fs/f2fs/super.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * fs/f2fs/super.c
   4  *
   5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   6  *             http://www.samsung.com/
   7  */
   8 #include <linux/module.h>
   9 #include <linux/init.h>
  10 #include <linux/fs.h>
  11 #include <linux/fs_context.h>
  12 #include <linux/sched/mm.h>
  13 #include <linux/statfs.h>
  14 #include <linux/buffer_head.h>
  15 #include <linux/kthread.h>
  16 #include <linux/parser.h>
  17 #include <linux/mount.h>
  18 #include <linux/seq_file.h>
  19 #include <linux/proc_fs.h>
  20 #include <linux/random.h>
  21 #include <linux/exportfs.h>
  22 #include <linux/blkdev.h>
  23 #include <linux/quotaops.h>
  24 #include <linux/f2fs_fs.h>
  25 #include <linux/sysfs.h>
  26 #include <linux/quota.h>
  27 #include <linux/unicode.h>
  28 #include <linux/part_stat.h>
  29 #include <linux/zstd.h>
  30 #include <linux/lz4.h>
  31
  32 #include "f2fs.h"
  33 #include "node.h"
  34 #include "segment.h"
  35 #include "xattr.h"
  36 #include "gc.h"
  37 #include "iostat.h"
  38
  39 #define CREATE_TRACE_POINTS
  40 #include <trace/events/f2fs.h>
  41
  42 static struct kmem_cache *f2fs_inode_cachep;
  43
  44 #ifdef CONFIG_F2FS_FAULT_INJECTION
  45
  46 const char *f2fs_fault_name[FAULT_MAX] = {
  47         [FAULT_KMALLOC]         = "kmalloc",
  48         [FAULT_KVMALLOC]        = "kvmalloc",
  49         [FAULT_PAGE_ALLOC]      = "page alloc",
  50         [FAULT_PAGE_GET]        = "page get",
  51         [FAULT_ALLOC_NID]       = "alloc nid",
  52         [FAULT_ORPHAN]          = "orphan",
  53         [FAULT_BLOCK]           = "no more block",
  54         [FAULT_DIR_DEPTH]       = "too big dir depth",
  55         [FAULT_EVICT_INODE]     = "evict_inode fail",
  56         [FAULT_TRUNCATE]        = "truncate fail",
  57         [FAULT_READ_IO]         = "read IO error",
  58         [FAULT_CHECKPOINT]      = "checkpoint error",
  59         [FAULT_DISCARD]         = "discard error",
  60         [FAULT_WRITE_IO]        = "write IO error",
  61         [FAULT_SLAB_ALLOC]      = "slab alloc",
  62         [FAULT_DQUOT_INIT]      = "dquot initialize",
  63         [FAULT_LOCK_OP]         = "lock_op",
  64         [FAULT_BLKADDR]         = "invalid blkaddr",
  65 };
  66
  67 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  68                                                         unsigned int type)
  69 {
  70         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  71
  72         if (rate) {
  73                 atomic_set(&ffi->inject_ops, 0);
  74                 ffi->inject_rate = rate;
  75         }
  76
  77         if (type)
  78                 ffi->inject_type = type;
  79
  80         if (!rate && !type)
  81                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  82 }
  83 #endif
  84
  85 /* f2fs-wide shrinker description */
  86 static struct shrinker f2fs_shrinker_info = {
  87         .scan_objects = f2fs_shrink_scan,
  88         .count_objects = f2fs_shrink_count,
  89         .seeks = DEFAULT_SEEKS,
  90 };
  91
  92 enum {
  93         Opt_gc_background,
  94         Opt_disable_roll_forward,
  95         Opt_norecovery,
  96         Opt_discard,
  97         Opt_nodiscard,
  98         Opt_noheap,
  99         Opt_heap,
 100         Opt_user_xattr,
 101         Opt_nouser_xattr,
 102         Opt_acl,
 103         Opt_noacl,
 104         Opt_active_logs,
 105         Opt_disable_ext_identify,
 106         Opt_inline_xattr,
 107         Opt_noinline_xattr,
 108         Opt_inline_xattr_size,
 109         Opt_inline_data,
 110         Opt_inline_dentry,
 111         Opt_noinline_dentry,
 112         Opt_flush_merge,
 113         Opt_noflush_merge,
 114         Opt_barrier,
 115         Opt_nobarrier,
 116         Opt_fastboot,
 117         Opt_extent_cache,
 118         Opt_noextent_cache,
 119         Opt_noinline_data,
 120         Opt_data_flush,
 121         Opt_reserve_root,
 122         Opt_resgid,
 123         Opt_resuid,
 124         Opt_mode,
 125         Opt_io_size_bits,
 126         Opt_fault_injection,
 127         Opt_fault_type,
 128         Opt_lazytime,
 129         Opt_nolazytime,
 130         Opt_quota,
 131         Opt_noquota,
 132         Opt_usrquota,
 133         Opt_grpquota,
 134         Opt_prjquota,
 135         Opt_usrjquota,
