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 *f2fs_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         [FAULT_DISCARD]         = "discard error",
  59 };
  60
  61 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  62                                                         unsigned int type)
  63 {
  64         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  65
  66         if (rate) {
  67                 atomic_set(&ffi->inject_ops, 0);
  68                 ffi->inject_rate = rate;
  69         }
  70
  71         if (type)
  72                 ffi->inject_type = type;
  73
  74         if (!rate && !type)
  75                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  76 }
  77 #endif
  78
  79 /* f2fs-wide shrinker description */
  80 static struct shrinker f2fs_shrinker_info = {
  81         .scan_objects = f2fs_shrink_scan,
  82         .count_objects = f2fs_shrink_count,
  83         .seeks = DEFAULT_SEEKS,
  84 };
  85
  86 enum {
  87         Opt_gc_background,
  88         Opt_disable_roll_forward,
  89         Opt_norecovery,
  90         Opt_discard,
  91         Opt_nodiscard,
  92         Opt_noheap,
  93         Opt_heap,
  94         Opt_user_xattr,
  95         Opt_nouser_xattr,
  96         Opt_acl,
  97         Opt_noacl,
  98         Opt_active_logs,
  99         Opt_disable_ext_identify,
 100         Opt_inline_xattr,
 101         Opt_noinline_xattr,
 102         Opt_inline_xattr_size,
 103         Opt_inline_data,
 104         Opt_inline_dentry,
 105         Opt_noinline_dentry,
 106         Opt_flush_merge,
 107         Opt_noflush_merge,
 108         Opt_nobarrier,
 109         Opt_fastboot,
 110         Opt_extent_cache,
 111         Opt_noextent_cache,
 112         Opt_noinline_data,
 113         Opt_data_flush,
 114         Opt_reserve_root,
 115         Opt_resgid,
 116         Opt_resuid,
 117         Opt_mode,
 118         Opt_io_size_bits,
 119         Opt_fault_injection,
 120         Opt_fault_type,
 121         Opt_lazytime,
 122         Opt_nolazytime,
 123         Opt_quota,
 124         Opt_noquota,
 125         Opt_usrquota,
 126         Opt_grpquota,
 127         Opt_prjquota,
 128         Opt_usrjquota,
 129         Opt_grpjquota,
 130         Opt_prjjquota,
 131         Opt_offusrjquota,
 132         Opt_offgrpjquota,
 133         Opt_offprjjquota,
 134         Opt_jqfmt_vfsold,
 135         Opt_jqfmt_vfsv0,
 136         Opt_jqfmt_vfsv1,
 137         Opt_whint,
 138         Opt_alloc,
 139         Opt_fsync,
 140         Opt_test_dummy_encryption,
 141         Opt_err,
 142 };
 143
 144 static match_table_t f2fs_tokens = {
 145         {Opt_gc_background, "background_gc=%s"},
 146         {Opt_disable_roll_forward, "disable_roll_forward"},
 147         {Opt_norecovery, "norecovery"},
 148         {Opt_discard, "discard"},
 149         {Opt_nodiscard, "nodiscard"},
 150         {Opt_noheap, "no_heap"},
 151         {Opt_heap, "heap"},
 152         {Opt_user_xattr, "user_xattr"},
 153         {Opt_nouser_xattr, "nouser_xattr"},
 154         {Opt_acl, "acl"},
 155         {Opt_noacl, "noacl"},
 156         {Opt_active_logs, "active_logs=%u"},
 157         {Opt_disable_ext_identify, "disable_ext_identify"},
 158         {Opt_inline_xattr, "inline_xattr"},
 159         {Opt_noinline_xattr, "noinline_xattr"},
 160         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 161         {Opt_inline_data, "inline_data"},
 162         {Opt_inline_dentry, "inline_dentry"},
 163         {Opt_noinline_dentry, "noinline_dentry"},
 164         {Opt_flush_merge, "flush_merge"},
 165         {Opt_noflush_merge, "noflush_merge"},
 166         {Opt_nobarrier, "nobarrier"},
 167         {Opt_fastboot, "fastboot"},
 168         {Opt_extent_cache, "extent_cache"},
 169         {Opt_noextent_cache, "noextent_cache"},
 170         {Opt_noinline_data, "noinline_data"},
 171         {Opt_data_flush, "data_flush"},
 172         {Opt_reserve_root, "reserve_root=%u"},
 173         {Opt_resgid, "resgid=%u"},
 174         {Opt_resuid, "resuid=%u"},
 175         {Opt_mode, "mode=%s"},
 176         {Opt_io_size_bits, "io_bits=%u"},
 177         {Opt_fault_injection, "fault_injection=%u"},
 178         {Opt_fault_type, "fault_type=%u"},
 179         {Opt_lazytime, "lazytime"},
 180         {Opt_nolazytime, "nolazytime"},
 181         {Opt_quota, "quota"},
 182         {Opt_noquota, "noquota"},
 183         {Opt_usrquota, "usrquota"},
 184         {Opt_grpquota, "grpquota"},
 185         {Opt_prjquota, "prjquota"},
 186         {Opt_usrjquota, "usrjquota=%s"},
 187         {Opt_grpjquota, "grpjquota=%s"},
 188         {Opt_prjjquota, "prjjquota=%s"},
 189         {Opt_offusrjquota, "usrjquota="},
 190         {Opt_offgrpjquota, "grpjquota="},
 191         {Opt_offprjjquota, "prjjquota="},
 192         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 193         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 194         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 195         {Opt_whint, "whint_mode=%s"},
 196         {Opt_alloc, "alloc_mode=%s"},
 197         {Opt_fsync, "fsync_mode=%s"},
 198         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 199         {Opt_err, NULL},
 200 };
 201
 202 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 203 {
 204         struct va_format vaf;
 205         va_list args;
 206
 207         va_start(args, fmt);
 208         vaf.fmt = fmt;
 209         vaf.va = &args;
 210         printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 211         va_end(args);
 212 }
 213
 214 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 215 {
 216         block_t limit = (sbi->user_block_count << 1) / 1000;
 217
 218         /* limit is 0.2% */
 219         if (test_opt(sbi, RESERVE_ROOT) &&
 220                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 221                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 222                 f2fs_msg(sbi->sb, KERN_INFO,
 223                         "Reduce reserved blocks for root = %u",
 224                         F2FS_OPTION(sbi).root_reserved_blocks);
 225         }
 226         if (!test_opt(sbi, RESERVE_ROOT) &&
 227                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 228                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 229                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 230                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 231                 f2fs_msg(sbi->sb, KERN_INFO,
 232                         "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 233                                 from_kuid_munged(&init_user_ns,
 234                                         F2FS_OPTION(sbi).s_resuid),
 235                                 from_kgid_munged(&init_user_ns,
 236                                         F2FS_OPTION(sbi).s_resgid));
 237 }
 238
 239 static void init_once(void *foo)
 240 {
 241         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 242
 243         inode_init_once(&fi->vfs_inode);
 244 }
 245
 246 #ifdef CONFIG_QUOTA
 247 static const char * const quotatypes[] = INITQFNAMES;
 248 #define QTYPE2NAME(t) (quotatypes[t])
 249 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 250                                                         substring_t *args)
 251 {
 252         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 253         char *qname;
 254         int ret = -EINVAL;
 255
 256         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 257                 f2fs_msg(sb, KERN_ERR,
 258                         "Cannot change journaled "
 259                         "quota options when quota turned on");
 260                 return -EINVAL;
 261         }
 262         if (f2fs_sb_has_quota_ino(sb)) {
 263                 f2fs_msg(sb, KERN_INFO,
 264                         "QUOTA feature is enabled, so ignore qf_name");
 265                 return 0;
 266         }
 267
 268         qname = match_strdup(args);
 269         if (!qname) {
 270                 f2fs_msg(sb, KERN_ERR,
 271                         "Not enough memory for storing quotafile name");
 272                 return -EINVAL;
 273         }
 274         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 275                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 276                         ret = 0;
 277                 else
 278                         f2fs_msg(sb, KERN_ERR,
 279                                  "%s quota file already specified",
 280                                  QTYPE2NAME(qtype));
 281                 goto errout;
 282         }
 283         if (strchr(qname, '/')) {
 284                 f2fs_msg(sb, KERN_ERR,
 285                         "quotafile must be on filesystem root");
 286                 goto errout;
 287         }
 288         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 289         set_opt(sbi, QUOTA);
 290         return 0;
 291 errout:
 292         kfree(qname);
 293         return ret;
 294 }
 295
 296 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 297 {
 298         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 299
 300         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 301                 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
 302                         " when quota turned on");
 303                 return -EINVAL;
 304         }
 305         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 306         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 307         return 0;
 308 }
 309
 310 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 311 {
 312         /*
 313          * We do the test below only for project quotas. 'usrquota' and
 314          * 'grpquota' mount options are allowed even without quota feature
 315          * to support legacy quotas in quota files.
 316          */
 317         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
 318                 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
 319                          "Cannot enable project quota enforcement.");
 320                 return -1;
 321         }
 322         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 323                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 324                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 325                 if (test_opt(sbi, USRQUOTA) &&
 326                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 327                         clear_opt(sbi, USRQUOTA);
 328
 329                 if (test_opt(sbi, GRPQUOTA) &&
 330                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 331                         clear_opt(sbi, GRPQUOTA);
 332
 333                 if (test_opt(sbi, PRJQUOTA) &&
 334                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 335                         clear_opt(sbi, PRJQUOTA);
 336
 337                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 338                                 test_opt(sbi, PRJQUOTA)) {
 339                         f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
 340                                         "format mixing");
 341                         return -1;
 342                 }
 343
 344                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 345                         f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
 346                                         "not specified");
 347                         return -1;
 348                 }
 349         }
 350
 351         if (f2fs_sb_has_quota_ino(sbi->sb) && F2FS_OPTION(sbi).s_jquota_fmt) {
 352                 f2fs_msg(sbi->sb, KERN_INFO,
 353                         "QUOTA feature is enabled, so ignore jquota_fmt");
 354                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 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         substring_t args[MAX_OPT_ARGS];
 364         char *p, *name;
 365         int arg = 0;
 366         kuid_t uid;
 367         kgid_t gid;
 368 #ifdef CONFIG_QUOTA
 369         int ret;
 370 #endif
 371
 372         if (!options)
 373                 return 0;
 374
 375         while ((p = strsep(&options, ",")) != NULL) {
 376                 int token;
 377                 if (!*p)
 378                         continue;
 379                 /*
 380                  * Initialize args struct so we know whether arg was
 381                  * found; some options take optional arguments.
