fs/f2fs/super.c

   1 /*
   2  * fs/f2fs/super.c
   3  *
   4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   5  *             http://www.samsung.com/
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  */
  11 #include <linux/module.h>
  12 #include <linux/init.h>
  13 #include <linux/fs.h>
  14 #include <linux/statfs.h>
  15 #include <linux/buffer_head.h>
  16 #include <linux/backing-dev.h>
  17 #include <linux/kthread.h>
  18 #include <linux/parser.h>
  19 #include <linux/mount.h>
  20 #include <linux/seq_file.h>
  21 #include <linux/proc_fs.h>
  22 #include <linux/random.h>
  23 #include <linux/exportfs.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/quotaops.h>
  26 #include <linux/f2fs_fs.h>
  27 #include <linux/sysfs.h>
  28 #include <linux/quota.h>
  29
  30 #include "f2fs.h"
  31 #include "node.h"
  32 #include "segment.h"
  33 #include "xattr.h"
  34 #include "gc.h"
  35 #include "trace.h"
  36
  37 #define CREATE_TRACE_POINTS
  38 #include <trace/events/f2fs.h>
  39
  40 static struct kmem_cache *f2fs_inode_cachep;
  41
  42 #ifdef CONFIG_F2FS_FAULT_INJECTION
  43
  44 char *fault_name[FAULT_MAX] = {
  45         [FAULT_KMALLOC]         = "kmalloc",
  46         [FAULT_KVMALLOC]        = "kvmalloc",
  47         [FAULT_PAGE_ALLOC]      = "page alloc",
  48         [FAULT_PAGE_GET]        = "page get",
  49         [FAULT_ALLOC_BIO]       = "alloc bio",
  50         [FAULT_ALLOC_NID]       = "alloc nid",
  51         [FAULT_ORPHAN]          = "orphan",
  52         [FAULT_BLOCK]           = "no more block",
  53         [FAULT_DIR_DEPTH]       = "too big dir depth",
  54         [FAULT_EVICT_INODE]     = "evict_inode fail",
  55         [FAULT_TRUNCATE]        = "truncate fail",
  56         [FAULT_IO]              = "IO error",
  57         [FAULT_CHECKPOINT]      = "checkpoint error",
  58 };
  59
  60 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
  61                                                 unsigned int rate)
  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                 ffi->inject_type = (1 << FAULT_MAX) - 1;
  69         } else {
  70                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  71         }
  72 }
  73 #endif
  74
  75 /* f2fs-wide shrinker description */
  76 static struct shrinker f2fs_shrinker_info = {
  77         .scan_objects = f2fs_shrink_scan,
  78         .count_objects = f2fs_shrink_count,
  79         .seeks = DEFAULT_SEEKS,
  80 };
  81
  82 enum {
  83         Opt_gc_background,
  84         Opt_disable_roll_forward,
  85         Opt_norecovery,
  86         Opt_discard,
  87         Opt_nodiscard,
  88         Opt_noheap,
  89         Opt_heap,
  90         Opt_user_xattr,
  91         Opt_nouser_xattr,
  92         Opt_acl,
  93         Opt_noacl,
  94         Opt_active_logs,
  95         Opt_disable_ext_identify,
  96         Opt_inline_xattr,
  97         Opt_noinline_xattr,
  98         Opt_inline_xattr_size,
  99         Opt_inline_data,
 100         Opt_inline_dentry,
 101         Opt_noinline_dentry,
 102         Opt_flush_merge,
 103         Opt_noflush_merge,
 104         Opt_nobarrier,
 105         Opt_fastboot,
 106         Opt_extent_cache,
 107         Opt_noextent_cache,
 108         Opt_noinline_data,
 109         Opt_data_flush,
 110         Opt_reserve_root,
 111         Opt_resgid,
 112         Opt_resuid,
 113         Opt_mode,
 114         Opt_io_size_bits,
 115         Opt_fault_injection,
 116         Opt_lazytime,
 117         Opt_nolazytime,
 118         Opt_quota,
 119         Opt_noquota,
 120         Opt_usrquota,
 121         Opt_grpquota,
 122         Opt_prjquota,
 123         Opt_usrjquota,
 124         Opt_grpjquota,
 125         Opt_prjjquota,
 126         Opt_offusrjquota,
 127         Opt_offgrpjquota,
 128         Opt_offprjjquota,
 129         Opt_jqfmt_vfsold,
 130         Opt_jqfmt_vfsv0,
 131         Opt_jqfmt_vfsv1,
 132         Opt_whint,
 133         Opt_alloc,
 134         Opt_fsync,
 135         Opt_test_dummy_encryption,
 136         Opt_err,
 137 };
 138
 139 static match_table_t f2fs_tokens = {
 140         {Opt_gc_background, "background_gc=%s"},
 141         {Opt_disable_roll_forward, "disable_roll_forward"},
 142         {Opt_norecovery, "norecovery"},
 143         {Opt_discard, "discard"},
 144         {Opt_nodiscard, "nodiscard"},
 145         {Opt_noheap, "no_heap"},
 146         {Opt_heap, "heap"},
 147         {Opt_user_xattr, "user_xattr"},
 148         {Opt_nouser_xattr, "nouser_xattr"},
 149         {Opt_acl, "acl"},
 150         {Opt_noacl, "noacl"},
 151         {Opt_active_logs, "active_logs=%u"},
 152         {Opt_disable_ext_identify, "disable_ext_identify"},
 153         {Opt_inline_xattr, "inline_xattr"},
 154         {Opt_noinline_xattr, "noinline_xattr"},
 155         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 156         {Opt_inline_data, "inline_data"},
 157         {Opt_inline_dentry, "inline_dentry"},
 158         {Opt_noinline_dentry, "noinline_dentry"},
 159         {Opt_flush_merge, "flush_merge"},
 160         {Opt_noflush_merge, "noflush_merge"},
 161         {Opt_nobarrier, "nobarrier"},
 162         {Opt_fastboot, "fastboot"},
 163         {Opt_extent_cache, "extent_cache"},
 164         {Opt_noextent_cache, "noextent_cache"},
 165         {Opt_noinline_data, "noinline_data"},
 166         {Opt_data_flush, "data_flush"},
 167         {Opt_reserve_root, "reserve_root=%u"},
 168         {Opt_resgid, "resgid=%u"},
 169         {Opt_resuid, "resuid=%u"},
 170         {Opt_mode, "mode=%s"},
 171         {Opt_io_size_bits, "io_bits=%u"},
 172         {Opt_fault_injection, "fault_injection=%u"},
 173         {Opt_lazytime, "lazytime"},
 174         {Opt_nolazytime, "nolazytime"},
 175         {Opt_quota, "quota"},
 176         {Opt_noquota, "noquota"},
 177         {Opt_usrquota, "usrquota"},
 178         {Opt_grpquota, "grpquota"},
 179         {Opt_prjquota, "prjquota"},
 180         {Opt_usrjquota, "usrjquota=%s"},
 181         {Opt_grpjquota, "grpjquota=%s"},
 182         {Opt_prjjquota, "prjjquota=%s"},
 183         {Opt_offusrjquota, "usrjquota="},
 184         {Opt_offgrpjquota, "grpjquota="},
 185         {Opt_offprjjquota, "prjjquota="},
 186         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 187         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 188         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 189         {Opt_whint, "whint_mode=%s"},
 190         {Opt_alloc, "alloc_mode=%s"},
 191         {Opt_fsync, "fsync_mode=%s"},
 192         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 193         {Opt_err, NULL},
 194 };
 195
 196 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 197 {
 198         struct va_format vaf;
 199         va_list args;
 200
 201         va_start(args, fmt);
 202         vaf.fmt = fmt;
 203         vaf.va = &args;
 204         printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 205         va_end(args);
 206 }
 207
 208 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 209 {
 210         block_t limit = (sbi->user_block_count << 1) / 1000;
 211
 212         /* limit is 0.2% */
 213         if (test_opt(sbi, RESERVE_ROOT) &&
 214                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 215                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 216                 f2fs_msg(sbi->sb, KERN_INFO,
 217                         "Reduce reserved blocks for root = %u",
 218                         F2FS_OPTION(sbi).root_reserved_blocks);
 219         }
 220         if (!test_opt(sbi, RESERVE_ROOT) &&
 221                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 222                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 223                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 224                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 225                 f2fs_msg(sbi->sb, KERN_INFO,
 226                         "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 227                                 from_kuid_munged(&init_user_ns,
 228                                         F2FS_OPTION(sbi).s_resuid),
 229                                 from_kgid_munged(&init_user_ns,
 230                                         F2FS_OPTION(sbi).s_resgid));
 231 }
 232
 233 static void init_once(void *foo)
 234 {
 235         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 236
 237         inode_init_once(&fi->vfs_inode);
 238 }
 239
 240 #ifdef CONFIG_QUOTA
 241 static const char * const quotatypes[] = INITQFNAMES;
 242 #define QTYPE2NAME(t) (quotatypes[t])
 243 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 244                                                         substring_t *args)
 245 {
 246         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 247         char *qname;
 248         int ret = -EINVAL;
 249
 250         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 251                 f2fs_msg(sb, KERN_ERR,
 252                         "Cannot change journaled "
 253                         "quota options when quota turned on");
 254                 return -EINVAL;
 255         }
 256         if (f2fs_sb_has_quota_ino(sb)) {
 257                 f2fs_msg(sb, KERN_INFO,
 258                         "QUOTA feature is enabled, so ignore qf_name");
 259                 return 0;
 260         }
 261
 262         qname = match_strdup(args);
 263         if (!qname) {
 264                 f2fs_msg(sb, KERN_ERR,
 265                         "Not enough memory for storing quotafile name");
 266                 return -EINVAL;
 267         }
 268         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 269                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 270                         ret = 0;
 271                 else
 272                         f2fs_msg(sb, KERN_ERR,
 273                                  "%s quota file already specified",
 274                                  QTYPE2NAME(qtype));
 275                 goto errout;
 276         }
 277         if (strchr(qname, '/')) {
 278                 f2fs_msg(sb, KERN_ERR,
 279                         "quotafile must be on filesystem root");
 280                 goto errout;
 281         }
 282         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 283         set_opt(sbi, QUOTA);
 284         return 0;
 285 errout:
 286         kfree(qname);
 287         return ret;
 288 }
 289
 290 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 291 {
 292         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 293
 294         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 295                 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
 296                         " when quota turned on");
 297                 return -EINVAL;
 298         }
 299         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 300         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 301         return 0;
 302 }
 303
 304 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 305 {
 306         /*
 307          * We do the test below only for project quotas. 'usrquota' and
 308          * 'grpquota' mount options are allowed even without quota feature
 309          * to support legacy quotas in quota files.
 310          */
 311         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
 312                 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
 313                          "Cannot enable project quota enforcement.");
 314                 return -1;
 315         }
 316         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 317                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 318                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 319                 if (test_opt(sbi, USRQUOTA) &&
 320                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 321                         clear_opt(sbi, USRQUOTA);
 322
 323                 if (test_opt(sbi, GRPQUOTA) &&
 324                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 325                         clear_opt(sbi, GRPQUOTA);
 326
 327                 if (test_opt(sbi, PRJQUOTA) &&
 328                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 329                         clear_opt(sbi, PRJQUOTA);
 330
 331                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 332                                 test_opt(sbi, PRJQUOTA)) {
 333                         f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
 334                                         "format mixing");
 335                         return -1;
 336                 }
 337
 338                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 339                         f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
 340                                         "not specified");
 341                         return -1;
 342                 }
 343         }
 344
 345         if (f2fs_sb_has_quota_ino(sbi->sb) && F2FS_OPTION(sbi).s_jquota_fmt) {
 346                 f2fs_msg(sbi->sb, KERN_INFO,
 347                         "QUOTA feature is enabled, so ignore jquota_fmt");
 348                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 349         }
 350         if (f2fs_sb_has_quota_ino(sbi->sb) && f2fs_readonly(sbi->sb)) {
 351                 f2fs_msg(sbi->sb, KERN_INFO,
 352                          "Filesystem with quota feature cannot be mounted RDWR "
 353                          "without CONFIG_QUOTA");
 354                 return -1;
 355         }
 356         return 0;
 357 }
 358 #endif
 359
 360 static int parse_options(struct super_block *sb, char *options)
 361 {
 362         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 363         struct request_queue *q;
 364         substring_t args[MAX_OPT_ARGS];
 365         char *p, *name;
 366         int arg = 0;
 367         kuid_t uid;
 368         kgid_t gid;
 369 #ifdef CONFIG_QUOTA
 370         int ret;
 371 #endif
 372
 373         if (!options)
 374                 return 0;
 375
 376         while ((p = strsep(&options, ",")) != NULL) {
 377                 int token;
 378                 if (!*p)
 379                         continue;
 380                 /*
 381                  * Initialize args struct so we know whether arg was
 382                  * found; some options take optional arguments.
