fs/f2fs/super.c

   1 /*
   2  * fs/f2fs/super.c
   3  *
   4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   5  *             http://www.samsung.com/
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  */
  11 #include <linux/module.h>
  12 #include <linux/init.h>
  13 #include <linux/fs.h>
  14 #include <linux/statfs.h>
  15 #include <linux/buffer_head.h>
  16 #include <linux/backing-dev.h>
  17 #include <linux/kthread.h>
  18 #include <linux/parser.h>
  19 #include <linux/mount.h>
  20 #include <linux/seq_file.h>
  21 #include <linux/proc_fs.h>
  22 #include <linux/random.h>
  23 #include <linux/exportfs.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/quotaops.h>
  26 #include <linux/f2fs_fs.h>
  27 #include <linux/sysfs.h>
  28 #include <linux/quota.h>
  29
  30 #include "f2fs.h"
  31 #include "node.h"
  32 #include "segment.h"
  33 #include "xattr.h"
  34 #include "gc.h"
  35 #include "trace.h"
  36
  37 #define CREATE_TRACE_POINTS
  38 #include <trace/events/f2fs.h>
  39
  40 static struct kmem_cache *f2fs_inode_cachep;
  41
  42 #ifdef CONFIG_F2FS_FAULT_INJECTION
  43
  44 char *fault_name[FAULT_MAX] = {
  45         [FAULT_KMALLOC]         = "kmalloc",
  46         [FAULT_PAGE_ALLOC]      = "page alloc",
  47         [FAULT_ALLOC_NID]       = "alloc nid",
  48         [FAULT_ORPHAN]          = "orphan",
  49         [FAULT_BLOCK]           = "no more block",
  50         [FAULT_DIR_DEPTH]       = "too big dir depth",
  51         [FAULT_EVICT_INODE]     = "evict_inode fail",
  52         [FAULT_TRUNCATE]        = "truncate fail",
  53         [FAULT_IO]              = "IO error",
  54         [FAULT_CHECKPOINT]      = "checkpoint error",
  55 };
  56
  57 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi,
  58                                                 unsigned int rate)
  59 {
  60         struct f2fs_fault_info *ffi = &sbi->fault_info;
  61
  62         if (rate) {
  63                 atomic_set(&ffi->inject_ops, 0);
  64                 ffi->inject_rate = rate;
  65                 ffi->inject_type = (1 << FAULT_MAX) - 1;
  66         } else {
  67                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  68         }
  69 }
  70 #endif
  71
  72 /* f2fs-wide shrinker description */
  73 static struct shrinker f2fs_shrinker_info = {
  74         .scan_objects = f2fs_shrink_scan,
  75         .count_objects = f2fs_shrink_count,
  76         .seeks = DEFAULT_SEEKS,
  77 };
  78
  79 enum {
  80         Opt_gc_background,
  81         Opt_disable_roll_forward,
  82         Opt_norecovery,
  83         Opt_discard,
  84         Opt_nodiscard,
  85         Opt_noheap,
  86         Opt_heap,
  87         Opt_user_xattr,
  88         Opt_nouser_xattr,
  89         Opt_acl,
  90         Opt_noacl,
  91         Opt_active_logs,
  92         Opt_disable_ext_identify,
  93         Opt_inline_xattr,
  94         Opt_noinline_xattr,
  95         Opt_inline_data,
  96         Opt_inline_dentry,
  97         Opt_noinline_dentry,
  98         Opt_flush_merge,
  99         Opt_noflush_merge,
 100         Opt_nobarrier,
 101         Opt_fastboot,
 102         Opt_extent_cache,
 103         Opt_noextent_cache,
 104         Opt_noinline_data,
 105         Opt_data_flush,
 106         Opt_mode,
 107         Opt_io_size_bits,
 108         Opt_fault_injection,
 109         Opt_lazytime,
 110         Opt_nolazytime,
 111         Opt_quota,
 112         Opt_noquota,
 113         Opt_usrquota,
 114         Opt_grpquota,
 115         Opt_prjquota,
 116         Opt_usrjquota,
 117         Opt_grpjquota,
 118         Opt_prjjquota,
 119         Opt_offusrjquota,
 120         Opt_offgrpjquota,
 121         Opt_offprjjquota,
 122         Opt_jqfmt_vfsold,
 123         Opt_jqfmt_vfsv0,
 124         Opt_jqfmt_vfsv1,
 125         Opt_err,
 126 };
 127
 128 static match_table_t f2fs_tokens = {
 129         {Opt_gc_background, "background_gc=%s"},
 130         {Opt_disable_roll_forward, "disable_roll_forward"},
 131         {Opt_norecovery, "norecovery"},
 132         {Opt_discard, "discard"},
 133         {Opt_nodiscard, "nodiscard"},
 134         {Opt_noheap, "no_heap"},
 135         {Opt_heap, "heap"},
 136         {Opt_user_xattr, "user_xattr"},
 137         {Opt_nouser_xattr, "nouser_xattr"},
 138         {Opt_acl, "acl"},
 139         {Opt_noacl, "noacl"},
 140         {Opt_active_logs, "active_logs=%u"},
 141         {Opt_disable_ext_identify, "disable_ext_identify"},
 142         {Opt_inline_xattr, "inline_xattr"},
 143         {Opt_noinline_xattr, "noinline_xattr"},
 144         {Opt_inline_data, "inline_data"},
 145         {Opt_inline_dentry, "inline_dentry"},
 146         {Opt_noinline_dentry, "noinline_dentry"},
 147         {Opt_flush_merge, "flush_merge"},
 148         {Opt_noflush_merge, "noflush_merge"},
 149         {Opt_nobarrier, "nobarrier"},
 150         {Opt_fastboot, "fastboot"},
 151         {Opt_extent_cache, "extent_cache"},
 152         {Opt_noextent_cache, "noextent_cache"},
 153         {Opt_noinline_data, "noinline_data"},
 154         {Opt_data_flush, "data_flush"},
 155         {Opt_mode, "mode=%s"},
 156         {Opt_io_size_bits, "io_bits=%u"},
 157         {Opt_fault_injection, "fault_injection=%u"},
 158         {Opt_lazytime, "lazytime"},
 159         {Opt_nolazytime, "nolazytime"},
 160         {Opt_quota, "quota"},
 161         {Opt_noquota, "noquota"},
 162         {Opt_usrquota, "usrquota"},
 163         {Opt_grpquota, "grpquota"},
 164         {Opt_prjquota, "prjquota"},
 165         {Opt_usrjquota, "usrjquota=%s"},
 166         {Opt_grpjquota, "grpjquota=%s"},
 167         {Opt_prjjquota, "prjjquota=%s"},
 168         {Opt_offusrjquota, "usrjquota="},
 169         {Opt_offgrpjquota, "grpjquota="},
 170         {Opt_offprjjquota, "prjjquota="},
 171         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 172         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 173         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 174         {Opt_err, NULL},
 175 };
 176
 177 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
 178 {
 179         struct va_format vaf;
 180         va_list args;
 181
 182         va_start(args, fmt);
 183         vaf.fmt = fmt;
 184         vaf.va = &args;
 185         printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
 186         va_end(args);
 187 }
 188
 189 static void init_once(void *foo)
 190 {
 191         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 192
 193         inode_init_once(&fi->vfs_inode);
 194 }
 195
 196 #ifdef CONFIG_QUOTA
 197 static const char * const quotatypes[] = INITQFNAMES;
 198 #define QTYPE2NAME(t) (quotatypes[t])
 199 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 200                                                         substring_t *args)
 201 {
 202         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 203         char *qname;
 204         int ret = -EINVAL;
 205
 206         if (sb_any_quota_loaded(sb) && !sbi->s_qf_names[qtype]) {
 207                 f2fs_msg(sb, KERN_ERR,
 208                         "Cannot change journaled "
 209                         "quota options when quota turned on");
 210                 return -EINVAL;
 211         }
 212         qname = match_strdup(args);
 213         if (!qname) {
 214                 f2fs_msg(sb, KERN_ERR,
 215                         "Not enough memory for storing quotafile name");
 216                 return -EINVAL;
 217         }
 218         if (sbi->s_qf_names[qtype]) {
 219                 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
 220                         ret = 0;
 221                 else
 222                         f2fs_msg(sb, KERN_ERR,
 223                                  "%s quota file already specified",
 224                                  QTYPE2NAME(qtype));
 225                 goto errout;
 226         }
 227         if (strchr(qname, '/')) {
 228                 f2fs_msg(sb, KERN_ERR,
 229                         "quotafile must be on filesystem root");
 230                 goto errout;
 231         }
 232         sbi->s_qf_names[qtype] = qname;
 233         set_opt(sbi, QUOTA);
 234         return 0;
 235 errout:
 236         kfree(qname);
 237         return ret;
 238 }
 239
 240 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 241 {
 242         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 243
 244         if (sb_any_quota_loaded(sb) && sbi->s_qf_names[qtype]) {
 245                 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options"
 246                         " when quota turned on");
 247                 return -EINVAL;
 248         }
 249         kfree(sbi->s_qf_names[qtype]);
 250         sbi->s_qf_names[qtype] = NULL;
 251         return 0;
 252 }
 253
 254 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 255 {
 256         /*
 257          * We do the test below only for project quotas. 'usrquota' and
 258          * 'grpquota' mount options are allowed even without quota feature
 259          * to support legacy quotas in quota files.
 260          */
 261         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) {
 262                 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. "
 263                          "Cannot enable project quota enforcement.");
 264                 return -1;
 265         }
 266         if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA] ||
 267                         sbi->s_qf_names[PRJQUOTA]) {
 268                 if (test_opt(sbi, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
 269                         clear_opt(sbi, USRQUOTA);
 270
 271                 if (test_opt(sbi, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
 272                         clear_opt(sbi, GRPQUOTA);
 273
 274                 if (test_opt(sbi, PRJQUOTA) && sbi->s_qf_names[PRJQUOTA])
 275                         clear_opt(sbi, PRJQUOTA);
 276
 277                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 278                                 test_opt(sbi, PRJQUOTA)) {
 279                         f2fs_msg(sbi->sb, KERN_ERR, "old and new quota "
 280                                         "format mixing");
 281                         return -1;
 282                 }
 283
 284                 if (!sbi->s_jquota_fmt) {
 285                         f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format "
 286                                         "not specified");
 287                         return -1;
 288                 }
 289         }
 290         return 0;
 291 }
 292 #endif
 293
 294 static int parse_options(struct super_block *sb, char *options)
 295 {
 296         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 297         struct request_queue *q;
 298         substring_t args[MAX_OPT_ARGS];
 299         char *p, *name;
 300         int arg = 0;
 301 #ifdef CONFIG_QUOTA
 302         int ret;
 303 #endif
 304
 305         if (!options)
 306                 return 0;
 307
 308         while ((p = strsep(&options, ",")) != NULL) {
 309                 int token;
 310                 if (!*p)
 311                         continue;
 312                 /*
 313                  * Initialize args struct so we know whether arg was
 314                  * found; some options take optional arguments.
