f2fs_export_ops;
3392         sb->s_magic = F2FS_SUPER_MAGIC;
3393         sb->s_time_gran = 1;
3394         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3395                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3396         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3397         sb->s_iflags |= SB_I_CGROUPWB;
3398
3399         /* init f2fs-specific super block info */
3400         sbi->valid_super_block = valid_super_block;
3401         init_rwsem(&sbi->gc_lock);
3402         mutex_init(&sbi->writepages);
3403         mutex_init(&sbi->cp_mutex);
3404         mutex_init(&sbi->resize_mutex);
3405         init_rwsem(&sbi->node_write);
3406         init_rwsem(&sbi->node_change);
3407
3408         /* disallow all the data/node/meta page writes */
3409         set_sbi_flag(sbi, SBI_POR_DOING);
3410         spin_lock_init(&sbi->stat_lock);
3411
3412         /* init iostat info */
3413         spin_lock_init(&sbi->iostat_lock);
3414         sbi->iostat_enable = false;
3415
3416         for (i = 0; i < NR_PAGE_TYPE; i++) {
3417                 int n = (i == META) ? 1: NR_TEMP_TYPE;
3418                 int j;
3419
3420                 sbi->write_io[i] =
3421                         f2fs_kmalloc(sbi,
3422                                      array_size(n,
3423                                                 sizeof(struct f2fs_bio_info)),
3424                                      GFP_KERNEL);
3425                 if (!sbi->write_io[i]) {
3426                         err = -ENOMEM;
3427                         goto free_bio_info;
3428                 }
3429
3430                 for (j = HOT; j < n; j++) {
3431                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
3432                         sbi->write_io[i][j].sbi = sbi;
3433                         sbi->write_io[i][j].bio = NULL;
3434                         spin_lock_init(&sbi->write_io[i][j].io_lock);
3435                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3436                         INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3437                         init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3438                 }
3439         }
3440
3441         init_rwsem(&sbi->cp_rwsem);
3442         init_rwsem(&sbi->quota_sem);
3443         init_waitqueue_head(&sbi->cp_wait);
3444         init_sb_info(sbi);
3445
3446         err = init_percpu_info(sbi);
3447         if (err)
3448                 goto free_bio_info;
3449
3450         if (F2FS_IO_ALIGNED(sbi)) {
3451                 sbi->write_io_dummy =
3452                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3453                 if (!sbi->write_io_dummy) {
3454                         err = -ENOMEM;
3455                         goto free_percpu;
3456                 }
3457         }
3458
3459         /* get an inode for meta space */
3460         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3461         if (IS_ERR(sbi->meta_inode)) {
3462                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
3463                 err = PTR_ERR(sbi->meta_inode);
3464                 goto free_io_dummy;
3465         }
3466
3467         err = f2fs_get_valid_checkpoint(sbi);
3468         if (err) {
3469                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3470                 goto free_meta_inode;
3471         }
3472
3473         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3474                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3475         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3476                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3477                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3478         }
3479
3480         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3481                 set_sbi_flag(sbi, SBI_NEED_FSCK);
3482
3483         /* Initialize device list */
3484         err = f2fs_scan_devices(sbi);
3485         if (err) {
3486                 f2fs_err(sbi, "Failed to find devices");
3487                 goto free_devices;
3488         }
3489
3490         err = f2fs_init_post_read_wq(sbi);
3491         if (err) {
3492                 f2fs_err(sbi, "Failed to initialize post read workqueue");
3493                 goto free_devices;
3494         }
3495
3496         sbi->total_valid_node_count =
3497                                 le32_to_cpu(sbi->ckpt->valid_node_count);
3498         percpu_counter_set(&sbi->total_valid_inode_count,
3499                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
3500         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3501         sbi->total_valid_block_count =
3502                                 le64_to_cpu(sbi->ckpt->valid_block_count);
3503         sbi->last_valid_block_count = sbi->total_valid_block_count;
3504         sbi->reserved_blocks = 0;
3505         sbi->current_reserved_blocks = 0;
3506         limit_reserve_root(sbi);
3507
3508         for (i = 0; i < NR_INODE_TYPE; i++) {
3509                 INIT_LIST_HEAD(&sbi->inode_list[i]);
3510                 spin_lock_init(&sbi->inode_lock[i]);
3511         }
3512         mutex_init(&sbi->flush_lock);
3513
3514         f2fs_init_extent_cache_info(sbi);
3515
3516         f2fs_init_ino_entry_info(sbi);
3517
3518         f2fs_init_fsync_node_info(sbi);
3519
3520         /* setup f2fs internal modules */
3521         err = f2fs_build_segment_manager(sbi);
3522         if (err) {
3523                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3524                          err);
3525                 goto free_sm;
3526         }