 383                  */
 384                 args[0].to = args[0].from = NULL;
 385                 token = match_token(p, f2fs_tokens, args);
 386
 387                 switch (token) {
 388                 case Opt_gc_background:
 389                         name = match_strdup(&args[0]);
 390
 391                         if (!name)
 392                                 return -ENOMEM;
 393                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 394                                 set_opt(sbi, BG_GC);
 395                                 clear_opt(sbi, FORCE_FG_GC);
 396                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 397                                 clear_opt(sbi, BG_GC);
 398                                 clear_opt(sbi, FORCE_FG_GC);
 399                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 400                                 set_opt(sbi, BG_GC);
 401                                 set_opt(sbi, FORCE_FG_GC);
 402                         } else {
 403                                 kfree(name);
 404                                 return -EINVAL;
 405                         }
 406                         kfree(name);
 407                         break;
 408                 case Opt_disable_roll_forward:
 409                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 410                         break;
 411                 case Opt_norecovery:
 412                         /* this option mounts f2fs with ro */
 413                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 414                         if (!f2fs_readonly(sb))
 415                                 return -EINVAL;
 416                         break;
 417                 case Opt_discard:
 418                         q = bdev_get_queue(sb->s_bdev);
 419                         if (blk_queue_discard(q)) {
 420                                 set_opt(sbi, DISCARD);
 421                         } else if (!f2fs_sb_has_blkzoned(sb)) {
 422                                 f2fs_msg(sb, KERN_WARNING,
 423                                         "mounting with \"discard\" option, but "
 424                                         "the device does not support discard");
 425                         }
 426                         break;
 427                 case Opt_nodiscard:
 428                         if (f2fs_sb_has_blkzoned(sb)) {
 429                                 f2fs_msg(sb, KERN_WARNING,
 430                                         "discard is required for zoned block devices");
 431                                 return -EINVAL;
 432                         }
 433                         clear_opt(sbi, DISCARD);
 434                         break;
 435                 case Opt_noheap:
 436                         set_opt(sbi, NOHEAP);
 437                         break;
 438                 case Opt_heap:
 439                         clear_opt(sbi, NOHEAP);
 440                         break;
 441 #ifdef CONFIG_F2FS_FS_XATTR
 442                 case Opt_user_xattr:
 443                         set_opt(sbi, XATTR_USER);
 444                         break;
 445                 case Opt_nouser_xattr:
 446                         clear_opt(sbi, XATTR_USER);
 447                         break;
 448                 case Opt_inline_xattr:
 449                         set_opt(sbi, INLINE_XATTR);
 450                         break;
 451                 case Opt_noinline_xattr:
 452                         clear_opt(sbi, INLINE_XATTR);
 453                         break;
 454                 case Opt_inline_xattr_size:
 455                         if (args->from && match_int(args, &arg))
 456                                 return -EINVAL;
 457                         set_opt(sbi, INLINE_XATTR_SIZE);
 458                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 459                         break;
 460 #else
 461                 case Opt_user_xattr:
 462                         f2fs_msg(sb, KERN_INFO,
 463                                 "user_xattr options not supported");
 464                         break;
 465                 case Opt_nouser_xattr:
 466                         f2fs_msg(sb, KERN_INFO,
 467                                 "nouser_xattr options not supported");
 468                         break;
 469                 case Opt_inline_xattr:
 470                         f2fs_msg(sb, KERN_INFO,
 471                                 "inline_xattr options not supported");
 472                         break;
 473                 case Opt_noinline_xattr:
 474                         f2fs_msg(sb, KERN_INFO,
 475                                 "noinline_xattr options not supported");
 476                         break;
 477 #endif
 478 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 479                 case Opt_acl:
 480                         set_opt(sbi, POSIX_ACL);
 481                         break;
 482                 case Opt_noacl:
 483                         clear_opt(sbi, POSIX_ACL);
 484                         break;
 485 #else
 486                 case Opt_acl:
 487                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 488                         break;
 489                 case Opt_noacl:
 490                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 491                         break;
 492 #endif
 493                 case Opt_active_logs:
 494                         if (args->from && match_int(args, &arg))
 495                                 return -EINVAL;
 496                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 497                                 return -EINVAL;
 498                         F2FS_OPTION(sbi).active_logs = arg;
 499                         break;
 500                 case Opt_disable_ext_identify:
 501                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 502                         break;
 503                 case Opt_inline_data:
 504                         set_opt(sbi, INLINE_DATA);
 505                         break;
 506                 case Opt_inline_dentry:
 507                         set_opt(sbi, INLINE_DENTRY);
 508                         break;
 509                 case Opt_noinline_dentry:
 510                         clear_opt(sbi, INLINE_DENTRY);
 511                         break;
 512                 case Opt_flush_merge:
 513                         set_opt(sbi, FLUSH_MERGE);
 514                         break;
 515                 case Opt_noflush_merge:
 516                         clear_opt(sbi, FLUSH_MERGE);
 517                         break;
 518                 case Opt_nobarrier:
 519                         set_opt(sbi, NOBARRIER);
 520                         break;
 521                 case Opt_fastboot:
 522                         set_opt(sbi, FASTBOOT);
 523                         break;
 524                 case Opt_extent_cache:
 525                         set_opt(sbi, EXTENT_CACHE);
 526                         break;
 527                 case Opt_noextent_cache:
 528                         clear_opt(sbi, EXTENT_CACHE);
 529                         break;
 530                 case Opt_noinline_data:
 531                         clear_opt(sbi, INLINE_DATA);
 532                         break;
 533                 case Opt_data_flush:
 534                         set_opt(sbi, DATA_FLUSH);
 535                         break;
 536                 case Opt_reserve_root:
 537                         if (args->from && match_int(args, &arg))
 538                                 return -EINVAL;
 539                         if (test_opt(sbi, RESERVE_ROOT)) {
 540                                 f2fs_msg(sb, KERN_INFO,
 541                                         "Preserve previous reserve_root=%u",
 542                                         F2FS_OPTION(sbi).root_reserved_blocks);
 543                         } else {
 544                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 545                                 set_opt(sbi, RESERVE_ROOT);
 546                         }
 547                         break;
 548                 case Opt_resuid:
 549                         if (args->from && match_int(args, &arg))
 550                                 return -EINVAL;
 551                         uid = make_kuid(current_user_ns(), arg);
 552                         if (!uid_valid(uid)) {
 553                                 f2fs_msg(sb, KERN_ERR,
 554                                         "Invalid uid value %d", arg);
 555                                 return -EINVAL;
 556                         }
 557                         F2FS_OPTION(sbi).s_resuid = uid;
 558                         break;
 559                 case Opt_resgid:
 560                         if (args->from && match_int(args, &arg))
 561                                 return -EINVAL;
 562                         gid = make_kgid(current_user_ns(), arg);
 563                         if (!gid_valid(gid)) {
 564                                 f2fs_msg(sb, KERN_ERR,
 565                                         "Invalid gid value %d", arg);
 566                                 return -EINVAL;
 567                         }
 568                         F2FS_OPTION(sbi).s_resgid = gid;
 569                         break;
 570                 case Opt_mode:
 571                         name = match_strdup(&args[0]);
 572
 573                         if (!name)
 574                                 return -ENOMEM;
 575                         if (strlen(name) == 8 &&
 576                                         !strncmp(name, "adaptive", 8)) {
 577                                 if (f2fs_sb_has_blkzoned(sb)) {
 578                                         f2fs_msg(sb, KERN_WARNING,
 579                                                  "adaptive mode is not allowed with "
 580                                                  "zoned block device feature");
 581                                         kfree(name);
 582                                         return -EINVAL;
 583                                 }
 584                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 585                         } else if (strlen(name) == 3 &&
 586                                         !strncmp(name, "lfs", 3)) {
 587                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 588                         } else {
 589                                 kfree(name);
 590                                 return -EINVAL;
 591                         }
 592                         kfree(name);
 593                         break;
 594                 case Opt_io_size_bits:
 595                         if (args->from && match_int(args, &arg))
 596                                 return -EINVAL;
 597                         if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
 598                                 f2fs_msg(sb, KERN_WARNING,
 599                                         "Not support %d, larger than %d",
 600                                         1 << arg, BIO_MAX_PAGES);
 601                                 return -EINVAL;
 602                         }
 603                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 604                         break;
 605                 case Opt_fault_injection:
 606                         if (args->from && match_int(args, &arg))
 607                                 return -EINVAL;
 608 #ifdef CONFIG_F2FS_FAULT_INJECTION
 609                         f2fs_build_fault_attr(sbi, arg);
 610                         set_opt(sbi, FAULT_INJECTION);
 611 #else
 612                         f2fs_msg(sb, KERN_INFO,
 613                                 "FAULT_INJECTION was not selected");
 614 #endif
 615                         break;
 616                 case Opt_lazytime:
 617                         sb->s_flags |= SB_LAZYTIME;
 618                         break;
 619                 case Opt_nolazytime:
 620                         sb->s_flags &= ~SB_LAZYTIME;
 621                         break;
 622 #ifdef CONFIG_QUOTA
 623                 case Opt_quota:
 624                 case Opt_usrquota:
 625                         set_opt(sbi, USRQUOTA);
 626                         break;
 627                 case Opt_grpquota:
 628                         set_opt(sbi, GRPQUOTA);
 629                         break;
 630                 case Opt_prjquota:
 631                         set_opt(sbi, PRJQUOTA);
 632                         break;
 633                 case Opt_usrjquota:
 634                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 635                         if (ret)
 636                                 return ret;
 637                         break;
 638                 case Opt_grpjquota:
 639                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 640                         if (ret)
 641                                 return ret;
 642                         break;
 643                 case Opt_prjjquota:
 644                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 645                         if (ret)
 646                                 return ret;
 647                         break;
 648                 case Opt_offusrjquota:
 649                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 650                         if (ret)
 651                                 return ret;
 652                         break;
 653                 case Opt_offgrpjquota:
 654                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 655                         if (ret)
 656                                 return ret;
 657                         break;
 658                 case Opt_offprjjquota:
 659                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 660                         if (ret)
 661                                 return ret;
 662                         break;
 663                 case Opt_jqfmt_vfsold:
 664                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 665                         break;
 666                 case Opt_jqfmt_vfsv0:
 667                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 668                         break;
 669                 case Opt_jqfmt_vfsv1:
 670                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 671                         break;
 672                 case Opt_noquota:
 673                         clear_opt(sbi, QUOTA);
 674                         clear_opt(sbi, USRQUOTA);
 675                         clear_opt(sbi, GRPQUOTA);
 676                         clear_opt(sbi, PRJQUOTA);
 677                         break;
 678 #else
 679                 case Opt_quota:
 680                 case Opt_usrquota:
 681                 case Opt_grpquota:
 682                 case Opt_prjquota:
 683                 case Opt_usrjquota:
 684                 case Opt_grpjquota:
 685                 case Opt_prjjquota:
 686                 case Opt_offusrjquota:
 687                 case Opt_offgrpjquota:
 688                 case Opt_offprjjquota:
 689                 case Opt_jqfmt_vfsold:
 690                 case Opt_jqfmt_vfsv0:
 691                 case Opt_jqfmt_vfsv1:
 692                 case Opt_noquota:
 693                         f2fs_msg(sb, KERN_INFO,
 694                                         "quota operations not supported");
 695                         break;
 696 #endif
 697                 case Opt_whint:
 698                         name = match_strdup(&args[0]);
 699                         if (!name)
 700                                 return -ENOMEM;
 701                         if (strlen(name) == 10 &&
 702                                         !strncmp(name, "user-based", 10)) {
 703                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 704                         } else if (strlen(name) == 3 &&
 705                                         !strncmp(name, "off", 3)) {
 706                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 707                         } else if (strlen(name) == 8 &&
 708                                         !strncmp(name, "fs-based", 8)) {
 709                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 710                         } else {
 711                                 kfree(name);
 712                                 return -EINVAL;
 713                         }
 714                         kfree(name);
 715                         break;
 716                 case Opt_alloc:
 717                         name = match_strdup(&args[0]);
 718                         if (!name)
 719                                 return -ENOMEM;
 720
 721                         if (strlen(name) == 7 &&
 722                                         !strncmp(name, "default", 7)) {
 723                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 724                         } else if (strlen(name) == 5 &&
 725                                         !strncmp(name, "reuse", 5)) {
 726                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 727                         } else {
 728                                 kfree(name);
 729                                 return -EINVAL;
 730                         }
 731                         kfree(name);
 732                         break;
 733                 case Opt_fsync:
 734                         name = match_strdup(&args[0]);
 735                         if (!name)
 736                                 return -ENOMEM;
 737                         if (strlen(name) == 5 &&
 738                                         !strncmp(name, "posix", 5)) {
 739                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 740                         } else if (strlen(name) == 6 &&
 741                                         !strncmp(name, "strict", 6)) {
 742                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 743                         } else if (strlen(name) == 9 &&
 744                                         !strncmp(name, "nobarrier", 9)) {
 745                                 F2FS_OPTION(sbi).fsync_mode =
 746                                                         FSYNC_MODE_NOBARRIER;
 747                         } else {
 748                                 kfree(name);
 749                                 return -EINVAL;
 750                         }
 751                         kfree(name);
 752                         break;
 753                 case Opt_test_dummy_encryption:
 754 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 755                         if (!f2fs_sb_has_encrypt(sb)) {
 756                                 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
 757                                 return -EINVAL;
 758                         }
 759
 760                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 761                         f2fs_msg(sb, KERN_INFO,
 762                                         "Test dummy encryption mode enabled");
 763 #else
 764                         f2fs_msg(sb, KERN_INFO,
 765                                         "Test dummy encryption mount option ignored");
 766 #endif
 767                         break;
 768                 default:
 769                         f2fs_msg(sb, KERN_ERR,
 770                                 "Unrecognized mount option \"%s\" or missing value",
 771                                 p);
 772                         return -EINVAL;
 773                 }
 774         }
 775 #ifdef CONFIG_QUOTA
 776         if (f2fs_check_quota_options(sbi))
 777                 return -EINVAL;
 778 #endif
 779
 780         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 781                 f2fs_msg(sb, KERN_ERR,
 782                                 "Should set mode=lfs with %uKB-sized IO",
 783                                 F2FS_IO_SIZE_KB(sbi));
 784                 return -EINVAL;
 785         }
 786
 787         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 788                 if (!f2fs_sb_has_extra_attr(sb) ||
 789                         !f2fs_sb_has_flexible_inline_xattr(sb)) {
 790                         f2fs_msg(sb, KERN_ERR,
 791                                         "extra_attr or flexible_inline_xattr "
 792                                         "feature is off");
 793                         return -EINVAL;
 794                 }
 795                 if (!test_opt(sbi, INLINE_XATTR)) {
 796                         f2fs_msg(sb, KERN_ERR,
 797                                         "inline_xattr_size option should be "
 798                                         "set with inline_xattr option");
 799                         return -EINVAL;
 800                 }
 801                 if (!F2FS_OPTION(sbi).inline_xattr_size ||
 802                         F2FS_OPTION(sbi).inline_xattr_size >=
 803                                         DEF_ADDRS_PER_INODE -
 804                                         F2FS_TOTAL_EXTRA_ATTR_SIZE -
 805                                         DEF_INLINE_RESERVED_SIZE -
 806                                         DEF_MIN_INLINE_SIZE) {
 807                         f2fs_msg(sb, KERN_ERR,
 808                                         "inline xattr size is out of range");
 809                         return -EINVAL;
 810                 }
 811         }
 812
 813         /* Not pass down write hints if the number of active logs is lesser
 814          * than NR_CURSEG_TYPE.