 315                  */
 316                 args[0].to = args[0].from = NULL;
 317                 token = match_token(p, f2fs_tokens, args);
 318
 319                 switch (token) {
 320                 case Opt_gc_background:
 321                         name = match_strdup(&args[0]);
 322
 323                         if (!name)
 324                                 return -ENOMEM;
 325                         if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 326                                 set_opt(sbi, BG_GC);
 327                                 clear_opt(sbi, FORCE_FG_GC);
 328                         } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 329                                 clear_opt(sbi, BG_GC);
 330                                 clear_opt(sbi, FORCE_FG_GC);
 331                         } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 332                                 set_opt(sbi, BG_GC);
 333                                 set_opt(sbi, FORCE_FG_GC);
 334                         } else {
 335                                 kfree(name);
 336                                 return -EINVAL;
 337                         }
 338                         kfree(name);
 339                         break;
 340                 case Opt_disable_roll_forward:
 341                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 342                         break;
 343                 case Opt_norecovery:
 344                         /* this option mounts f2fs with ro */
 345                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 346                         if (!f2fs_readonly(sb))
 347                                 return -EINVAL;
 348                         break;
 349                 case Opt_discard:
 350                         q = bdev_get_queue(sb->s_bdev);
 351                         if (blk_queue_discard(q)) {
 352                                 set_opt(sbi, DISCARD);
 353                         } else if (!f2fs_sb_mounted_blkzoned(sb)) {
 354                                 f2fs_msg(sb, KERN_WARNING,
 355                                         "mounting with \"discard\" option, but "
 356                                         "the device does not support discard");
 357                         }
 358                         break;
 359                 case Opt_nodiscard:
 360                         if (f2fs_sb_mounted_blkzoned(sb)) {
 361                                 f2fs_msg(sb, KERN_WARNING,
 362                                         "discard is required for zoned block devices");
 363                                 return -EINVAL;
 364                         }
 365                         clear_opt(sbi, DISCARD);
 366                         break;
 367                 case Opt_noheap:
 368                         set_opt(sbi, NOHEAP);
 369                         break;
 370                 case Opt_heap:
 371                         clear_opt(sbi, NOHEAP);
 372                         break;
 373 #ifdef CONFIG_F2FS_FS_XATTR
 374                 case Opt_user_xattr:
 375                         set_opt(sbi, XATTR_USER);
 376                         break;
 377                 case Opt_nouser_xattr:
 378                         clear_opt(sbi, XATTR_USER);
 379                         break;
 380                 case Opt_inline_xattr:
 381                         set_opt(sbi, INLINE_XATTR);
 382                         break;
 383                 case Opt_noinline_xattr:
 384                         clear_opt(sbi, INLINE_XATTR);
 385                         break;
 386 #else
 387                 case Opt_user_xattr:
 388                         f2fs_msg(sb, KERN_INFO,
 389                                 "user_xattr options not supported");
 390                         break;
 391                 case Opt_nouser_xattr:
 392                         f2fs_msg(sb, KERN_INFO,
 393                                 "nouser_xattr options not supported");
 394                         break;
 395                 case Opt_inline_xattr:
 396                         f2fs_msg(sb, KERN_INFO,
 397                                 "inline_xattr options not supported");
 398                         break;
 399                 case Opt_noinline_xattr:
 400                         f2fs_msg(sb, KERN_INFO,
 401                                 "noinline_xattr options not supported");
 402                         break;
 403 #endif
 404 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 405                 case Opt_acl:
 406                         set_opt(sbi, POSIX_ACL);
 407                         break;
 408                 case Opt_noacl:
 409                         clear_opt(sbi, POSIX_ACL);
 410                         break;
 411 #else
 412                 case Opt_acl:
 413                         f2fs_msg(sb, KERN_INFO, "acl options not supported");
 414                         break;
 415                 case Opt_noacl:
 416                         f2fs_msg(sb, KERN_INFO, "noacl options not supported");
 417                         break;
 418 #endif
 419                 case Opt_active_logs:
 420                         if (args->from && match_int(args, &arg))
 421                                 return -EINVAL;
 422                         if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 423                                 return -EINVAL;
 424                         sbi->active_logs = arg;
 425                         break;
 426                 case Opt_disable_ext_identify:
 427                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 428                         break;
 429                 case Opt_inline_data:
 430                         set_opt(sbi, INLINE_DATA);
 431                         break;
 432                 case Opt_inline_dentry:
 433                         set_opt(sbi, INLINE_DENTRY);
 434                         break;
 435                 case Opt_noinline_dentry:
 436                         clear_opt(sbi, INLINE_DENTRY);
 437                         break;
 438                 case Opt_flush_merge:
 439                         set_opt(sbi, FLUSH_MERGE);
 440                         break;
 441                 case Opt_noflush_merge:
 442                         clear_opt(sbi, FLUSH_MERGE);
 443                         break;
 444                 case Opt_nobarrier:
 445                         set_opt(sbi, NOBARRIER);
 446                         break;
 447                 case Opt_fastboot:
 448                         set_opt(sbi, FASTBOOT);
 449                         break;
 450                 case Opt_extent_cache:
 451                         set_opt(sbi, EXTENT_CACHE);
 452                         break;
 453                 case Opt_noextent_cache:
 454                         clear_opt(sbi, EXTENT_CACHE);
 455                         break;
 456                 case Opt_noinline_data:
 457                         clear_opt(sbi, INLINE_DATA);
 458                         break;
 459                 case Opt_data_flush:
 460                         set_opt(sbi, DATA_FLUSH);
 461                         break;
 462                 case Opt_mode:
 463                         name = match_strdup(&args[0]);
 464
 465                         if (!name)
 466                                 return -ENOMEM;
 467                         if (strlen(name) == 8 &&
 468                                         !strncmp(name, "adaptive", 8)) {
 469                                 if (f2fs_sb_mounted_blkzoned(sb)) {
 470                                         f2fs_msg(sb, KERN_WARNING,
 471                                                  "adaptive mode is not allowed with "
 472                                                  "zoned block device feature");
 473                                         kfree(name);
 474                                         return -EINVAL;
 475                                 }
 476                                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
 477                         } else if (strlen(name) == 3 &&
 478                                         !strncmp(name, "lfs", 3)) {
 479                                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
 480                         } else {
 481                                 kfree(name);
 482                                 return -EINVAL;
 483                         }
 484                         kfree(name);
 485                         break;
 486                 case Opt_io_size_bits:
 487                         if (args->from && match_int(args, &arg))
 488                                 return -EINVAL;
 489                         if (arg > __ilog2_u32(BIO_MAX_PAGES)) {
 490                                 f2fs_msg(sb, KERN_WARNING,
 491                                         "Not support %d, larger than %d",
 492                                         1 << arg, BIO_MAX_PAGES);
 493                                 return -EINVAL;
 494                         }
 495                         sbi->write_io_size_bits = arg;
 496                         break;
 497                 case Opt_fault_injection:
 498                         if (args->from && match_int(args, &arg))
 499                                 return -EINVAL;
 500 #ifdef CONFIG_F2FS_FAULT_INJECTION
 501                         f2fs_build_fault_attr(sbi, arg);
 502                         set_opt(sbi, FAULT_INJECTION);
 503 #else
 504                         f2fs_msg(sb, KERN_INFO,
 505                                 "FAULT_INJECTION was not selected");
 506 #endif
 507                         break;
 508                 case Opt_lazytime:
 509                         sb->s_flags |= MS_LAZYTIME;
 510                         break;
 511                 case Opt_nolazytime:
 512                         sb->s_flags &= ~MS_LAZYTIME;
 513                         break;
 514 #ifdef CONFIG_QUOTA
 515                 case Opt_quota:
 516                 case Opt_usrquota:
 517                         set_opt(sbi, USRQUOTA);
 518                         break;
 519                 case Opt_grpquota:
 520                         set_opt(sbi, GRPQUOTA);
 521                         break;
 522                 case Opt_prjquota:
 523                         set_opt(sbi, PRJQUOTA);
 524                         break;
 525                 case Opt_usrjquota:
 526                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 527                         if (ret)
 528                                 return ret;
 529                         break;
 530                 case Opt_grpjquota:
 531                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 532                         if (ret)
 533                                 return ret;
 534                         break;
 535                 case Opt_prjjquota:
 536                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 537                         if (ret)
 538                                 return ret;
 539                         break;
 540                 case Opt_offusrjquota:
 541                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 542                         if (ret)
 543                                 return ret;
 544                         break;
 545                 case Opt_offgrpjquota:
 546                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 547                         if (ret)
 548                                 return ret;
 549                         break;
 550                 case Opt_offprjjquota:
 551                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 552                         if (ret)
 553                                 return ret;
 554                         break;
 555                 case Opt_jqfmt_vfsold:
 556                         sbi->s_jquota_fmt = QFMT_VFS_OLD;
 557                         break;
 558                 case Opt_jqfmt_vfsv0:
 559                         sbi->s_jquota_fmt = QFMT_VFS_V0;
 560                         break;
 561                 case Opt_jqfmt_vfsv1:
 562                         sbi->s_jquota_fmt = QFMT_VFS_V1;
 563                         break;
 564                 case Opt_noquota:
 565                         clear_opt(sbi, QUOTA);
 566                         clear_opt(sbi, USRQUOTA);
 567                         clear_opt(sbi, GRPQUOTA);
 568                         clear_opt(sbi, PRJQUOTA);
 569                         break;
 570 #else
 571                 case Opt_quota:
 572                 case Opt_usrquota:
 573                 case Opt_grpquota:
 574                 case Opt_prjquota:
 575                 case Opt_usrjquota:
 576                 case Opt_grpjquota:
 577                 case Opt_prjjquota:
 578                 case Opt_offusrjquota:
 579                 case Opt_offgrpjquota:
 580                 case Opt_offprjjquota:
 581                 case Opt_jqfmt_vfsold:
 582                 case Opt_jqfmt_vfsv0:
 583                 case Opt_jqfmt_vfsv1:
 584                 case Opt_noquota:
 585                         f2fs_msg(sb, KERN_INFO,
 586                                         "quota operations not supported");
 587                         break;
 588 #endif
 589                 default:
 590                         f2fs_msg(sb, KERN_ERR,
 591                                 "Unrecognized mount option \"%s\" or missing value",
 592                                 p);
 593                         return -EINVAL;
 594                 }
 595         }
 596 #ifdef CONFIG_QUOTA
 597         if (f2fs_check_quota_options(sbi))
 598                 return -EINVAL;
 599 #endif
 600
 601         if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) {
 602                 f2fs_msg(sb, KERN_ERR,
 603                                 "Should set mode=lfs with %uKB-sized IO",
 604                                 F2FS_IO_SIZE_KB(sbi));
 605                 return -EINVAL;
 606         }
 607         return 0;
 608 }
 609
 610 static struct inode *f2fs_alloc_inode(struct super_block *sb)
 611 {
 612         struct f2fs_inode_info *fi;
 613
 614         fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 615         if (!fi)
 616                 return NULL;
 617
 618         init_once((void *) fi);
 619
 620         /* Initialize f2fs-specific inode info */
 621         fi->vfs_inode.i_version = 1;
 622         atomic_set(&fi->dirty_pages, 0);
 623         fi->i_current_depth = 1;
 624         fi->i_advise = 0;
 625         init_rwsem(&fi->i_sem);
 626         INIT_LIST_HEAD(&fi->dirty_list);
 627         INIT_LIST_HEAD(&fi->gdirty_list);
 628         INIT_LIST_HEAD(&fi->inmem_pages);
 629         mutex_init(&fi->inmem_lock);
 630         init_rwsem(&fi->dio_rwsem[READ]);
 631         init_rwsem(&fi->dio_rwsem[WRITE]);
 632         init_rwsem(&fi->i_mmap_sem);
 633         init_rwsem(&fi->i_xattr_sem);
 634
 635 #ifdef CONFIG_QUOTA
 636         memset(&fi->i_dquot, 0, sizeof(fi->i_dquot));
 637         fi->i_reserved_quota = 0;
 638 #endif
 639         /* Will be used by directory only */
 640         fi->i_dir_level = F2FS_SB(sb)->dir_level;
 641
 642         return &fi->vfs_inode;
 643 }
 644
 645 static int f2fs_drop_inode(struct inode *inode)
 646 {
 647         int ret;
 648         /*
 649          * This is to avoid a deadlock condition like below.
 650          * writeback_single_inode(inode)
 651          *  - f2fs_write_data_page
 652          *    - f2fs_gc -> iput -> evict
 653          *       - inode_wait_for_writeback(inode)
 654          */
 655         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
 656                 if (!inode->i_nlink && !is_bad_inode(inode)) {
 657                         /* to avoid evict_inode call simultaneously */
 658                         atomic_inc(&inode->i_count);
 659                         spin_unlock(&inode->i_lock);
 660
 661                         /* some remained atomic pages should discarded */
 662                         if (f2fs_is_atomic_file(inode))
 663                                 drop_inmem_pages(inode);
 664
 665                         /* should remain fi->extent_tree for writepage */
 666                         f2fs_destroy_extent_node(inode);
 667
 668                         sb_start_intwrite(inode->i_sb);
 669                         f2fs_i_size_write(inode, 0);
 670
 671                         if (F2FS_HAS_BLOCKS(inode))
 672                                 f2fs_truncate(inode);
 673
 674                         sb_end_intwrite(inode->i_sb);
 675
 676                         fscrypt_put_encryption_info(inode, NULL);
 677                         spin_lock(&inode->i_lock);
 678                         atomic_dec(&inode->i_count);
 679                 }
 680                 trace_f2fs_drop_inode(inode, 0);
 681                 return 0;
 682         }
 683         ret = generic_drop_inode(inode);
 684         trace_f2fs_drop_inode(inode, ret);
 685         return ret;
 686 }
 687
 688 int f2fs_inode_dirtied(struct inode *inode, bool sync)
 689 {
 690         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 691         int ret = 0;
 692
 693         spin_lock(&sbi->inode_lock[DIRTY_META]);
 694         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 695                 ret = 1;
 696         } else {
 697                 set_inode_flag(inode, FI_DIRTY_INODE);
 698                 stat_inc_dirty_inode(sbi, DIRTY_META);
 699         }
 700         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
 701                 list_add_tail(&F2FS_I(inode)->gdirty_list,
 702                                 &sbi->inode_list[DIRTY_META]);
 703                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
 704         }
 705         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 706         return ret;
 707 }
 708
 709 void f2fs_inode_synced(struct inode *inode)
 710 {
 711         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 712
 713         spin_lock(&sbi->inode_lock[DIRTY_META]);
 714         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
 715                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
 716                 return;
 717         }
 718         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
 719                 list_del_init(&F2FS_I(inode)->gdirty_list);
 720                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
 721         }
 722         clear_inode_flag(inode, FI_DIRTY_INODE);
 723         clear_inode_flag(inode, FI_AUTO_RECOVER);
 724         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
 725         spin_unlock(&sbi->inode_lock[DIRTY_META]);
 726 }
 727
 728 /*
 729  * f2fs_dirty_inode() is called from __mark_inode_dirty()
 730  *
 731  * We should call set_dirty_inode to write the dirty inode through write_inode.