 382                  */
 383                 args[0].to = args[0].from = NULL;
 384                 token = match_token(p, f2fs_tokens, args);
 385
 386                 switch (token) {
 387                 case Opt_gc_background:
 388                         name = match_strdup(&args[0]);
 389
 390                         if (!name)
 391                                 return -ENOMEM;
 392                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 393                                 set_opt(sbi, BG_GC);
 394                                 clear_opt(sbi, FORCE_FG_GC);
 395                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 396                                 clear_opt(sbi, BG_GC);
 397                                 clear_opt(sbi, FORCE_FG_GC);
 398                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 399                                 set_opt(sbi, BG_GC);
 400                                 set_opt(sbi, FORCE_FG_GC);
 401                         } else {
 402                                 kfree(name);
 403                                 return -EINVAL;
 404                         }
 405                         kfree(name);
 406                         break;
 407                 case Opt_disable_roll_forward:
 408                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 409                         break;
 410                 case Opt_norecovery:
 411                         /* this option mounts f2fs with ro */
 412                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 413                         if (!f2fs_readonly(sb))
 414                                 return -EINVAL;
 415                         break;
 416                 case Opt_discard:
 417                         set_opt(sbi, DISCARD);
 418                         break;
 419                 case Opt_nodiscard:
 420                         if (f2fs_sb_has_blkzoned(sb)) {
 421                                 f2fs_msg(sb, KERN_WARNING,
 422                                         "discard is required for zoned block devices");
 423                                 return -EINVAL;
 424                         }
 425                         clear_opt(sbi, DISCARD);
 426                         break;
 427                 case Opt_noheap:
 428                         set_opt(sbi, NOHEAP);
 429                         break;
 430                 case Opt_heap:
 431                         clear_opt(sbi, NOHEAP);
 432                         break;
 433 #ifdef CONFIG_F2FS_FS_XATTR
 434                 case Opt_user_xattr:
 435                         set_opt(sbi, XATTR_USER);
 436                         break;
 437                 case Opt_nouser_xattr:
 438                         clear_opt(sbi, XATTR_USER);
 439                         break;
 440                 case Opt_inline_xattr:
 441                         set_opt(sbi, INLINE_XATTR);
 442                         break;
 443                 case Opt_noinline_xattr:
 444                         clear_opt(sbi, INLINE_XATTR);
 445                         break;
 446                 case Opt_inline_xattr_size:
 447                         if (args->from && match_int(args, &arg))
 448                                 return -EINVAL;
 449                         set_opt(sbi, INLINE_XATTR_SIZE);
 450                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 451                         break;
 452 #else
 453                 case Opt_user_xattr:
 454                         f2fs_msg(sb, KERN_INFO,
 455                                 "user_xattr options not supported");
 456                         break;
 457                 case Opt_nouser_xattr:
 458                         f2fs_msg(sb, KERN_INFO,
 459                                 "nouser_xattr options not supported");
 460                         break;
 461                 case Opt_inline_xattr:
 462                         f2fs_msg(sb, KERN_INFO,
 463                                 "inline_xattr options not supported");
 464                         break;
 465                 case Opt_noinline_xattr:
 466                         f2fs_msg(sb, KERN_INFO,
 467                                 "noinline_xattr options not supported");
 468                         break;
 469 #endif
 470 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 471                 case Opt_acl:
 472                         set_opt(sbi, POSIX_ACL);
 473                         break;
 474                 case Opt_noacl:
 475                         clear_opt(sbi, POSIX_ACL);
 476                         break;
 477 #else
 478                 case Opt_acl:
 479                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 480                         break;
 481                 case Opt_noacl:
 482                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 483                         break;
 484 #endif
 485                 case Opt_active_logs:
 486                         if (args->from && match_int(args, &arg))
 487                                 return -EINVAL;
 488                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 489                                 return -EINVAL;
 490                         F2FS_OPTION(sbi).active_logs = arg;
 491                         break;
 492                 case Opt_disable_ext_identify:
 493                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 494                         break;
 495                 case Opt_inline_data:
 496                         set_opt(sbi, INLINE_DATA);
 497                         break;
 498                 case Opt_inline_dentry:
 499                         set_opt(sbi, INLINE_DENTRY);
 500                         break;
 501                 case Opt_noinline_dentry:
 502                         clear_opt(sbi, INLINE_DENTRY);
 503                         break;
 504                 case Opt_flush_merge:
 505                         set_opt(sbi, FLUSH_MERGE);
 506                         break;
 507                 case Opt_noflush_merge:
 508                         clear_opt(sbi, FLUSH_MERGE);
 509                         break;
 510                 case Opt_nobarrier:
 511                         set_opt(sbi, NOBARRIER);
 512                         break;
 513                 case Opt_fastboot:
 514                         set_opt(sbi, FASTBOOT);
 515                         break;
 516                 case Opt_extent_cache:
 517                         set_opt(sbi, EXTENT_CACHE);
 518                         break;
 519                 case Opt_noextent_cache:
 520                         clear_opt(sbi, EXTENT_CACHE);
 521                         break;
 522                 case Opt_noinline_data:
 523                         clear_opt(sbi, INLINE_DATA);
 524                         break;
 525                 case Opt_data_flush:
 526                         set_opt(sbi, DATA_FLUSH);
 527                         break;
 528                 case Opt_reserve_root:
 529                         if (args->from && match_int(args, &arg))
 530                                 return -EINVAL;
 531                         if (test_opt(sbi, RESERVE_ROOT)) {
 532                                 f2fs_msg(sb, KERN_INFO,
 533                                         "Preserve previous reserve_root=%u",
 534                                         F2FS_OPTION(sbi).root_reserved_blocks);
 535                         } else {
 536                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 537                                 set_opt(sbi, RESERVE_ROOT);
 538                         }
 539                         break;
 540                 case Opt_resuid:
 541                         if (args->from && match_int(args, &arg))
 542                                 return -EINVAL;
 543                         uid = make_kuid(current_user_ns(), arg);
 544                         if (!uid_valid(uid)) {
 545                                 f2fs_msg(sb, KERN_ERR,
 546                                         "Invalid uid value %d", arg);
 547                                 return -EINVAL;
 548                         }
 549                         F2FS_OPTION(sbi).s_resuid = uid;
 550                         break;
 551                 case Opt_resgid:
 552                         if (args->from && match_int(args, &arg))
 553                                 return -EINVAL;
 554                         gid = make_kgid(current_user_ns(), arg);
 555                         if (!gid_valid(gid)) {
 556                                 f2fs_msg(sb, KERN_ERR,
 557                                         "Invalid gid value %d", arg);
 558                                 return -EINVAL;
 559                         }
 560                         F2FS_OPTION(sbi).s_resgid = gid;
 561                         break;
 562                 case Opt_mode:
 563                         name = match_strdup(&args[0]);
 564
 565                         if (!name)
 566                                 return -ENOMEM;
 567                         if (strlen(name) == 8 &&
 568                                         !strncmp(name, "adaptive", 8)) {
 569                                 if (f2fs_sb_has_blkzoned(sb)) {
 570                                         f2fs_msg(sb, KERN_WARNING,
 571                                                  "adaptive mode is not allowed with "
 572                                                  "zoned block device feature");
 573                                         kfree(name);
 574                                         return -EINVAL;
 575                                 }
 576                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 577                         } else if (strlen(name) == 3 &&
 578                                         !strncmp(name, "lfs", 3)) {
 579                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 580                         } else {
 581                                 kfree(name);
 582                                 return -EINVAL;
 583                         }
 584                         kfree(name);
 585                         break;
 586                 case Opt_io_size_bits:
 587                         if (args->from && match_int(args, &arg))
 588                                 return -EINVAL;
 589                         if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
 590                                 f2fs_msg(sb, KERN_WARNING,
 591                                         "Not support %d, larger than %d",
 592                                         1 << arg, BIO_MAX_PAGES);
 593                                 return -EINVAL;
 594                         }
 595                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 596                         break;
 597                 case Opt_fault_injection:
 598                         if (args->from && match_int(args, &arg))
 599                                 return -EINVAL;
 600 #ifdef CONFIG_F2FS_FAULT_INJECTION
 601                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 602                         set_opt(sbi, FAULT_INJECTION);
 603 #else
 604                         f2fs_msg(sb, KERN_INFO,
 605                                 "FAULT_INJECTION was not selected");
 606 #endif
 607                         break;
 608                 case Opt_fault_type:
 609                         if (args->from && match_int(args, &arg))
 610                                 return -EINVAL;
 611 #ifdef CONFIG_F2FS_FAULT_INJECTION
 612                         f2fs_build_fault_attr(sbi, 0, arg);
 613                         set_opt(sbi, FAULT_INJECTION);
 614 #else
 615                         f2fs_msg(sb, KERN_INFO,
 616                                 "FAULT_INJECTION was not selected");
 617 #endif
 618                         break;
 619                 case Opt_lazytime:
 620                         sb->s_flags |= SB_LAZYTIME;
 621                         break;
 622                 case Opt_nolazytime:
 623                         sb->s_flags &= ~SB_LAZYTIME;
 624                         break;
 625 #ifdef CONFIG_QUOTA
 626                 case Opt_quota:
 627                 case Opt_usrquota:
 628                         set_opt(sbi, USRQUOTA);
 629                         break;
 630                 case Opt_grpquota:
 631                         set_opt(sbi, GRPQUOTA);
 632                         break;
 633                 case Opt_prjquota:
 634                         set_opt(sbi, PRJQUOTA);
 635                         break;
 636                 case Opt_usrjquota:
 637                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 638                         if (ret)
 639                                 return ret;
 640                         break;
 641                 case Opt_grpjquota:
 642                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 643                         if (ret)
 644                                 return ret;
 645                         break;
 646                 case Opt_prjjquota:
 647                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 648                         if (ret)
 649                                 return ret;
 650                         break;
 651                 case Opt_offusrjquota:
 652                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 653                         if (ret)
 654                                 return ret;
 655                         break;
 656                 case Opt_offgrpjquota:
 657                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 658                         if (ret)
 659                                 return ret;
 660                         break;
 661                 case Opt_offprjjquota:
 662                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 663                         if (ret)
 664                                 return ret;
 665                         break;
 666                 case Opt_jqfmt_vfsold:
 667                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 668                         break;
 669                 case Opt_jqfmt_vfsv0:
 670                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 671                         break;
 672                 case Opt_jqfmt_vfsv1:
 673                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 674                         break;
 675                 case Opt_noquota:
 676                         clear_opt(sbi, QUOTA);
 677                         clear_opt(sbi, USRQUOTA);
 678                         clear_opt(sbi, GRPQUOTA);
 679                         clear_opt(sbi, PRJQUOTA);
 680                         break;
 681 #else
 682                 case Opt_quota:
 683                 case Opt_usrquota:
 684                 case Opt_grpquota:
 685                 case Opt_prjquota:
 686                 case Opt_usrjquota:
 687                 case Opt_grpjquota:
 688                 case Opt_prjjquota:
 689                 case Opt_offusrjquota:
 690                 case Opt_offgrpjquota:
 691                 case Opt_offprjjquota:
 692                 case Opt_jqfmt_vfsold:
 693                 case Opt_jqfmt_vfsv0:
 694                 case Opt_jqfmt_vfsv1:
 695                 case Opt_noquota:
 696                         f2fs_msg(sb, KERN_INFO,
 697                                         "quota operations not supported");
 698                         break;
 699 #endif
 700                 case Opt_whint:
 701                         name = match_strdup(&args[0]);
 702                         if (!name)
 703                                 return -ENOMEM;
 704                         if (strlen(name) == 10 &&
 705                                         !strncmp(name, "user-based", 10)) {
 706                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 707                         } else if (strlen(name) == 3 &&
 708                                         !strncmp(name, "off", 3)) {
 709                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 710                         } else if (strlen(name) == 8 &&
 711                                         !strncmp(name, "fs-based", 8)) {
 712                                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 713                         } else {
 714                                 kfree(name);
 715                                 return -EINVAL;
 716                         }
 717                         kfree(name);
 718                         break;
 719                 case Opt_alloc:
 720                         name = match_strdup(&args[0]);
 721                         if (!name)
 722                                 return -ENOMEM;
 723
 724                         if (strlen(name) == 7 &&
 725                                         !strncmp(name, "default", 7)) {
 726                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 727                         } else if (strlen(name) == 5 &&
 728                                         !strncmp(name, "reuse", 5)) {
 729                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 730                         } else {
 731                                 kfree(name);
 732                                 return -EINVAL;
 733                         }
 734                         kfree(name);
 735                         break;
 736                 case Opt_fsync:
 737                         name = match_strdup(&args[0]);
 738                         if (!name)
 739                                 return -ENOMEM;
 740                         if (strlen(name) == 5 &&
 741                                         !strncmp(name, "posix", 5)) {
 742                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 743                         } else if (strlen(name) == 6 &&
 744                                         !strncmp(name, "strict", 6)) {
 745                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 746                         } else if (strlen(name) == 9 &&
 747                                         !strncmp(name, "nobarrier", 9)) {
 748                                 F2FS_OPTION(sbi).fsync_mode =
 749                                                         FSYNC_MODE_NOBARRIER;
 750                         } else {
 751                                 kfree(name);
 752                                 return -EINVAL;
 753                         }
 754                         kfree(name);
 755                         break;
 756                 case Opt_test_dummy_encryption:
 757 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 758                         if (!f2fs_sb_has_encrypt(sb)) {
 759                                 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off");
 760                                 return -EINVAL;
 761                         }
 762
 763                         F2FS_OPTION(sbi).test_dummy_encryption = true;
 764                         f2fs_msg(sb, KERN_INFO,
 765                                         "Test dummy encryption mode enabled");
 766 #else
 767                         f2fs_msg(sb, KERN_INFO,
 768                                         "Test dummy encryption mount option ignored");
 769 #endif
 770                         break;
 771                 default:
 772                         f2fs_msg(sb, KERN_ERR,
 773                                 "Unrecognized mount option \"%s\" or missing value",
 774                                 p);
 775                         return -EINVAL;
 776                 }
 777         }
 778 #ifdef CONFIG_QUOTA
 779         if (f2fs_check_quota_options(sbi))
 780                 return -EINVAL;
 781 #else
 782         if (f2fs_sb_has_quota_ino(sbi->sb) && !f2fs_readonly(sbi->sb)) {
 783                 f2fs_msg(sbi->sb, KERN_INFO,
 784                          "Filesystem with quota feature cannot be mounted RDWR "
 785                          "without CONFIG_QUOTA");
 786                 return -EINVAL;
 787         }
 788         if (f2fs_sb_has_project_quota(sbi->sb) && !f2fs_readonly(sbi->sb)) {
 789                 f2fs_msg(sb, KERN_ERR,
 790                         "Filesystem with project quota feature cannot be "
 791                         "mounted RDWR without CONFIG_QUOTA");
 792                 return -EINVAL;
 793         }
 794 #endif
 795
 796         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 797                 f2fs_msg(sb, KERN_ERR,
 798                                 "Should set mode=lfs with %uKB-sized IO",
 799                                 F2FS_IO_SIZE_KB(sbi));
 800                 return -EINVAL;
 801         }
 802
 803         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 804                 if (!f2fs_sb_has_extra_attr(sb) ||
 805                         !f2fs_sb_has_flexible_inline_xattr(sb)) {
 806                         f2fs_msg(sb, KERN_ERR,
 807                                         "extra_attr or flexible_inline_xattr "
 808                                         "feature is off");
 809                         return -EINVAL;
 810                 }
 811                 if (!test_opt(sbi, INLINE_XATTR)) {
 812                         f2fs_msg(sb, KERN_ERR,
 813                                         "inline_xattr_size option should be "
 814                                         "set with inline_xattr option");
 815                         return -EINVAL;
 816                 }
 817                 if (F2FS_OPTION(sbi).inline_xattr_size <
 818                         sizeof(struct f2fs_xattr_header) / sizeof(__le32) ||
 819                         F2FS_OPTION(sbi).inline_xattr_size >
 820                         DEF_ADDRS_PER_INODE -
 821                         F2FS_TOTAL_EXTRA_ATTR_SIZE / sizeof(__le32) -
 822                         DEF_INLINE_RESERVED_SIZE -
 823                         MIN_INLINE_DENTRY_SIZE / sizeof(__le32)) {
 824                         f2fs_msg(sb, KERN_ERR,
 825                                         "inline xattr size is out of range");
 826                         return -EINVAL;
 827                 }
 828         }
 829
 830         /* Not pass down write hints if the number of active logs is lesser
 831          * than NR_CURSEG_TYPE.