 815          */
 816         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 817                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 818         return 0;
 819 }
 820
 821 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 822 {
 823         struct f2fs_inode_info *fi;
 824
 825         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 826         if (!fi)
 827                 return NULL;
 828
 829         init_once((void *) fi);
 830
 831         /* Initialize f2fs-specific inode info */
 832         atomic_set(&fi->dirty_pages, 0);
 833         init_rwsem(&fi->i_sem);
 834         INIT_LIST_HEAD(&fi->dirty_list);
 835         INIT_LIST_HEAD(&fi->gdirty_list);
 836         INIT_LIST_HEAD(&fi->inmem_ilist);
 837         INIT_LIST_HEAD(&fi->inmem_pages);
 838         mutex_init(&fi->inmem_lock);
 839         init_rwsem(&fi->i_gc_rwsem[READ]);
 840         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 841         init_rwsem(&fi->i_mmap_sem);
 842         init_rwsem(&fi->i_xattr_sem);
 843
 844         /* Will be used by directory only */
 845         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 846
 847         return &fi->vfs_inode;
 848 }
 849
 850 static int f2fs_drop_inode(struct inode *inode)
 851 {
 852         int ret;
 853         /*
 854          * This is to avoid a deadlock condition like below.
 855          * writeback_single_inode(inode)
 856          *  - f2fs_write_data_page
 857          *    - f2fs_gc -> iput -> evict
 858          *       - inode_wait_for_writeback(inode)
 859          */
 860         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 861                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 862                         /* to avoid evict_inode call simultaneously */
 863                         atomic_inc(&inode->i_count);
 864                         spin_unlock(&inode->i_lock);
 865
 866                         /* some remained atomic pages should discarded */
 867                         if (f2fs_is_atomic_file(inode))
 868                                 f2fs_drop_inmem_pages(inode);
 869
 870                         /* should remain fi->extent_tree for writepage */
 871                         f2fs_destroy_extent_node(inode);
 872
 873                         sb_start_intwrite(inode->i_sb);
 874                         f2fs_i_size_write(inode, 0);
 875
 876                         if (F2FS_HAS_BLOCKS(inode))
 877                                 f2fs_truncate(inode);
 878
 879                         sb_end_intwrite(inode->i_sb);
 880
 881                         spin_lock(&inode->i_lock);
 882                         atomic_dec(&inode->i_count);
 883                 }
 884                 trace_f2fs_drop_inode(inode, 0);
 885                 return 0;
 886         }
 887         ret = generic_drop_inode(inode);
 888         trace_f2fs_drop_inode(inode, ret);
 889         return ret;
 890 }
 891
 892 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 893 {
 894         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 895         int ret = 0;
 896
 897         spin_lock(&sbi->inode_lock[DIRTY_META]);
 898         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 899                 ret = 1;
 900         } else {
 901                 set_inode_flag(inode, FI_DIRTY_INODE);
 902                 stat_inc_dirty_inode(sbi, DIRTY_META);
 903         }
 904         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 905                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 906                                 &sbi->inode_list[DIRTY_META]);
 907                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 908         }
 909         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 910         return ret;
 911 }
 912
 913 void f2fs_inode_synced(struct inode *inode)
 914 {
 915         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 916
 917         spin_lock(&sbi->inode_lock[DIRTY_META]);
 918         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 919                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 920                 return;
 921         }
 922         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 923                 list_del_init(&F2FS_I(inode)->gdirty_list);
 924                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 925         }
 926         clear_inode_flag(inode, FI_DIRTY_INODE);
 927         clear_inode_flag(inode, FI_AUTO_RECOVER);
 928         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 929         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 930 }
 931
 932 /*
 933  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 934  *
 935  * We should call set_dirty_inode to write the dirty inode through write_inode.
 936  */
 937 static void f2fs_dirty_inode(struct inode *inode, int flags)
 938 {
 939         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 940
 941         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 942                         inode->i_ino == F2FS_META_INO(sbi))
 943                 return;
 944
 945         if (flags == I_DIRTY_TIME)
 946                 return;
 947
 948         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 949                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 950
 951         f2fs_inode_dirtied(inode, false);
 952 }
 953
 954 static void f2fs_i_callback(struct rcu_head *head)
 955 {
 956         struct inode *inode = container_of(head, struct inode, i_rcu);
 957         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 958 }
 959
 960 static void f2fs_destroy_inode(struct inode *inode)
 961 {
 962         call_rcu(&inode->i_rcu, f2fs_i_callback);
 963 }
 964
 965 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 966 {
 967         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 968         percpu_counter_destroy(&sbi->total_valid_inode_count);
 969 }
 970
 971 static void destroy_device_list(struct f2fs_sb_info *sbi)
 972 {
 973         int i;
 974
 975         for (i = 0; i < sbi->s_ndevs; i++) {
 976                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
 977 #ifdef CONFIG_BLK_DEV_ZONED
 978                 kfree(FDEV(i).blkz_type);
 979 #endif
 980         }
 981         kfree(sbi->devs);
 982 }
 983
 984 static void f2fs_put_super(struct super_block *sb)
 985 {
 986         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 987         int i;
 988         bool dropped;
 989
 990         f2fs_quota_off_umount(sb);
 991
 992         /* prevent remaining shrinker jobs */
 993         mutex_lock(&sbi->umount_mutex);
 994
 995         /*
 996          * We don't need to do checkpoint when superblock is clean.
 997          * But, the previous checkpoint was not done by umount, it needs to do
 998          * clean checkpoint again.
 999          */
1000         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1001                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
1002                 struct cp_control cpc = {
1003                         .reason = CP_UMOUNT,
1004                 };
1005                 f2fs_write_checkpoint(sbi, &cpc);
1006         }
1007
1008         /* be sure to wait for any on-going discard commands */
1009         dropped = f2fs_wait_discard_bios(sbi);
1010
1011         if (f2fs_discard_en(sbi) && !sbi->discard_blks && !dropped) {
1012                 struct cp_control cpc = {
1013                         .reason = CP_UMOUNT | CP_TRIMMED,
1014                 };
1015                 f2fs_write_checkpoint(sbi, &cpc);
1016         }
1017
1018         /* f2fs_write_checkpoint can update stat informaion */
1019         f2fs_destroy_stats(sbi);
1020
1021         /*
1022          * normally superblock is clean, so we need to release this.
1023          * In addition, EIO will skip do checkpoint, we need this as well.
1024          */
1025         f2fs_release_ino_entry(sbi, true);
1026
1027         f2fs_leave_shrinker(sbi);
1028         mutex_unlock(&sbi->umount_mutex);
1029
1030         /* our cp_error case, we can wait for any writeback page */
1031         f2fs_flush_merged_writes(sbi);
1032
1033         iput(sbi->node_inode);
1034         iput(sbi->meta_inode);
1035
1036         /* destroy f2fs internal modules */
1037         f2fs_destroy_node_manager(sbi);
1038         f2fs_destroy_segment_manager(sbi);
1039
1040         kfree(sbi->ckpt);
1041
1042         f2fs_unregister_sysfs(sbi);
1043
1044         sb->s_fs_info = NULL;
1045         if (sbi->s_chksum_driver)
1046                 crypto_free_shash(sbi->s_chksum_driver);
1047         kfree(sbi->raw_super);
1048
1049         destroy_device_list(sbi);
1050         mempool_destroy(sbi->write_io_dummy);
1051 #ifdef CONFIG_QUOTA
1052         for (i = 0; i < MAXQUOTAS; i++)
1053                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1054 #endif
1055         destroy_percpu_info(sbi);
1056         for (i = 0; i < NR_PAGE_TYPE; i++)
1057                 kfree(sbi->write_io[i]);
1058         kfree(sbi);
1059 }
1060
1061 int f2fs_sync_fs(struct super_block *sb, int sync)
1062 {
1063         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1064         int err = 0;
1065
1066         if (unlikely(f2fs_cp_error(sbi)))
1067                 return 0;
1068
1069         trace_f2fs_sync_fs(sb, sync);
1070
1071         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1072                 return -EAGAIN;
1073
1074         if (sync) {
1075                 struct cp_control cpc;
1076
1077                 cpc.reason = __get_cp_reason(sbi);
1078
1079                 mutex_lock(&sbi->gc_mutex);
1080                 err = f2fs_write_checkpoint(sbi, &cpc);
1081                 mutex_unlock(&sbi->gc_mutex);
1082         }
1083         f2fs_trace_ios(NULL, 1);
1084
1085         return err;
1086 }
1087
1088 static int f2fs_freeze(struct super_block *sb)
1089 {
1090         if (f2fs_readonly(sb))
1091                 return 0;
1092
1093         /* IO error happened before */
1094         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1095                 return -EIO;
1096
1097         /* must be clean, since sync_filesystem() was already called */
1098         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1099                 return -EINVAL;
1100         return 0;
1101 }
1102
1103 static int f2fs_unfreeze(struct super_block *sb)
1104 {
1105         return 0;
1106 }
1107
1108 #ifdef CONFIG_QUOTA
1109 static int f2fs_statfs_project(struct super_block *sb,
1110                                 kprojid_t projid, struct kstatfs *buf)
1111 {
1112         struct kqid qid;
1113         struct dquot *dquot;
1114         u64 limit;
1115         u64 curblock;
1116
1117         qid = make_kqid_projid(projid);
1118         dquot = dqget(sb, qid);
1119         if (IS_ERR(dquot))
1120                 return PTR_ERR(dquot);
1121         spin_lock(&dq_data_lock);
1122
1123         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1124                  dquot->dq_dqb.dqb_bsoftlimit :
1125                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1126         if (limit && buf->f_blocks > limit) {
1127                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1128                 buf->f_blocks = limit;
1129                 buf->f_bfree = buf->f_bavail =
1130                         (buf->f_blocks > curblock) ?
1131                          (buf->f_blocks - curblock) : 0;
1132         }
1133
1134         limit = dquot->dq_dqb.dqb_isoftlimit ?