 732  */
 733 static void f2fs_dirty_inode(struct inode *inode, int flags)
 734 {
 735         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 736
 737         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 738                         inode->i_ino == F2FS_META_INO(sbi))
 739                 return;
 740
 741         if (flags == I_DIRTY_TIME)
 742                 return;
 743
 744         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
 745                 clear_inode_flag(inode, FI_AUTO_RECOVER);
 746
 747         f2fs_inode_dirtied(inode, false);
 748 }
 749
 750 static void f2fs_i_callback(struct rcu_head *head)
 751 {
 752         struct inode *inode = container_of(head, struct inode, i_rcu);
 753         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
 754 }
 755
 756 static void f2fs_destroy_inode(struct inode *inode)
 757 {
 758         call_rcu(&inode->i_rcu, f2fs_i_callback);
 759 }
 760
 761 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
 762 {
 763         percpu_counter_destroy(&sbi->alloc_valid_block_count);
 764         percpu_counter_destroy(&sbi->total_valid_inode_count);
 765 }
 766
 767 static void destroy_device_list(struct f2fs_sb_info *sbi)
 768 {
 769         int i;
 770
 771         for (i = 0; i < sbi->s_ndevs; i++) {
 772                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
 773 #ifdef CONFIG_BLK_DEV_ZONED
 774                 kfree(FDEV(i).blkz_type);
 775 #endif
 776         }
 777         kfree(sbi->devs);
 778 }
 779
 780 static void f2fs_put_super(struct super_block *sb)
 781 {
 782         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 783         int i;
 784
 785         /* unregister procfs/sysfs entries in advance to avoid race case */
 786         f2fs_unregister_sysfs(sbi);
 787
 788         f2fs_quota_off_umount(sb);
 789
 790         /* prevent remaining shrinker jobs */
 791         mutex_lock(&sbi->umount_mutex);
 792
 793         /*
 794          * We don't need to do checkpoint when superblock is clean.
 795          * But, the previous checkpoint was not done by umount, it needs to do
 796          * clean checkpoint again.
 797          */
 798         if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
 799                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
 800                 struct cp_control cpc = {
 801                         .reason = CP_UMOUNT,
 802                 };
 803                 write_checkpoint(sbi, &cpc);
 804         }
 805
 806         /* be sure to wait for any on-going discard commands */
 807         f2fs_wait_discard_bios(sbi, true);
 808
 809         if (f2fs_discard_en(sbi) && !sbi->discard_blks) {
 810                 struct cp_control cpc = {
 811                         .reason = CP_UMOUNT | CP_TRIMMED,
 812                 };
 813                 write_checkpoint(sbi, &cpc);
 814         }
 815
 816         /* write_checkpoint can update stat informaion */
 817         f2fs_destroy_stats(sbi);
 818
 819         /*
 820          * normally superblock is clean, so we need to release this.
 821          * In addition, EIO will skip do checkpoint, we need this as well.
 822          */
 823         release_ino_entry(sbi, true);
 824
 825         f2fs_leave_shrinker(sbi);
 826         mutex_unlock(&sbi->umount_mutex);
 827
 828         /* our cp_error case, we can wait for any writeback page */
 829         f2fs_flush_merged_writes(sbi);
 830
 831         iput(sbi->node_inode);
 832         iput(sbi->meta_inode);
 833
 834         /* destroy f2fs internal modules */
 835         destroy_node_manager(sbi);
 836         destroy_segment_manager(sbi);
 837
 838         kfree(sbi->ckpt);
 839
 840         sb->s_fs_info = NULL;
 841         if (sbi->s_chksum_driver)
 842                 crypto_free_shash(sbi->s_chksum_driver);
 843         kfree(sbi->raw_super);
 844
 845         destroy_device_list(sbi);
 846         mempool_destroy(sbi->write_io_dummy);
 847 #ifdef CONFIG_QUOTA
 848         for (i = 0; i < MAXQUOTAS; i++)
 849                 kfree(sbi->s_qf_names[i]);
 850 #endif
 851         destroy_percpu_info(sbi);
 852         for (i = 0; i < NR_PAGE_TYPE; i++)
 853                 kfree(sbi->write_io[i]);
 854         kfree(sbi);
 855 }
 856
 857 int f2fs_sync_fs(struct super_block *sb, int sync)
 858 {
 859         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 860         int err = 0;
 861
 862         trace_f2fs_sync_fs(sb, sync);
 863
 864         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
 865                 return -EAGAIN;
 866
 867         if (sync) {
 868                 struct cp_control cpc;
 869
 870                 cpc.reason = __get_cp_reason(sbi);
 871
 872                 mutex_lock(&sbi->gc_mutex);
 873                 err = write_checkpoint(sbi, &cpc);
 874                 mutex_unlock(&sbi->gc_mutex);
 875         }
 876         f2fs_trace_ios(NULL, 1);
 877
 878         return err;
 879 }
 880
 881 static int f2fs_freeze(struct super_block *sb)
 882 {
 883         if (f2fs_readonly(sb))
 884                 return 0;
 885
 886         /* IO error happened before */
 887         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
 888                 return -EIO;
 889
 890         /* must be clean, since sync_filesystem() was already called */
 891         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
 892                 return -EINVAL;
 893         return 0;
 894 }
 895
 896 static int f2fs_unfreeze(struct super_block *sb)
 897 {
 898         return 0;
 899 }
 900
 901 #ifdef CONFIG_QUOTA
 902 static int f2fs_statfs_project(struct super_block *sb,
 903                                 kprojid_t projid, struct kstatfs *buf)
 904 {
 905         struct kqid qid;
 906         struct dquot *dquot;
 907         u64 limit;
 908         u64 curblock;
 909
 910         qid = make_kqid_projid(projid);
 911         dquot = dqget(sb, qid);
 912         if (IS_ERR(dquot))
 913                 return PTR_ERR(dquot);
 914         spin_lock(&dq_data_lock);
 915
 916         limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
 917                                         dquot->dq_dqb.dqb_bhardlimit);
 918         if (limit)
 919                 limit >>= sb->s_blocksize_bits;
 920
 921         if (limit && buf->f_blocks > limit) {
 922                 curblock = (dquot->dq_dqb.dqb_curspace +
 923                             dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
 924                 buf->f_blocks = limit;
 925                 buf->f_bfree = buf->f_bavail =
 926                         (buf->f_blocks > curblock) ?
 927                          (buf->f_blocks - curblock) : 0;
 928         }
 929
 930         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
 931                                         dquot->dq_dqb.dqb_ihardlimit);
 932
 933         if (limit && buf->f_files > limit) {
 934                 buf->f_files = limit;
 935                 buf->f_ffree =
 936                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
 937                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
 938         }
 939
 940         spin_unlock(&dq_data_lock);
 941         dqput(dquot);
 942         return 0;
 943 }
 944 #endif
 945
 946 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
 947 {
 948         struct super_block *sb = dentry->d_sb;
 949         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 950         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 951         block_t total_count, user_block_count, start_count, ovp_count;
 952         u64 avail_node_count;
 953
 954         total_count = le64_to_cpu(sbi->raw_super->block_count);
 955         user_block_count = sbi->user_block_count;
 956         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
 957         ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
 958         buf->f_type = F2FS_SUPER_MAGIC;
 959         buf->f_bsize = sbi->blocksize;
 960
 961         buf->f_blocks = total_count - start_count;
 962         buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count;
 963         buf->f_bavail = user_block_count - valid_user_blocks(sbi) -
 964                                                 sbi->reserved_blocks;
 965
 966         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
 967
 968         if (avail_node_count > user_block_count) {
 969                 buf->f_files = user_block_count;
 970                 buf->f_ffree = buf->f_bavail;
 971         } else {
 972                 buf->f_files = avail_node_count;
 973                 buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
 974                                         buf->f_bavail);
 975         }
 976
 977         buf->f_namelen = F2FS_NAME_LEN;
 978         buf->f_fsid.val[0] = (u32)id;
 979         buf->f_fsid.val[1] = (u32)(id >> 32);
 980
 981 #ifdef CONFIG_QUOTA
 982         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
 983                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
 984                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
 985         }
 986 #endif
 987         return 0;
 988 }
 989
 990 static inline void f2fs_show_quota_options(struct seq_file *seq,
 991                                            struct super_block *sb)
 992 {
 993 #ifdef CONFIG_QUOTA
 994         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 995
 996         if (sbi->s_jquota_fmt) {
 997                 char *fmtname = "";
 998
 999                 switch (sbi->s_jquota_fmt) {
1000                 case QFMT_VFS_OLD:
1001                         fmtname = "vfsold";
1002                         break;
1003                 case QFMT_VFS_V0:
1004                         fmtname = "vfsv0";
1005                         break;
1006                 case QFMT_VFS_V1:
1007                         fmtname = "vfsv1";
1008                         break;
1009                 }
1010                 seq_printf(seq, ",jqfmt=%s", fmtname);
1011         }
1012
1013         if (sbi->s_qf_names[USRQUOTA])
1014                 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1015
1016         if (sbi->s_qf_names[GRPQUOTA])
1017                 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1018
1019         if (sbi->s_qf_names[PRJQUOTA])
1020                 seq_show_option(seq, "prjjquota", sbi->s_qf_names[PRJQUOTA]);
1021 #endif
1022 }
1023
1024 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1025 {
1026         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1027
1028         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) {
1029                 if (test_opt(sbi, FORCE_FG_GC))
1030                         seq_printf(seq, ",background_gc=%s", "sync");
1031                 else
1032                         seq_printf(seq, ",background_gc=%s", "on");
1033         } else {
1034                 seq_printf(seq, ",background_gc=%s", "off");
1035         }
1036         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1037                 seq_puts(seq, ",disable_roll_forward");
1038         if (test_opt(sbi, DISCARD))
1039                 seq_puts(seq, ",discard");
1040         if (test_opt(sbi, NOHEAP))
1041                 seq_puts(seq, ",no_heap");
1042         else
1043                 seq_puts(seq, ",heap");
1044 #ifdef CONFIG_F2FS_FS_XATTR
1045         if (test_opt(sbi, XATTR_USER))
1046                 seq_puts(seq, ",user_xattr");
1047         else
1048                 seq_puts(seq, ",nouser_xattr");
1049         if (test_opt(sbi, INLINE_XATTR))
1050                 seq_puts(seq, ",inline_xattr");
1051         else
1052                 seq_puts(seq, ",noinline_xattr");
1053 #endif
1054 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1055         if (test_opt(sbi, POSIX_ACL))
1056                 seq_puts(seq, ",acl");
1057         else
1058                 seq_puts(seq, ",noacl");
1059 #endif
1060         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1061                 seq_puts(seq, ",disable_ext_identify");
1062         if (test_opt(sbi, INLINE_DATA))
1063                 seq_puts(seq, ",inline_data");
1064         else
1065                 seq_puts(seq, ",noinline_data");
1066         if (test_opt(sbi, INLINE_DENTRY))
1067                 seq_puts(seq, ",inline_dentry");
1068         else
1069                 seq_puts(seq, ",noinline_dentry");
1070         if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1071                 seq_puts(seq, ",flush_merge");
1072         if (test_opt(sbi, NOBARRIER))
1073                 seq_puts(seq, ",nobarrier");
1074         if (test_opt(sbi, FASTBOOT))
1075                 seq_puts(seq, ",fastboot");
1076         if (test_opt(sbi, EXTENT_CACHE))
1077                 seq_puts(seq, ",extent_cache");
1078         else
1079                 seq_puts(seq, ",noextent_cache");
1080         if (test_opt(sbi, DATA_FLUSH))
1081                 seq_puts(seq, ",data_flush");
1082
1083         seq_puts(seq, ",mode=");
1084         if (test_opt(sbi, ADAPTIVE))
1085                 seq_puts(seq, "adaptive");
1086         else if (test_opt(sbi, LFS))
1087                 seq_puts(seq, "lfs");