 136         Opt_grpjquota,
 137         Opt_prjjquota,
 138         Opt_offusrjquota,
 139         Opt_offgrpjquota,
 140         Opt_offprjjquota,
 141         Opt_jqfmt_vfsold,
 142         Opt_jqfmt_vfsv0,
 143         Opt_jqfmt_vfsv1,
 144         Opt_alloc,
 145         Opt_fsync,
 146         Opt_test_dummy_encryption,
 147         Opt_inlinecrypt,
 148         Opt_checkpoint_disable,
 149         Opt_checkpoint_disable_cap,
 150         Opt_checkpoint_disable_cap_perc,
 151         Opt_checkpoint_enable,
 152         Opt_checkpoint_merge,
 153         Opt_nocheckpoint_merge,
 154         Opt_compress_algorithm,
 155         Opt_compress_log_size,
 156         Opt_compress_extension,
 157         Opt_nocompress_extension,
 158         Opt_compress_chksum,
 159         Opt_compress_mode,
 160         Opt_compress_cache,
 161         Opt_atgc,
 162         Opt_gc_merge,
 163         Opt_nogc_merge,
 164         Opt_discard_unit,
 165         Opt_memory_mode,
 166         Opt_age_extent_cache,
 167         Opt_errors,
 168         Opt_err,
 169 };
 170
 171 static match_table_t f2fs_tokens = {
 172         {Opt_gc_background, "background_gc=%s"},
 173         {Opt_disable_roll_forward, "disable_roll_forward"},
 174         {Opt_norecovery, "norecovery"},
 175         {Opt_discard, "discard"},
 176         {Opt_nodiscard, "nodiscard"},
 177         {Opt_noheap, "no_heap"},
 178         {Opt_heap, "heap"},
 179         {Opt_user_xattr, "user_xattr"},
 180         {Opt_nouser_xattr, "nouser_xattr"},
 181         {Opt_acl, "acl"},
 182         {Opt_noacl, "noacl"},
 183         {Opt_active_logs, "active_logs=%u"},
 184         {Opt_disable_ext_identify, "disable_ext_identify"},
 185         {Opt_inline_xattr, "inline_xattr"},
 186         {Opt_noinline_xattr, "noinline_xattr"},
 187         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 188         {Opt_inline_data, "inline_data"},
 189         {Opt_inline_dentry, "inline_dentry"},
 190         {Opt_noinline_dentry, "noinline_dentry"},
 191         {Opt_flush_merge, "flush_merge"},
 192         {Opt_noflush_merge, "noflush_merge"},
 193         {Opt_barrier, "barrier"},
 194         {Opt_nobarrier, "nobarrier"},
 195         {Opt_fastboot, "fastboot"},
 196         {Opt_extent_cache, "extent_cache"},
 197         {Opt_noextent_cache, "noextent_cache"},
 198         {Opt_noinline_data, "noinline_data"},
 199         {Opt_data_flush, "data_flush"},
 200         {Opt_reserve_root, "reserve_root=%u"},
 201         {Opt_resgid, "resgid=%u"},
 202         {Opt_resuid, "resuid=%u"},
 203         {Opt_mode, "mode=%s"},
 204         {Opt_io_size_bits, "io_bits=%u"},
 205         {Opt_fault_injection, "fault_injection=%u"},
 206         {Opt_fault_type, "fault_type=%u"},
 207         {Opt_lazytime, "lazytime"},
 208         {Opt_nolazytime, "nolazytime"},
 209         {Opt_quota, "quota"},
 210         {Opt_noquota, "noquota"},
 211         {Opt_usrquota, "usrquota"},
 212         {Opt_grpquota, "grpquota"},
 213         {Opt_prjquota, "prjquota"},
 214         {Opt_usrjquota, "usrjquota=%s"},
 215         {Opt_grpjquota, "grpjquota=%s"},
 216         {Opt_prjjquota, "prjjquota=%s"},
 217         {Opt_offusrjquota, "usrjquota="},
 218         {Opt_offgrpjquota, "grpjquota="},
 219         {Opt_offprjjquota, "prjjquota="},
 220         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 221         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 222         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 223         {Opt_alloc, "alloc_mode=%s"},
 224         {Opt_fsync, "fsync_mode=%s"},
 225         {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
 226         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 227         {Opt_inlinecrypt, "inlinecrypt"},
 228         {Opt_checkpoint_disable, "checkpoint=disable"},
 229         {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 230         {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 231         {Opt_checkpoint_enable, "checkpoint=enable"},
 232         {Opt_checkpoint_merge, "checkpoint_merge"},
 233         {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
 234         {Opt_compress_algorithm, "compress_algorithm=%s"},
 235         {Opt_compress_log_size, "compress_log_size=%u"},
 236         {Opt_compress_extension, "compress_extension=%s"},