1135                 dquot->dq_dqb.dqb_isoftlimit :
1136                 dquot->dq_dqb.dqb_ihardlimit;
1137         if (limit && buf->f_files > limit) {
1138                 buf->f_files = limit;
1139                 buf->f_ffree =
1140                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1141                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1142         }
1143
1144         spin_unlock(&dq_data_lock);
1145         dqput(dquot);
1146         return 0;
1147 }
1148 #endif
1149
1150 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1151 {
1152         struct super_block *sb = dentry->d_sb;
1153         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1154         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1155         block_t total_count, user_block_count, start_count;
1156         u64 avail_node_count;
1157
1158         total_count = le64_to_cpu(sbi->raw_super->block_count);
1159         user_block_count = sbi->user_block_count;
1160         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1161         buf->f_type = F2FS_SUPER_MAGIC;
1162         buf->f_bsize = sbi->blocksize;
1163
1164         buf->f_blocks = total_count - start_count;
1165         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1166                                                 sbi->current_reserved_blocks;
1167         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1168                 buf->f_bavail = buf->f_bfree -
1169                                 F2FS_OPTION(sbi).root_reserved_blocks;
1170         else
1171                 buf->f_bavail = 0;
1172
1173         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1174                                                 F2FS_RESERVED_NODE_NUM;
1175
1176         if (avail_node_count > user_block_count) {
1177                 buf->f_files = user_block_count;
1178                 buf->f_ffree = buf->f_bavail;
1179         } else {
1180                 buf->f_files = avail_node_count;
1181                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1182                                         buf->f_bavail);
1183         }
1184
1185         buf->f_namelen = F2FS_NAME_LEN;
1186         buf->f_fsid.val[0] = (u32)id;
1187         buf->f_fsid.val[1] = (u32)(id >> 32);
1188
1189 #ifdef CONFIG_QUOTA
1190         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1191                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1192                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1193         }
1194 #endif
1195         return 0;
1196 }
1197
1198 static inline void f2fs_show_quota_options(struct seq_file *seq,
1199                                            struct super_block *sb)
1200 {
1201 #ifdef CONFIG_QUOTA
1202         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1203
1204         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1205                 char *fmtname = "";
1206
1207                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1208                 case QFMT_VFS_OLD:
1209                         fmtname = "vfsold";
1210                         break;
1211                 case QFMT_VFS_V0:
1212                         fmtname = "vfsv0";
1213                         break;
1214                 case QFMT_VFS_V1:
1215                         fmtname = "vfsv1";
1216                         break;
1217                 }
1218                 seq_printf(seq, ",jqfmt=%s", fmtname);
1219         }
1220
1221         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1222                 seq_show_option(seq, "usrjquota",
1223                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1224
1225         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1226                 seq_show_option(seq, "grpjquota",
1227                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1228
1229         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1230                 seq_show_option(seq, "prjjquota",
1231                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1232 #endif
1233 }
1234
1235 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1236 {
1237         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1238
1239         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1240                 if (test_opt(sbi, FORCE_FG_GC))
1241                         seq_printf(seq, ",background_gc=%s", "sync");
1242                 else
1243                         seq_printf(seq, ",background_gc=%s", "on");
1244         } else {
1245                 seq_printf(seq, ",background_gc=%s", "off");
1246         }
1247         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1248                 seq_puts(seq, ",disable_roll_forward");
1249         if (test_opt(sbi, DISCARD))
1250                 seq_puts(seq, ",discard");
1251         if (test_opt(sbi, NOHEAP))
1252                 seq_puts(seq, ",no_heap");
1253         else
1254                 seq_puts(seq, ",heap");
1255 #ifdef CONFIG_F2FS_FS_XATTR
1256         if (test_opt(sbi, XATTR_USER))
1257                 seq_puts(seq, ",user_xattr");
1258         else
1259                 seq_puts(seq, ",nouser_xattr");
1260         if (test_opt(sbi, INLINE_XATTR))
1261                 seq_puts(seq, ",inline_xattr");
1262         else
1263                 seq_puts(seq, ",noinline_xattr");
1264         if (test_opt(sbi, INLINE_XATTR_SIZE))
1265                 seq_printf(seq, ",inline_xattr_size=%u",
1266                                         F2FS_OPTION(sbi).inline_xattr_size);
1267 #endif
1268 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1269         if (test_opt(sbi, POSIX_ACL))
1270                 seq_puts(seq, ",acl");
1271         else
1272                 seq_puts(seq, ",noacl");
1273 #endif
1274         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1275                 seq_puts(seq, ",disable_ext_identify");
1276         if (test_opt(sbi, INLINE_DATA))
1277                 seq_puts(seq, ",inline_data");
1278         else
1279                 seq_puts(seq, ",noinline_data");
1280         if (test_opt(sbi, INLINE_DENTRY))
1281                 seq_puts(seq, ",inline_dentry");
1282         else
1283                 seq_puts(seq, ",noinline_dentry");
1284         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1285                 seq_puts(seq, ",flush_merge");
1286         if (test_opt(sbi, NOBARRIER))
1287                 seq_puts(seq, ",nobarrier");
1288         if (test_opt(sbi, FASTBOOT))
1289                 seq_puts(seq, ",fastboot");
1290         if (test_opt(sbi, EXTENT_CACHE))
1291                 seq_puts(seq, ",extent_cache");
1292         else
1293                 seq_puts(seq, ",noextent_cache");
1294         if (test_opt(sbi, DATA_FLUSH))
1295                 seq_puts(seq, ",data_flush");
1296
1297         seq_puts(seq, ",mode=");
1298         if (test_opt(sbi, ADAPTIVE))
1299                 seq_puts(seq, "adaptive");
1300         else if (test_opt(sbi, LFS))
1301                 seq_puts(seq, "lfs");
1302         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1303         if (test_opt(sbi, RESERVE_ROOT))
1304                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1305                                 F2FS_OPTION(sbi).root_reserved_blocks,
1306                                 from_kuid_munged(&init_user_ns,
1307                                         F2FS_OPTION(sbi).s_resuid),
1308                                 from_kgid_munged(&init_user_ns,
1309                                         F2FS_OPTION(sbi).s_resgid));
1310         if (F2FS_IO_SIZE_BITS(sbi))
1311                 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1312 #ifdef CONFIG_F2FS_FAULT_INJECTION
1313         if (test_opt(sbi, FAULT_INJECTION))
1314                 seq_printf(seq, ",fault_injection=%u",
1315                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1316 #endif
1317 #ifdef CONFIG_QUOTA
1318         if (test_opt(sbi, QUOTA))
1319                 seq_puts(seq, ",quota");
1320         if (test_opt(sbi, USRQUOTA))
1321                 seq_puts(seq, ",usrquota");
1322         if (test_opt(sbi, GRPQUOTA))
1323                 seq_puts(seq, ",grpquota");
1324         if (test_opt(sbi, PRJQUOTA))
1325                 seq_puts(seq, ",prjquota");
1326 #endif
1327         f2fs_show_quota_options(seq, sbi->sb);
1328         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1329                 seq_printf(seq, ",whint_mode=%s", "user-based");
1330         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1331                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1332 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1333         if (F2FS_OPTION(sbi).test_dummy_encryption)
1334                 seq_puts(seq, ",test_dummy_encryption");
1335 #endif
1336
1337         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1338                 seq_printf(seq, ",alloc_mode=%s", "default");
1339         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1340                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1341
1342         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1343                 seq_printf(seq, ",fsync_mode=%s", "posix");
1344         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1345                 seq_printf(seq, ",fsync_mode=%s", "strict");
1346         return 0;
1347 }
1348
1349 static void default_options(struct f2fs_sb_info *sbi)
1350 {
1351         /* init some FS parameters */
1352         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1353         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1354         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1355         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1356         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1357         F2FS_OPTION(sbi).test_dummy_encryption = false;
1358         sbi->readdir_ra = 1;
1359
1360         set_opt(sbi, BG_GC);
1361         set_opt(sbi, INLINE_XATTR);
1362         set_opt(sbi, INLINE_DATA);
1363         set_opt(sbi, INLINE_DENTRY);
1364         set_opt(sbi, EXTENT_CACHE);
1365         set_opt(sbi, NOHEAP);
1366         sbi->sb->s_flags |= SB_LAZYTIME;
1367         set_opt(sbi, FLUSH_MERGE);
1368         if (f2fs_sb_has_blkzoned(sbi->sb)) {
1369                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1370                 set_opt(sbi, DISCARD);
1371         } else {
1372                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1373         }
1374
1375 #ifdef CONFIG_F2FS_FS_XATTR
1376         set_opt(sbi, XATTR_USER);
1377 #endif
1378 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1379         set_opt(sbi, POSIX_ACL);
1380 #endif
1381
1382 #ifdef CONFIG_F2FS_FAULT_INJECTION
1383         f2fs_build_fault_attr(sbi, 0);
1384 #endif
1385 }
1386
1387 #ifdef CONFIG_QUOTA
1388 static int f2fs_enable_quotas(struct super_block *sb);
1389 #endif
1390 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1391 {
1392         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1393         struct f2fs_mount_info org_mount_opt;
1394         unsigned long old_sb_flags;
1395         int err;
1396         bool need_restart_gc = false;
1397         bool need_stop_gc = false;
1398         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1399 #ifdef CONFIG_QUOTA
1400         int i, j;
1401 #endif
1402
1403         /*
1404          * Save the old mount options in case we
1405          * need to restore them.
1406          */
1407         org_mount_opt = sbi->mount_opt;
1408         old_sb_flags = sb->s_flags;
1409
1410 #ifdef CONFIG_QUOTA
1411         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1412         for (i = 0; i < MAXQUOTAS; i++) {
1413                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1414                         org_mount_opt.s_qf_names[i] =
1415                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1416                                 GFP_KERNEL);
1417                         if (!org_mount_opt.s_qf_names[i]) {
1418                                 for (j = 0; j < i; j++)
1419                                         kfree(org_mount_opt.s_qf_names[j]);
1420                                 return -ENOMEM;
1421                         }
1422                 } else {
1423                         org_mount_opt.s_qf_names[i] = NULL;
1424                 }
1425         }
1426 #endif
1427
1428         /* recover superblocks we couldn't write due to previous RO mount */
1429         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1430                 err = f2fs_commit_super(sbi, false);
1431                 f2fs_msg(sb, KERN_INFO,
1432                         "Try to recover all the superblocks, ret: %d", err);
1433                 if (!err)
1434                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1435         }
1436
1437         default_options(sbi);
1438
1439         /* parse mount options */
1440         err = parse_options(sb, data);
1441         if (err)
1442                 goto restore_opts;
1443
1444         /*
1445          * Previous and new state of filesystem is RO,
1446          * so skip checking GC and FLUSH_MERGE conditions.
1447          */
1448         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1449                 goto skip;
1450
1451 #ifdef CONFIG_QUOTA
1452         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1453                 err = dquot_suspend(sb, -1);
1454                 if (err < 0)
1455                         goto restore_opts;
1456         } else if (f2fs_readonly(sb) && !(*flags & MS_RDONLY)) {
1457                 /* dquot_resume needs RW */
1458                 sb->s_flags &= ~SB_RDONLY;
1459                 if (sb_any_quota_suspended(sb)) {
1460                         dquot_resume(sb, -1);
1461                 } else if (f2fs_sb_has_quota_ino(sb)) {
1462                         err = f2fs_enable_quotas(sb);
1463                         if (err)
1464                                 goto restore_opts;
1465                 }
1466         }
1467 #endif
1468         /* disallow enable/disable extent_cache dynamically */
1469         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1470                 err = -EINVAL;
1471                 f2fs_msg(sbi->sb, KERN_WARNING,
1472                                 "switch extent_cache option is not allowed");
1473                 goto restore_opts;
1474         }
1475
1476         /*
1477          * We stop the GC thread if FS is mounted as RO
1478          * or if background_gc = off is passed in mount
1479          * option. Also sync the filesystem.
1480          */
1481         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1482                 if (sbi->gc_thread) {
1483                         f2fs_stop_gc_thread(sbi);
1484                         need_restart_gc = true;
1485                 }
1486         } else if (!sbi->gc_thread) {
1487                 err = f2fs_start_gc_thread(sbi);
1488                 if (err)
1489                         goto restore_opts;
1490                 need_stop_gc = true;
1491         }
1492
1493         if (*flags & SB_RDONLY ||
1494                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1495                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1496                 sync_inodes_sb(sb);
1497
1498                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1499                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1500                 f2fs_sync_fs(sb, 1);
1501                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1502         }
1503
1504         /*
1505          * We stop issue flush thread if FS is mounted as RO
1506          * or if flush_merge is not passed in mount option.