1088         seq_printf(seq, ",active_logs=%u", sbi->active_logs);
1089         if (F2FS_IO_SIZE_BITS(sbi))
1090                 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi));
1091 #ifdef CONFIG_F2FS_FAULT_INJECTION
1092         if (test_opt(sbi, FAULT_INJECTION))
1093                 seq_printf(seq, ",fault_injection=%u",
1094                                 sbi->fault_info.inject_rate);
1095 #endif
1096 #ifdef CONFIG_QUOTA
1097         if (test_opt(sbi, QUOTA))
1098                 seq_puts(seq, ",quota");
1099         if (test_opt(sbi, USRQUOTA))
1100                 seq_puts(seq, ",usrquota");
1101         if (test_opt(sbi, GRPQUOTA))
1102                 seq_puts(seq, ",grpquota");
1103         if (test_opt(sbi, PRJQUOTA))
1104                 seq_puts(seq, ",prjquota");
1105 #endif
1106         f2fs_show_quota_options(seq, sbi->sb);
1107
1108         return 0;
1109 }
1110
1111 static void default_options(struct f2fs_sb_info *sbi)
1112 {
1113         /* init some FS parameters */
1114         sbi->active_logs = NR_CURSEG_TYPE;
1115
1116         set_opt(sbi, BG_GC);
1117         set_opt(sbi, INLINE_XATTR);
1118         set_opt(sbi, INLINE_DATA);
1119         set_opt(sbi, INLINE_DENTRY);
1120         set_opt(sbi, EXTENT_CACHE);
1121         set_opt(sbi, NOHEAP);
1122         sbi->sb->s_flags |= MS_LAZYTIME;
1123         set_opt(sbi, FLUSH_MERGE);
1124         if (f2fs_sb_mounted_blkzoned(sbi->sb)) {
1125                 set_opt_mode(sbi, F2FS_MOUNT_LFS);
1126                 set_opt(sbi, DISCARD);
1127         } else {
1128                 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
1129         }
1130
1131 #ifdef CONFIG_F2FS_FS_XATTR
1132         set_opt(sbi, XATTR_USER);
1133 #endif
1134 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1135         set_opt(sbi, POSIX_ACL);
1136 #endif
1137
1138 #ifdef CONFIG_F2FS_FAULT_INJECTION
1139         f2fs_build_fault_attr(sbi, 0);
1140 #endif
1141 }
1142
1143 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1144 {
1145         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1146         struct f2fs_mount_info org_mount_opt;
1147         unsigned long old_sb_flags;
1148         int err, active_logs;
1149         bool need_restart_gc = false;
1150         bool need_stop_gc = false;
1151         bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1152 #ifdef CONFIG_F2FS_FAULT_INJECTION
1153         struct f2fs_fault_info ffi = sbi->fault_info;
1154 #endif
1155 #ifdef CONFIG_QUOTA
1156         int s_jquota_fmt;
1157         char *s_qf_names[MAXQUOTAS];
1158         int i, j;
1159 #endif
1160
1161         /*
1162          * Save the old mount options in case we
1163          * need to restore them.
1164          */
1165         org_mount_opt = sbi->mount_opt;
1166         old_sb_flags = sb->s_flags;
1167         active_logs = sbi->active_logs;
1168
1169 #ifdef CONFIG_QUOTA
1170         s_jquota_fmt = sbi->s_jquota_fmt;
1171         for (i = 0; i < MAXQUOTAS; i++) {
1172                 if (sbi->s_qf_names[i]) {
1173                         s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
1174                                                          GFP_KERNEL);
1175                         if (!s_qf_names[i]) {
1176                                 for (j = 0; j < i; j++)
1177                                         kfree(s_qf_names[j]);
1178                                 return -ENOMEM;
1179                         }
1180                 } else {
1181                         s_qf_names[i] = NULL;
1182                 }
1183         }
1184 #endif
1185
1186         /* recover superblocks we couldn't write due to previous RO mount */
1187         if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1188                 err = f2fs_commit_super(sbi, false);
1189                 f2fs_msg(sb, KERN_INFO,
1190                         "Try to recover all the superblocks, ret: %d", err);
1191                 if (!err)
1192                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1193         }
1194
1195         default_options(sbi);
1196
1197         /* parse mount options */
1198         err = parse_options(sb, data);
1199         if (err)
1200                 goto restore_opts;
1201
1202         /*
1203          * Previous and new state of filesystem is RO,
1204          * so skip checking GC and FLUSH_MERGE conditions.
1205          */
1206         if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
1207                 goto skip;
1208
1209         if (!f2fs_readonly(sb) && (*flags & MS_RDONLY)) {
1210                 err = dquot_suspend(sb, -1);
1211                 if (err < 0)
1212                         goto restore_opts;
1213         } else {
1214                 /* dquot_resume needs RW */
1215                 sb->s_flags &= ~MS_RDONLY;
1216                 dquot_resume(sb, -1);
1217         }
1218
1219         /* disallow enable/disable extent_cache dynamically */
1220         if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1221                 err = -EINVAL;
1222                 f2fs_msg(sbi->sb, KERN_WARNING,
1223                                 "switch extent_cache option is not allowed");
1224                 goto restore_opts;
1225         }
1226
1227         /*
1228          * We stop the GC thread if FS is mounted as RO
1229          * or if background_gc = off is passed in mount
1230          * option. Also sync the filesystem.
1231          */
1232         if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
1233                 if (sbi->gc_thread) {
1234                         stop_gc_thread(sbi);
1235                         need_restart_gc = true;
1236                 }
1237         } else if (!sbi->gc_thread) {
1238                 err = start_gc_thread(sbi);
1239                 if (err)
1240                         goto restore_opts;
1241                 need_stop_gc = true;
1242         }
1243
1244         if (*flags & MS_RDONLY) {
1245                 writeback_inodes_sb(sb, WB_REASON_SYNC);
1246                 sync_inodes_sb(sb);
1247
1248                 set_sbi_flag(sbi, SBI_IS_DIRTY);
1249                 set_sbi_flag(sbi, SBI_IS_CLOSE);
1250                 f2fs_sync_fs(sb, 1);
1251                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
1252         }
1253
1254         /*
1255          * We stop issue flush thread if FS is mounted as RO
1256          * or if flush_merge is not passed in mount option.
1257          */
1258         if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1259                 clear_opt(sbi, FLUSH_MERGE);
1260                 destroy_flush_cmd_control(sbi, false);
1261         } else {
1262                 err = create_flush_cmd_control(sbi);
1263                 if (err)
1264                         goto restore_gc;
1265         }
1266 skip:
1267 #ifdef CONFIG_QUOTA
1268         /* Release old quota file names */
1269         for (i = 0; i < MAXQUOTAS; i++)
1270                 kfree(s_qf_names[i]);
1271 #endif
1272         /* Update the POSIXACL Flag */
1273         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1274                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1275
1276         return 0;
1277 restore_gc:
1278         if (need_restart_gc) {
1279                 if (start_gc_thread(sbi))
1280                         f2fs_msg(sbi->sb, KERN_WARNING,
1281                                 "background gc thread has stopped");
1282         } else if (need_stop_gc) {
1283                 stop_gc_thread(sbi);
1284         }
1285 restore_opts:
1286 #ifdef CONFIG_QUOTA
1287         sbi->s_jquota_fmt = s_jquota_fmt;
1288         for (i = 0; i < MAXQUOTAS; i++) {
1289                 kfree(sbi->s_qf_names[i]);
1290                 sbi->s_qf_names[i] = s_qf_names[i];
1291         }
1292 #endif
1293         sbi->mount_opt = org_mount_opt;
1294         sbi->active_logs = active_logs;
1295         sb->s_flags = old_sb_flags;
1296 #ifdef CONFIG_F2FS_FAULT_INJECTION
1297         sbi->fault_info = ffi;
1298 #endif
1299         return err;
1300 }
1301
1302 #ifdef CONFIG_QUOTA
1303 /* Read data from quotafile */
1304 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1305                                size_t len, loff_t off)
1306 {
1307         struct inode *inode = sb_dqopt(sb)->files[type];
1308         struct address_space *mapping = inode->i_mapping;
1309         block_t blkidx = F2FS_BYTES_TO_BLK(off);
1310         int offset = off & (sb->s_blocksize - 1);
1311         int tocopy;
1312         size_t toread;
1313         loff_t i_size = i_size_read(inode);
1314         struct page *page;
1315         char *kaddr;
1316
1317         if (off > i_size)
1318                 return 0;
1319
1320         if (off + len > i_size)
1321                 len = i_size - off;
1322         toread = len;
1323         while (toread > 0) {
1324                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1325 repeat:
1326                 page = read_mapping_page(mapping, blkidx, NULL);
1327                 if (IS_ERR(page))
1328                         return PTR_ERR(page);
1329
1330                 lock_page(page);
1331
1332                 if (unlikely(page->mapping != mapping)) {
1333                         f2fs_put_page(page, 1);
1334                         goto repeat;
1335                 }
1336                 if (unlikely(!PageUptodate(page))) {
1337                         f2fs_put_page(page, 1);
1338                         return -EIO;
1339                 }
1340
1341                 kaddr = kmap_atomic(page);
1342                 memcpy(data, kaddr + offset, tocopy);
1343                 kunmap_atomic(kaddr);
1344                 f2fs_put_page(page, 1);
1345
1346                 offset = 0;
1347                 toread -= tocopy;
1348                 data += tocopy;
1349                 blkidx++;
1350         }
1351         return len;
1352 }
1353
1354 /* Write to quotafile */
1355 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1356                                 const char *data, size_t len, loff_t off)
1357 {
1358         struct inode *inode = sb_dqopt(sb)->files[type];
1359         struct address_space *mapping = inode->i_mapping;
1360         const struct address_space_operations *a_ops = mapping->a_ops;
1361         int offset = off & (sb->s_blocksize - 1);
1362         size_t towrite = len;
1363         struct page *page;
1364         char *kaddr;
1365         int err = 0;
1366         int tocopy;
1367
1368         while (towrite > 0) {
1369                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1370                                                                 towrite);
1371
1372                 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1373                                                         &page, NULL);
1374                 if (unlikely(err))
1375                         break;
1376
1377                 kaddr = kmap_atomic(page);
1378                 memcpy(kaddr + offset, data, tocopy);
1379                 kunmap_atomic(kaddr);
1380                 flush_dcache_page(page);
1381
1382                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1383                                                 page, NULL);
1384                 offset = 0;
1385                 towrite -= tocopy;
1386                 off += tocopy;
1387                 data += tocopy;
1388                 cond_resched();
1389         }
1390
1391         if (len == towrite)
1392                 return 0;
1393         inode->i_version++;
1394         inode->i_mtime = inode->i_ctime = current_time(inode);
1395         f2fs_mark_inode_dirty_sync(inode, false);
1396         return len - towrite;
1397 }
1398
1399 static struct dquot **f2fs_get_dquots(struct inode *inode)
1400 {
1401         return F2FS_I(inode)->i_dquot;
1402 }
1403
1404 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1405 {
1406         return &F2FS_I(inode)->i_reserved_quota;
1407 }
1408
1409 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1410 {
1411         return dquot_quota_on_mount(sbi->sb, sbi->s_qf_names[type],
1412                                                 sbi->s_jquota_fmt, type);
1413 }
1414
1415 void f2fs_enable_quota_files(struct f2fs_sb_info *sbi)
1416 {
1417         int i, ret;
1418
1419         for (i = 0; i < MAXQUOTAS; i++) {
1420                 if (sbi->s_qf_names[i]) {
1421                         ret = f2fs_quota_on_mount(sbi, i);
1422                         if (ret < 0)
1423                                 f2fs_msg(sbi->sb, KERN_ERR,
1424                                         "Cannot turn on journaled "
1425                                         "quota: error %d", ret);
1426                 }
1427         }
1428 }
1429
1430 static int f2fs_quota_sync(struct super_block *sb, int type)
1431 {
1432         struct quota_info *dqopt = sb_dqopt(sb);
1433         int cnt;
1434         int ret;
1435
1436         ret = dquot_writeback_dquots(sb, type);
1437         if (ret)
1438                 return ret;
1439
1440         /*
1441          * Now when everything is written we can discard the pagecache so
1442          * that userspace sees the changes.