 832          */
 833         if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 834                 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 835         return 0;
 836 }
 837
 838 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 839 {
 840         struct f2fs_inode_info *fi;
 841
 842         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 843         if (!fi)
 844                 return NULL;
 845
 846         init_once((void *) fi);
 847
 848         /* Initialize f2fs-specific inode info */
 849         atomic_set(&fi->dirty_pages, 0);
 850         init_rwsem(&fi->i_sem);
 851         INIT_LIST_HEAD(&fi->dirty_list);
 852         INIT_LIST_HEAD(&fi->gdirty_list);
 853         INIT_LIST_HEAD(&fi->inmem_ilist);
 854         INIT_LIST_HEAD(&fi->inmem_pages);
 855         mutex_init(&fi->inmem_lock);
 856         init_rwsem(&fi->i_gc_rwsem[READ]);
 857         init_rwsem(&fi->i_gc_rwsem[WRITE]);
 858         init_rwsem(&fi->i_mmap_sem);
 859         init_rwsem(&fi->i_xattr_sem);
 860
 861         /* Will be used by directory only */
 862         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 863
 864         return &fi->vfs_inode;
 865 }
 866
 867 static int f2fs_drop_inode(struct inode *inode)
 868 {
 869         int ret;
 870         /*
 871          * This is to avoid a deadlock condition like below.
 872          * writeback_single_inode(inode)
 873          *  - f2fs_write_data_page
 874          *    - f2fs_gc -> iput -> evict
 875          *       - inode_wait_for_writeback(inode)
 876          */
 877         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 878                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 879                         /* to avoid evict_inode call simultaneously */
 880                         atomic_inc(&inode->i_count);
 881                         spin_unlock(&inode->i_lock);
 882
 883                         /* some remained atomic pages should discarded */
 884                         if (f2fs_is_atomic_file(inode))
 885                                 f2fs_drop_inmem_pages(inode);
 886
 887                         /* should remain fi->extent_tree for writepage */
 888                         f2fs_destroy_extent_node(inode);
 889
 890                         sb_start_intwrite(inode->i_sb);
 891                         f2fs_i_size_write(inode, 0);
 892
 893                         if (F2FS_HAS_BLOCKS(inode))
 894                                 f2fs_truncate(inode);
 895
 896                         sb_end_intwrite(inode->i_sb);
 897
 898                         spin_lock(&inode->i_lock);
 899                         atomic_dec(&inode->i_count);
 900                 }
 901                 trace_f2fs_drop_inode(inode, 0);
 902                 return 0;
 903         }
 904         ret = generic_drop_inode(inode);
 905         trace_f2fs_drop_inode(inode, ret);
 906         return ret;
 907 }
 908
 909 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 910 {
 911         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 912         int ret = 0;
 913
 914         spin_lock(&sbi->inode_lock[DIRTY_META]);
 915         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 916                 ret = 1;
 917         } else {
 918                 set_inode_flag(inode, FI_DIRTY_INODE);
 919                 stat_inc_dirty_inode(sbi, DIRTY_META);
 920         }
 921         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 922                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 923                                 &sbi->inode_list[DIRTY_META]);
 924                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 925         }
 926         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 927         return ret;
 928 }
 929
 930 void f2fs_inode_synced(struct inode *inode)
 931 {
 932         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 933
 934         spin_lock(&sbi->inode_lock[DIRTY_META]);
 935         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 936                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 937                 return;
 938         }
 939         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 940                 list_del_init(&F2FS_I(inode)->gdirty_list);
 941                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 942         }
 943         clear_inode_flag(inode, FI_DIRTY_INODE);
 944         clear_inode_flag(inode, FI_AUTO_RECOVER);
 945         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 946         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 947 }
 948
 949 /*
 950  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 951  *
 952  * We should call set_dirty_inode to write the dirty inode through write_inode.
 953  */
 954 static void f2fs_dirty_inode(struct inode *inode, int flags)
 955 {
 956         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 957
 958         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 959                         inode->i_ino == F2FS_META_INO(sbi))
 960                 return;
 961
 962         if (flags == I_DIRTY_TIME)
 963                 return;
 964
 965         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 966                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 967
 968         f2fs_inode_dirtied(inode, false);
 969 }
 970
 971 static void f2fs_i_callback(struct rcu_head *head)
 972 {
 973         struct inode *inode = container_of(head, struct inode, i_rcu);
 974         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 975 }
 976
 977 static void f2fs_destroy_inode(struct inode *inode)
 978 {
 979         call_rcu(&inode->i_rcu, f2fs_i_callback);
 980 }
 981
 982 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 983 {
 984         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 985         percpu_counter_destroy(&sbi->total_valid_inode_count);
 986 }
 987
 988 static void destroy_device_list(struct f2fs_sb_info *sbi)
 989 {
 990         int i;
 991
 992         for (i = 0; i < sbi->s_ndevs; i++) {
 993                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
 994 #ifdef CONFIG_BLK_DEV_ZONED
 995                 kfree(FDEV(i).blkz_type);
 996 #endif
 997         }
 998         kfree(sbi->devs);
 999 }
1000
1001 static void f2fs_put_super(struct super_block *sb)
1002 {
1003         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1004         int i;
1005         bool dropped;
1006
1007         f2fs_quota_off_umount(sb);
1008
1009         /* prevent remaining shrinker jobs */
1010         mutex_lock(&sbi->umount_mutex);
1011
1012         /*
1013          * We don't need to do checkpoint when superblock is clean.
1014          * But, the previous checkpoint was not done by umount, it needs to do
1015          * clean checkpoint again.
1016          */
1017         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1018                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
1019                 struct cp_control cpc = {
1020                         .reason = CP_UMOUNT,
1021                 };
1022                 f2fs_write_checkpoint(sbi, &cpc);
1023         }
1024
1025         /* be sure to wait for any on-going discard commands */
1026         dropped = f2fs_wait_discard_bios(sbi);
1027
1028         if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1029                                         !sbi->discard_blks && !dropped) {
1030                 struct cp_control cpc = {
1031                         .reason = CP_UMOUNT | CP_TRIMMED,
1032                 };
1033                 f2fs_write_checkpoint(sbi, &cpc);
1034         }
1035
1036         /*
1037          * normally superblock is clean, so we need to release this.
1038          * In addition, EIO will skip do checkpoint, we need this as well.
1039          */
1040         f2fs_release_ino_entry(sbi, true);
1041
1042         f2fs_leave_shrinker(sbi);
1043         mutex_unlock(&sbi->umount_mutex);
1044
1045         /* our cp_error case, we can wait for any writeback page */
1046         f2fs_flush_merged_writes(sbi);
1047
1048         f2fs_wait_on_all_pages_writeback(sbi);
1049
1050         f2fs_bug_on(sbi, sbi->fsync_node_num);
1051
1052         iput(sbi->node_inode);
1053         sbi->node_inode = NULL;
1054
1055         iput(sbi->meta_inode);
1056         sbi->meta_inode = NULL;
1057
1058         /*
1059          * iput() can update stat information, if f2fs_write_checkpoint()
1060          * above failed with error.
1061          */
1062         f2fs_destroy_stats(sbi);
1063
1064         /* destroy f2fs internal modules */
1065         f2fs_destroy_node_manager(sbi);
1066         f2fs_destroy_segment_manager(sbi);
1067
1068         kfree(sbi->ckpt);
1069
1070         f2fs_unregister_sysfs(sbi);
1071
1072         sb->s_fs_info = NULL;
1073         if (sbi->s_chksum_driver)
1074                 crypto_free_shash(sbi->s_chksum_driver);
1075         kfree(sbi->raw_super);
1076
1077         destroy_device_list(sbi);
1078         mempool_destroy(sbi->write_io_dummy);
1079 #ifdef CONFIG_QUOTA
1080         for (i = 0; i < MAXQUOTAS; i++)
1081                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1082 #endif
1083         destroy_percpu_info(sbi);
1084         for (i = 0; i < NR_PAGE_TYPE; i++)
1085                 kfree(sbi->write_io[i]);
1086         kfree(sbi);
1087 }
1088
1089 int f2fs_sync_fs(struct super_block *sb, int sync)
1090 {
1091         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1092         int err = 0;
1093
1094         if (unlikely(f2fs_cp_error(sbi)))
1095                 return 0;
1096
1097         trace_f2fs_sync_fs(sb, sync);
1098
1099         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1100                 return -EAGAIN;
1101
1102         if (sync) {
1103                 struct cp_control cpc;
1104
1105                 cpc.reason = __get_cp_reason(sbi);
1106
1107                 mutex_lock(&sbi->gc_mutex);
1108                 err = f2fs_write_checkpoint(sbi, &cpc);
1109                 mutex_unlock(&sbi->gc_mutex);
1110         }
1111         f2fs_trace_ios(NULL, 1);
1112
1113         return err;
1114 }
1115
1116 static int f2fs_freeze(struct super_block *sb)
1117 {
1118         if (f2fs_readonly(sb))
1119                 return 0;
1120
1121         /* IO error happened before */
1122         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1123                 return -EIO;
1124
1125         /* must be clean, since sync_filesystem() was already called */
1126         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1127                 return -EINVAL;
1128         return 0;
1129 }
1130
1131 static int f2fs_unfreeze(struct super_block *sb)
1132 {
1133         return 0;
1134 }
1135
1136 #ifdef CONFIG_QUOTA
1137 static int f2fs_statfs_project(struct super_block *sb,
1138                                 kprojid_t projid, struct kstatfs *buf)
1139 {
1140         struct kqid qid;
1141         struct dquot *dquot;
1142         u64 limit;
1143         u64 curblock;
1144
1145         qid = make_kqid_projid(projid);
1146         dquot = dqget(sb, qid);
1147         if (IS_ERR(dquot))
1148                 return PTR_ERR(dquot);
1149         spin_lock(&dquot->dq_dqb_lock);
1150
1151         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
1152                  dquot->dq_dqb.dqb_bsoftlimit :
1153                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
1154         if (limit && buf->f_blocks > limit) {
1155                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1156                 buf->f_blocks = limit;
1157                 buf->f_bfree = buf->f_bavail =
1158                         (buf->f_blocks > curblock) ?
1159                          (buf->f_blocks - curblock) : 0;
1160         }
1161
1162         limit = dquot->dq_dqb.dqb_isoftlimit ?