1507          */
1508         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1509                 clear_opt(sbi, FLUSH_MERGE);
1510                 f2fs_destroy_flush_cmd_control(sbi, false);
1511         } else {
1512                 err = f2fs_create_flush_cmd_control(sbi);
1513                 if (err)
1514                         goto restore_gc;
1515         }
1516 skip:
1517 #ifdef CONFIG_QUOTA
1518         /* Release old quota file names */
1519         for (i = 0; i < MAXQUOTAS; i++)
1520                 kfree(org_mount_opt.s_qf_names[i]);
1521 #endif
1522         /* Update the POSIXACL Flag */
1523         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1524                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1525
1526         limit_reserve_root(sbi);
1527         return 0;
1528 restore_gc:
1529         if (need_restart_gc) {
1530                 if (f2fs_start_gc_thread(sbi))
1531                         f2fs_msg(sbi->sb, KERN_WARNING,
1532                                 "background gc thread has stopped");
1533         } else if (need_stop_gc) {
1534                 f2fs_stop_gc_thread(sbi);
1535         }
1536 restore_opts:
1537 #ifdef CONFIG_QUOTA
1538         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1539         for (i = 0; i < MAXQUOTAS; i++) {
1540                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1541                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1542         }
1543 #endif
1544         sbi->mount_opt = org_mount_opt;
1545         sb->s_flags = old_sb_flags;
1546         return err;
1547 }
1548
1549 #ifdef CONFIG_QUOTA
1550 /* Read data from quotafile */
1551 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1552                                size_t len, loff_t off)
1553 {
1554         struct inode *inode = sb_dqopt(sb)->files[type];
1555         struct address_space *mapping = inode->i_mapping;
1556         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1557         int offset = off & (sb->s_blocksize - 1);
1558         int tocopy;
1559         size_t toread;
1560         loff_t i_size = i_size_read(inode);
1561         struct page *page;
1562         char *kaddr;
1563
1564         if (off > i_size)
1565                 return 0;
1566
1567         if (off + len > i_size)
1568                 len = i_size - off;
1569         toread = len;
1570         while (toread > 0) {
1571                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1572 repeat:
1573                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1574                 if (IS_ERR(page)) {
1575                         if (PTR_ERR(page) == -ENOMEM) {
1576                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1577                                 goto repeat;
1578                         }
1579                         return PTR_ERR(page);
1580                 }
1581
1582                 lock_page(page);
1583
1584                 if (unlikely(page->mapping != mapping)) {
1585                         f2fs_put_page(page, 1);
1586                         goto repeat;
1587                 }
1588                 if (unlikely(!PageUptodate(page))) {
1589                         f2fs_put_page(page, 1);
1590                         return -EIO;
1591                 }
1592
1593                 kaddr = kmap_atomic(page);
1594                 memcpy(data, kaddr + offset, tocopy);
1595                 kunmap_atomic(kaddr);
1596                 f2fs_put_page(page, 1);
1597
1598                 offset = 0;
1599                 toread -= tocopy;
1600                 data += tocopy;
1601                 blkidx++;
1602         }
1603         return len;
1604 }
1605
1606 /* Write to quotafile */
1607 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1608                                 const char *data, size_t len, loff_t off)
1609 {
1610         struct inode *inode = sb_dqopt(sb)->files[type];
1611         struct address_space *mapping = inode->i_mapping;
1612         const struct address_space_operations *a_ops = mapping->a_ops;
1613         int offset = off & (sb->s_blocksize - 1);
1614         size_t towrite = len;
1615         struct page *page;
1616         char *kaddr;
1617         int err = 0;
1618         int tocopy;
1619
1620         while (towrite > 0) {
1621                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1622                                                                 towrite);
1623 retry:
1624                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1625                                                         &page, NULL);
1626                 if (unlikely(err)) {
1627                         if (err == -ENOMEM) {
1628                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1629                                 goto retry;
1630                         }
1631                         break;
1632                 }
1633
1634                 kaddr = kmap_atomic(page);
1635                 memcpy(kaddr + offset, data, tocopy);
1636                 kunmap_atomic(kaddr);
1637                 flush_dcache_page(page);
1638
1639                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1640                                                 page, NULL);
1641                 offset = 0;
1642                 towrite -= tocopy;
1643                 off += tocopy;
1644                 data += tocopy;
1645                 cond_resched();
1646         }
1647
1648         if (len == towrite)
1649                 return err;
1650         inode->i_mtime = inode->i_ctime = current_time(inode);
1651         f2fs_mark_inode_dirty_sync(inode, false);
1652         return len - towrite;
1653 }
1654
1655 static struct dquot **f2fs_get_dquots(struct inode *inode)
1656 {
1657         return F2FS_I(inode)->i_dquot;
1658 }
1659
1660 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1661 {
1662         return &F2FS_I(inode)->i_reserved_quota;
1663 }
1664
1665 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1666 {
1667         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1668                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1669 }
1670
1671 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1672 {
1673         int enabled = 0;
1674         int i, err;
1675
1676         if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) {
1677                 err = f2fs_enable_quotas(sbi->sb);
1678                 if (err) {
1679                         f2fs_msg(sbi->sb, KERN_ERR,
1680                                         "Cannot turn on quota_ino: %d", err);
1681                         return 0;
1682                 }
1683                 return 1;
1684         }
1685
1686         for (i = 0; i < MAXQUOTAS; i++) {
1687                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1688                         err = f2fs_quota_on_mount(sbi, i);
1689                         if (!err) {
1690                                 enabled = 1;
1691                                 continue;
1692                         }
1693                         f2fs_msg(sbi->sb, KERN_ERR,
1694                                 "Cannot turn on quotas: %d on %d", err, i);
1695                 }
1696         }
1697         return enabled;
1698 }
1699
1700 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1701                              unsigned int flags)
1702 {
1703         struct inode *qf_inode;
1704         unsigned long qf_inum;
1705         int err;
1706
1707         BUG_ON(!f2fs_sb_has_quota_ino(sb));
1708
1709         qf_inum = f2fs_qf_ino(sb, type);
1710         if (!qf_inum)
1711                 return -EPERM;
1712
1713         qf_inode = f2fs_iget(sb, qf_inum);
1714         if (IS_ERR(qf_inode)) {
1715                 f2fs_msg(sb, KERN_ERR,
1716                         "Bad quota inode %u:%lu", type, qf_inum);
1717                 return PTR_ERR(qf_inode);
1718         }
1719
1720         /* Don't account quota for quota files to avoid recursion */
1721         qf_inode->i_flags |= S_NOQUOTA;
1722         err = dquot_enable(qf_inode, type, format_id, flags);
1723         iput(qf_inode);
1724         return err;
1725 }
1726
1727 static int f2fs_enable_quotas(struct super_block *sb)
1728 {
1729         int type, err = 0;
1730         unsigned long qf_inum;
1731         bool quota_mopt[MAXQUOTAS] = {
1732                 test_opt(F2FS_SB(sb), USRQUOTA),
1733                 test_opt(F2FS_SB(sb), GRPQUOTA),
1734                 test_opt(F2FS_SB(sb), PRJQUOTA),
1735         };
1736
1737         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
1738         for (type = 0; type < MAXQUOTAS; type++) {
1739                 qf_inum = f2fs_qf_ino(sb, type);
1740                 if (qf_inum) {
1741                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1742                                 DQUOT_USAGE_ENABLED |
1743                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1744                         if (err) {
1745                                 f2fs_msg(sb, KERN_ERR,
1746                                         "Failed to enable quota tracking "
1747                                         "(type=%d, err=%d). Please run "
1748                                         "fsck to fix.", type, err);
1749                                 for (type--; type >= 0; type--)
1750                                         dquot_quota_off(sb, type);
1751                                 return err;
1752                         }
1753                 }
1754         }
1755         return 0;
1756 }
1757
1758 static int f2fs_quota_sync(struct super_block *sb, int type)
1759 {
1760         struct quota_info *dqopt = sb_dqopt(sb);
1761         int cnt;
1762         int ret;
1763
1764         ret = dquot_writeback_dquots(sb, type);
1765         if (ret)
1766                 return ret;
1767
1768         /*
1769          * Now when everything is written we can discard the pagecache so
1770          * that userspace sees the changes.
1771          */
1772         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1773                 if (type != -1 && cnt != type)
1774                         continue;
1775                 if (!sb_has_quota_active(sb, cnt))
1776                         continue;
1777
1778                 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1779                 if (ret)
1780                         return ret;
1781
1782                 inode_lock(dqopt->files[cnt]);
1783                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1784                 inode_unlock(dqopt->files[cnt]);
1785         }
1786         return 0;
1787 }
1788
1789 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1790                                                         const struct path *path)
1791 {
1792         struct inode *inode;
1793         int err;
1794
1795         err = f2fs_quota_sync(sb, type);
1796         if (err)
1797                 return err;
1798
1799         err = dquot_quota_on(sb, type, format_id, path);
1800         if (err)
1801                 return err;
1802
1803         inode = d_inode(path->dentry);
1804
1805         inode_lock(inode);
1806         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1807         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1808                                         S_NOATIME | S_IMMUTABLE);
1809         inode_unlock(inode);
1810         f2fs_mark_inode_dirty_sync(inode, false);
1811
1812         return 0;
1813 }
1814
1815 static int f2fs_quota_off(struct super_block *sb, int type)
1816 {
1817         struct inode *inode = sb_dqopt(sb)->files[type];
1818         int err;
1819
1820         if (!inode || !igrab(inode))
1821                 return dquot_quota_off(sb, type);
1822
1823         f2fs_quota_sync(sb, type);
1824
1825         err = dquot_quota_off(sb, type);
1826         if (err || f2fs_sb_has_quota_ino(sb))
1827                 goto out_put;
1828
1829         inode_lock(inode);
1830         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
1831         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1832         inode_unlock(inode);
1833         f2fs_mark_inode_dirty_sync(inode, false);
1834 out_put:
1835         iput(inode);
1836         return err;
1837 }
1838
1839 void f2fs_quota_off_umount(struct super_block *sb)
1840 {
1841         int type;
1842
1843         for (type = 0; type < MAXQUOTAS; type++)
1844                 f2fs_quota_off(sb, type);
1845 }
1846
1847 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1848 {
1849         *projid = F2FS_I(inode)->i_projid;
1850         return 0;
1851 }
1852
1853 static const struct dquot_operations f2fs_quota_operations = {
1854         .get_reserved_space = f2fs_get_reserved_space,
1855         .write_dquot    = dquot_commit,
1856         .acquire_dquot  = dquot_acquire,
1857         .release_dquot  = dquot_release,
1858         .mark_dirty     = dquot_mark_dquot_dirty,
1859         .write_info     = dquot_commit_info,
1860         .alloc_dquot    = dquot_alloc,
1861         .destroy_dquot  = dquot_destroy,
1862         .get_projid     = f2fs_get_projid,
1863         .get_next_id    = dquot_get_next_id,
1864 };
1865
1866 static const struct quotactl_ops f2fs_quotactl_ops = {
1867         .quota_on       = f2fs_quota_on,
1868         .quota_off      = f2fs_quota_off,
1869         .quota_sync     = f2fs_quota_sync,
1870         .get_state      = dquot_get_state,
1871         .set_info       = dquot_set_dqinfo,
1872         .get_dqblk      = dquot_get_dqblk,
1873         .set_dqblk      = dquot_set_dqblk,
1874         .get_nextdqblk  = dquot_get_next_dqblk,
1875 };
1876 #else
1877 void f2fs_quota_off_umount(struct super_block *sb)
1878 {
1879 }
1880 #endif
1881
1882 static const struct super_operations f2fs_sops = {
1883         .alloc_inode    = f2fs_alloc_inode,
1884         .drop_inode     = f2fs_drop_inode,
1885         .destroy_inode  = f2fs_destroy_inode,
1886         .write_inode    = f2fs_write_inode,
1887         .dirty_inode    = f2fs_dirty_inode,
1888         .show_options   = f2fs_show_options,
1889 #ifdef CONFIG_QUOTA
1890         .quota_read     = f2fs_quota_read,
1891         .quota_write    = f2fs_quota_write,
1892         .get_dquots     = f2fs_get_dquots,
1893 #endif
1894         .evict_inode    = f2fs_evict_inode,
1895         .put_super      = f2fs_put_super,
1896         .sync_fs        = f2fs_sync_fs,
1897         .freeze_fs      = f2fs_freeze,
1898         .unfreeze_fs    = f2fs_unfreeze,
1899         .statfs         = f2fs_statfs,
1900         .remount_fs     = f2fs_remount,
1901 };
1902
1903 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1904 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1905 {
1906         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1907                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1908                                 ctx, len, NULL);
1909 }
1910
1911 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1912                                                         void *fs_data)
1913 {
1914         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1915
1916         /*
1917          * Encrypting the root directory is not allowed because fsck
1918          * expects lost+found directory to exist and remain unencrypted
1919          * if LOST_FOUND feature is enabled.
1920          *
1921          */
1922         if (f2fs_sb_has_lost_found(sbi->sb) &&
1923                         inode->i_ino == F2FS_ROOT_INO(sbi))
1924                 return -EPERM;
1925
1926         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1927                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1928                                 ctx, len, fs_data, XATTR_CREATE);
1929 }
1930
1931 static bool f2fs_dummy_context(struct inode *inode)
1932 {
1933         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
1934 }
1935
1936 static const struct fscrypt_operations f2fs_cryptops = {
1937         .key_prefix     = "f2fs:",
1938         .get_context    = f2fs_get_context,
1939         .set_context    = f2fs_set_context,
1940         .dummy_context  = f2fs_dummy_context,
1941         .empty_dir      = f2fs_empty_dir,
1942         .max_namelen    = F2FS_NAME_LEN,
1943 };
1944 #endif
1945
1946 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1947                 u64 ino, u32 generation)
1948 {
1949         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1950         struct inode *inode;
1951
1952         if (f2fs_check_nid_range(sbi, ino))
1953                 return ERR_PTR(-ESTALE);
1954
1955         /*
1956          * f2fs_iget isn't quite right if the inode is currently unallocated!
1957          * However f2fs_iget currently does appropriate checks to handle stale
1958          * inodes so everything is OK.
1959          */
1960         inode = f2fs_iget(sb, ino);
1961         if (IS_ERR(inode))
1962                 return ERR_CAST(inode);
1963         if (unlikely(generation && inode->i_generation != generation)) {
1964                 /* we didn't find the right inode.. */
1965                 iput(inode);
1966                 return ERR_PTR(-ESTALE);
1967         }
1968         return inode;
1969 }
1970
1971 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1972                 int fh_len, int fh_type)
1973 {
1974         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1975                                     f2fs_nfs_get_inode);
1976 }
1977
1978 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1979                 int fh_len, int fh_type)
1980 {
1981         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1982                                     f2fs_nfs_get_inode);
1983 }
1984
1985 static const struct export_operations f2fs_export_ops = {
1986         .fh_to_dentry = f2fs_fh_to_dentry,
1987         .fh_to_parent = f2fs_fh_to_parent,
1988         .get_parent = f2fs_get_parent,
1989 };
1990
1991 static loff_t max_file_blocks(void)
1992 {
1993         loff_t result = 0;
1994         loff_t leaf_count = ADDRS_PER_BLOCK;
1995
1996         /*
1997          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1998          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1999          * space in inode.i_addr, it will be more safe to reassign
2000          * result as zero.