3527         err = f2fs_build_node_manager(sbi);
3528         if (err) {
3529                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3530                          err);
3531                 goto free_nm;
3532         }
3533
3534         /* For write statistics */
3535         if (sb->s_bdev->bd_part)
3536                 sbi->sectors_written_start =
3537                         (u64)part_stat_read(sb->s_bdev->bd_part,
3538                                             sectors[STAT_WRITE]);
3539
3540         /* Read accumulated write IO statistics if exists */
3541         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3542         if (__exist_node_summaries(sbi))
3543                 sbi->kbytes_written =
3544                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3545
3546         f2fs_build_gc_manager(sbi);
3547
3548         err = f2fs_build_stats(sbi);
3549         if (err)
3550                 goto free_nm;
3551
3552         /* get an inode for node space */
3553         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3554         if (IS_ERR(sbi->node_inode)) {
3555                 f2fs_err(sbi, "Failed to read node inode");
3556                 err = PTR_ERR(sbi->node_inode);
3557                 goto free_stats;
3558         }
3559
3560         /* read root inode and dentry */
3561         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3562         if (IS_ERR(root)) {
3563                 f2fs_err(sbi, "Failed to read root inode");
3564                 err = PTR_ERR(root);
3565                 goto free_node_inode;
3566         }
3567         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3568                         !root->i_size || !root->i_nlink) {
3569                 iput(root);
3570                 err = -EINVAL;
3571                 goto free_node_inode;
3572         }
3573
3574         sb->s_root = d_make_root(root); /* allocate root dentry */
3575         if (!sb->s_root) {
3576                 err = -ENOMEM;
3577                 goto free_node_inode;
3578         }
3579
3580         err = f2fs_register_sysfs(sbi);
3581         if (err)
3582                 goto free_root_inode;
3583
3584 #ifdef CONFIG_QUOTA
3585         /* Enable quota usage during mount */
3586         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3587                 err = f2fs_enable_quotas(sb);
3588                 if (err)
3589                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3590         }
3591 #endif
3592         /* if there are nt orphan nodes free them */
3593         err = f2fs_recover_orphan_inodes(sbi);
3594         if (err)
3595                 goto free_meta;
3596
3597         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3598                 goto reset_checkpoint;
3599
3600         /* recover fsynced data */
3601         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3602                 /*
3603                  * mount should be failed, when device has readonly mode, and
3604                  * previous checkpoint was not done by clean system shutdown.
3605                  */
3606                 if (f2fs_hw_is_readonly(sbi)) {
3607                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3608                                 err = -EROFS;
3609                                 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3610                                 goto free_meta;
3611                         }
3612                         f2fs_info(sbi, "write access unavailable, skipping recovery");
3613                         goto reset_checkpoint;
3614                 }
3615
3616                 if (need_fsck)
3617                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3618
3619                 if (skip_recovery)
3620                         goto reset_checkpoint;
3621
3622                 err = f2fs_recover_fsync_data(sbi, false);
3623                 if (err < 0) {
3624                         if (err != -ENOMEM)
3625                                 skip_recovery = true;
3626                         need_fsck = true;
3627                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3628                                  err);
3629                         goto free_meta;
3630                 }
3631         } else {
3632                 err = f2fs_recover_fsync_data(sbi, true);
3633
3634                 if (!f2fs_readonly(sb) && err > 0) {
3635                         err = -EINVAL;
3636                         f2fs_err(sbi, "Need to recover fsync data");
3637                         goto free_meta;
3638                 }
3639         }
3640
3641         /*
3642          * If the f2fs is not readonly and fsync data recovery succeeds,
3643          * check zoned block devices' write pointer consistency.
3644          */
3645         if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3646                 err = f2fs_check_write_pointer(sbi);
3647                 if (err)
3648                         goto free_meta;
3649         }
3650
3651 reset_checkpoint:
3652         /* f2fs_recover_fsync_data() cleared this already */
3653         clear_sbi_flag(sbi, SBI_POR_DOING);
3654
3655         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3656                 err = f2fs_disable_checkpoint(sbi);
3657                 if (err)
3658                         goto sync_free_meta;
3659         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3660                 f2fs_enable_checkpoint(sbi);
3661         }
3662
3663         /*
3664          * If filesystem is not mounted as read-only then
3665          * do start the gc_thread.