1163                 dquot->dq_dqb.dqb_isoftlimit :
1164                 dquot->dq_dqb.dqb_ihardlimit;
1165         if (limit && buf->f_files > limit) {
1166                 buf->f_files = limit;
1167                 buf->f_ffree =
1168                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1169                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1170         }
1171
1172         spin_unlock(&dquot->dq_dqb_lock);
1173         dqput(dquot);
1174         return 0;
1175 }
1176 #endif
1177
1178 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1179 {
1180         struct super_block *sb = dentry->d_sb;
1181         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1182         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1183         block_t total_count, user_block_count, start_count;
1184         u64 avail_node_count;
1185
1186         total_count = le64_to_cpu(sbi->raw_super->block_count);
1187         user_block_count = sbi->user_block_count;
1188         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1189         buf->f_type = F2FS_SUPER_MAGIC;
1190         buf->f_bsize = sbi->blocksize;
1191
1192         buf->f_blocks = total_count - start_count;
1193         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1194                                                 sbi->current_reserved_blocks;
1195         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1196                 buf->f_bavail = buf->f_bfree -
1197                                 F2FS_OPTION(sbi).root_reserved_blocks;
1198         else
1199                 buf->f_bavail = 0;
1200
1201         avail_node_count = sbi->total_node_count - sbi->nquota_files -
1202                                                 F2FS_RESERVED_NODE_NUM;
1203
1204         if (avail_node_count > user_block_count) {
1205                 buf->f_files = user_block_count;
1206                 buf->f_ffree = buf->f_bavail;
1207         } else {
1208                 buf->f_files = avail_node_count;
1209                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1210                                         buf->f_bavail);
1211         }
1212
1213         buf->f_namelen = F2FS_NAME_LEN;
1214         buf->f_fsid.val[0] = (u32)id;
1215         buf->f_fsid.val[1] = (u32)(id >> 32);
1216
1217 #ifdef CONFIG_QUOTA
1218         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1219                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1220                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1221         }
1222 #endif
1223         return 0;
1224 }
1225
1226 static inline void f2fs_show_quota_options(struct seq_file *seq,
1227                                            struct super_block *sb)
1228 {
1229 #ifdef CONFIG_QUOTA
1230         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1231
1232         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1233                 char *fmtname = "";
1234
1235                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1236                 case QFMT_VFS_OLD:
1237                         fmtname = "vfsold";
1238                         break;
1239                 case QFMT_VFS_V0:
1240                         fmtname = "vfsv0";
1241                         break;
1242                 case QFMT_VFS_V1:
1243                         fmtname = "vfsv1";
1244                         break;
1245                 }
1246                 seq_printf(seq, ",jqfmt=%s", fmtname);
1247         }
1248
1249         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1250                 seq_show_option(seq, "usrjquota",
1251                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1252
1253         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1254                 seq_show_option(seq, "grpjquota",
1255                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1256
1257         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1258                 seq_show_option(seq, "prjjquota",
1259                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1260 #endif
1261 }
1262
1263 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1264 {
1265         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1266
1267         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1268                 if (test_opt(sbi, FORCE_FG_GC))
1269                         seq_printf(seq, ",background_gc=%s", "sync");
1270                 else
1271                         seq_printf(seq, ",background_gc=%s", "on");
1272         } else {
1273                 seq_printf(seq, ",background_gc=%s", "off");
1274         }
1275         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1276                 seq_puts(seq, ",disable_roll_forward");
1277         if (test_opt(sbi, DISCARD))
1278                 seq_puts(seq, ",discard");
1279         if (test_opt(sbi, NOHEAP))
1280                 seq_puts(seq, ",no_heap");
1281         else
1282                 seq_puts(seq, ",heap");
1283 #ifdef CONFIG_F2FS_FS_XATTR
1284         if (test_opt(sbi, XATTR_USER))
1285                 seq_puts(seq, ",user_xattr");
1286         else
1287                 seq_puts(seq, ",nouser_xattr");
1288         if (test_opt(sbi, INLINE_XATTR))
1289                 seq_puts(seq, ",inline_xattr");
1290         else
1291                 seq_puts(seq, ",noinline_xattr");
1292         if (test_opt(sbi, INLINE_XATTR_SIZE))
1293                 seq_printf(seq, ",inline_xattr_size=%u",
1294                                         F2FS_OPTION(sbi).inline_xattr_size);
1295 #endif
1296 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1297         if (test_opt(sbi, POSIX_ACL))
1298                 seq_puts(seq, ",acl");
1299         else
1300                 seq_puts(seq, ",noacl");
1301 #endif
1302         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1303                 seq_puts(seq, ",disable_ext_identify");
1304         if (test_opt(sbi, INLINE_DATA))
1305                 seq_puts(seq, ",inline_data");
1306         else
1307                 seq_puts(seq, ",noinline_data");
1308         if (test_opt(sbi, INLINE_DENTRY))
1309                 seq_puts(seq, ",inline_dentry");
1310         else
1311                 seq_puts(seq, ",noinline_dentry");
1312         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1313                 seq_puts(seq, ",flush_merge");
1314         if (test_opt(sbi, NOBARRIER))
1315                 seq_puts(seq, ",nobarrier");
1316         if (test_opt(sbi, FASTBOOT))
1317                 seq_puts(seq, ",fastboot");
1318         if (test_opt(sbi, EXTENT_CACHE))
1319                 seq_puts(seq, ",extent_cache");
1320         else
1321                 seq_puts(seq, ",noextent_cache");
1322         if (test_opt(sbi, DATA_FLUSH))
1323                 seq_puts(seq, ",data_flush");
1324
1325         seq_puts(seq, ",mode=");
1326         if (test_opt(sbi, ADAPTIVE))
1327                 seq_puts(seq, "adaptive");
1328         else if (test_opt(sbi, LFS))
1329                 seq_puts(seq, "lfs");
1330         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1331         if (test_opt(sbi, RESERVE_ROOT))
1332                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1333                                 F2FS_OPTION(sbi).root_reserved_blocks,
1334                                 from_kuid_munged(&init_user_ns,
1335                                         F2FS_OPTION(sbi).s_resuid),
1336                                 from_kgid_munged(&init_user_ns,
1337                                         F2FS_OPTION(sbi).s_resgid));
1338         if (F2FS_IO_SIZE_BITS(sbi))
1339                 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1340 #ifdef CONFIG_F2FS_FAULT_INJECTION
1341         if (test_opt(sbi, FAULT_INJECTION)) {
1342                 seq_printf(seq, ",fault_injection=%u",
1343                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1344                 seq_printf(seq, ",fault_type=%u",
1345                                 F2FS_OPTION(sbi).fault_info.inject_type);
1346         }
1347 #endif
1348 #ifdef CONFIG_QUOTA
1349         if (test_opt(sbi, QUOTA))
1350                 seq_puts(seq, ",quota");
1351         if (test_opt(sbi, USRQUOTA))
1352                 seq_puts(seq, ",usrquota");
1353         if (test_opt(sbi, GRPQUOTA))
1354                 seq_puts(seq, ",grpquota");
1355         if (test_opt(sbi, PRJQUOTA))
1356                 seq_puts(seq, ",prjquota");
1357 #endif
1358         f2fs_show_quota_options(seq, sbi->sb);
1359         if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1360                 seq_printf(seq, ",whint_mode=%s", "user-based");
1361         else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1362                 seq_printf(seq, ",whint_mode=%s", "fs-based");
1363 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1364         if (F2FS_OPTION(sbi).test_dummy_encryption)
1365                 seq_puts(seq, ",test_dummy_encryption");
1366 #endif
1367
1368         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1369                 seq_printf(seq, ",alloc_mode=%s", "default");
1370         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1371                 seq_printf(seq, ",alloc_mode=%s", "reuse");
1372
1373         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1374                 seq_printf(seq, ",fsync_mode=%s", "posix");
1375         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1376                 seq_printf(seq, ",fsync_mode=%s", "strict");
1377         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1378                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1379         return 0;
1380 }
1381
1382 static void default_options(struct f2fs_sb_info *sbi)
1383 {
1384         /* init some FS parameters */
1385         F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1386         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1387         F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1388         F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1389         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1390         F2FS_OPTION(sbi).test_dummy_encryption = false;
1391         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1392         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1393
1394         set_opt(sbi, BG_GC);
1395         set_opt(sbi, INLINE_XATTR);
1396         set_opt(sbi, INLINE_DATA);
1397         set_opt(sbi, INLINE_DENTRY);
1398         set_opt(sbi, EXTENT_CACHE);
1399         set_opt(sbi, NOHEAP);
1400         sbi->sb->s_flags |= SB_LAZYTIME;
1401         set_opt(sbi, FLUSH_MERGE);
1402         set_opt(sbi, DISCARD);
1403         if (f2fs_sb_has_blkzoned(sbi->sb))
1404                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1405         else
1406                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1407
1408 #ifdef CONFIG_F2FS_FS_XATTR
1409         set_opt(sbi, XATTR_USER);
1410 #endif
1411 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1412         set_opt(sbi, POSIX_ACL);
1413 #endif
1414
1415         f2fs_build_fault_attr(sbi, 0, 0);
1416 }
1417
1418 #ifdef CONFIG_QUOTA
1419 static int f2fs_enable_quotas(struct super_block *sb);
1420 #endif
1421 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1422 {
1423         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1424         struct f2fs_mount_info org_mount_opt;
1425         unsigned long old_sb_flags;
1426         int err;
1427         bool need_restart_gc = false;
1428         bool need_stop_gc = false;
1429         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1430 #ifdef CONFIG_QUOTA
1431         int i, j;
1432 #endif
1433
1434         /*
1435          * Save the old mount options in case we
1436          * need to restore them.
1437          */
1438         org_mount_opt = sbi->mount_opt;
1439         old_sb_flags = sb->s_flags;
1440
1441 #ifdef CONFIG_QUOTA
1442         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1443         for (i = 0; i < MAXQUOTAS; i++) {
1444                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1445                         org_mount_opt.s_qf_names[i] =
1446                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1447                                 GFP_KERNEL);
1448                         if (!org_mount_opt.s_qf_names[i]) {
1449                                 for (j = 0; j < i; j++)
1450                                         kfree(org_mount_opt.s_qf_names[j]);
1451                                 return -ENOMEM;
1452                         }
1453                 } else {
1454                         org_mount_opt.s_qf_names[i] = NULL;
1455                 }
1456         }
1457 #endif
1458
1459         /* recover superblocks we couldn't write due to previous RO mount */
1460         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1461                 err = f2fs_commit_super(sbi, false);
1462                 f2fs_msg(sb, KERN_INFO,
1463                         "Try to recover all the superblocks, ret: %d", err);
1464                 if (!err)
1465                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1466         }
1467
1468         default_options(sbi);
1469
1470         /* parse mount options */
1471         err = parse_options(sb, data);
1472         if (err)
1473                 goto restore_opts;
1474
1475         /*
1476          * Previous and new state of filesystem is RO,
1477          * so skip checking GC and FLUSH_MERGE conditions.
1478          */
1479         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1480                 goto skip;
1481
1482 #ifdef CONFIG_QUOTA
1483         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1484                 err = dquot_suspend(sb, -1);
1485                 if (err < 0)
1486                         goto restore_opts;
1487         } else if (f2fs_readonly(sb) && !(*flags & MS_RDONLY)) {
1488                 /* dquot_resume needs RW */
1489                 sb->s_flags &= ~SB_RDONLY;
1490                 if (sb_any_quota_suspended(sb)) {
1491                         dquot_resume(sb, -1);
1492                 } else if (f2fs_sb_has_quota_ino(sb)) {
1493                         err = f2fs_enable_quotas(sb);
1494                         if (err)
1495                                 goto restore_opts;
1496                 }
1497         }
1498 #endif
1499         /* disallow enable/disable extent_cache dynamically */
1500         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1501                 err = -EINVAL;
1502                 f2fs_msg(sbi->sb, KERN_WARNING,
1503                                 "switch extent_cache option is not allowed");
1504                 goto restore_opts;
1505         }
1506
1507         /*
1508          * We stop the GC thread if FS is mounted as RO
1509          * or if background_gc = off is passed in mount
1510          * option. Also sync the filesystem.
1511          */
1512         if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) {
1513                 if (sbi->gc_thread) {
1514                         f2fs_stop_gc_thread(sbi);
1515                         need_restart_gc = true;
1516                 }
1517         } else if (!sbi->gc_thread) {
1518                 err = f2fs_start_gc_thread(sbi);
1519                 if (err)
1520                         goto restore_opts;
1521                 need_stop_gc = true;
1522         }
1523
1524         if (*flags & SB_RDONLY ||
1525                 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1526                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1527                 sync_inodes_sb(sb);
1528
1529                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1530                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1531                 f2fs_sync_fs(sb, 1);
1532                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1533         }
1534
1535         /*
1536          * We stop issue flush thread if FS is mounted as RO
1537          * or if flush_merge is not passed in mount option.
1538          */
1539         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1540                 clear_opt(sbi, FLUSH_MERGE);
1541                 f2fs_destroy_flush_cmd_control(sbi, false);
1542         } else {
1543                 err = f2fs_create_flush_cmd_control(sbi);
1544                 if (err)
1545                         goto restore_gc;
1546         }
1547 skip:
1548 #ifdef CONFIG_QUOTA
1549         /* Release old quota file names */
1550         for (i = 0; i < MAXQUOTAS; i++)
1551                 kfree(org_mount_opt.s_qf_names[i]);
1552 #endif
1553         /* Update the POSIXACL Flag */
1554         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1555                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1556
1557         limit_reserve_root(sbi);
1558         return 0;
1559 restore_gc:
1560         if (need_restart_gc) {
1561                 if (f2fs_start_gc_thread(sbi))
1562                         f2fs_msg(sbi->sb, KERN_WARNING,
1563                                 "background gc thread has stopped");
1564         } else if (need_stop_gc) {
1565                 f2fs_stop_gc_thread(sbi);
1566         }
1567 restore_opts:
1568 #ifdef CONFIG_QUOTA
1569         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1570         for (i = 0; i < MAXQUOTAS; i++) {
1571                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1572                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1573         }
1574 #endif
1575         sbi->mount_opt = org_mount_opt;
1576         sb->s_flags = old_sb_flags;
1577         return err;
1578 }
1579
1580 #ifdef CONFIG_QUOTA
1581 /* Read data from quotafile */
1582 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1583                                size_t len, loff_t off)
1584 {
1585         struct inode *inode = sb_dqopt(sb)->files[type];
1586         struct address_space *mapping = inode->i_mapping;
1587         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1588         int offset = off & (sb->s_blocksize - 1);
1589         int tocopy;
1590         size_t toread;
1591         loff_t i_size = i_size_read(inode);
1592         struct page *page;
1593         char *kaddr;
1594
1595         if (off > i_size)
1596                 return 0;
1597
1598         if (off + len > i_size)
1599                 len = i_size - off;
1600         toread = len;
1601         while (toread > 0) {
1602                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1603 repeat:
1604                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1605                 if (IS_ERR(page)) {
1606                         if (PTR_ERR(page) == -ENOMEM) {
1607                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1608                                 goto repeat;
1609                         }
1610                         return PTR_ERR(page);
1611                 }
1612
1613                 lock_page(page);
1614
1615                 if (unlikely(page->mapping != mapping)) {
1616                         f2fs_put_page(page, 1);
1617                         goto repeat;
1618                 }
1619                 if (unlikely(!PageUptodate(page))) {
1620                         f2fs_put_page(page, 1);
1621                         return -EIO;
1622                 }
1623
1624                 kaddr = kmap_atomic(page);
1625                 memcpy(data, kaddr + offset, tocopy);
1626                 kunmap_atomic(kaddr);
1627                 f2fs_put_page(page, 1);
1628
1629                 offset = 0;
1630                 toread -= tocopy;
1631                 data += tocopy;
1632                 blkidx++;
1633         }
1634         return len;
1635 }
1636
1637 /* Write to quotafile */
1638 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1639                                 const char *data, size_t len, loff_t off)
1640 {
1641         struct inode *inode = sb_dqopt(sb)->files[type];
1642         struct address_space *mapping = inode->i_mapping;
1643         const struct address_space_operations *a_ops = mapping->a_ops;
1644         int offset = off & (sb->s_blocksize - 1);
1645         size_t towrite = len;
1646         struct page *page;
1647         char *kaddr;
1648         int err = 0;
1649         int tocopy;
1650
1651         while (towrite > 0) {
1652                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1653                                                                 towrite);
1654 retry:
1655                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1656                                                         &page, NULL);
1657                 if (unlikely(err)) {
1658                         if (err == -ENOMEM) {
1659                                 congestion_wait(BLK_RW_ASYNC, HZ/50);
1660                                 goto retry;
1661                         }
1662                         break;
1663                 }
1664
1665                 kaddr = kmap_atomic(page);
1666                 memcpy(kaddr + offset, data, tocopy);
1667                 kunmap_atomic(kaddr);
1668                 flush_dcache_page(page);
1669
1670                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1671                                                 page, NULL);
1672                 offset = 0;
1673                 towrite -= tocopy;
1674                 off += tocopy;
1675                 data += tocopy;
1676                 cond_resched();
1677         }
1678
1679         if (len == towrite)
1680                 return err;
1681         inode->i_mtime = inode->i_ctime = current_time(inode);
1682         f2fs_mark_inode_dirty_sync(inode, false);
1683         return len - towrite;
1684 }
1685
1686 static struct dquot **f2fs_get_dquots(struct inode *inode)
1687 {
1688         return F2FS_I(inode)->i_dquot;
1689 }
1690
1691 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1692 {
1693         return &F2FS_I(inode)->i_reserved_quota;
1694 }
1695
1696 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1697 {
1698         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
1699                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
1700 }
1701
1702 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
1703 {
1704         int enabled = 0;
1705         int i, err;
1706
1707         if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) {
1708                 err = f2fs_enable_quotas(sbi->sb);
1709                 if (err) {
1710                         f2fs_msg(sbi->sb, KERN_ERR,
1711                                         "Cannot turn on quota_ino: %d", err);
1712                         return 0;
1713                 }
1714                 return 1;
1715         }
1716
1717         for (i = 0; i < MAXQUOTAS; i++) {
1718                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
1719                         err = f2fs_quota_on_mount(sbi, i);
1720                         if (!err) {
1721                                 enabled = 1;
1722                                 continue;
1723                         }
1724                         f2fs_msg(sbi->sb, KERN_ERR,
1725                                 "Cannot turn on quotas: %d on %d", err, i);
1726                 }
1727         }
1728         return enabled;
1729 }
1730
1731 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
1732                              unsigned int flags)
1733 {
1734         struct inode *qf_inode;
1735         unsigned long qf_inum;
1736         int err;
1737
1738         BUG_ON(!f2fs_sb_has_quota_ino(sb));
1739
1740         qf_inum = f2fs_qf_ino(sb, type);
1741         if (!qf_inum)
1742                 return -EPERM;
1743
1744         qf_inode = f2fs_iget(sb, qf_inum);
1745         if (IS_ERR(qf_inode)) {
1746                 f2fs_msg(sb, KERN_ERR,
1747                         "Bad quota inode %u:%lu", type, qf_inum);
1748                 return PTR_ERR(qf_inode);
1749         }
1750
1751         /* Don't account quota for quota files to avoid recursion */
1752         qf_inode->i_flags |= S_NOQUOTA;
1753         err = dquot_enable(qf_inode, type, format_id, flags);
1754         iput(qf_inode);
1755         return err;
1756 }
1757
1758 static int f2fs_enable_quotas(struct super_block *sb)
1759 {
1760         int type, err = 0;
1761         unsigned long qf_inum;
1762         bool quota_mopt[MAXQUOTAS] = {
1763                 test_opt(F2FS_SB(sb), USRQUOTA),
1764                 test_opt(F2FS_SB(sb), GRPQUOTA),
1765                 test_opt(F2FS_SB(sb), PRJQUOTA),
1766         };
1767
1768         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
1769         for (type = 0; type < MAXQUOTAS; type++) {
1770                 qf_inum = f2fs_qf_ino(sb, type);
1771                 if (qf_inum) {
1772                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
1773                                 DQUOT_USAGE_ENABLED |
1774                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
1775                         if (err) {
1776                                 f2fs_msg(sb, KERN_ERR,
1777                                         "Failed to enable quota tracking "
1778                                         "(type=%d, err=%d). Please run "
1779                                         "fsck to fix.", type, err);
1780                                 for (type--; type >= 0; type--)
1781                                         dquot_quota_off(sb, type);
1782                                 return err;
1783                         }
1784                 }
1785         }
1786         return 0;
1787 }
1788
1789 static int f2fs_quota_sync(struct super_block *sb, int type)
1790 {
1791         struct quota_info *dqopt = sb_dqopt(sb);
1792         int cnt;
1793         int ret;
1794
1795         ret = dquot_writeback_dquots(sb, type);
1796         if (ret)
1797                 return ret;
1798
1799         /*
1800          * Now when everything is written we can discard the pagecache so
1801          * that userspace sees the changes.