 237         {Opt_nocompress_extension, "nocompress_extension=%s"},
 238         {Opt_compress_chksum, "compress_chksum"},
 239         {Opt_compress_mode, "compress_mode=%s"},
 240         {Opt_compress_cache, "compress_cache"},
 241         {Opt_atgc, "atgc"},
 242         {Opt_gc_merge, "gc_merge"},
 243         {Opt_nogc_merge, "nogc_merge"},
 244         {Opt_discard_unit, "discard_unit=%s"},
 245         {Opt_memory_mode, "memory=%s"},
 246         {Opt_age_extent_cache, "age_extent_cache"},
 247         {Opt_errors, "errors=%s"},
 248         {Opt_err, NULL},
 249 };
 250
 251 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 252 {
 253         struct va_format vaf;
 254         va_list args;
 255         int level;
 256
 257         va_start(args, fmt);
 258
 259         level = printk_get_level(fmt);
 260         vaf.fmt = printk_skip_level(fmt);
 261         vaf.va = &args;
 262         printk("%c%cF2FS-fs (%s): %pV\n",
 263                KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 264
 265         va_end(args);
 266 }
 267
 268 #if IS_ENABLED(CONFIG_UNICODE)
 269 static const struct f2fs_sb_encodings {
 270         __u16 magic;
 271         char *name;
 272         unsigned int version;
 273 } f2fs_sb_encoding_map[] = {
 274         {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
 275 };
 276
 277 static const struct f2fs_sb_encodings *
 278 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
 279 {
 280         __u16 magic = le16_to_cpu(sb->s_encoding);
 281         int i;
 282
 283         for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
 284                 if (magic == f2fs_sb_encoding_map[i].magic)
 285                         return &f2fs_sb_encoding_map[i];
 286
 287         return NULL;
 288 }
 289
 290 struct kmem_cache *f2fs_cf_name_slab;
 291 static int __init f2fs_create_casefold_cache(void)
 292 {
 293         f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
 294                                                         F2FS_NAME_LEN);
 295         return f2fs_cf_name_slab ? 0 : -ENOMEM;
 296 }
 297
 298 static void f2fs_destroy_casefold_cache(void)
 299 {
 300         kmem_cache_destroy(f2fs_cf_name_slab);
 301 }
 302 #else
 303 static int __init f2fs_create_casefold_cache(void) { return 0; }
 304 static void f2fs_destroy_casefold_cache(void) { }
 305 #endif
 306
 307 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 308 {
 309         block_t limit = min((sbi->user_block_count >> 3),
 310                         sbi->user_block_count - sbi->reserved_blocks);
 311
 312         /* limit is 12.5% */
 313         if (test_opt(sbi, RESERVE_ROOT) &&
 314                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 315                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 316                 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 317                           F2FS_OPTION(sbi).root_reserved_blocks);
 318         }
 319         if (!test_opt(sbi, RESERVE_ROOT) &&
 320                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 321                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 322                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 323                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 324                 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 325                           from_kuid_munged(&init_user_ns,
 326                                            F2FS_OPTION(sbi).s_resuid),
 327                           from_kgid_munged(&init_user_ns,
 328                                            F2FS_OPTION(sbi).s_resgid));
 329 }
 330
 331 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
 332 {
 333         unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
 334         unsigned int avg_vblocks;
 335         unsigned int wanted_reserved_segments;
 336         block_t avail_user_block_count;
 337
 338         if (!F2FS_IO_ALIGNED(sbi))
 339                 return 0;
 340
 341         /* average valid block count in section in worst case */
 342         avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
 343
 344         /*
 345          * we need enough free space when migrating one section in worst case
 346          */
 347         wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