1443          */
1444         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
1445                 if (type != -1 && cnt != type)
1446                         continue;
1447                 if (!sb_has_quota_active(sb, cnt))
1448                         continue;
1449
1450                 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping);
1451                 if (ret)
1452                         return ret;
1453
1454                 inode_lock(dqopt->files[cnt]);
1455                 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
1456                 inode_unlock(dqopt->files[cnt]);
1457         }
1458         return 0;
1459 }
1460
1461 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
1462                                                         const struct path *path)
1463 {
1464         struct inode *inode;
1465         int err;
1466
1467         err = f2fs_quota_sync(sb, type);
1468         if (err)
1469                 return err;
1470
1471         err = dquot_quota_on(sb, type, format_id, path);
1472         if (err)
1473                 return err;
1474
1475         inode = d_inode(path->dentry);
1476
1477         inode_lock(inode);
1478         F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL;
1479         inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1480                                         S_NOATIME | S_IMMUTABLE);
1481         inode_unlock(inode);
1482         f2fs_mark_inode_dirty_sync(inode, false);
1483
1484         return 0;
1485 }
1486
1487 static int f2fs_quota_off(struct super_block *sb, int type)
1488 {
1489         struct inode *inode = sb_dqopt(sb)->files[type];
1490         int err;
1491
1492         if (!inode || !igrab(inode))
1493                 return dquot_quota_off(sb, type);
1494
1495         err = f2fs_quota_sync(sb, type);
1496         if (err)
1497                 goto out_put;
1498
1499         err = dquot_quota_off(sb, type);
1500         if (err)
1501                 goto out_put;
1502
1503         inode_lock(inode);
1504         F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL);
1505         inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1506         inode_unlock(inode);
1507         f2fs_mark_inode_dirty_sync(inode, false);
1508 out_put:
1509         iput(inode);
1510         return err;
1511 }
1512
1513 void f2fs_quota_off_umount(struct super_block *sb)
1514 {
1515         int type;
1516         int err;
1517
1518         for (type = 0; type < MAXQUOTAS; type++) {
1519                 err = f2fs_quota_off(sb, type);
1520                 if (err) {
1521                         int ret = dquot_quota_off(sb, type);
1522
1523                         f2fs_msg(sb, KERN_ERR,
1524                                 "Fail to turn off disk quota "
1525                                 "(type: %d, err: %d, ret:%d), Please "
1526                                 "run fsck to fix it.", type, err, ret);
1527                         set_sbi_flag(F2FS_SB(sb), SBI_NEED_FSCK);
1528                 }
1529         }
1530         /*
1531          * In case of checkpoint=disable, we must flush quota blocks.
1532          * This can cause NULL exception for node_inode in end_io, since
1533          * put_super already dropped it.
1534          */
1535         sync_filesystem(sb);
1536 }
1537
1538 int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
1539 {
1540         *projid = F2FS_I(inode)->i_projid;
1541         return 0;
1542 }
1543
1544 static const struct dquot_operations f2fs_quota_operations = {
1545         .get_reserved_space = f2fs_get_reserved_space,
1546         .write_dquot    = dquot_commit,
1547         .acquire_dquot  = dquot_acquire,
1548         .release_dquot  = dquot_release,
1549         .mark_dirty     = dquot_mark_dquot_dirty,
1550         .write_info     = dquot_commit_info,
1551         .alloc_dquot    = dquot_alloc,
1552         .destroy_dquot  = dquot_destroy,
1553         .get_projid     = f2fs_get_projid,
1554         .get_next_id    = dquot_get_next_id,
1555 };
1556
1557 static const struct quotactl_ops f2fs_quotactl_ops = {
1558         .quota_on       = f2fs_quota_on,
1559         .quota_off      = f2fs_quota_off,
1560         .quota_sync     = f2fs_quota_sync,
1561         .get_state      = dquot_get_state,
1562         .set_info       = dquot_set_dqinfo,
1563         .get_dqblk      = dquot_get_dqblk,
1564         .set_dqblk      = dquot_set_dqblk,
1565         .get_nextdqblk  = dquot_get_next_dqblk,
1566 };
1567 #else
1568 void f2fs_quota_off_umount(struct super_block *sb)
1569 {
1570 }
1571 #endif
1572
1573 static const struct super_operations f2fs_sops = {
1574         .alloc_inode    = f2fs_alloc_inode,
1575         .drop_inode     = f2fs_drop_inode,
1576         .destroy_inode  = f2fs_destroy_inode,
1577         .write_inode    = f2fs_write_inode,
1578         .dirty_inode    = f2fs_dirty_inode,
1579         .show_options   = f2fs_show_options,
1580 #ifdef CONFIG_QUOTA
1581         .quota_read     = f2fs_quota_read,
1582         .quota_write    = f2fs_quota_write,
1583         .get_dquots     = f2fs_get_dquots,
1584 #endif
1585         .evict_inode    = f2fs_evict_inode,
1586         .put_super      = f2fs_put_super,
1587         .sync_fs        = f2fs_sync_fs,
1588         .freeze_fs      = f2fs_freeze,
1589         .unfreeze_fs    = f2fs_unfreeze,
1590         .statfs         = f2fs_statfs,
1591         .remount_fs     = f2fs_remount,
1592 };
1593
1594 #ifdef CONFIG_F2FS_FS_ENCRYPTION
1595 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
1596 {
1597         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1598                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1599                                 ctx, len, NULL);
1600 }
1601
1602 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
1603                                                         void *fs_data)
1604 {
1605         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
1606                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
1607                                 ctx, len, fs_data, XATTR_CREATE);
1608 }
1609
1610 static unsigned f2fs_max_namelen(struct inode *inode)
1611 {
1612         return S_ISLNK(inode->i_mode) ?
1613                         inode->i_sb->s_blocksize : F2FS_NAME_LEN;
1614 }
1615
1616 static const struct fscrypt_operations f2fs_cryptops = {
1617         .key_prefix     = "f2fs:",
1618         .get_context    = f2fs_get_context,
1619         .set_context    = f2fs_set_context,
1620         .is_encrypted   = f2fs_encrypted_inode,
1621         .empty_dir      = f2fs_empty_dir,
1622         .max_namelen    = f2fs_max_namelen,
1623 };
1624 #else
1625 static const struct fscrypt_operations f2fs_cryptops = {
1626         .is_encrypted   = f2fs_encrypted_inode,
1627 };
1628 #endif
1629
1630 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
1631                 u64 ino, u32 generation)
1632 {
1633         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1634         struct inode *inode;
1635
1636         if (check_nid_range(sbi, ino))
1637                 return ERR_PTR(-ESTALE);
1638
1639         /*
1640          * f2fs_iget isn't quite right if the inode is currently unallocated!
1641          * However f2fs_iget currently does appropriate checks to handle stale
1642          * inodes so everything is OK.
1643          */
1644         inode = f2fs_iget(sb, ino);
1645         if (IS_ERR(inode))
1646                 return ERR_CAST(inode);
1647         if (unlikely(generation && inode->i_generation != generation)) {
1648                 /* we didn't find the right inode.. */
1649                 iput(inode);
1650                 return ERR_PTR(-ESTALE);
1651         }
1652         return inode;
1653 }
1654
1655 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1656                 int fh_len, int fh_type)
1657 {
1658         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1659                                     f2fs_nfs_get_inode);
1660 }
1661
1662 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
1663                 int fh_len, int fh_type)
1664 {
1665         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1666                                     f2fs_nfs_get_inode);
1667 }
1668
1669 static const struct export_operations f2fs_export_ops = {
1670         .fh_to_dentry = f2fs_fh_to_dentry,
1671         .fh_to_parent = f2fs_fh_to_parent,
1672         .get_parent = f2fs_get_parent,
1673 };
1674
1675 static loff_t max_file_blocks(void)
1676 {
1677         loff_t result = 0;
1678         loff_t leaf_count = ADDRS_PER_BLOCK;
1679
1680         /*
1681          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
1682          * F2FS_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
1683          * space in inode.i_addr, it will be more safe to reassign
1684          * result as zero.