1802          */
1803         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1804                 if (type != -1 && cnt != type)
1805                         continue;
1806                 if (!sb_has_quota_active(sb, cnt))
1807                         continue;
1808
1809                 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1810                 if (ret)
1811                         return ret;
1812
1813                 inode_lock(dqopt->files[cnt]);
1814                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1815                 inode_unlock(dqopt->files[cnt]);
1816         }
1817         return 0;
1818 }
1819
1820 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1821                                                         const struct path *path)
1822 {
1823         struct inode *inode;
1824         int err;
1825
1826         err = f2fs_quota_sync(sb, type);
1827         if (err)
1828                 return err;
1829
1830         err = dquot_quota_on(sb, type, format_id, path);
1831         if (err)
1832                 return err;
1833
1834         inode = d_inode(path->dentry);
1835
1836         inode_lock(inode);
1837         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
1838         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1839                                         S_NOATIME | S_IMMUTABLE);
1840         inode_unlock(inode);
1841         f2fs_mark_inode_dirty_sync(inode, false);
1842
1843         return 0;
1844 }
1845
1846 static int f2fs_quota_off(struct super_block *sb, int type)
1847 {
1848         struct inode *inode = sb_dqopt(sb)->files[type];
1849         int err;
1850
1851         if (!inode || !igrab(inode))
1852                 return dquot_quota_off(sb, type);
1853
1854         err = f2fs_quota_sync(sb, type);
1855         if (err)
1856                 goto out_put;
1857
1858         err = dquot_quota_off(sb, type);
1859         if (err || f2fs_sb_has_quota_ino(sb))
1860                 goto out_put;
1861
1862         inode_lock(inode);
1863         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
1864         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1865         inode_unlock(inode);
1866         f2fs_mark_inode_dirty_sync(inode, false);
1867 out_put:
1868         iput(inode);
1869         return err;
1870 }
1871
1872 void f2fs_quota_off_umount(struct super_block *sb)
1873 {
1874         int type;
1875         int err;
1876
1877         for (type = 0; type < MAXQUOTAS; type++) {
1878                 err = f2fs_quota_off(sb, type);
1879                 if (err) {
1880                         int ret = dquot_quota_off(sb, type);
1881
1882                         f2fs_msg(sb, KERN_ERR,
1883                                 "Fail to turn off disk quota "
1884                                 "(type: %d, err: %d, ret:%d), Please "
1885                                 "run fsck to fix it.", type, err, ret);
1886                         set_sbi_flag(F2FS_SB(sb), SBI_NEED_FSCK);
1887                 }
1888         }
1889 }
1890
1891 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
1892 {
1893         struct quota_info *dqopt = sb_dqopt(sb);
1894         int type;
1895
1896         for (type = 0; type < MAXQUOTAS; type++) {
1897                 if (!dqopt->files[type])
1898                         continue;
1899                 f2fs_inode_synced(dqopt->files[type]);
1900         }
1901 }
1902
1903
1904 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1905 {
1906         *projid = F2FS_I(inode)->i_projid;
1907         return 0;
1908 }
1909
1910 static const struct dquot_operations f2fs_quota_operations = {
1911         .get_reserved_space = f2fs_get_reserved_space,
1912         .write_dquot    = dquot_commit,
1913         .acquire_dquot  = dquot_acquire,
1914         .release_dquot  = dquot_release,
1915         .mark_dirty     = dquot_mark_dquot_dirty,
1916         .write_info     = dquot_commit_info,
1917         .alloc_dquot    = dquot_alloc,
1918         .destroy_dquot  = dquot_destroy,
1919         .get_projid     = f2fs_get_projid,
1920         .get_next_id    = dquot_get_next_id,
1921 };
1922
1923 static const struct quotactl_ops f2fs_quotactl_ops = {
1924         .quota_on       = f2fs_quota_on,
1925         .quota_off      = f2fs_quota_off,
1926         .quota_sync     = f2fs_quota_sync,
1927         .get_state      = dquot_get_state,
1928         .set_info       = dquot_set_dqinfo,
1929         .get_dqblk      = dquot_get_dqblk,
1930         .set_dqblk      = dquot_set_dqblk,
1931         .get_nextdqblk  = dquot_get_next_dqblk,
1932 };
1933 #else
1934 void f2fs_quota_off_umount(struct super_block *sb)
1935 {
1936 }
1937 #endif
1938
1939 static const struct super_operations f2fs_sops = {
1940         .alloc_inode    = f2fs_alloc_inode,
1941         .drop_inode     = f2fs_drop_inode,
1942         .destroy_inode  = f2fs_destroy_inode,
1943         .write_inode    = f2fs_write_inode,
1944         .dirty_inode    = f2fs_dirty_inode,
1945         .show_options   = f2fs_show_options,
1946 #ifdef CONFIG_QUOTA
1947         .quota_read     = f2fs_quota_read,
1948         .quota_write    = f2fs_quota_write,
1949         .get_dquots     = f2fs_get_dquots,
1950 #endif
1951         .evict_inode    = f2fs_evict_inode,
1952         .put_super      = f2fs_put_super,
1953         .sync_fs        = f2fs_sync_fs,
1954         .freeze_fs      = f2fs_freeze,
1955         .unfreeze_fs    = f2fs_unfreeze,
1956         .statfs         = f2fs_statfs,
1957         .remount_fs     = f2fs_remount,
1958 };
1959
1960 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1961 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1962 {
1963         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1964                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1965                                 ctx, len, NULL);
1966 }
1967
1968 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1969                                                         void *fs_data)
1970 {
1971         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1972
1973         /*
1974          * Encrypting the root directory is not allowed because fsck
1975          * expects lost+found directory to exist and remain unencrypted
1976          * if LOST_FOUND feature is enabled.
1977          *
1978          */
1979         if (f2fs_sb_has_lost_found(sbi->sb) &&
1980                         inode->i_ino == F2FS_ROOT_INO(sbi))
1981                 return -EPERM;
1982
1983         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1984                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1985                                 ctx, len, fs_data, XATTR_CREATE);
1986 }
1987
1988 static bool f2fs_dummy_context(struct inode *inode)
1989 {
1990         return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
1991 }
1992
1993 static const struct fscrypt_operations f2fs_cryptops = {
1994         .key_prefix     = "f2fs:",
1995         .get_context    = f2fs_get_context,
1996         .set_context    = f2fs_set_context,
1997         .dummy_context  = f2fs_dummy_context,
1998         .empty_dir      = f2fs_empty_dir,
1999         .max_namelen    = F2FS_NAME_LEN,
2000 };
2001 #endif
2002
2003 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2004                 u64 ino, u32 generation)
2005 {
2006         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2007         struct inode *inode;
2008
2009         if (f2fs_check_nid_range(sbi, ino))
2010                 return ERR_PTR(-ESTALE);
2011
2012         /*
2013          * f2fs_iget isn't quite right if the inode is currently unallocated!
2014          * However f2fs_iget currently does appropriate checks to handle stale
2015          * inodes so everything is OK.
2016          */
2017         inode = f2fs_iget(sb, ino);
2018         if (IS_ERR(inode))
2019                 return ERR_CAST(inode);
2020         if (unlikely(generation && inode->i_generation != generation)) {
2021                 /* we didn't find the right inode.. */
2022                 iput(inode);
2023                 return ERR_PTR(-ESTALE);
2024         }
2025         return inode;
2026 }
2027
2028 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2029                 int fh_len, int fh_type)
2030 {
2031         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2032                                     f2fs_nfs_get_inode);
2033 }
2034
2035 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2036                 int fh_len, int fh_type)
2037 {
2038         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2039                                     f2fs_nfs_get_inode);
2040 }
2041
2042 static const struct export_operations f2fs_export_ops = {
2043         .fh_to_dentry = f2fs_fh_to_dentry,
2044         .fh_to_parent = f2fs_fh_to_parent,
2045         .get_parent = f2fs_get_parent,
2046 };
2047
2048 static loff_t max_file_blocks(void)
2049 {
2050         loff_t result = 0;
2051         loff_t leaf_count = ADDRS_PER_BLOCK;
2052
2053         /*
2054          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2055          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2056          * space in inode.i_addr, it will be more safe to reassign
2057          * result as zero.