 348                                                 reserved_segments(sbi);
 349         wanted_reserved_segments -= reserved_segments(sbi);
 350
 351         avail_user_block_count = sbi->user_block_count -
 352                                 sbi->current_reserved_blocks -
 353                                 F2FS_OPTION(sbi).root_reserved_blocks;
 354
 355         if (wanted_reserved_segments * sbi->blocks_per_seg >
 356                                         avail_user_block_count) {
 357                 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
 358                         wanted_reserved_segments,
 359                         avail_user_block_count >> sbi->log_blocks_per_seg);
 360                 return -ENOSPC;
 361         }
 362
 363         SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
 364
 365         f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
 366                          wanted_reserved_segments);
 367
 368         return 0;
 369 }
 370
 371 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
 372 {
 373         if (!F2FS_OPTION(sbi).unusable_cap_perc)
 374                 return;
 375
 376         if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
 377                 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
 378         else
 379                 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
 380                                         F2FS_OPTION(sbi).unusable_cap_perc;
 381
 382         f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
 383                         F2FS_OPTION(sbi).unusable_cap,
 384                         F2FS_OPTION(sbi).unusable_cap_perc);
 385 }
 386
 387 static void init_once(void *foo)
 388 {
 389         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 390
 391         inode_init_once(&fi->vfs_inode);
 392 }
 393
 394 #ifdef CONFIG_QUOTA
 395 static const char * const quotatypes[] = INITQFNAMES;
 396 #define QTYPE2NAME(t) (quotatypes[t])
 397 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 398                                                         substring_t *args)
 399 {
 400         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 401         char *qname;
 402         int ret = -EINVAL;
 403
 404         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 405                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 406                 return -EINVAL;
 407         }
 408         if (f2fs_sb_has_quota_ino(sbi)) {
 409                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 410                 return 0;
 411         }
 412
 413         qname = match_strdup(args);
 414         if (!qname) {
 415                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 416                 return -ENOMEM;
 417         }
 418         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 419                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 420                         ret = 0;
 421                 else
 422                         f2fs_err(sbi, "%s quota file already specified",
 423                                  QTYPE2NAME(qtype));
 424                 goto errout;
 425         }
 426         if (strchr(qname, '/')) {
 427                 f2fs_err(sbi, "quotafile must be on filesystem root");
 428                 goto errout;
 429         }
 430         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 431         set_opt(sbi, QUOTA);
 432         return 0;
 433 errout:
 434         kfree(qname);
 435         return ret;
 436 }
 437
 438 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 439 {
 440         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 441
 442         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 443                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 444                 return -EINVAL;
 445         }
 446         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 447         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 448         return 0;
 449 }
 450
 451 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 452 {
 453         /*
 454          * We do the test below only for project quotas. 'usrquota' and
 455          * 'grpquota' mount options are allowed even without quota feature
 456          * to support legacy quotas in quota files.