2001          */
2002
2003         /* two direct node blocks */
2004         result += (leaf_count * 2);
2005
2006         /* two indirect node blocks */
2007         leaf_count *= NIDS_PER_BLOCK;
2008         result += (leaf_count * 2);
2009
2010         /* one double indirect node block */
2011         leaf_count *= NIDS_PER_BLOCK;
2012         result += leaf_count;
2013
2014         return result;
2015 }
2016
2017 static int __f2fs_commit_super(struct buffer_head *bh,
2018                         struct f2fs_super_block *super)
2019 {
2020         lock_buffer(bh);
2021         if (super)
2022                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2023         set_buffer_dirty(bh);
2024         unlock_buffer(bh);
2025
2026         /* it's rare case, we can do fua all the time */
2027         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2028 }
2029
2030 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2031                                         struct buffer_head *bh)
2032 {
2033         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2034                                         (bh->b_data + F2FS_SUPER_OFFSET);
2035         struct super_block *sb = sbi->sb;
2036         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2037         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2038         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2039         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2040         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2041         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2042         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2043         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2044         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2045         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2046         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2047         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2048         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2049         u64 main_end_blkaddr = main_blkaddr +
2050                                 (segment_count_main << log_blocks_per_seg);
2051         u64 seg_end_blkaddr = segment0_blkaddr +
2052                                 (segment_count << log_blocks_per_seg);
2053
2054         if (segment0_blkaddr != cp_blkaddr) {
2055                 f2fs_msg(sb, KERN_INFO,
2056                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2057                         segment0_blkaddr, cp_blkaddr);
2058                 return true;
2059         }
2060
2061         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2062                                                         sit_blkaddr) {
2063                 f2fs_msg(sb, KERN_INFO,
2064                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2065                         cp_blkaddr, sit_blkaddr,
2066                         segment_count_ckpt << log_blocks_per_seg);
2067                 return true;
2068         }
2069
2070         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2071                                                         nat_blkaddr) {
2072                 f2fs_msg(sb, KERN_INFO,
2073                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2074                         sit_blkaddr, nat_blkaddr,
2075                         segment_count_sit << log_blocks_per_seg);
2076                 return true;
2077         }
2078
2079         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2080                                                         ssa_blkaddr) {
2081                 f2fs_msg(sb, KERN_INFO,
2082                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2083                         nat_blkaddr, ssa_blkaddr,
2084                         segment_count_nat << log_blocks_per_seg);
2085                 return true;
2086         }
2087
2088         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2089                                                         main_blkaddr) {
2090                 f2fs_msg(sb, KERN_INFO,
2091                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2092                         ssa_blkaddr, main_blkaddr,
2093                         segment_count_ssa << log_blocks_per_seg);
2094                 return true;
2095         }
2096
2097         if (main_end_blkaddr > seg_end_blkaddr) {
2098                 f2fs_msg(sb, KERN_INFO,
2099                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2100                         main_blkaddr,
2101                         segment0_blkaddr +
2102                                 (segment_count << log_blocks_per_seg),
2103                         segment_count_main << log_blocks_per_seg);
2104                 return true;
2105         } else if (main_end_blkaddr < seg_end_blkaddr) {
2106                 int err = 0;
2107                 char *res;
2108
2109                 /* fix in-memory information all the time */
2110                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2111                                 segment0_blkaddr) >> log_blocks_per_seg);
2112
2113                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2114                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2115                         res = "internally";
2116                 } else {
2117                         err = __f2fs_commit_super(bh, NULL);
2118                         res = err ? "failed" : "done";
2119                 }
2120                 f2fs_msg(sb, KERN_INFO,
2121                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
2122                         res, main_blkaddr,
2123                         segment0_blkaddr +
2124                                 (segment_count << log_blocks_per_seg),
2125                         segment_count_main << log_blocks_per_seg);
2126                 if (err)
2127                         return true;
2128         }
2129         return false;
2130 }
2131
2132 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2133                                 struct buffer_head *bh)
2134 {
2135         block_t segment_count, segs_per_sec, secs_per_zone;
2136         block_t total_sections, blocks_per_seg;
2137         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2138                                         (bh->b_data + F2FS_SUPER_OFFSET);
2139         struct super_block *sb = sbi->sb;
2140         unsigned int blocksize;
2141
2142         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2143                 f2fs_msg(sb, KERN_INFO,
2144                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
2145                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2146                 return 1;
2147         }
2148
2149         /* Currently, support only 4KB page cache size */
2150         if (F2FS_BLKSIZE != PAGE_SIZE) {
2151                 f2fs_msg(sb, KERN_INFO,
2152                         "Invalid page_cache_size (%lu), supports only 4KB\n",
2153                         PAGE_SIZE);
2154                 return 1;
2155         }
2156
2157         /* Currently, support only 4KB block size */
2158         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2159         if (blocksize != F2FS_BLKSIZE) {
2160                 f2fs_msg(sb, KERN_INFO,
2161                         "Invalid blocksize (%u), supports only 4KB\n",
2162                         blocksize);
2163                 return 1;
2164         }
2165
2166         /* check log blocks per segment */
2167         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2168                 f2fs_msg(sb, KERN_INFO,
2169                         "Invalid log blocks per segment (%u)\n",
2170                         le32_to_cpu(raw_super->log_blocks_per_seg));
2171                 return 1;
2172         }
2173
2174         /* Currently, support 512/1024/2048/4096 bytes sector size */
2175         if (le32_to_cpu(raw_super->log_sectorsize) >
2176                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2177                 le32_to_cpu(raw_super->log_sectorsize) <
2178                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2179                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2180                         le32_to_cpu(raw_super->log_sectorsize));
2181                 return 1;
2182         }
2183         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2184                 le32_to_cpu(raw_super->log_sectorsize) !=
2185                         F2FS_MAX_LOG_SECTOR_SIZE) {
2186                 f2fs_msg(sb, KERN_INFO,
2187                         "Invalid log sectors per block(%u) log sectorsize(%u)",
2188                         le32_to_cpu(raw_super->log_sectors_per_block),
2189                         le32_to_cpu(raw_super->log_sectorsize));
2190                 return 1;
2191         }
2192
2193         segment_count = le32_to_cpu(raw_super->segment_count);
2194         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2195         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2196         total_sections = le32_to_cpu(raw_super->section_count);
2197
2198         /* blocks_per_seg should be 512, given the above check */
2199         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2200
2201         if (segment_count > F2FS_MAX_SEGMENT ||
2202                                 segment_count < F2FS_MIN_SEGMENTS) {
2203                 f2fs_msg(sb, KERN_INFO,
2204                         "Invalid segment count (%u)",
2205                         segment_count);
2206                 return 1;
2207         }
2208
2209         if (total_sections > segment_count ||
2210                         total_sections < F2FS_MIN_SEGMENTS ||
2211                         segs_per_sec > segment_count || !segs_per_sec) {
2212                 f2fs_msg(sb, KERN_INFO,
2213                         "Invalid segment/section count (%u, %u x %u)",
2214                         segment_count, total_sections, segs_per_sec);
2215                 return 1;
2216         }
2217
2218         if ((segment_count / segs_per_sec) < total_sections) {
2219                 f2fs_msg(sb, KERN_INFO,
2220                         "Small segment_count (%u < %u * %u)",
2221                         segment_count, segs_per_sec, total_sections);
2222                 return 1;
2223         }
2224
2225         if (segment_count > (le32_to_cpu(raw_super->block_count) >> 9)) {
2226                 f2fs_msg(sb, KERN_INFO,
2227                         "Wrong segment_count / block_count (%u > %u)",
2228                         segment_count, le32_to_cpu(raw_super->block_count));
2229                 return 1;
2230         }
2231
2232         if (secs_per_zone > total_sections) {
2233                 f2fs_msg(sb, KERN_INFO,
2234                         "Wrong secs_per_zone (%u > %u)",
2235                         secs_per_zone, total_sections);
2236                 return 1;
2237         }
2238         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2239                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2240                         (le32_to_cpu(raw_super->extension_count) +
2241                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2242                 f2fs_msg(sb, KERN_INFO,
2243                         "Corrupted extension count (%u + %u > %u)",
2244                         le32_to_cpu(raw_super->extension_count),
2245                         raw_super->hot_ext_count,
2246                         F2FS_MAX_EXTENSION);
2247                 return 1;
2248         }
2249
2250         if (le32_to_cpu(raw_super->cp_payload) >
2251                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2252                 f2fs_msg(sb, KERN_INFO,
2253                         "Insane cp_payload (%u > %u)",
2254                         le32_to_cpu(raw_super->cp_payload),
2255                         blocks_per_seg - F2FS_CP_PACKS);
2256                 return 1;
2257         }
2258
2259         /* check reserved ino info */
2260         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2261                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2262                 le32_to_cpu(raw_super->root_ino) != 3) {
2263                 f2fs_msg(sb, KERN_INFO,
2264                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2265                         le32_to_cpu(raw_super->node_ino),
2266                         le32_to_cpu(raw_super->meta_ino),
2267                         le32_to_cpu(raw_super->root_ino));
2268                 return 1;
2269         }
2270
2271         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2272         if (sanity_check_area_boundary(sbi, bh))
2273                 return 1;
2274
2275         return 0;
2276 }
2277
2278 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2279 {
2280         unsigned int total, fsmeta;
2281         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2282         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2283         unsigned int ovp_segments, reserved_segments;
2284         unsigned int main_segs, blocks_per_seg;
2285         int i;
2286
2287         total = le32_to_cpu(raw_super->segment_count);
2288         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2289         fsmeta += le32_to_cpu(raw_super->segment_count_sit);
2290         fsmeta += le32_to_cpu(raw_super->segment_count_nat);
2291         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2292         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2293
2294         if (unlikely(fsmeta >= total))
2295                 return 1;
2296
2297         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2298         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2299
2300         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2301                         ovp_segments == 0 || reserved_segments == 0)) {
2302                 f2fs_msg(sbi->sb, KERN_ERR,
2303                         "Wrong layout: check mkfs.f2fs version");
2304                 return 1;
2305         }
2306
2307         main_segs = le32_to_cpu(raw_super->segment_count_main);
2308         blocks_per_seg = sbi->blocks_per_seg;
2309
2310         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2311                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2312                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2313                         return 1;
2314         }
2315         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2316                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2317                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2318                         return 1;
2319         }
2320
2321         if (unlikely(f2fs_cp_error(sbi))) {
2322                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2323                 return 1;
2324         }
2325         return 0;
2326 }
2327
2328 static void init_sb_info(struct f2fs_sb_info *sbi)
2329 {
2330         struct f2fs_super_block *raw_super = sbi->raw_super;
2331         int i, j;
2332
2333         sbi->log_sectors_per_block =
2334                 le32_to_cpu(raw_super->log_sectors_per_block);
2335         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2336         sbi->blocksize = 1 << sbi->log_blocksize;
2337         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2338         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2339         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2340         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2341         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2342         sbi->total_node_count =
2343                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2344                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2345         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2346         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2347         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2348         sbi->cur_victim_sec = NULL_SECNO;
2349         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2350
2351         sbi->dir_level = DEF_DIR_LEVEL;
2352         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2353         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2354         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2355
2356         for (i = 0; i < NR_COUNT_TYPE; i++)
2357                 atomic_set(&sbi->nr_pages[i], 0);
2358
2359         for (i = 0; i < META; i++)
2360                 atomic_set(&sbi->wb_sync_req[i], 0);
2361
2362         INIT_LIST_HEAD(&sbi->s_list);
2363         mutex_init(&sbi->umount_mutex);
2364         for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2365                 for (j = HOT; j < NR_TEMP_TYPE; j++)
2366                         mutex_init(&sbi->wio_mutex[i][j]);
2367         init_rwsem(&sbi->io_order_lock);
2368         spin_lock_init(&sbi->cp_lock);
2369
2370         sbi->dirty_device = 0;
2371         spin_lock_init(&sbi->dev_lock);
2372
2373         init_rwsem(&sbi->sb_lock);
2374 }
2375
2376 static int init_percpu_info(struct f2fs_sb_info *sbi)
2377 {
2378         int err;
2379
2380         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2381         if (err)
2382                 return err;
2383
2384         return percpu_counter_init(&sbi->total_valid_inode_count, 0,
2385                                                                 GFP_KERNEL);
2386 }
2387
2388 #ifdef CONFIG_BLK_DEV_ZONED
2389 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2390 {
2391         struct block_device *bdev = FDEV(devi).bdev;
2392         sector_t nr_sectors = bdev->bd_part->nr_sects;
2393         sector_t sector = 0;
2394         struct blk_zone *zones;
2395         unsigned int i, nr_zones;
2396         unsigned int n = 0;
2397         int err = -EIO;
2398
2399         if (!f2fs_sb_has_blkzoned(sbi->sb))
2400                 return 0;
2401
2402         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2403                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2404                 return -EINVAL;
2405         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2406         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2407                                 __ilog2_u32(sbi->blocks_per_blkz))
2408                 return -EINVAL;
2409         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2410         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2411                                         sbi->log_blocks_per_blkz;
2412         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2413                 FDEV(devi).nr_blkz++;
2414
2415         FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2416                                                                 GFP_KERNEL);
2417         if (!FDEV(devi).blkz_type)
2418                 return -ENOMEM;
2419
2420 #define F2FS_REPORT_NR_ZONES   4096
2421
2422         zones = f2fs_kzalloc(sbi, sizeof(struct blk_zone) *
2423                                 F2FS_REPORT_NR_ZONES, GFP_KERNEL);
2424         if (!zones)
2425                 return -ENOMEM;
2426
2427         /* Get block zones type */
2428         while (zones && sector < nr_sectors) {
2429
2430                 nr_zones = F2FS_REPORT_NR_ZONES;
2431                 err = blkdev_report_zones(bdev, sector,
2432                                           zones, &nr_zones,
2433                                           GFP_KERNEL);
2434                 if (err)
2435                         break;
2436                 if (!nr_zones) {
2437                         err = -EIO;
2438                         break;
2439                 }
2440
2441                 for (i = 0; i < nr_zones; i++) {
2442                         FDEV(devi).blkz_type[n] = zones[i].type;
2443                         sector += zones[i].len;
2444                         n++;
2445                 }
2446         }
2447
2448         kfree(zones);
2449
2450         return err;
2451 }
2452 #endif
2453
2454 /*
2455  * Read f2fs raw super block.