2058          */
2059
2060         /* two direct node blocks */
2061         result += (leaf_count * 2);
2062
2063         /* two indirect node blocks */
2064         leaf_count *= NIDS_PER_BLOCK;
2065         result += (leaf_count * 2);
2066
2067         /* one double indirect node block */
2068         leaf_count *= NIDS_PER_BLOCK;
2069         result += leaf_count;
2070
2071         return result;
2072 }
2073
2074 static int __f2fs_commit_super(struct buffer_head *bh,
2075                         struct f2fs_super_block *super)
2076 {
2077         lock_buffer(bh);
2078         if (super)
2079                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2080         set_buffer_dirty(bh);
2081         unlock_buffer(bh);
2082
2083         /* it's rare case, we can do fua all the time */
2084         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2085 }
2086
2087 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2088                                         struct buffer_head *bh)
2089 {
2090         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2091                                         (bh->b_data + F2FS_SUPER_OFFSET);
2092         struct super_block *sb = sbi->sb;
2093         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2094         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2095         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2096         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2097         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2098         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2099         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2100         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2101         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2102         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2103         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2104         u32 segment_count = le32_to_cpu(raw_super->segment_count);
2105         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2106         u64 main_end_blkaddr = main_blkaddr +
2107                                 (segment_count_main << log_blocks_per_seg);
2108         u64 seg_end_blkaddr = segment0_blkaddr +
2109                                 (segment_count << log_blocks_per_seg);
2110
2111         if (segment0_blkaddr != cp_blkaddr) {
2112                 f2fs_msg(sb, KERN_INFO,
2113                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2114                         segment0_blkaddr, cp_blkaddr);
2115                 return true;
2116         }
2117
2118         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2119                                                         sit_blkaddr) {
2120                 f2fs_msg(sb, KERN_INFO,
2121                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2122                         cp_blkaddr, sit_blkaddr,
2123                         segment_count_ckpt << log_blocks_per_seg);
2124                 return true;
2125         }
2126
2127         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2128                                                         nat_blkaddr) {
2129                 f2fs_msg(sb, KERN_INFO,
2130                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2131                         sit_blkaddr, nat_blkaddr,
2132                         segment_count_sit << log_blocks_per_seg);
2133                 return true;
2134         }
2135
2136         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2137                                                         ssa_blkaddr) {
2138                 f2fs_msg(sb, KERN_INFO,
2139                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2140                         nat_blkaddr, ssa_blkaddr,
2141                         segment_count_nat << log_blocks_per_seg);
2142                 return true;
2143         }
2144
2145         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2146                                                         main_blkaddr) {
2147                 f2fs_msg(sb, KERN_INFO,
2148                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2149                         ssa_blkaddr, main_blkaddr,
2150                         segment_count_ssa << log_blocks_per_seg);
2151                 return true;
2152         }
2153
2154         if (main_end_blkaddr > seg_end_blkaddr) {
2155                 f2fs_msg(sb, KERN_INFO,
2156                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2157                         main_blkaddr,
2158                         segment0_blkaddr +
2159                                 (segment_count << log_blocks_per_seg),
2160                         segment_count_main << log_blocks_per_seg);
2161                 return true;
2162         } else if (main_end_blkaddr < seg_end_blkaddr) {
2163                 int err = 0;
2164                 char *res;
2165
2166                 /* fix in-memory information all the time */
2167                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2168                                 segment0_blkaddr) >> log_blocks_per_seg);
2169
2170                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2171                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2172                         res = "internally";
2173                 } else {
2174                         err = __f2fs_commit_super(bh, NULL);
2175                         res = err ? "failed" : "done";
2176                 }
2177                 f2fs_msg(sb, KERN_INFO,
2178                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
2179                         res, main_blkaddr,
2180                         segment0_blkaddr +
2181                                 (segment_count << log_blocks_per_seg),
2182                         segment_count_main << log_blocks_per_seg);
2183                 if (err)
2184                         return true;
2185         }
2186         return false;
2187 }
2188
2189 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2190                                 struct buffer_head *bh)
2191 {
2192         block_t segment_count, segs_per_sec, secs_per_zone;
2193         block_t total_sections, blocks_per_seg;
2194         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2195                                         (bh->b_data + F2FS_SUPER_OFFSET);
2196         struct super_block *sb = sbi->sb;
2197         unsigned int blocksize;
2198
2199         if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
2200                 f2fs_msg(sb, KERN_INFO,
2201                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
2202                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2203                 return 1;
2204         }
2205
2206         /* Currently, support only 4KB page cache size */
2207         if (F2FS_BLKSIZE != PAGE_SIZE) {
2208                 f2fs_msg(sb, KERN_INFO,
2209                         "Invalid page_cache_size (%lu), supports only 4KB\n",
2210                         PAGE_SIZE);
2211                 return 1;
2212         }
2213
2214         /* Currently, support only 4KB block size */
2215         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2216         if (blocksize != F2FS_BLKSIZE) {
2217                 f2fs_msg(sb, KERN_INFO,
2218                         "Invalid blocksize (%u), supports only 4KB\n",
2219                         blocksize);
2220                 return 1;
2221         }
2222
2223         /* check log blocks per segment */
2224         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2225                 f2fs_msg(sb, KERN_INFO,
2226                         "Invalid log blocks per segment (%u)\n",
2227                         le32_to_cpu(raw_super->log_blocks_per_seg));
2228                 return 1;
2229         }
2230
2231         /* Currently, support 512/1024/2048/4096 bytes sector size */
2232         if (le32_to_cpu(raw_super->log_sectorsize) >
2233                                 F2FS_MAX_LOG_SECTOR_SIZE ||
2234                 le32_to_cpu(raw_super->log_sectorsize) <
2235                                 F2FS_MIN_LOG_SECTOR_SIZE) {
2236                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
2237                         le32_to_cpu(raw_super->log_sectorsize));
2238                 return 1;
2239         }
2240         if (le32_to_cpu(raw_super->log_sectors_per_block) +
2241                 le32_to_cpu(raw_super->log_sectorsize) !=
2242                         F2FS_MAX_LOG_SECTOR_SIZE) {
2243                 f2fs_msg(sb, KERN_INFO,
2244                         "Invalid log sectors per block(%u) log sectorsize(%u)",
2245                         le32_to_cpu(raw_super->log_sectors_per_block),
2246                         le32_to_cpu(raw_super->log_sectorsize));
2247                 return 1;
2248         }
2249
2250         segment_count = le32_to_cpu(raw_super->segment_count);
2251         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2252         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2253         total_sections = le32_to_cpu(raw_super->section_count);
2254
2255         /* blocks_per_seg should be 512, given the above check */
2256         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2257
2258         if (segment_count > F2FS_MAX_SEGMENT ||
2259                                 segment_count < F2FS_MIN_SEGMENTS) {
2260                 f2fs_msg(sb, KERN_INFO,
2261                         "Invalid segment count (%u)",
2262                         segment_count);
2263                 return 1;
2264         }
2265
2266         if (total_sections > segment_count ||
2267                         total_sections < F2FS_MIN_SEGMENTS ||
2268                         segs_per_sec > segment_count || !segs_per_sec) {
2269                 f2fs_msg(sb, KERN_INFO,
2270                         "Invalid segment/section count (%u, %u x %u)",
2271                         segment_count, total_sections, segs_per_sec);
2272                 return 1;
2273         }
2274
2275         if ((segment_count / segs_per_sec) < total_sections) {
2276                 f2fs_msg(sb, KERN_INFO,
2277                         "Small segment_count (%u < %u * %u)",
2278                         segment_count, segs_per_sec, total_sections);
2279                 return 1;
2280         }
2281
2282         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2283                 f2fs_msg(sb, KERN_INFO,
2284                         "Wrong segment_count / block_count (%u > %llu)",
2285                         segment_count, le64_to_cpu(raw_super->block_count));
2286                 return 1;
2287         }
2288
2289         if (secs_per_zone > total_sections || !secs_per_zone) {
2290                 f2fs_msg(sb, KERN_INFO,
2291                         "Wrong secs_per_zone / total_sections (%u, %u)",
2292                         secs_per_zone, total_sections);
2293                 return 1;
2294         }
2295         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2296                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2297                         (le32_to_cpu(raw_super->extension_count) +
2298                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2299                 f2fs_msg(sb, KERN_INFO,
2300                         "Corrupted extension count (%u + %u > %u)",
2301                         le32_to_cpu(raw_super->extension_count),
2302                         raw_super->hot_ext_count,
2303                         F2FS_MAX_EXTENSION);
2304                 return 1;
2305         }
2306
2307         if (le32_to_cpu(raw_super->cp_payload) >
2308                                 (blocks_per_seg - F2FS_CP_PACKS)) {
2309                 f2fs_msg(sb, KERN_INFO,
2310                         "Insane cp_payload (%u > %u)",
2311                         le32_to_cpu(raw_super->cp_payload),
2312                         blocks_per_seg - F2FS_CP_PACKS);
2313                 return 1;
2314         }
2315
2316         /* check reserved ino info */
2317         if (le32_to_cpu(raw_super->node_ino) != 1 ||
2318                 le32_to_cpu(raw_super->meta_ino) != 2 ||
2319                 le32_to_cpu(raw_super->root_ino) != 3) {
2320                 f2fs_msg(sb, KERN_INFO,
2321                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2322                         le32_to_cpu(raw_super->node_ino),
2323                         le32_to_cpu(raw_super->meta_ino),
2324                         le32_to_cpu(raw_super->root_ino));
2325                 return 1;
2326         }
2327
2328         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2329         if (sanity_check_area_boundary(sbi, bh))
2330                 return 1;
2331
2332         return 0;
2333 }
2334
2335 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2336 {
2337         unsigned int total, fsmeta;
2338         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2339         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2340         unsigned int ovp_segments, reserved_segments;
2341         unsigned int main_segs, blocks_per_seg;
2342         unsigned int sit_segs, nat_segs;
2343         unsigned int sit_bitmap_size, nat_bitmap_size;
2344         unsigned int log_blocks_per_seg;
2345         unsigned int segment_count_main;
2346         unsigned int cp_pack_start_sum, cp_payload;
2347         block_t user_block_count;
2348         int i, j;
2349
2350         total = le32_to_cpu(raw_super->segment_count);
2351         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2352         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2353         fsmeta += sit_segs;
2354         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2355         fsmeta += nat_segs;
2356         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2357         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2358
2359         if (unlikely(fsmeta >= total))
2360                 return 1;
2361
2362         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2363         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2364
2365         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2366                         ovp_segments == 0 || reserved_segments == 0)) {
2367                 f2fs_msg(sbi->sb, KERN_ERR,
2368                         "Wrong layout: check mkfs.f2fs version");
2369                 return 1;
2370         }
2371
2372         user_block_count = le64_to_cpu(ckpt->user_block_count);
2373         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2374         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2375         if (!user_block_count || user_block_count >=
2376                         segment_count_main << log_blocks_per_seg) {
2377                 f2fs_msg(sbi->sb, KERN_ERR,
2378                         "Wrong user_block_count: %u", user_block_count);
2379                 return 1;
2380         }
2381
2382         main_segs = le32_to_cpu(raw_super->segment_count_main);
2383         blocks_per_seg = sbi->blocks_per_seg;
2384
2385         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2386                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2387                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2388                         return 1;
2389                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2390                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2391                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2392                                 f2fs_msg(sbi->sb, KERN_ERR,
2393                                         "Node segment (%u, %u) has the same "
2394                                         "segno: %u", i, j,
2395                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2396                                 return 1;
2397                         }
2398                 }
2399         }
2400         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2401                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2402                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2403                         return 1;
2404                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2405                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2406                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2407                                 f2fs_msg(sbi->sb, KERN_ERR,
2408                                         "Data segment (%u, %u) has the same "
2409                                         "segno: %u", i, j,
2410                                         le32_to_cpu(ckpt->cur_data_segno[i]));
2411                                 return 1;
2412                         }
2413                 }
2414         }
2415         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2416                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2417                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2418                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2419                                 f2fs_msg(sbi->sb, KERN_ERR,
2420                                         "Node segment (%u) and Data segment (%u)"
2421                                         " has the same segno: %u", i, j,
2422                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2423                                 return 1;
2424                         }
2425                 }
2426         }
2427
2428         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2429         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2430
2431         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2432                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2433                 f2fs_msg(sbi->sb, KERN_ERR,
2434                         "Wrong bitmap size: sit: %u, nat:%u",
2435                         sit_bitmap_size, nat_bitmap_size);
2436                 return 1;
2437         }
2438
2439         cp_pack_start_sum = __start_sum_addr(sbi);
2440         cp_payload = __cp_payload(sbi);
2441         if (cp_pack_start_sum < cp_payload + 1 ||
2442                 cp_pack_start_sum > blocks_per_seg - 1 -
2443                         NR_CURSEG_TYPE) {
2444                 f2fs_msg(sbi->sb, KERN_ERR,
2445                         "Wrong cp_pack_start_sum: %u",
2446                         cp_pack_start_sum);
2447                 return 1;
2448         }
2449
2450         if (unlikely(f2fs_cp_error(sbi))) {
2451                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2452                 return 1;
2453         }
2454         return 0;
2455 }
2456
2457 static void init_sb_info(struct f2fs_sb_info *sbi)
2458 {
2459         struct f2fs_super_block *raw_super = sbi->raw_super;
2460         int i, j;
2461
2462         sbi->log_sectors_per_block =
2463                 le32_to_cpu(raw_super->log_sectors_per_block);
2464         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2465         sbi->blocksize = 1 << sbi->log_blocksize;
2466         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2467         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2468         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2469         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2470         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2471         sbi->total_node_count =
2472                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2473                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2474         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2475         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2476         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2477         sbi->cur_victim_sec = NULL_SECNO;
2478         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2479
2480         sbi->dir_level = DEF_DIR_LEVEL;
2481         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2482         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2483         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2484
2485         for (i = 0; i < NR_COUNT_TYPE; i++)
2486                 atomic_set(&sbi->nr_pages[i], 0);
2487
2488         for (i = 0; i < META; i++)
2489                 atomic_set(&sbi->wb_sync_req[i], 0);
2490
2491         INIT_LIST_HEAD(&sbi->s_list);
2492         mutex_init(&sbi->umount_mutex);
2493         for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2494                 for (j = HOT; j < NR_TEMP_TYPE; j++)
2495                         mutex_init(&sbi->wio_mutex[i][j]);
2496         init_rwsem(&sbi->io_order_lock);
2497         spin_lock_init(&sbi->cp_lock);
2498
2499         sbi->dirty_device = 0;
2500         spin_lock_init(&sbi->dev_lock);
2501
2502         init_rwsem(&sbi->sb_lock);
2503 }
2504
2505 static int init_percpu_info(struct f2fs_sb_info *sbi)
2506 {
2507         int err;
2508
2509         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2510         if (err)
2511                 return err;
2512
2513         return percpu_counter_init(&sbi->total_valid_inode_count, 0,
2514                                                                 GFP_KERNEL);
2515 }
2516
2517 #ifdef CONFIG_BLK_DEV_ZONED
2518 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2519 {
2520         struct block_device *bdev = FDEV(devi).bdev;
2521         sector_t nr_sectors = bdev->bd_part->nr_sects;
2522         sector_t sector = 0;
2523         struct blk_zone *zones;
2524         unsigned int i, nr_zones;
2525         unsigned int n = 0;
2526         int err = -EIO;
2527
2528         if (!f2fs_sb_has_blkzoned(sbi->sb))
2529                 return 0;
2530
2531         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2532                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2533                 return -EINVAL;
2534         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2535         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2536                                 __ilog2_u32(sbi->blocks_per_blkz))
2537                 return -EINVAL;
2538         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2539         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2540                                         sbi->log_blocks_per_blkz;
2541         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2542                 FDEV(devi).nr_blkz++;
2543
2544         FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz,
2545                                                                 GFP_KERNEL);
2546         if (!FDEV(devi).blkz_type)
2547                 return -ENOMEM;
2548
2549 #define F2FS_REPORT_NR_ZONES   4096
2550
2551         zones = f2fs_kzalloc(sbi,
2552                              array_size(F2FS_REPORT_NR_ZONES,
2553                                         sizeof(struct blk_zone)),
2554                              GFP_KERNEL);
2555         if (!zones)
2556                 return -ENOMEM;
2557
2558         /* Get block zones type */
2559         while (zones && sector < nr_sectors) {
2560
2561                 nr_zones = F2FS_REPORT_NR_ZONES;
2562                 err = blkdev_report_zones(bdev, sector,
2563                                           zones, &nr_zones,
2564                                           GFP_KERNEL);
2565                 if (err)
2566                         break;
2567                 if (!nr_zones) {
2568                         err = -EIO;
2569                         break;
2570                 }
2571
2572                 for (i = 0; i < nr_zones; i++) {
2573                         FDEV(devi).blkz_type[n] = zones[i].type;
2574                         sector += zones[i].len;
2575                         n++;
2576                 }
2577         }
2578
2579         kfree(zones);
2580
2581         return err;
2582 }
2583 #endif
2584
2585 /*
2586  * Read f2fs raw super block.