1685          */
1686
1687         /* two direct node blocks */
1688         result += (leaf_count * 2);
1689
1690         /* two indirect node blocks */
1691         leaf_count *= NIDS_PER_BLOCK;
1692         result += (leaf_count * 2);
1693
1694         /* one double indirect node block */
1695         leaf_count *= NIDS_PER_BLOCK;
1696         result += leaf_count;
1697
1698         return result;
1699 }
1700
1701 static int __f2fs_commit_super(struct buffer_head *bh,
1702                         struct f2fs_super_block *super)
1703 {
1704         lock_buffer(bh);
1705         if (super)
1706                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
1707         set_buffer_uptodate(bh);
1708         set_buffer_dirty(bh);
1709         unlock_buffer(bh);
1710
1711         /* it's rare case, we can do fua all the time */
1712         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
1713 }
1714
1715 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
1716                                         struct buffer_head *bh)
1717 {
1718         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1719                                         (bh->b_data + F2FS_SUPER_OFFSET);
1720         struct super_block *sb = sbi->sb;
1721         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
1722         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
1723         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
1724         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
1725         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
1726         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
1727         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
1728         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
1729         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
1730         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
1731         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1732         u32 segment_count = le32_to_cpu(raw_super->segment_count);
1733         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1734         u64 main_end_blkaddr = main_blkaddr +
1735                                 (segment_count_main << log_blocks_per_seg);
1736         u64 seg_end_blkaddr = segment0_blkaddr +
1737                                 (segment_count << log_blocks_per_seg);
1738
1739         if (segment0_blkaddr != cp_blkaddr) {
1740                 f2fs_msg(sb, KERN_INFO,
1741                         "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
1742                         segment0_blkaddr, cp_blkaddr);
1743                 return true;
1744         }
1745
1746         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
1747                                                         sit_blkaddr) {
1748                 f2fs_msg(sb, KERN_INFO,
1749                         "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
1750                         cp_blkaddr, sit_blkaddr,
1751                         segment_count_ckpt << log_blocks_per_seg);
1752                 return true;
1753         }
1754
1755         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
1756                                                         nat_blkaddr) {
1757                 f2fs_msg(sb, KERN_INFO,
1758                         "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
1759                         sit_blkaddr, nat_blkaddr,
1760                         segment_count_sit << log_blocks_per_seg);
1761                 return true;
1762         }
1763
1764         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
1765                                                         ssa_blkaddr) {
1766                 f2fs_msg(sb, KERN_INFO,
1767                         "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
1768                         nat_blkaddr, ssa_blkaddr,
1769                         segment_count_nat << log_blocks_per_seg);
1770                 return true;
1771         }
1772
1773         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
1774                                                         main_blkaddr) {
1775                 f2fs_msg(sb, KERN_INFO,
1776                         "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
1777                         ssa_blkaddr, main_blkaddr,
1778                         segment_count_ssa << log_blocks_per_seg);
1779                 return true;
1780         }
1781
1782         if (main_end_blkaddr > seg_end_blkaddr) {
1783                 f2fs_msg(sb, KERN_INFO,
1784                         "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
1785                         main_blkaddr,
1786                         segment0_blkaddr +
1787                                 (segment_count << log_blocks_per_seg),
1788                         segment_count_main << log_blocks_per_seg);
1789                 return true;
1790         } else if (main_end_blkaddr < seg_end_blkaddr) {
1791                 int err = 0;
1792                 char *res;
1793
1794                 /* fix in-memory information all the time */
1795                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
1796                                 segment0_blkaddr) >> log_blocks_per_seg);
1797
1798                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
1799                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1800                         res = "internally";
1801                 } else {
1802                         err = __f2fs_commit_super(bh, NULL);
1803                         res = err ? "failed" : "done";
1804                 }
1805                 f2fs_msg(sb, KERN_INFO,
1806                         "Fix alignment : %s, start(%u) end(%u) block(%u)",
1807                         res, main_blkaddr,
1808                         segment0_blkaddr +
1809                                 (segment_count << log_blocks_per_seg),
1810                         segment_count_main << log_blocks_per_seg);
1811                 if (err)
1812                         return true;
1813         }
1814         return false;
1815 }
1816
1817 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
1818                                 struct buffer_head *bh)
1819 {
1820         block_t segment_count, segs_per_sec, secs_per_zone;
1821         block_t total_sections, blocks_per_seg;
1822         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
1823                                         (bh->b_data + F2FS_SUPER_OFFSET);
1824         struct super_block *sb = sbi->sb;
1825         unsigned int blocksize;
1826
1827         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
1828                 f2fs_msg(sb, KERN_INFO,
1829                         "Magic Mismatch, valid(0x%x) - read(0x%x)",
1830                         F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
1831                 return -EINVAL;
1832         }
1833
1834         /* Currently, support only 4KB page cache size */
1835         if (F2FS_BLKSIZE != PAGE_SIZE) {
1836                 f2fs_msg(sb, KERN_INFO,
1837                         "Invalid page_cache_size (%lu), supports only 4KB\n",
1838                         PAGE_SIZE);
1839                 return -EFSCORRUPTED;
1840         }
1841
1842         /* Currently, support only 4KB block size */
1843         blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
1844         if (blocksize != F2FS_BLKSIZE) {
1845                 f2fs_msg(sb, KERN_INFO,
1846                         "Invalid blocksize (%u), supports only 4KB\n",
1847                         blocksize);
1848                 return -EFSCORRUPTED;
1849         }
1850
1851         /* check log blocks per segment */
1852         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
1853                 f2fs_msg(sb, KERN_INFO,
1854                         "Invalid log blocks per segment (%u)\n",
1855                         le32_to_cpu(raw_super->log_blocks_per_seg));
1856                 return -EFSCORRUPTED;
1857         }
1858
1859         /* Currently, support 512/1024/2048/4096 bytes sector size */
1860         if (le32_to_cpu(raw_super->log_sectorsize) >
1861                                 F2FS_MAX_LOG_SECTOR_SIZE ||
1862                 le32_to_cpu(raw_super->log_sectorsize) <
1863                                 F2FS_MIN_LOG_SECTOR_SIZE) {
1864                 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
1865                         le32_to_cpu(raw_super->log_sectorsize));
1866                 return -EFSCORRUPTED;
1867         }
1868         if (le32_to_cpu(raw_super->log_sectors_per_block) +
1869                 le32_to_cpu(raw_super->log_sectorsize) !=
1870                         F2FS_MAX_LOG_SECTOR_SIZE) {
1871                 f2fs_msg(sb, KERN_INFO,
1872                         "Invalid log sectors per block(%u) log sectorsize(%u)",
1873                         le32_to_cpu(raw_super->log_sectors_per_block),
1874                         le32_to_cpu(raw_super->log_sectorsize));
1875                 return -EFSCORRUPTED;
1876         }
1877
1878         segment_count = le32_to_cpu(raw_super->segment_count);
1879         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
1880         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
1881         total_sections = le32_to_cpu(raw_super->section_count);
1882
1883         /* blocks_per_seg should be 512, given the above check */
1884         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
1885
1886         if (segment_count > F2FS_MAX_SEGMENT ||
1887                                 segment_count < F2FS_MIN_SEGMENTS) {
1888                 f2fs_msg(sb, KERN_INFO,
1889                         "Invalid segment count (%u)",
1890                         segment_count);
1891                 return -EFSCORRUPTED;
1892         }
1893
1894         if (total_sections > segment_count ||
1895                         total_sections < F2FS_MIN_SEGMENTS ||
1896                         segs_per_sec > segment_count || !segs_per_sec) {
1897                 f2fs_msg(sb, KERN_INFO,
1898                         "Invalid segment/section count (%u, %u x %u)",
1899                         segment_count, total_sections, segs_per_sec);
1900                 return -EFSCORRUPTED;
1901         }
1902
1903         if ((segment_count / segs_per_sec) < total_sections) {
1904                 f2fs_msg(sb, KERN_INFO,
1905                         "Small segment_count (%u < %u * %u)",
1906                         segment_count, segs_per_sec, total_sections);
1907                 return -EFSCORRUPTED;
1908         }
1909
1910         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
1911                 f2fs_msg(sb, KERN_INFO,
1912                         "Wrong segment_count / block_count (%u > %llu)",
1913                         segment_count, le64_to_cpu(raw_super->block_count));
1914                 return -EFSCORRUPTED;
1915         }
1916
1917         if (secs_per_zone > total_sections || !secs_per_zone) {
1918                 f2fs_msg(sb, KERN_INFO,
1919                         "Wrong secs_per_zone / total_sections (%u, %u)",
1920                         secs_per_zone, total_sections);
1921                 return -EFSCORRUPTED;
1922         }
1923         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION) {
1924                 f2fs_msg(sb, KERN_INFO,
1925                         "Corrupted extension count (%u > %u)",
1926                         le32_to_cpu(raw_super->extension_count),
1927                         F2FS_MAX_EXTENSION);
1928                 return -EFSCORRUPTED;
1929         }
1930
1931         if (le32_to_cpu(raw_super->cp_payload) >
1932                                 (blocks_per_seg - F2FS_CP_PACKS)) {
1933                 f2fs_msg(sb, KERN_INFO,
1934                         "Insane cp_payload (%u > %u)",
1935                         le32_to_cpu(raw_super->cp_payload),
1936                         blocks_per_seg - F2FS_CP_PACKS);
1937                 return -EFSCORRUPTED;
1938         }
1939
1940         /* check reserved ino info */
1941         if (le32_to_cpu(raw_super->node_ino) != 1 ||
1942                 le32_to_cpu(raw_super->meta_ino) != 2 ||
1943                 le32_to_cpu(raw_super->root_ino) != 3) {
1944                 f2fs_msg(sb, KERN_INFO,
1945                         "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
1946                         le32_to_cpu(raw_super->node_ino),
1947                         le32_to_cpu(raw_super->meta_ino),
1948                         le32_to_cpu(raw_super->root_ino));
1949                 return -EFSCORRUPTED;
1950         }
1951
1952         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
1953         if (sanity_check_area_boundary(sbi, bh))
1954                 return -EFSCORRUPTED;
1955
1956         return 0;
1957 }
1958
1959 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1960 {
1961         unsigned int total, fsmeta;
1962         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
1963         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1964         unsigned int ovp_segments, reserved_segments;
1965         unsigned int main_segs, blocks_per_seg;
1966         unsigned int sit_segs, nat_segs;
1967         unsigned int sit_bitmap_size, nat_bitmap_size;
1968         unsigned int log_blocks_per_seg;
1969         unsigned int segment_count_main;
1970         unsigned int cp_pack_start_sum, cp_payload;
1971         block_t user_block_count;
1972         int i, j;
1973
1974         total = le32_to_cpu(raw_super->segment_count);
1975         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
1976         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
1977         fsmeta += sit_segs;
1978         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
1979         fsmeta += nat_segs;
1980         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
1981         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
1982
1983         if (unlikely(fsmeta >= total))
1984                 return 1;
1985
1986         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
1987         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
1988
1989         if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
1990                         ovp_segments == 0 || reserved_segments == 0)) {
1991                 f2fs_msg(sbi->sb, KERN_ERR,
1992                         "Wrong layout: check mkfs.f2fs version");
1993                 return 1;
1994         }
1995
1996         user_block_count = le64_to_cpu(ckpt->user_block_count);
1997         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
1998         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
1999         if (!user_block_count || user_block_count >=
2000                         segment_count_main << log_blocks_per_seg) {
2001                 f2fs_msg(sbi->sb, KERN_ERR,
2002                         "Wrong user_block_count: %u", user_block_count);
2003                 return 1;
2004         }
2005
2006         main_segs = le32_to_cpu(raw_super->segment_count_main);
2007         blocks_per_seg = sbi->blocks_per_seg;
2008
2009         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2010                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2011                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2012                         return 1;
2013                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2014                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2015                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
2016                                 f2fs_msg(sbi->sb, KERN_ERR,
2017                                         "Node segment (%u, %u) has the same "
2018                                         "segno: %u", i, j,
2019                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2020                                 return 1;
2021                         }
2022                 }
2023         }
2024         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2025                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2026                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2027                         return 1;
2028                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2029                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2030                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2031                                 f2fs_msg(sbi->sb, KERN_ERR,
2032                                         "Data segment (%u, %u) has the same "
2033                                         "segno: %u", i, j,
2034                                         le32_to_cpu(ckpt->cur_data_segno[i]));
2035                                 return 1;
2036                         }
2037                 }
2038         }
2039         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2040                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2041                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2042                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
2043                                 f2fs_msg(sbi->sb, KERN_ERR,
2044                                         "Node segment (%u) and Data segment (%u)"
2045                                         " has the same segno: %u", i, j,
2046                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2047                                 return 1;
2048                         }
2049                 }
2050         }
2051
2052         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2053         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2054
2055         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2056                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2057                 f2fs_msg(sbi->sb, KERN_ERR,
2058                         "Wrong bitmap size: sit: %u, nat:%u",
2059                         sit_bitmap_size, nat_bitmap_size);
2060                 return 1;
2061         }
2062
2063         cp_pack_start_sum = __start_sum_addr(sbi);
2064         cp_payload = __cp_payload(sbi);
2065         if (cp_pack_start_sum < cp_payload + 1 ||
2066                 cp_pack_start_sum > blocks_per_seg - 1 -
2067                         NR_CURSEG_TYPE) {
2068                 f2fs_msg(sbi->sb, KERN_ERR,
2069                         "Wrong cp_pack_start_sum: %u",
2070                         cp_pack_start_sum);
2071                 return 1;
2072         }
2073
2074         if (unlikely(f2fs_cp_error(sbi))) {
2075                 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
2076                 return 1;
2077         }
2078         return 0;
2079 }
2080
2081 static void init_sb_info(struct f2fs_sb_info *sbi)
2082 {
2083         struct f2fs_super_block *raw_super = sbi->raw_super;
2084         int i, j;
2085
2086         sbi->log_sectors_per_block =
2087                 le32_to_cpu(raw_super->log_sectors_per_block);
2088         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2089         sbi->blocksize = 1 << sbi->log_blocksize;
2090         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2091         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2092         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2093         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2094         sbi->total_sections = le32_to_cpu(raw_super->section_count);
2095         sbi->total_node_count =
2096                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
2097                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2098         sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2099         sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2100         sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2101         sbi->cur_victim_sec = NULL_SECNO;
2102         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2103
2104         sbi->dir_level = DEF_DIR_LEVEL;
2105         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2106         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2107         clear_sbi_flag(sbi, SBI_NEED_FSCK);
2108
2109         for (i = 0; i < NR_COUNT_TYPE; i++)
2110                 atomic_set(&sbi->nr_pages[i], 0);
2111
2112         atomic_set(&sbi->wb_sync_req, 0);
2113
2114         INIT_LIST_HEAD(&sbi->s_list);
2115         mutex_init(&sbi->umount_mutex);
2116         for (i = 0; i < NR_PAGE_TYPE - 1; i++)
2117                 for (j = HOT; j < NR_TEMP_TYPE; j++)
2118                         mutex_init(&sbi->wio_mutex[i][j]);
2119         spin_lock_init(&sbi->cp_lock);
2120 }
2121
2122 static int init_percpu_info(struct f2fs_sb_info *sbi)
2123 {
2124         int err;
2125
2126         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2127         if (err)
2128                 return err;
2129
2130         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2131                                                                 GFP_KERNEL);
2132         if (err)
2133                 percpu_counter_destroy(&sbi->alloc_valid_block_count);
2134
2135         return err;
2136 }
2137
2138 #ifdef CONFIG_BLK_DEV_ZONED
2139 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
2140 {
2141         struct block_device *bdev = FDEV(devi).bdev;
2142         sector_t nr_sectors = bdev->bd_part->nr_sects;
2143         sector_t sector = 0;
2144         struct blk_zone *zones;
2145         unsigned int i, nr_zones;
2146         unsigned int n = 0;
2147         int err = -EIO;
2148
2149         if (!f2fs_sb_mounted_blkzoned(sbi->sb))
2150                 return 0;
2151
2152         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
2153                                 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
2154                 return -EINVAL;
2155         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
2156         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
2157                                 __ilog2_u32(sbi->blocks_per_blkz))
2158                 return -EINVAL;
2159         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
2160         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
2161                                         sbi->log_blocks_per_blkz;
2162         if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
2163                 FDEV(devi).nr_blkz++;
2164
2165         FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL);
2166         if (!FDEV(devi).blkz_type)
2167                 return -ENOMEM;
2168
2169 #define F2FS_REPORT_NR_ZONES   4096
2170
2171         zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone),
2172                         GFP_KERNEL);
2173         if (!zones)
2174                 return -ENOMEM;
2175
2176         /* Get block zones type */
2177         while (zones && sector < nr_sectors) {
2178
2179                 nr_zones = F2FS_REPORT_NR_ZONES;
2180                 err = blkdev_report_zones(bdev, sector,
2181                                           zones, &nr_zones,
2182                                           GFP_KERNEL);
2183                 if (err)
2184                         break;
2185                 if (!nr_zones) {
2186                         err = -EIO;
2187                         break;
2188                 }
2189
2190                 for (i = 0; i < nr_zones; i++) {
2191                         FDEV(devi).blkz_type[n] = zones[i].type;
2192                         sector += zones[i].len;
2193                         n++;
2194                 }
2195         }
2196
2197         kfree(zones);
2198
2199         return err;
2200 }
2201 #endif
2202
2203 /*
2204  * Read f2fs raw super block.