2456  * Because we have two copies of super block, so read both of them
2457  * to get the first valid one. If any one of them is broken, we pass
2458  * them recovery flag back to the caller.
2459  */
2460 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2461                         struct f2fs_super_block **raw_super,
2462                         int *valid_super_block, int *recovery)
2463 {
2464         struct super_block *sb = sbi->sb;
2465         int block;
2466         struct buffer_head *bh;
2467         struct f2fs_super_block *super;
2468         int err = 0;
2469
2470         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2471         if (!super)
2472                 return -ENOMEM;
2473
2474         for (block = 0; block < 2; block++) {
2475                 bh = sb_bread(sb, block);
2476                 if (!bh) {
2477                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2478                                 block + 1);
2479                         err = -EIO;
2480                         continue;
2481                 }
2482
2483                 /* sanity checking of raw super */
2484                 if (sanity_check_raw_super(sbi, bh)) {
2485                         f2fs_msg(sb, KERN_ERR,
2486                                 "Can't find valid F2FS filesystem in %dth superblock",
2487                                 block + 1);
2488                         err = -EINVAL;
2489                         brelse(bh);
2490                         continue;
2491                 }
2492
2493                 if (!*raw_super) {
2494                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2495                                                         sizeof(*super));
2496                         *valid_super_block = block;
2497                         *raw_super = super;
2498                 }
2499                 brelse(bh);
2500         }
2501
2502         /* Fail to read any one of the superblocks*/
2503         if (err < 0)
2504                 *recovery = 1;
2505
2506         /* No valid superblock */
2507         if (!*raw_super)
2508                 kfree(super);
2509         else
2510                 err = 0;
2511
2512         return err;
2513 }
2514
2515 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2516 {
2517         struct buffer_head *bh;
2518         int err;
2519
2520         if ((recover && f2fs_readonly(sbi->sb)) ||
2521                                 bdev_read_only(sbi->sb->s_bdev)) {
2522                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2523                 return -EROFS;
2524         }
2525
2526         /* write back-up superblock first */
2527         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2528         if (!bh)
2529                 return -EIO;
2530         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2531         brelse(bh);
2532
2533         /* if we are in recovery path, skip writing valid superblock */
2534         if (recover || err)
2535                 return err;
2536
2537         /* write current valid superblock */
2538         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2539         if (!bh)
2540                 return -EIO;
2541         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2542         brelse(bh);
2543         return err;
2544 }
2545
2546 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2547 {
2548         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2549         unsigned int max_devices = MAX_DEVICES;
2550         int i;
2551
2552         /* Initialize single device information */
2553         if (!RDEV(0).path[0]) {
2554                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2555                         return 0;
2556                 max_devices = 1;
2557         }
2558
2559         /*
2560          * Initialize multiple devices information, or single
2561          * zoned block device information.
2562          */
2563         sbi->devs = f2fs_kzalloc(sbi, sizeof(struct f2fs_dev_info) *
2564                                                 max_devices, GFP_KERNEL);
2565         if (!sbi->devs)
2566                 return -ENOMEM;
2567
2568         for (i = 0; i < max_devices; i++) {
2569
2570                 if (i > 0 && !RDEV(i).path[0])
2571                         break;
2572
2573                 if (max_devices == 1) {
2574                         /* Single zoned block device mount */
2575                         FDEV(0).bdev =
2576                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2577                                         sbi->sb->s_mode, sbi->sb->s_type);
2578                 } else {
2579                         /* Multi-device mount */
2580                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2581                         FDEV(i).total_segments =
2582                                 le32_to_cpu(RDEV(i).total_segments);
2583                         if (i == 0) {
2584                                 FDEV(i).start_blk = 0;
2585                                 FDEV(i).end_blk = FDEV(i).start_blk +
2586                                     (FDEV(i).total_segments <<
2587                                     sbi->log_blocks_per_seg) - 1 +
2588                                     le32_to_cpu(raw_super->segment0_blkaddr);
2589                         } else {
2590                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2591                                 FDEV(i).end_blk = FDEV(i).start_blk +
2592                                         (FDEV(i).total_segments <<
2593                                         sbi->log_blocks_per_seg) - 1;
2594                         }
2595                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2596                                         sbi->sb->s_mode, sbi->sb->s_type);
2597                 }
2598                 if (IS_ERR(FDEV(i).bdev))
2599                         return PTR_ERR(FDEV(i).bdev);
2600
2601                 /* to release errored devices */
2602                 sbi->s_ndevs = i + 1;
2603
2604 #ifdef CONFIG_BLK_DEV_ZONED
2605                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2606                                 !f2fs_sb_has_blkzoned(sbi->sb)) {
2607                         f2fs_msg(sbi->sb, KERN_ERR,
2608                                 "Zoned block device feature not enabled\n");
2609                         return -EINVAL;
2610                 }
2611                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2612                         if (init_blkz_info(sbi, i)) {
2613                                 f2fs_msg(sbi->sb, KERN_ERR,
2614                                         "Failed to initialize F2FS blkzone information");
2615                                 return -EINVAL;
2616                         }
2617                         if (max_devices == 1)
2618                                 break;
2619                         f2fs_msg(sbi->sb, KERN_INFO,
2620                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2621                                 i, FDEV(i).path,
2622                                 FDEV(i).total_segments,
2623                                 FDEV(i).start_blk, FDEV(i).end_blk,
2624                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2625                                 "Host-aware" : "Host-managed");
2626                         continue;
2627                 }
2628 #endif
2629                 f2fs_msg(sbi->sb, KERN_INFO,
2630                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2631                                 i, FDEV(i).path,
2632                                 FDEV(i).total_segments,
2633                                 FDEV(i).start_blk, FDEV(i).end_blk);
2634         }
2635         f2fs_msg(sbi->sb, KERN_INFO,
2636                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2637         return 0;
2638 }
2639
2640 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
2641 {
2642         struct f2fs_sm_info *sm_i = SM_I(sbi);
2643
2644         /* adjust parameters according to the volume size */
2645         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
2646                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2647                 sm_i->dcc_info->discard_granularity = 1;
2648                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
2649         }
2650 }
2651
2652 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2653 {
2654         struct f2fs_sb_info *sbi;
2655         struct f2fs_super_block *raw_super;
2656         struct inode *root;
2657         int err;
2658         bool retry = true, need_fsck = false;
2659         char *options = NULL;
2660         int recovery, i, valid_super_block;
2661         struct curseg_info *seg_i;
2662
2663 try_onemore:
2664         err = -EINVAL;
2665         raw_super = NULL;
2666         valid_super_block = -1;
2667         recovery = 0;
2668
2669         /* allocate memory for f2fs-specific super block info */
2670         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2671         if (!sbi)
2672                 return -ENOMEM;
2673
2674         sbi->sb = sb;
2675
2676         /* Load the checksum driver */
2677         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2678         if (IS_ERR(sbi->s_chksum_driver)) {
2679                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2680                 err = PTR_ERR(sbi->s_chksum_driver);
2681                 sbi->s_chksum_driver = NULL;
2682                 goto free_sbi;
2683         }
2684
2685         /* set a block size */
2686         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2687                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2688                 goto free_sbi;
2689         }
2690
2691         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2692                                                                 &recovery);
2693         if (err)
2694                 goto free_sbi;
2695
2696         sb->s_fs_info = sbi;
2697         sbi->raw_super = raw_super;
2698
2699         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2700         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2701
2702         /* precompute checksum seed for metadata */
2703         if (f2fs_sb_has_inode_chksum(sb))
2704                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2705                                                 sizeof(raw_super->uuid));
2706
2707         /*
2708          * The BLKZONED feature indicates that the drive was formatted with
2709          * zone alignment optimization. This is optional for host-aware
2710          * devices, but mandatory for host-managed zoned block devices.
2711          */
2712 #ifndef CONFIG_BLK_DEV_ZONED
2713         if (f2fs_sb_has_blkzoned(sb)) {
2714                 f2fs_msg(sb, KERN_ERR,
2715                          "Zoned block device support is not enabled\n");
2716                 err = -EOPNOTSUPP;
2717                 goto free_sb_buf;
2718         }
2719 #endif
2720         default_options(sbi);
2721         /* parse mount options */
2722         options = kstrdup((const char *)data, GFP_KERNEL);
2723         if (data && !options) {
2724                 err = -ENOMEM;
2725                 goto free_sb_buf;
2726         }
2727
2728         err = parse_options(sb, options);
2729         if (err)
2730                 goto free_options;
2731
2732         sbi->max_file_blocks = max_file_blocks();
2733         sb->s_maxbytes = sbi->max_file_blocks <<
2734                                 le32_to_cpu(raw_super->log_blocksize);
2735         sb->s_max_links = F2FS_LINK_MAX;
2736         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2737
2738 #ifdef CONFIG_QUOTA
2739         sb->dq_op = &f2fs_quota_operations;
2740         if (f2fs_sb_has_quota_ino(sb))
2741                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2742         else
2743                 sb->s_qcop = &f2fs_quotactl_ops;
2744         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2745
2746         if (f2fs_sb_has_quota_ino(sbi->sb)) {
2747                 for (i = 0; i < MAXQUOTAS; i++) {
2748                         if (f2fs_qf_ino(sbi->sb, i))
2749                                 sbi->nquota_files++;
2750                 }
2751         }
2752 #endif
2753
2754         sb->s_op = &f2fs_sops;
2755 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2756         sb->s_cop = &f2fs_cryptops;
2757 #endif
2758         sb->s_xattr = f2fs_xattr_handlers;
2759         sb->s_export_op = &f2fs_export_ops;
2760         sb->s_magic = F2FS_SUPER_MAGIC;
2761         sb->s_time_gran = 1;
2762         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2763                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2764         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2765         sb->s_iflags |= SB_I_CGROUPWB;
2766
2767         /* init f2fs-specific super block info */
2768         sbi->valid_super_block = valid_super_block;
2769         mutex_init(&sbi->gc_mutex);
2770         mutex_init(&sbi->cp_mutex);
2771         init_rwsem(&sbi->node_write);
2772         init_rwsem(&sbi->node_change);
2773
2774         /* disallow all the data/node/meta page writes */
2775         set_sbi_flag(sbi, SBI_POR_DOING);
2776         spin_lock_init(&sbi->stat_lock);
2777
2778         /* init iostat info */
2779         spin_lock_init(&sbi->iostat_lock);
2780         sbi->iostat_enable = false;
2781
2782         for (i = 0; i < NR_PAGE_TYPE; i++) {
2783                 int n = (i == META) ? 1: NR_TEMP_TYPE;
2784                 int j;
2785
2786                 sbi->write_io[i] = f2fs_kmalloc(sbi,
2787                                         n * sizeof(struct f2fs_bio_info),
2788                                         GFP_KERNEL);
2789                 if (!sbi->write_io[i]) {
2790                         err = -ENOMEM;
2791                         goto free_options;
2792                 }
2793
2794                 for (j = HOT; j < n; j++) {
2795                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
2796                         sbi->write_io[i][j].sbi = sbi;
2797                         sbi->write_io[i][j].bio = NULL;
2798                         spin_lock_init(&sbi->write_io[i][j].io_lock);
2799                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2800                 }
2801         }
2802
2803         init_rwsem(&sbi->cp_rwsem);
2804         init_waitqueue_head(&sbi->cp_wait);
2805         init_sb_info(sbi);
2806
2807         err = init_percpu_info(sbi);
2808         if (err)
2809                 goto free_bio_info;
2810
2811         if (F2FS_IO_SIZE(sbi) > 1) {
2812                 sbi->write_io_dummy =
2813                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2814                 if (!sbi->write_io_dummy) {
2815                         err = -ENOMEM;
2816                         goto free_percpu;
2817                 }
2818         }
2819
2820         /* get an inode for meta space */
2821         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2822         if (IS_ERR(sbi->meta_inode)) {
2823                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2824                 err = PTR_ERR(sbi->meta_inode);
2825                 goto free_io_dummy;
2826         }
2827
2828         err = f2fs_get_valid_checkpoint(sbi);
2829         if (err) {
2830                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2831                 goto free_meta_inode;
2832         }
2833
2834         /* Initialize device list */
2835         err = f2fs_scan_devices(sbi);
2836         if (err) {
2837                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2838                 goto free_devices;
2839         }
2840
2841         sbi->total_valid_node_count =
2842                                 le32_to_cpu(sbi->ckpt->valid_node_count);
2843         percpu_counter_set(&sbi->total_valid_inode_count,
2844                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
2845         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2846         sbi->total_valid_block_count =
2847                                 le64_to_cpu(sbi->ckpt->valid_block_count);
2848         sbi->last_valid_block_count = sbi->total_valid_block_count;
2849         sbi->reserved_blocks = 0;
2850         sbi->current_reserved_blocks = 0;
2851         limit_reserve_root(sbi);
2852
2853         for (i = 0; i < NR_INODE_TYPE; i++) {
2854                 INIT_LIST_HEAD(&sbi->inode_list[i]);
2855                 spin_lock_init(&sbi->inode_lock[i]);
2856         }
2857
2858         f2fs_init_extent_cache_info(sbi);
2859
2860         f2fs_init_ino_entry_info(sbi);
2861
2862         /* setup f2fs internal modules */
2863         err = f2fs_build_segment_manager(sbi);
2864         if (err) {
2865                 f2fs_msg(sb, KERN_ERR,
2866                         "Failed to initialize F2FS segment manager");
2867                 goto free_sm;
2868         }
2869         err = f2fs_build_node_manager(sbi);
2870         if (err) {
2871                 f2fs_msg(sb, KERN_ERR,
2872                         "Failed to initialize F2FS node manager");
2873                 goto free_nm;
2874         }
2875
2876         /* For write statistics */
2877         if (sb->s_bdev->bd_part)
2878                 sbi->sectors_written_start =
2879                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2880
2881         /* Read accumulated write IO statistics if exists */
2882         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2883         if (__exist_node_summaries(sbi))
2884                 sbi->kbytes_written =
2885                         le64_to_cpu(seg_i->journal->info.kbytes_written);
2886
2887         f2fs_build_gc_manager(sbi);
2888
2889         /* get an inode for node space */
2890         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2891         if (IS_ERR(sbi->node_inode)) {
2892                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2893                 err = PTR_ERR(sbi->node_inode);
2894                 goto free_nm;
2895         }
2896
2897         err = f2fs_build_stats(sbi);
2898         if (err)
2899                 goto free_node_inode;
2900
2901         /* read root inode and dentry */
2902         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2903         if (IS_ERR(root)) {
2904                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2905                 err = PTR_ERR(root);
2906                 goto free_stats;
2907         }
2908         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2909                 iput(root);
2910                 err = -EINVAL;
2911                 goto free_node_inode;
2912         }
2913
2914         sb->s_root = d_make_root(root); /* allocate root dentry */
2915         if (!sb->s_root) {
2916                 err = -ENOMEM;
2917                 goto free_root_inode;
2918         }
2919
2920         err = f2fs_register_sysfs(sbi);
2921         if (err)
2922                 goto free_root_inode;
2923
2924 #ifdef CONFIG_QUOTA
2925         /*
2926          * Turn on quotas which were not enabled for read-only mounts if
2927          * filesystem has quota feature, so that they are updated correctly.