2587  * Because we have two copies of super block, so read both of them
2588  * to get the first valid one. If any one of them is broken, we pass
2589  * them recovery flag back to the caller.
2590  */
2591 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2592                         struct f2fs_super_block **raw_super,
2593                         int *valid_super_block, int *recovery)
2594 {
2595         struct super_block *sb = sbi->sb;
2596         int block;
2597         struct buffer_head *bh;
2598         struct f2fs_super_block *super;
2599         int err = 0;
2600
2601         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2602         if (!super)
2603                 return -ENOMEM;
2604
2605         for (block = 0; block < 2; block++) {
2606                 bh = sb_bread(sb, block);
2607                 if (!bh) {
2608                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2609                                 block + 1);
2610                         err = -EIO;
2611                         continue;
2612                 }
2613
2614                 /* sanity checking of raw super */
2615                 if (sanity_check_raw_super(sbi, bh)) {
2616                         f2fs_msg(sb, KERN_ERR,
2617                                 "Can't find valid F2FS filesystem in %dth superblock",
2618                                 block + 1);
2619                         err = -EFSCORRUPTED;
2620                         brelse(bh);
2621                         continue;
2622                 }
2623
2624                 if (!*raw_super) {
2625                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2626                                                         sizeof(*super));
2627                         *valid_super_block = block;
2628                         *raw_super = super;
2629                 }
2630                 brelse(bh);
2631         }
2632
2633         /* Fail to read any one of the superblocks*/
2634         if (err < 0)
2635                 *recovery = 1;
2636
2637         /* No valid superblock */
2638         if (!*raw_super)
2639                 kfree(super);
2640         else
2641                 err = 0;
2642
2643         return err;
2644 }
2645
2646 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2647 {
2648         struct buffer_head *bh;
2649         int err;
2650
2651         if ((recover && f2fs_readonly(sbi->sb)) ||
2652                                 bdev_read_only(sbi->sb->s_bdev)) {
2653                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2654                 return -EROFS;
2655         }
2656
2657         /* write back-up superblock first */
2658         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
2659         if (!bh)
2660                 return -EIO;
2661         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2662         brelse(bh);
2663
2664         /* if we are in recovery path, skip writing valid superblock */
2665         if (recover || err)
2666                 return err;
2667
2668         /* write current valid superblock */
2669         bh = sb_bread(sbi->sb, sbi->valid_super_block);
2670         if (!bh)
2671                 return -EIO;
2672         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2673         brelse(bh);
2674         return err;
2675 }
2676
2677 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2678 {
2679         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2680         unsigned int max_devices = MAX_DEVICES;
2681         int i;
2682
2683         /* Initialize single device information */
2684         if (!RDEV(0).path[0]) {
2685                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2686                         return 0;
2687                 max_devices = 1;
2688         }
2689
2690         /*
2691          * Initialize multiple devices information, or single
2692          * zoned block device information.
2693          */
2694         sbi->devs = f2fs_kzalloc(sbi,
2695                                  array_size(max_devices,
2696                                             sizeof(struct f2fs_dev_info)),
2697                                  GFP_KERNEL);
2698         if (!sbi->devs)
2699                 return -ENOMEM;
2700
2701         for (i = 0; i < max_devices; i++) {
2702
2703                 if (i > 0 && !RDEV(i).path[0])
2704                         break;
2705
2706                 if (max_devices == 1) {
2707                         /* Single zoned block device mount */
2708                         FDEV(0).bdev =
2709                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2710                                         sbi->sb->s_mode, sbi->sb->s_type);
2711                 } else {
2712                         /* Multi-device mount */
2713                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2714                         FDEV(i).total_segments =
2715                                 le32_to_cpu(RDEV(i).total_segments);
2716                         if (i == 0) {
2717                                 FDEV(i).start_blk = 0;
2718                                 FDEV(i).end_blk = FDEV(i).start_blk +
2719                                     (FDEV(i).total_segments <<
2720                                     sbi->log_blocks_per_seg) - 1 +
2721                                     le32_to_cpu(raw_super->segment0_blkaddr);
2722                         } else {
2723                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2724                                 FDEV(i).end_blk = FDEV(i).start_blk +
2725                                         (FDEV(i).total_segments <<
2726                                         sbi->log_blocks_per_seg) - 1;
2727                         }
2728                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2729                                         sbi->sb->s_mode, sbi->sb->s_type);
2730                 }
2731                 if (IS_ERR(FDEV(i).bdev))
2732                         return PTR_ERR(FDEV(i).bdev);
2733
2734                 /* to release errored devices */
2735                 sbi->s_ndevs = i + 1;
2736
2737 #ifdef CONFIG_BLK_DEV_ZONED
2738                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2739                                 !f2fs_sb_has_blkzoned(sbi->sb)) {
2740                         f2fs_msg(sbi->sb, KERN_ERR,
2741                                 "Zoned block device feature not enabled\n");
2742                         return -EINVAL;
2743                 }
2744                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2745                         if (init_blkz_info(sbi, i)) {
2746                                 f2fs_msg(sbi->sb, KERN_ERR,
2747                                         "Failed to initialize F2FS blkzone information");
2748                                 return -EINVAL;
2749                         }
2750                         if (max_devices == 1)
2751                                 break;
2752                         f2fs_msg(sbi->sb, KERN_INFO,
2753                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2754                                 i, FDEV(i).path,
2755                                 FDEV(i).total_segments,
2756                                 FDEV(i).start_blk, FDEV(i).end_blk,
2757                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2758                                 "Host-aware" : "Host-managed");
2759                         continue;
2760                 }
2761 #endif
2762                 f2fs_msg(sbi->sb, KERN_INFO,
2763                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2764                                 i, FDEV(i).path,
2765                                 FDEV(i).total_segments,
2766                                 FDEV(i).start_blk, FDEV(i).end_blk);
2767         }
2768         f2fs_msg(sbi->sb, KERN_INFO,
2769                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2770         return 0;
2771 }
2772
2773 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
2774 {
2775         struct f2fs_sm_info *sm_i = SM_I(sbi);
2776
2777         /* adjust parameters according to the volume size */
2778         if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
2779                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2780                 sm_i->dcc_info->discard_granularity = 1;
2781                 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
2782         }
2783
2784         sbi->readdir_ra = 1;
2785 }
2786
2787 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2788 {
2789         struct f2fs_sb_info *sbi;
2790         struct f2fs_super_block *raw_super;
2791         struct inode *root;
2792         int err;
2793         bool retry = true, need_fsck = false;
2794         char *options = NULL;
2795         int recovery, i, valid_super_block;
2796         struct curseg_info *seg_i;
2797
2798 try_onemore:
2799         err = -EINVAL;
2800         raw_super = NULL;
2801         valid_super_block = -1;
2802         recovery = 0;
2803
2804         /* allocate memory for f2fs-specific super block info */
2805         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2806         if (!sbi)
2807                 return -ENOMEM;
2808
2809         sbi->sb = sb;
2810
2811         /* Load the checksum driver */
2812         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2813         if (IS_ERR(sbi->s_chksum_driver)) {
2814                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2815                 err = PTR_ERR(sbi->s_chksum_driver);
2816                 sbi->s_chksum_driver = NULL;
2817                 goto free_sbi;
2818         }
2819
2820         /* set a block size */
2821         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2822                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2823                 goto free_sbi;
2824         }
2825
2826         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2827                                                                 &recovery);
2828         if (err)
2829                 goto free_sbi;
2830
2831         sb->s_fs_info = sbi;
2832         sbi->raw_super = raw_super;
2833
2834         /* precompute checksum seed for metadata */
2835         if (f2fs_sb_has_inode_chksum(sb))
2836                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2837                                                 sizeof(raw_super->uuid));
2838
2839         /*
2840          * The BLKZONED feature indicates that the drive was formatted with
2841          * zone alignment optimization. This is optional for host-aware
2842          * devices, but mandatory for host-managed zoned block devices.