2205  * Because we have two copies of super block, so read both of them
2206  * to get the first valid one. If any one of them is broken, we pass
2207  * them recovery flag back to the caller.
2208  */
2209 static int read_raw_super_block(struct f2fs_sb_info *sbi,
2210                         struct f2fs_super_block **raw_super,
2211                         int *valid_super_block, int *recovery)
2212 {
2213         struct super_block *sb = sbi->sb;
2214         int block;
2215         struct buffer_head *bh;
2216         struct f2fs_super_block *super;
2217         int err = 0;
2218
2219         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
2220         if (!super)
2221                 return -ENOMEM;
2222
2223         for (block = 0; block < 2; block++) {
2224                 bh = sb_bread(sb, block);
2225                 if (!bh) {
2226                         f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
2227                                 block + 1);
2228                         err = -EIO;
2229                         continue;
2230                 }
2231
2232                 /* sanity checking of raw super */
2233                 err = sanity_check_raw_super(sbi, bh);
2234                 if (err) {
2235                         f2fs_msg(sb, KERN_ERR,
2236                                 "Can't find valid F2FS filesystem in %dth superblock",
2237                                 block + 1);
2238                         brelse(bh);
2239                         continue;
2240                 }
2241
2242                 if (!*raw_super) {
2243                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
2244                                                         sizeof(*super));
2245                         *valid_super_block = block;
2246                         *raw_super = super;
2247                 }
2248                 brelse(bh);
2249         }
2250
2251         /* Fail to read any one of the superblocks*/
2252         if (err < 0)
2253                 *recovery = 1;
2254
2255         /* No valid superblock */
2256         if (!*raw_super)
2257                 kfree(super);
2258         else
2259                 err = 0;
2260
2261         return err;
2262 }
2263
2264 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
2265 {
2266         struct buffer_head *bh;
2267         int err;
2268
2269         if ((recover && f2fs_readonly(sbi->sb)) ||
2270                                 bdev_read_only(sbi->sb->s_bdev)) {
2271                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2272                 return -EROFS;
2273         }
2274
2275         /* write back-up superblock first */
2276         bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1);
2277         if (!bh)
2278                 return -EIO;
2279         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2280         brelse(bh);
2281
2282         /* if we are in recovery path, skip writing valid superblock */
2283         if (recover || err)
2284                 return err;
2285
2286         /* write current valid superblock */
2287         bh = sb_getblk(sbi->sb, sbi->valid_super_block);
2288         if (!bh)
2289                 return -EIO;
2290         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
2291         brelse(bh);
2292         return err;
2293 }
2294
2295 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
2296 {
2297         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2298         unsigned int max_devices = MAX_DEVICES;
2299         int i;
2300
2301         /* Initialize single device information */
2302         if (!RDEV(0).path[0]) {
2303                 if (!bdev_is_zoned(sbi->sb->s_bdev))
2304                         return 0;
2305                 max_devices = 1;
2306         }
2307
2308         /*
2309          * Initialize multiple devices information, or single
2310          * zoned block device information.
2311          */
2312         sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info),
2313                                 GFP_KERNEL);
2314         if (!sbi->devs)
2315                 return -ENOMEM;
2316
2317         for (i = 0; i < max_devices; i++) {
2318
2319                 if (i > 0 && !RDEV(i).path[0])
2320                         break;
2321
2322                 if (max_devices == 1) {
2323                         /* Single zoned block device mount */
2324                         FDEV(0).bdev =
2325                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
2326                                         sbi->sb->s_mode, sbi->sb->s_type);
2327                 } else {
2328                         /* Multi-device mount */
2329                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
2330                         FDEV(i).total_segments =
2331                                 le32_to_cpu(RDEV(i).total_segments);
2332                         if (i == 0) {
2333                                 FDEV(i).start_blk = 0;
2334                                 FDEV(i).end_blk = FDEV(i).start_blk +
2335                                     (FDEV(i).total_segments <<
2336                                     sbi->log_blocks_per_seg) - 1 +
2337                                     le32_to_cpu(raw_super->segment0_blkaddr);
2338                         } else {
2339                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
2340                                 FDEV(i).end_blk = FDEV(i).start_blk +
2341                                         (FDEV(i).total_segments <<
2342                                         sbi->log_blocks_per_seg) - 1;
2343                         }
2344                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
2345                                         sbi->sb->s_mode, sbi->sb->s_type);
2346                 }
2347                 if (IS_ERR(FDEV(i).bdev))
2348                         return PTR_ERR(FDEV(i).bdev);
2349
2350                 /* to release errored devices */
2351                 sbi->s_ndevs = i + 1;
2352
2353 #ifdef CONFIG_BLK_DEV_ZONED
2354                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
2355                                 !f2fs_sb_mounted_blkzoned(sbi->sb)) {
2356                         f2fs_msg(sbi->sb, KERN_ERR,
2357                                 "Zoned block device feature not enabled\n");
2358                         return -EINVAL;
2359                 }
2360                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
2361                         if (init_blkz_info(sbi, i)) {
2362                                 f2fs_msg(sbi->sb, KERN_ERR,
2363                                         "Failed to initialize F2FS blkzone information");
2364                                 return -EINVAL;
2365                         }
2366                         if (max_devices == 1)
2367                                 break;
2368                         f2fs_msg(sbi->sb, KERN_INFO,
2369                                 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
2370                                 i, FDEV(i).path,
2371                                 FDEV(i).total_segments,
2372                                 FDEV(i).start_blk, FDEV(i).end_blk,
2373                                 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
2374                                 "Host-aware" : "Host-managed");
2375                         continue;
2376                 }
2377 #endif
2378                 f2fs_msg(sbi->sb, KERN_INFO,
2379                         "Mount Device [%2d]: %20s, %8u, %8x - %8x",
2380                                 i, FDEV(i).path,
2381                                 FDEV(i).total_segments,
2382                                 FDEV(i).start_blk, FDEV(i).end_blk);
2383         }
2384         f2fs_msg(sbi->sb, KERN_INFO,
2385                         "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
2386         return 0;
2387 }
2388
2389 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
2390 {
2391         struct f2fs_sb_info *sbi;
2392         struct f2fs_super_block *raw_super;
2393         struct inode *root;
2394         int err;
2395         bool retry = true, need_fsck = false;
2396         char *options = NULL;
2397         int recovery, i, valid_super_block;
2398         struct curseg_info *seg_i;
2399
2400 try_onemore:
2401         err = -EINVAL;
2402         raw_super = NULL;
2403         valid_super_block = -1;
2404         recovery = 0;
2405
2406         /* allocate memory for f2fs-specific super block info */
2407         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
2408         if (!sbi)
2409                 return -ENOMEM;
2410
2411         sbi->sb = sb;
2412
2413         /* Load the checksum driver */
2414         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
2415         if (IS_ERR(sbi->s_chksum_driver)) {
2416                 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver.");
2417                 err = PTR_ERR(sbi->s_chksum_driver);
2418                 sbi->s_chksum_driver = NULL;
2419                 goto free_sbi;
2420         }
2421
2422         /* set a block size */
2423         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
2424                 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
2425                 goto free_sbi;
2426         }
2427
2428         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
2429                                                                 &recovery);
2430         if (err)
2431                 goto free_sbi;
2432
2433         sb->s_fs_info = sbi;
2434         sbi->raw_super = raw_super;
2435
2436         /* precompute checksum seed for metadata */
2437         if (f2fs_sb_has_inode_chksum(sb))
2438                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
2439                                                 sizeof(raw_super->uuid));
2440
2441         /*
2442          * The BLKZONED feature indicates that the drive was formatted with
2443          * zone alignment optimization. This is optional for host-aware
2444          * devices, but mandatory for host-managed zoned block devices.