3666          */
3667         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3668                 /* After POR, we can run background GC thread.*/
3669                 err = f2fs_start_gc_thread(sbi);
3670                 if (err)
3671                         goto sync_free_meta;
3672         }
3673         kvfree(options);
3674
3675         /* recover broken superblock */
3676         if (recovery) {
3677                 err = f2fs_commit_super(sbi, true);
3678                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3679                           sbi->valid_super_block ? 1 : 2, err);
3680         }
3681
3682         f2fs_join_shrinker(sbi);
3683
3684         f2fs_tuning_parameters(sbi);
3685
3686         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3687                     cur_cp_version(F2FS_CKPT(sbi)));
3688         f2fs_update_time(sbi, CP_TIME);
3689         f2fs_update_time(sbi, REQ_TIME);
3690         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3691         return 0;
3692
3693 sync_free_meta:
3694         /* safe to flush all the data */
3695         sync_filesystem(sbi->sb);
3696         retry_cnt = 0;
3697
3698 free_meta:
3699 #ifdef CONFIG_QUOTA
3700         f2fs_truncate_quota_inode_pages(sb);
3701         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3702                 f2fs_quota_off_umount(sbi->sb);
3703 #endif
3704         /*
3705          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3706          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3707          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3708          * falls into an infinite loop in f2fs_sync_meta_pages().
3709          */
3710         truncate_inode_pages_final(META_MAPPING(sbi));
3711         /* evict some inodes being cached by GC */
3712         evict_inodes(sb);
3713         f2fs_unregister_sysfs(sbi);
3714 free_root_inode:
3715         dput(sb->s_root);
3716         sb->s_root = NULL;
3717 free_node_inode:
3718         f2fs_release_ino_entry(sbi, true);
3719         truncate_inode_pages_final(NODE_MAPPING(sbi));
3720         iput(sbi->node_inode);
3721         sbi->node_inode = NULL;
3722 free_stats:
3723         f2fs_destroy_stats(sbi);
3724 free_nm:
3725         f2fs_destroy_node_manager(sbi);
3726 free_sm:
3727         f2fs_destroy_segment_manager(sbi);
3728         f2fs_destroy_post_read_wq(sbi);
3729 free_devices:
3730         destroy_device_list(sbi);
3731         kvfree(sbi->ckpt);
3732 free_meta_inode:
3733         make_bad_inode(sbi->meta_inode);
3734         iput(sbi->meta_inode);
3735         sbi->meta_inode = NULL;
3736 free_io_dummy:
3737         mempool_destroy(sbi->write_io_dummy);
3738 free_percpu:
3739         destroy_percpu_info(sbi);
3740 free_bio_info:
3741         for (i = 0; i < NR_PAGE_TYPE; i++)
3742                 kvfree(sbi->write_io[i]);
3743
3744 #ifdef CONFIG_UNICODE
3745         utf8_unload(sbi->s_encoding);
3746 #endif
3747 free_options:
3748 #ifdef CONFIG_QUOTA
3749         for (i = 0; i < MAXQUOTAS; i++)
3750                 kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3751 #endif
3752         kvfree(options);
3753 free_sb_buf:
3754         kvfree(raw_super);
3755 free_sbi:
3756         if (sbi->s_chksum_driver)
3757                 crypto_free_shash(sbi->s_chksum_driver);
3758         kvfree(sbi);
3759
3760         /* give only one another chance */
3761         if (retry_cnt > 0 && skip_recovery) {
3762                 retry_cnt--;
3763                 shrink_dcache_sb(sb);
3764                 goto try_onemore;
3765         }
3766         return err;
3767 }
3768
3769 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3770                         const char *dev_name, void *data)
3771 {
3772         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3773 }
3774
3775 static void kill_f2fs_super(struct super_block *sb)
3776 {
3777         if (sb->s_root) {
3778                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3779
3780                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3781                 f2fs_stop_gc_thread(sbi);
3782                 f2fs_stop_discard_thread(sbi);
3783
3784                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3785                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3786                         struct cp_control cpc = {
3787                                 .reason = CP_UMOUNT,
3788                         };
3789                         f2fs_write_checkpoint(sbi, &cpc);
3790                 }
3791
3792                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3793                         sb->s_flags &= ~SB_RDONLY;
3794         }
3795         kill_block_super(sb);
3796 }
3797
3798 static struct file_system_type f2fs_fs_type = {
3799         .owner          = THIS_MODULE,
3800         .name           = "f2fs",
3801         .mount          = f2fs_mount,
3802         .kill_sb        = kill_f2fs_super,
3803         .fs_flags       = FS_REQUIRES_DEV,
3804 };
3805 MODULE_ALIAS_FS("f2fs");
3806
3807 static int __init init_inodecache(void)
3808 {
3809         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3810                         sizeof(struct f2fs_inode_info), 0,
3811                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3812         if (!f2fs_inode_cachep)
3813                 return -ENOMEM;
3814         return 0;
3815 }
3816
3817 static void destroy_inodecache(void)
3818 {
3819         /*
3820          * Make sure all delayed rcu free inodes are flushed before we
3821          * destroy cache.