2843          */
2844 #ifndef CONFIG_BLK_DEV_ZONED
2845         if (f2fs_sb_has_blkzoned(sb)) {
2846                 f2fs_msg(sb, KERN_ERR,
2847                          "Zoned block device support is not enabled\n");
2848                 err = -EOPNOTSUPP;
2849                 goto free_sb_buf;
2850         }
2851 #endif
2852         default_options(sbi);
2853         /* parse mount options */
2854         options = kstrdup((const char *)data, GFP_KERNEL);
2855         if (data && !options) {
2856                 err = -ENOMEM;
2857                 goto free_sb_buf;
2858         }
2859
2860         err = parse_options(sb, options);
2861         if (err)
2862                 goto free_options;
2863
2864         sbi->max_file_blocks = max_file_blocks();
2865         sb->s_maxbytes = sbi->max_file_blocks <<
2866                                 le32_to_cpu(raw_super->log_blocksize);
2867         sb->s_max_links = F2FS_LINK_MAX;
2868         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2869
2870 #ifdef CONFIG_QUOTA
2871         sb->dq_op = &f2fs_quota_operations;
2872         if (f2fs_sb_has_quota_ino(sb))
2873                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2874         else
2875                 sb->s_qcop = &f2fs_quotactl_ops;
2876         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2877
2878         if (f2fs_sb_has_quota_ino(sbi->sb)) {
2879                 for (i = 0; i < MAXQUOTAS; i++) {
2880                         if (f2fs_qf_ino(sbi->sb, i))
2881                                 sbi->nquota_files++;
2882                 }
2883         }
2884 #endif
2885
2886         sb->s_op = &f2fs_sops;
2887 #ifdef CONFIG_F2FS_FS_ENCRYPTION
2888         sb->s_cop = &f2fs_cryptops;
2889 #endif
2890         sb->s_xattr = f2fs_xattr_handlers;
2891         sb->s_export_op = &f2fs_export_ops;
2892         sb->s_magic = F2FS_SUPER_MAGIC;
2893         sb->s_time_gran = 1;
2894         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2895                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2896         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2897         sb->s_iflags |= SB_I_CGROUPWB;
2898
2899         /* init f2fs-specific super block info */
2900         sbi->valid_super_block = valid_super_block;
2901         mutex_init(&sbi->gc_mutex);
2902         mutex_init(&sbi->writepages);
2903         mutex_init(&sbi->cp_mutex);
2904         init_rwsem(&sbi->node_write);
2905         init_rwsem(&sbi->node_change);
2906
2907         /* disallow all the data/node/meta page writes */
2908         set_sbi_flag(sbi, SBI_POR_DOING);
2909         spin_lock_init(&sbi->stat_lock);
2910
2911         /* init iostat info */
2912         spin_lock_init(&sbi->iostat_lock);
2913         sbi->iostat_enable = false;
2914
2915         for (i = 0; i < NR_PAGE_TYPE; i++) {
2916                 int n = (i == META) ? 1: NR_TEMP_TYPE;
2917                 int j;
2918
2919                 sbi->write_io[i] =
2920                         f2fs_kmalloc(sbi,
2921                                      array_size(n,
2922                                                 sizeof(struct f2fs_bio_info)),
2923                                      GFP_KERNEL);
2924                 if (!sbi->write_io[i]) {
2925                         err = -ENOMEM;
2926                         goto free_options;
2927                 }
2928
2929                 for (j = HOT; j < n; j++) {
2930                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
2931                         sbi->write_io[i][j].sbi = sbi;
2932                         sbi->write_io[i][j].bio = NULL;
2933                         spin_lock_init(&sbi->write_io[i][j].io_lock);
2934                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2935                 }
2936         }
2937
2938         init_rwsem(&sbi->cp_rwsem);
2939         init_waitqueue_head(&sbi->cp_wait);
2940         init_sb_info(sbi);
2941
2942         err = init_percpu_info(sbi);
2943         if (err)
2944                 goto free_bio_info;
2945
2946         if (F2FS_IO_SIZE(sbi) > 1) {
2947                 sbi->write_io_dummy =
2948                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2949                 if (!sbi->write_io_dummy) {
2950                         err = -ENOMEM;
2951                         goto free_percpu;
2952                 }
2953         }
2954
2955         /* get an inode for meta space */
2956         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2957         if (IS_ERR(sbi->meta_inode)) {
2958                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2959                 err = PTR_ERR(sbi->meta_inode);
2960                 goto free_io_dummy;
2961         }
2962
2963         err = f2fs_get_valid_checkpoint(sbi);
2964         if (err) {
2965                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2966                 goto free_meta_inode;
2967         }
2968
2969         /* Initialize device list */
2970         err = f2fs_scan_devices(sbi);
2971         if (err) {
2972                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2973                 goto free_devices;
2974         }
2975
2976         sbi->total_valid_node_count =
2977                                 le32_to_cpu(sbi->ckpt->valid_node_count);
2978         percpu_counter_set(&sbi->total_valid_inode_count,
2979                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
2980         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2981         sbi->total_valid_block_count =
2982                                 le64_to_cpu(sbi->ckpt->valid_block_count);
2983         sbi->last_valid_block_count = sbi->total_valid_block_count;
2984         sbi->reserved_blocks = 0;
2985         sbi->current_reserved_blocks = 0;
2986         limit_reserve_root(sbi);
2987
2988         for (i = 0; i < NR_INODE_TYPE; i++) {
2989                 INIT_LIST_HEAD(&sbi->inode_list[i]);
2990                 spin_lock_init(&sbi->inode_lock[i]);
2991         }
2992
2993         f2fs_init_extent_cache_info(sbi);
2994
2995         f2fs_init_ino_entry_info(sbi);
2996
2997         f2fs_init_fsync_node_info(sbi);
2998
2999         /* setup f2fs internal modules */
3000         err = f2fs_build_segment_manager(sbi);
3001         if (err) {
3002                 f2fs_msg(sb, KERN_ERR,
3003                         "Failed to initialize F2FS segment manager");
3004                 goto free_sm;
3005         }
3006         err = f2fs_build_node_manager(sbi);
3007         if (err) {
3008                 f2fs_msg(sb, KERN_ERR,
3009                         "Failed to initialize F2FS node manager");
3010                 goto free_nm;
3011         }
3012
3013         /* For write statistics */
3014         if (sb->s_bdev->bd_part)
3015                 sbi->sectors_written_start =
3016                         (u64)part_stat_read(sb->s_bdev->bd_part,
3017                                             sectors[STAT_WRITE]);
3018
3019         /* Read accumulated write IO statistics if exists */
3020         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3021         if (__exist_node_summaries(sbi))
3022                 sbi->kbytes_written =
3023                         le64_to_cpu(seg_i->journal->info.kbytes_written);
3024
3025         f2fs_build_gc_manager(sbi);
3026
3027         err = f2fs_build_stats(sbi);
3028         if (err)
3029                 goto free_nm;
3030
3031         /* get an inode for node space */
3032         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3033         if (IS_ERR(sbi->node_inode)) {
3034                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
3035                 err = PTR_ERR(sbi->node_inode);
3036                 goto free_stats;
3037         }
3038
3039         /* read root inode and dentry */
3040         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3041         if (IS_ERR(root)) {
3042                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
3043                 err = PTR_ERR(root);
3044                 goto free_node_inode;
3045         }
3046         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3047                         !root->i_size || !root->i_nlink) {
3048                 iput(root);
3049                 err = -EINVAL;
3050                 goto free_node_inode;
3051         }
3052
3053         sb->s_root = d_make_root(root); /* allocate root dentry */
3054         if (!sb->s_root) {
3055                 err = -ENOMEM;
3056                 goto free_root_inode;
3057         }
3058
3059         err = f2fs_register_sysfs(sbi);
3060         if (err)
3061                 goto free_root_inode;
3062
3063 #ifdef CONFIG_QUOTA
3064         /* Enable quota usage during mount */
3065         if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb)) {
3066                 err = f2fs_enable_quotas(sb);
3067                 if (err) {
3068                         f2fs_msg(sb, KERN_ERR,
3069                                 "Cannot turn on quotas: error %d", err);
3070                         goto free_sysfs;
3071                 }
3072         }
3073 #endif
3074         /* if there are nt orphan nodes free them */
3075         err = f2fs_recover_orphan_inodes(sbi);
3076         if (err)
3077                 goto free_meta;
3078
3079         /* recover fsynced data */
3080         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
3081                 /*
3082                  * mount should be failed, when device has readonly mode, and
3083                  * previous checkpoint was not done by clean system shutdown.
3084                  */
3085                 if (bdev_read_only(sb->s_bdev) &&
3086                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3087                         err = -EROFS;
3088                         goto free_meta;
3089                 }
3090
3091                 if (need_fsck)
3092                         set_sbi_flag(sbi, SBI_NEED_FSCK);
3093
3094                 if (!retry)
3095                         goto skip_recovery;
3096
3097                 err = f2fs_recover_fsync_data(sbi, false);
3098                 if (err < 0) {
3099                         need_fsck = true;
3100                         f2fs_msg(sb, KERN_ERR,
3101                                 "Cannot recover all fsync data errno=%d", err);
3102                         goto free_meta;
3103                 }
3104         } else {
3105                 err = f2fs_recover_fsync_data(sbi, true);
3106
3107                 if (!f2fs_readonly(sb) && err > 0) {
3108                         err = -EINVAL;
3109                         f2fs_msg(sb, KERN_ERR,
3110                                 "Need to recover fsync data");
3111                         goto free_meta;
3112                 }
3113         }
3114 skip_recovery:
3115         /* f2fs_recover_fsync_data() cleared this already */
3116         clear_sbi_flag(sbi, SBI_POR_DOING);
3117
3118         /*
3119          * If filesystem is not mounted as read-only then
3120          * do start the gc_thread.
3121          */
3122         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
3123                 /* After POR, we can run background GC thread.*/
3124                 err = f2fs_start_gc_thread(sbi);
3125                 if (err)
3126                         goto free_meta;
3127         }
3128         kfree(options);
3129
3130         /* recover broken superblock */
3131         if (recovery) {
3132                 err = f2fs_commit_super(sbi, true);
3133                 f2fs_msg(sb, KERN_INFO,
3134                         "Try to recover %dth superblock, ret: %d",
3135                         sbi->valid_super_block ? 1 : 2, err);
3136         }
3137
3138         f2fs_join_shrinker(sbi);
3139
3140         f2fs_tuning_parameters(sbi);
3141
3142         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
3143                                 cur_cp_version(F2FS_CKPT(sbi)));
3144         f2fs_update_time(sbi, CP_TIME);
3145         f2fs_update_time(sbi, REQ_TIME);
3146         return 0;
3147
3148 free_meta:
3149 #ifdef CONFIG_QUOTA
3150         f2fs_truncate_quota_inode_pages(sb);
3151         if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb))
3152                 f2fs_quota_off_umount(sbi->sb);
3153 #endif
3154         /*
3155          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3156          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3157          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3158          * falls into an infinite loop in f2fs_sync_meta_pages().
3159          */
3160         truncate_inode_pages_final(META_MAPPING(sbi));
3161 #ifdef CONFIG_QUOTA
3162 free_sysfs:
3163 #endif
3164         f2fs_unregister_sysfs(sbi);
3165 free_root_inode:
3166         dput(sb->s_root);
3167         sb->s_root = NULL;
3168 free_node_inode:
3169         f2fs_release_ino_entry(sbi, true);
3170         truncate_inode_pages_final(NODE_MAPPING(sbi));
3171         iput(sbi->node_inode);
3172         sbi->node_inode = NULL;
3173 free_stats:
3174         f2fs_destroy_stats(sbi);
3175 free_nm:
3176         f2fs_destroy_node_manager(sbi);
3177 free_sm:
3178         f2fs_destroy_segment_manager(sbi);
3179 free_devices:
3180         destroy_device_list(sbi);
3181         kfree(sbi->ckpt);
3182 free_meta_inode:
3183         make_bad_inode(sbi->meta_inode);
3184         iput(sbi->meta_inode);
3185         sbi->meta_inode = NULL;
3186 free_io_dummy:
3187         mempool_destroy(sbi->write_io_dummy);
3188 free_percpu:
3189         destroy_percpu_info(sbi);
3190 free_bio_info:
3191         for (i = 0; i < NR_PAGE_TYPE; i++)
3192                 kfree(sbi->write_io[i]);
3193 free_options:
3194 #ifdef CONFIG_QUOTA
3195         for (i = 0; i < MAXQUOTAS; i++)
3196                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
3197 #endif
3198         kfree(options);
3199 free_sb_buf:
3200         kfree(raw_super);
3201 free_sbi:
3202         if (sbi->s_chksum_driver)
3203                 crypto_free_shash(sbi->s_chksum_driver);
3204         kfree(sbi);
3205
3206         /* give only one another chance */
3207         if (retry) {
3208                 retry = false;
3209                 shrink_dcache_sb(sb);
3210                 goto try_onemore;
3211         }
3212         return err;
3213 }
3214
3215 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3216                         const char *dev_name, void *data)
3217 {
3218         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3219 }
3220
3221 static void kill_f2fs_super(struct super_block *sb)
3222 {
3223         if (sb->s_root) {
3224                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
3225
3226                 set_sbi_flag(sbi, SBI_IS_CLOSE);
3227                 f2fs_stop_gc_thread(sbi);
3228                 f2fs_stop_discard_thread(sbi);
3229
3230                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3231                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3232                         struct cp_control cpc = {
3233                                 .reason = CP_UMOUNT,
3234                         };
3235                         f2fs_write_checkpoint(sbi, &cpc);
3236                 }
3237
3238                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3239                         sb->s_flags &= ~SB_RDONLY;
3240         }
3241         kill_block_super(sb);
3242 }
3243
3244 static struct file_system_type f2fs_fs_type = {
3245         .owner          = THIS_MODULE,
3246         .name           = "f2fs",
3247         .mount          = f2fs_mount,
3248         .kill_sb        = kill_f2fs_super,
3249         .fs_flags       = FS_REQUIRES_DEV,
3250 };
3251 MODULE_ALIAS_FS("f2fs");
3252
3253 static int __init init_inodecache(void)
3254 {
3255         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3256                         sizeof(struct f2fs_inode_info), 0,
3257                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3258         if (!f2fs_inode_cachep)
3259                 return -ENOMEM;
3260         return 0;
3261 }
3262
3263 static void destroy_inodecache(void)
3264 {
3265         /*
3266          * Make sure all delayed rcu free inodes are flushed before we
3267          * destroy cache.
3268          */
3269         rcu_barrier();
3270         kmem_cache_destroy(f2fs_inode_cachep);
3271 }
3272
3273 static int __init init_f2fs_fs(void)
3274 {
3275         int err;
3276
3277         if (PAGE_SIZE != F2FS_BLKSIZE) {
3278                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3279                                 PAGE_SIZE, F2FS_BLKSIZE);
3280                 return -EINVAL;
3281         }
3282
3283         f2fs_build_trace_ios();
3284
3285         err = init_inodecache();
3286         if (err)
3287                 goto fail;
3288         err = f2fs_create_node_manager_caches();
3289         if (err)
3290                 goto free_inodecache;
3291         err = f2fs_create_segment_manager_caches();
3292         if (err)
3293                 goto free_node_manager_caches;
3294         err = f2fs_create_checkpoint_caches();
3295         if (err)
3296                 goto free_segment_manager_caches;
3297         err = f2fs_create_extent_cache();
3298         if (err)
3299                 goto free_checkpoint_caches;
3300         err = f2fs_init_sysfs();
3301         if (err)
3302                 goto free_extent_cache;
3303         err = register_shrinker(&f2fs_shrinker_info);
3304         if (err)
3305                 goto free_sysfs;
3306         err = register_filesystem(&f2fs_fs_type);
3307         if (err)
3308                 goto free_shrinker;
3309         err = f2fs_create_root_stats();
3310         if (err)
3311                 goto free_filesystem;
3312         err = f2fs_init_post_read_processing();
3313         if (err)
3314                 goto free_root_stats;
3315         return 0;
3316
3317 free_root_stats:
3318         f2fs_destroy_root_stats();
3319 free_filesystem:
3320         unregister_filesystem(&f2fs_fs_type);
3321 free_shrinker:
3322         unregister_shrinker(&f2fs_shrinker_info);
3323 free_sysfs:
3324         f2fs_exit_sysfs();
3325 free_extent_cache:
3326         f2fs_destroy_extent_cache();
3327 free_checkpoint_caches:
3328         f2fs_destroy_checkpoint_caches();
3329 free_segment_manager_caches:
3330         f2fs_destroy_segment_manager_caches();
3331 free_node_manager_caches:
3332         f2fs_destroy_node_manager_caches();
3333 free_inodecache:
3334         destroy_inodecache();
3335 fail:
3336         return err;
3337 }
3338
3339 static void __exit exit_f2fs_fs(void)
3340 {
3341         f2fs_destroy_post_read_processing();
3342         f2fs_destroy_root_stats();
3343         unregister_filesystem(&f2fs_fs_type);
3344         unregister_shrinker(&f2fs_shrinker_info);
3345         f2fs_exit_sysfs();
3346         f2fs_destroy_extent_cache();
3347         f2fs_destroy_checkpoint_caches();
3348         f2fs_destroy_segment_manager_caches();
3349         f2fs_destroy_node_manager_caches();
3350         destroy_inodecache();
3351         f2fs_destroy_trace_ios();
3352 }
3353
3354 module_init(init_f2fs_fs)
3355 module_exit(exit_f2fs_fs)
3356
3357 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3358 MODULE_DESCRIPTION("Flash Friendly File System");
3359 MODULE_LICENSE("GPL");
3360