2445          */
2446 #ifndef CONFIG_BLK_DEV_ZONED
2447         if (f2fs_sb_mounted_blkzoned(sb)) {
2448                 f2fs_msg(sb, KERN_ERR,
2449                          "Zoned block device support is not enabled\n");
2450                 err = -EOPNOTSUPP;
2451                 goto free_sb_buf;
2452         }
2453 #endif
2454         default_options(sbi);
2455         /* parse mount options */
2456         options = kstrdup((const char *)data, GFP_KERNEL);
2457         if (data && !options) {
2458                 err = -ENOMEM;
2459                 goto free_sb_buf;
2460         }
2461
2462         err = parse_options(sb, options);
2463         if (err)
2464                 goto free_options;
2465
2466         sbi->max_file_blocks = max_file_blocks();
2467         sb->s_maxbytes = sbi->max_file_blocks <<
2468                                 le32_to_cpu(raw_super->log_blocksize);
2469         sb->s_max_links = F2FS_LINK_MAX;
2470         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2471
2472 #ifdef CONFIG_QUOTA
2473         sb->dq_op = &f2fs_quota_operations;
2474         sb->s_qcop = &f2fs_quotactl_ops;
2475         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
2476 #endif
2477
2478         sb->s_op = &f2fs_sops;
2479         sb->s_cop = &f2fs_cryptops;
2480         sb->s_xattr = f2fs_xattr_handlers;
2481         sb->s_export_op = &f2fs_export_ops;
2482         sb->s_magic = F2FS_SUPER_MAGIC;
2483         sb->s_time_gran = 1;
2484         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2485                 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
2486         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
2487
2488         /* init f2fs-specific super block info */
2489         sbi->valid_super_block = valid_super_block;
2490         mutex_init(&sbi->gc_mutex);
2491         mutex_init(&sbi->cp_mutex);
2492         init_rwsem(&sbi->node_write);
2493         init_rwsem(&sbi->node_change);
2494
2495         /* disallow all the data/node/meta page writes */
2496         set_sbi_flag(sbi, SBI_POR_DOING);
2497         spin_lock_init(&sbi->stat_lock);
2498
2499         /* init iostat info */
2500         spin_lock_init(&sbi->iostat_lock);
2501         sbi->iostat_enable = false;
2502
2503         for (i = 0; i < NR_PAGE_TYPE; i++) {
2504                 int n = (i == META) ? 1: NR_TEMP_TYPE;
2505                 int j;
2506
2507                 sbi->write_io[i] = kmalloc(n * sizeof(struct f2fs_bio_info),
2508                                                                 GFP_KERNEL);
2509                 if (!sbi->write_io[i]) {
2510                         err = -ENOMEM;
2511                         goto free_options;
2512                 }
2513
2514                 for (j = HOT; j < n; j++) {
2515                         init_rwsem(&sbi->write_io[i][j].io_rwsem);
2516                         sbi->write_io[i][j].sbi = sbi;
2517                         sbi->write_io[i][j].bio = NULL;
2518                         spin_lock_init(&sbi->write_io[i][j].io_lock);
2519                         INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
2520                 }
2521         }
2522
2523         init_rwsem(&sbi->cp_rwsem);
2524         init_waitqueue_head(&sbi->cp_wait);
2525         init_sb_info(sbi);
2526
2527         err = init_percpu_info(sbi);
2528         if (err)
2529                 goto free_options;
2530
2531         if (F2FS_IO_SIZE(sbi) > 1) {
2532                 sbi->write_io_dummy =
2533                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
2534                 if (!sbi->write_io_dummy) {
2535                         err = -ENOMEM;
2536                         goto free_options;
2537                 }
2538         }
2539
2540         /* get an inode for meta space */
2541         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
2542         if (IS_ERR(sbi->meta_inode)) {
2543                 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
2544                 err = PTR_ERR(sbi->meta_inode);
2545                 goto free_io_dummy;
2546         }
2547
2548         err = get_valid_checkpoint(sbi);
2549         if (err) {
2550                 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
2551                 goto free_meta_inode;
2552         }
2553
2554         /* Initialize device list */
2555         err = f2fs_scan_devices(sbi);
2556         if (err) {
2557                 f2fs_msg(sb, KERN_ERR, "Failed to find devices");
2558                 goto free_devices;
2559         }
2560
2561         sbi->total_valid_node_count =
2562                                 le32_to_cpu(sbi->ckpt->valid_node_count);
2563         percpu_counter_set(&sbi->total_valid_inode_count,
2564                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
2565         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
2566         sbi->total_valid_block_count =
2567                                 le64_to_cpu(sbi->ckpt->valid_block_count);
2568         sbi->last_valid_block_count = sbi->total_valid_block_count;
2569         sbi->reserved_blocks = 0;
2570
2571         for (i = 0; i < NR_INODE_TYPE; i++) {
2572                 INIT_LIST_HEAD(&sbi->inode_list[i]);
2573                 spin_lock_init(&sbi->inode_lock[i]);
2574         }
2575
2576         init_extent_cache_info(sbi);
2577
2578         init_ino_entry_info(sbi);
2579
2580         /* setup f2fs internal modules */
2581         err = build_segment_manager(sbi);
2582         if (err) {
2583                 f2fs_msg(sb, KERN_ERR,
2584                         "Failed to initialize F2FS segment manager");
2585                 goto free_sm;
2586         }
2587         err = build_node_manager(sbi);
2588         if (err) {
2589                 f2fs_msg(sb, KERN_ERR,
2590                         "Failed to initialize F2FS node manager");
2591                 goto free_nm;
2592         }
2593
2594         /* For write statistics */
2595         if (sb->s_bdev->bd_part)
2596                 sbi->sectors_written_start =
2597                         (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2598
2599         /* Read accumulated write IO statistics if exists */
2600         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
2601         if (__exist_node_summaries(sbi))
2602                 sbi->kbytes_written =
2603                         le64_to_cpu(seg_i->journal->info.kbytes_written);
2604
2605         build_gc_manager(sbi);
2606
2607         /* get an inode for node space */
2608         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
2609         if (IS_ERR(sbi->node_inode)) {
2610                 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
2611                 err = PTR_ERR(sbi->node_inode);
2612                 goto free_nm;
2613         }
2614
2615         f2fs_join_shrinker(sbi);
2616
2617         err = f2fs_build_stats(sbi);
2618         if (err)
2619                 goto free_nm;
2620
2621         /* read root inode and dentry */
2622         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
2623         if (IS_ERR(root)) {
2624                 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
2625                 err = PTR_ERR(root);
2626                 goto free_node_inode;
2627         }
2628         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2629                 iput(root);
2630                 err = -EINVAL;
2631                 goto free_node_inode;
2632         }
2633
2634         sb->s_root = d_make_root(root); /* allocate root dentry */
2635         if (!sb->s_root) {
2636                 err = -ENOMEM;
2637                 goto free_root_inode;
2638         }
2639
2640         err = f2fs_register_sysfs(sbi);
2641         if (err)
2642                 goto free_root_inode;
2643
2644         /* if there are nt orphan nodes free them */
2645         err = recover_orphan_inodes(sbi);
2646         if (err)
2647                 goto free_sysfs;
2648
2649         /* recover fsynced data */
2650         if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
2651                 /*
2652                  * mount should be failed, when device has readonly mode, and
2653                  * previous checkpoint was not done by clean system shutdown.
2654                  */
2655                 if (bdev_read_only(sb->s_bdev) &&
2656                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
2657                         err = -EROFS;
2658                         goto free_meta;
2659                 }
2660
2661                 if (need_fsck)
2662                         set_sbi_flag(sbi, SBI_NEED_FSCK);
2663
2664                 if (!retry)
2665                         goto skip_recovery;
2666
2667                 err = recover_fsync_data(sbi, false);
2668                 if (err < 0) {
2669                         need_fsck = true;
2670                         f2fs_msg(sb, KERN_ERR,
2671                                 "Cannot recover all fsync data errno=%d", err);
2672                         goto free_meta;
2673                 }
2674         } else {
2675                 err = recover_fsync_data(sbi, true);
2676
2677                 if (!f2fs_readonly(sb) && err > 0) {
2678                         err = -EINVAL;
2679                         f2fs_msg(sb, KERN_ERR,
2680                                 "Need to recover fsync data");
2681                         goto free_sysfs;
2682                 }
2683         }
2684 skip_recovery:
2685         /* recover_fsync_data() cleared this already */
2686         clear_sbi_flag(sbi, SBI_POR_DOING);
2687
2688         /*
2689          * If filesystem is not mounted as read-only then
2690          * do start the gc_thread.
2691          */
2692         if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
2693                 /* After POR, we can run background GC thread.*/
2694                 err = start_gc_thread(sbi);
2695                 if (err)
2696                         goto free_meta;
2697         }
2698         kfree(options);
2699
2700         /* recover broken superblock */
2701         if (recovery) {
2702                 err = f2fs_commit_super(sbi, true);
2703                 f2fs_msg(sb, KERN_INFO,
2704                         "Try to recover %dth superblock, ret: %d",
2705                         sbi->valid_super_block ? 1 : 2, err);
2706         }
2707
2708         f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx",
2709                                 cur_cp_version(F2FS_CKPT(sbi)));
2710         f2fs_update_time(sbi, CP_TIME);
2711         f2fs_update_time(sbi, REQ_TIME);
2712         return 0;
2713
2714 free_meta:
2715         f2fs_sync_inode_meta(sbi);
2716         /*
2717          * Some dirty meta pages can be produced by recover_orphan_inodes()
2718          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
2719          * followed by write_checkpoint() through f2fs_write_node_pages(), which
2720          * falls into an infinite loop in sync_meta_pages().
2721          */
2722         truncate_inode_pages_final(META_MAPPING(sbi));
2723 free_sysfs:
2724         f2fs_unregister_sysfs(sbi);
2725 free_root_inode:
2726         dput(sb->s_root);
2727         sb->s_root = NULL;
2728 free_node_inode:
2729         truncate_inode_pages_final(NODE_MAPPING(sbi));
2730         mutex_lock(&sbi->umount_mutex);
2731         release_ino_entry(sbi, true);
2732         f2fs_leave_shrinker(sbi);
2733         iput(sbi->node_inode);
2734         mutex_unlock(&sbi->umount_mutex);
2735         f2fs_destroy_stats(sbi);
2736 free_nm:
2737         destroy_node_manager(sbi);
2738 free_sm:
2739         destroy_segment_manager(sbi);
2740 free_devices:
2741         destroy_device_list(sbi);
2742         kfree(sbi->ckpt);
2743 free_meta_inode:
2744         make_bad_inode(sbi->meta_inode);
2745         iput(sbi->meta_inode);
2746 free_io_dummy:
2747         mempool_destroy(sbi->write_io_dummy);
2748 free_options:
2749         for (i = 0; i < NR_PAGE_TYPE; i++)
2750                 kfree(sbi->write_io[i]);
2751         destroy_percpu_info(sbi);
2752 #ifdef CONFIG_QUOTA
2753         for (i = 0; i < MAXQUOTAS; i++)
2754                 kfree(sbi->s_qf_names[i]);
2755 #endif
2756         kfree(options);
2757 free_sb_buf:
2758         kfree(raw_super);
2759 free_sbi:
2760         if (sbi->s_chksum_driver)
2761                 crypto_free_shash(sbi->s_chksum_driver);
2762         kfree(sbi);
2763
2764         /* give only one another chance */
2765         if (retry) {
2766                 retry = false;
2767                 shrink_dcache_sb(sb);
2768                 goto try_onemore;
2769         }
2770         return err;
2771 }
2772
2773 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
2774                         const char *dev_name, void *data)
2775 {
2776         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
2777 }
2778
2779 static void kill_f2fs_super(struct super_block *sb)
2780 {
2781         if (sb->s_root) {
2782                 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
2783                 stop_gc_thread(F2FS_SB(sb));
2784                 stop_discard_thread(F2FS_SB(sb));
2785         }
2786         kill_block_super(sb);
2787 }
2788
2789 static struct file_system_type f2fs_fs_type = {
2790         .owner          = THIS_MODULE,
2791         .name           = "f2fs",
2792         .mount          = f2fs_mount,
2793         .kill_sb        = kill_f2fs_super,
2794         .fs_flags       = FS_REQUIRES_DEV,
2795 };
2796 MODULE_ALIAS_FS("f2fs");
2797
2798 static int __init init_inodecache(void)
2799 {
2800         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
2801                         sizeof(struct f2fs_inode_info), 0,
2802                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
2803         if (!f2fs_inode_cachep)
2804                 return -ENOMEM;
2805         return 0;
2806 }
2807
2808 static void destroy_inodecache(void)
2809 {
2810         /*
2811          * Make sure all delayed rcu free inodes are flushed before we
2812          * destroy cache.
2813          */
2814         rcu_barrier();
2815         kmem_cache_destroy(f2fs_inode_cachep);
2816 }
2817
2818 static int __init init_f2fs_fs(void)
2819 {
2820         int err;
2821
2822         if (PAGE_SIZE != F2FS_BLKSIZE) {
2823                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
2824                                 PAGE_SIZE, F2FS_BLKSIZE);
2825                 return -EINVAL;
2826         }
2827
2828         f2fs_build_trace_ios();
2829
2830         err = init_inodecache();
2831         if (err)
2832                 goto fail;
2833         err = create_node_manager_caches();
2834         if (err)
2835                 goto free_inodecache;
2836         err = create_segment_manager_caches();
2837         if (err)
2838                 goto free_node_manager_caches;
2839         err = create_checkpoint_caches();
2840         if (err)
2841                 goto free_segment_manager_caches;
2842         err = create_extent_cache();
2843         if (err)
2844                 goto free_checkpoint_caches;
2845         err = f2fs_init_sysfs();
2846         if (err)
2847                 goto free_extent_cache;
2848         err = register_shrinker(&f2fs_shrinker_info);
2849         if (err)
2850                 goto free_sysfs;
2851         err = register_filesystem(&f2fs_fs_type);
2852         if (err)
2853                 goto free_shrinker;
2854         err = f2fs_create_root_stats();
2855         if (err)
2856                 goto free_filesystem;
2857         return 0;
2858
2859 free_filesystem:
2860         unregister_filesystem(&f2fs_fs_type);
2861 free_shrinker:
2862         unregister_shrinker(&f2fs_shrinker_info);
2863 free_sysfs:
2864         f2fs_exit_sysfs();
2865 free_extent_cache:
2866         destroy_extent_cache();
2867 free_checkpoint_caches:
2868         destroy_checkpoint_caches();
2869 free_segment_manager_caches:
2870         destroy_segment_manager_caches();
2871 free_node_manager_caches:
2872         destroy_node_manager_caches();
2873 free_inodecache:
2874         destroy_inodecache();
2875 fail:
2876         return err;
2877 }
2878
2879 static void __exit exit_f2fs_fs(void)
2880 {
2881         f2fs_destroy_root_stats();
2882         unregister_filesystem(&f2fs_fs_type);
2883         unregister_shrinker(&f2fs_shrinker_info);
2884         f2fs_exit_sysfs();
2885         destroy_extent_cache();
2886         destroy_checkpoint_caches();
2887         destroy_segment_manager_caches();
2888         destroy_node_manager_caches();
2889         destroy_inodecache();
2890         f2fs_destroy_trace_ios();
2891 }
2892
2893 module_init(init_f2fs_fs)
2894 module_exit(exit_f2fs_fs)
2895
2896 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
2897 MODULE_DESCRIPTION("Flash Friendly File System");
2898 MODULE_LICENSE("GPL");
2899 MODULE_SOFTDEP("pre: crc32");
2900