3822          */
3823         rcu_barrier();
3824         kmem_cache_destroy(f2fs_inode_cachep);
3825 }
3826
3827 static int __init init_f2fs_fs(void)
3828 {
3829         int err;
3830
3831         if (PAGE_SIZE != F2FS_BLKSIZE) {
3832                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3833                                 PAGE_SIZE, F2FS_BLKSIZE);
3834                 return -EINVAL;
3835         }
3836
3837         f2fs_build_trace_ios();
3838
3839         err = init_inodecache();
3840         if (err)
3841                 goto fail;
3842         err = f2fs_create_node_manager_caches();
3843         if (err)
3844                 goto free_inodecache;
3845         err = f2fs_create_segment_manager_caches();
3846         if (err)
3847                 goto free_node_manager_caches;
3848         err = f2fs_create_checkpoint_caches();
3849         if (err)
3850                 goto free_segment_manager_caches;
3851         err = f2fs_create_extent_cache();
3852         if (err)
3853                 goto free_checkpoint_caches;
3854         err = f2fs_init_sysfs();
3855         if (err)
3856                 goto free_extent_cache;
3857         err = register_shrinker(&f2fs_shrinker_info);
3858         if (err)
3859                 goto free_sysfs;
3860         err = register_filesystem(&f2fs_fs_type);
3861         if (err)
3862                 goto free_shrinker;
3863         f2fs_create_root_stats();
3864         err = f2fs_init_post_read_processing();
3865         if (err)
3866                 goto free_root_stats;
3867         err = f2fs_init_bio_entry_cache();
3868         if (err)
3869                 goto free_post_read;
3870         err = f2fs_init_bioset();
3871         if (err)
3872                 goto free_bio_enrty_cache;
3873         return 0;
3874 free_bio_enrty_cache:
3875         f2fs_destroy_bio_entry_cache();
3876 free_post_read:
3877         f2fs_destroy_post_read_processing();
3878 free_root_stats:
3879         f2fs_destroy_root_stats();
3880         unregister_filesystem(&f2fs_fs_type);
3881 free_shrinker:
3882         unregister_shrinker(&f2fs_shrinker_info);
3883 free_sysfs:
3884         f2fs_exit_sysfs();
3885 free_extent_cache:
3886         f2fs_destroy_extent_cache();
3887 free_checkpoint_caches:
3888         f2fs_destroy_checkpoint_caches();
3889 free_segment_manager_caches:
3890         f2fs_destroy_segment_manager_caches();
3891 free_node_manager_caches:
3892         f2fs_destroy_node_manager_caches();
3893 free_inodecache:
3894         destroy_inodecache();
3895 fail:
3896         return err;
3897 }
3898
3899 static void __exit exit_f2fs_fs(void)
3900 {
3901         f2fs_destroy_bioset();
3902         f2fs_destroy_bio_entry_cache();
3903         f2fs_destroy_post_read_processing();
3904         f2fs_destroy_root_stats();
3905         unregister_filesystem(&f2fs_fs_type);
3906         unregister_shrinker(&f2fs_shrinker_info);
3907         f2fs_exit_sysfs();
3908         f2fs_destroy_extent_cache();
3909         f2fs_destroy_checkpoint_caches();
3910         f2fs_destroy_segment_manager_caches();
3911         f2fs_destroy_node_manager_caches();
3912         destroy_inodecache();
3913         f2fs_destroy_trace_ios();
3914 }
3915
3916 module_init(init_f2fs_fs)
3917 module_exit(exit_f2fs_fs)
3918
3919 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3920 MODULE_DESCRIPTION("Flash Friendly File System");
3921 MODULE_LICENSE("GPL");
3922