2928          */
2929         if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb)) {
2930                 err = f2fs_enable_quotas(sb);
2931                 if (err) {
2932                         f2fs_msg(sb, KERN_ERR,
2933                                 "Cannot turn on quotas: error %d", err);
2934                         goto free_sysfs;
2935                 }
2936         }
2937 #endif
2938         /* if there are nt orphan nodes free them */
2939         err = f2fs_recover_orphan_inodes(sbi);
2940         if (err)
2941                 goto free_meta;
2942
2943         /* recover fsynced data */
2944         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2945                 /*
2946                  * mount should be failed, when device has readonly mode, and
2947                  * previous checkpoint was not done by clean system shutdown.
2948                  */
2949                 if (bdev_read_only(sb->s_bdev) &&
2950                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2951                         err = -EROFS;
2952                         goto free_meta;
2953                 }
2954
2955                 if (need_fsck)
2956                         set_sbi_flag(sbi, SBI_NEED_FSCK);
2957
2958                 if (!retry)
2959                         goto skip_recovery;
2960
2961                 err = f2fs_recover_fsync_data(sbi, false);
2962                 if (err < 0) {
2963                         need_fsck = true;
2964                         f2fs_msg(sb, KERN_ERR,
2965                                 "Cannot recover all fsync data errno=%d", err);
2966                         goto free_meta;
2967                 }
2968         } else {
2969                 err = f2fs_recover_fsync_data(sbi, true);
2970
2971                 if (!f2fs_readonly(sb) && err > 0) {
2972                         err = -EINVAL;
2973                         f2fs_msg(sb, KERN_ERR,
2974                                 "Need to recover fsync data");
2975                         goto free_meta;
2976                 }
2977         }
2978 skip_recovery:
2979         /* f2fs_recover_fsync_data() cleared this already */
2980         clear_sbi_flag(sbi, SBI_POR_DOING);
2981
2982         /*
2983          * If filesystem is not mounted as read-only then
2984          * do start the gc_thread.
2985          */
2986         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2987                 /* After POR, we can run background GC thread.*/
2988                 err = f2fs_start_gc_thread(sbi);
2989                 if (err)
2990                         goto free_meta;
2991         }
2992         kfree(options);
2993
2994         /* recover broken superblock */
2995         if (recovery) {
2996                 err = f2fs_commit_super(sbi, true);
2997                 f2fs_msg(sb, KERN_INFO,
2998                         "Try to recover %dth superblock, ret: %d",
2999                         sbi->valid_super_block ? 1 : 2, err);
3000         }
3001
3002         f2fs_join_shrinker(sbi);
3003
3004         f2fs_tuning_parameters(sbi);
3005
3006         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3007                                 cur_cp_version(F2FS_CKPT(sbi)));
3008         f2fs_update_time(sbi, CP_TIME);
3009         f2fs_update_time(sbi, REQ_TIME);
3010         return 0;
3011
3012 free_meta:
3013 #ifdef CONFIG_QUOTA
3014         if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb))
3015                 f2fs_quota_off_umount(sbi->sb);
3016 #endif
3017         f2fs_sync_inode_meta(sbi);
3018         /*
3019          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3020          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3021          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3022          * falls into an infinite loop in f2fs_sync_meta_pages().
3023          */
3024         truncate_inode_pages_final(META_MAPPING(sbi));
3025 #ifdef CONFIG_QUOTA
3026 free_sysfs:
3027 #endif
3028         f2fs_unregister_sysfs(sbi);
3029 free_root_inode:
3030         dput(sb->s_root);
3031         sb->s_root = NULL;
3032 free_stats:
3033         f2fs_destroy_stats(sbi);
3034 free_node_inode:
3035         f2fs_release_ino_entry(sbi, true);
3036         truncate_inode_pages_final(NODE_MAPPING(sbi));
3037         iput(sbi->node_inode);
3038 free_nm:
3039         f2fs_destroy_node_manager(sbi);
3040 free_sm:
3041         f2fs_destroy_segment_manager(sbi);
3042 free_devices:
3043         destroy_device_list(sbi);
3044         kfree(sbi->ckpt);
3045 free_meta_inode:
3046         make_bad_inode(sbi->meta_inode);
3047         iput(sbi->meta_inode);
3048 free_io_dummy:
3049         mempool_destroy(sbi->write_io_dummy);
3050 free_percpu:
3051         destroy_percpu_info(sbi);
3052 free_bio_info:
3053         for (i = 0; i < NR_PAGE_TYPE; i++)
3054                 kfree(sbi->write_io[i]);
3055 free_options:
3056 #ifdef CONFIG_QUOTA
3057         for (i = 0; i < MAXQUOTAS; i++)
3058                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
3059 #endif
3060         kfree(options);
3061 free_sb_buf:
3062         kfree(raw_super);
3063 free_sbi:
3064         if (sbi->s_chksum_driver)
3065                 crypto_free_shash(sbi->s_chksum_driver);
3066         kfree(sbi);
3067
3068         /* give only one another chance */
3069         if (retry) {
3070                 retry = false;
3071                 shrink_dcache_sb(sb);
3072                 goto try_onemore;
3073         }
3074         return err;
3075 }
3076
3077 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3078                         const char *dev_name, void *data)
3079 {
3080         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3081 }
3082
3083 static void kill_f2fs_super(struct super_block *sb)
3084 {
3085         if (sb->s_root) {
3086                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
3087                 f2fs_stop_gc_thread(F2FS_SB(sb));
3088                 f2fs_stop_discard_thread(F2FS_SB(sb));
3089         }
3090         kill_block_super(sb);
3091 }
3092
3093 static struct file_system_type f2fs_fs_type = {
3094         .owner          = THIS_MODULE,
3095         .name           = "f2fs",
3096         .mount          = f2fs_mount,
3097         .kill_sb        = kill_f2fs_super,
3098         .fs_flags       = FS_REQUIRES_DEV,
3099 };
3100 MODULE_ALIAS_FS("f2fs");
3101
3102 static int __init init_inodecache(void)
3103 {
3104         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3105                         sizeof(struct f2fs_inode_info), 0,
3106                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3107         if (!f2fs_inode_cachep)
3108                 return -ENOMEM;
3109         return 0;
3110 }
3111
3112 static void destroy_inodecache(void)
3113 {
3114         /*
3115          * Make sure all delayed rcu free inodes are flushed before we
3116          * destroy cache.
3117          */
3118         rcu_barrier();
3119         kmem_cache_destroy(f2fs_inode_cachep);
3120 }
3121
3122 static int __init init_f2fs_fs(void)
3123 {
3124         int err;
3125
3126         if (PAGE_SIZE != F2FS_BLKSIZE) {
3127                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3128                                 PAGE_SIZE, F2FS_BLKSIZE);
3129                 return -EINVAL;
3130         }
3131
3132         f2fs_build_trace_ios();
3133
3134         err = init_inodecache();
3135         if (err)
3136                 goto fail;
3137         err = f2fs_create_node_manager_caches();
3138         if (err)
3139                 goto free_inodecache;
3140         err = f2fs_create_segment_manager_caches();
3141         if (err)
3142                 goto free_node_manager_caches;
3143         err = f2fs_create_checkpoint_caches();
3144         if (err)
3145                 goto free_segment_manager_caches;
3146         err = f2fs_create_extent_cache();
3147         if (err)
3148                 goto free_checkpoint_caches;
3149         err = f2fs_init_sysfs();
3150         if (err)
3151                 goto free_extent_cache;
3152         err = register_shrinker(&f2fs_shrinker_info);
3153         if (err)
3154                 goto free_sysfs;
3155         err = register_filesystem(&f2fs_fs_type);
3156         if (err)
3157                 goto free_shrinker;
3158         err = f2fs_create_root_stats();
3159         if (err)
3160                 goto free_filesystem;
3161         err = f2fs_init_post_read_processing();
3162         if (err)
3163                 goto free_root_stats;
3164         return 0;
3165
3166 free_root_stats:
3167         f2fs_destroy_root_stats();
3168 free_filesystem:
3169         unregister_filesystem(&f2fs_fs_type);
3170 free_shrinker:
3171         unregister_shrinker(&f2fs_shrinker_info);
3172 free_sysfs:
3173         f2fs_exit_sysfs();
3174 free_extent_cache:
3175         f2fs_destroy_extent_cache();
3176 free_checkpoint_caches:
3177         f2fs_destroy_checkpoint_caches();
3178 free_segment_manager_caches:
3179         f2fs_destroy_segment_manager_caches();
3180 free_node_manager_caches:
3181         f2fs_destroy_node_manager_caches();
3182 free_inodecache:
3183         destroy_inodecache();
3184 fail:
3185         return err;
3186 }
3187
3188 static void __exit exit_f2fs_fs(void)
3189 {
3190         f2fs_destroy_post_read_processing();
3191         f2fs_destroy_root_stats();
3192         unregister_filesystem(&f2fs_fs_type);
3193         unregister_shrinker(&f2fs_shrinker_info);
3194         f2fs_exit_sysfs();
3195         f2fs_destroy_extent_cache();
3196         f2fs_destroy_checkpoint_caches();
3197         f2fs_destroy_segment_manager_caches();
3198         f2fs_destroy_node_manager_caches();
3199         destroy_inodecache();
3200         f2fs_destroy_trace_ios();
3201 }
3202
3203 module_init(init_f2fs_fs)
3204 module_exit(exit_f2fs_fs)
3205
3206 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3207 MODULE_DESCRIPTION("Flash Friendly File System");
3208 MODULE_LICENSE("GPL");
3209