 457          */
 458         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 459                 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 460                 return -1;
 461         }
 462         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 463                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 464                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 465                 if (test_opt(sbi, USRQUOTA) &&
 466                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 467                         clear_opt(sbi, USRQUOTA);
 468
 469                 if (test_opt(sbi, GRPQUOTA) &&
 470                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 471                         clear_opt(sbi, GRPQUOTA);
 472
 473                 if (test_opt(sbi, PRJQUOTA) &&
 474                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 475                         clear_opt(sbi, PRJQUOTA);
 476
 477                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 478                                 test_opt(sbi, PRJQUOTA)) {
 479                         f2fs_err(sbi, "old and new quota format mixing");
 480                         return -1;
 481                 }
 482
 483                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 484                         f2fs_err(sbi, "journaled quota format not specified");
 485                         return -1;
 486                 }
 487         }
 488
 489         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 490                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 491                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 492         }
 493         return 0;
 494 }
 495 #endif
 496
 497 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
 498                                           const char *opt,
 499                                           const substring_t *arg,
 500                                           bool is_remount)
 501 {
 502         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 503         struct fs_parameter param = {
 504                 .type = fs_value_is_string,
 505                 .string = arg->from ? arg->from : "",
 506         };
 507         struct fscrypt_dummy_policy *policy =
 508                 &F2FS_OPTION(sbi).dummy_enc_policy;
 509         int err;
 510
 511         if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
 512                 f2fs_warn(sbi, "test_dummy_encryption option not supported");
 513                 return -EINVAL;
 514         }
 515
 516         if (!f2fs_sb_has_encrypt(sbi)) {
 517                 f2fs_err(sbi, "Encrypt feature is off");
 518                 return -EINVAL;
 519         }
 520
 521         /*
 522          * This mount option is just for testing, and it's not worthwhile to
 523          * implement the extra complexity (e.g. RCU protection) that would be
 524          * needed to allow it to be set or changed during remount.  We do allow
 525          * it to be specified during remount, but only if there is no change.
 526          */
 527         if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
 528                 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
 529                 return -EINVAL;
 530         }
 531
 532         err = fscrypt_parse_test_dummy_encryption(&param, policy);
 533         if (err) {
 534                 if (err == -EEXIST)
 535                         f2fs_warn(sbi,
 536                                   "Can't change test_dummy_encryption on remount");
 537                 else if (err == -EINVAL)
 538                         f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
 539                                   opt);
 540                 else
 541                         f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
 542                                   opt, err);
 543                 return -EINVAL;
 544         }
 545         f2fs_warn(sbi, "Test dummy encryption mode enabled");
 546         return 0;
 547 }
 548
 549 #ifdef CONFIG_F2FS_FS_COMPRESSION
 550 /*
 551  * 1. The same extension name cannot not appear in both compress and non-compress extension
 552  * at the same time.
 553  * 2. If the compress extension specifies all files, the types specified by the non-compress
 554  * extension will be treated as special cases and will not be compressed.
 555  * 3. Don't allow the non-compress extension specifies all files.
 556  */
 557 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
 558 {
 559         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 560         unsigned char (*noext)[F2FS_EXTENSION_LEN];
 561         int ext_cnt, noext_cnt, index = 0, no_index = 0;
 562
 563         ext = F2FS_OPTION(sbi).extensions;
 564         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
 565         noext = F2FS_OPTION(sbi).noextensions;
 566         noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
 567
 568         if (!noext_cnt)
 569                 return 0;
 570
 571         for (no_index = 0; no_index < noext_cnt; no_index++) {
 572                 if (!strcasecmp("*", noext[no_index])) {
 573                         f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
 574                         return -EINVAL;
 575                 }
 576                 for (index = 0; index < ext_cnt; index++) {
 577                         if (!strcasecmp(ext[index], noext[no_index])) {
 578                                 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
 579                                                 ext[index]);
 580                                 return -EINVAL;
 581                         }
 582                 }
 583         }
 584         return 0;
 585 }
 586
 587 #ifdef CONFIG_F2FS_FS_LZ4
 588 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
 589 {
 590 #ifdef CONFIG_F2FS_FS_LZ4HC
 591         unsigned int level;
 592
 593         if (strlen(str) == 3) {
 594                 F2FS_OPTION(sbi).compress_level = LZ4HC_DEFAULT_CLEVEL;
 595                 return 0;
 596         }
 597
 598         str += 3;
 599
 600         if (str[0] != ':') {
 601                 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
 602                 return -EINVAL;
 603         }
 604         if (kstrtouint(str + 1, 10, &level))
 605                 return -EINVAL;
 606
 607         if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
 608                 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
 609                 return -EINVAL;
 610         }
 611
 612         F2FS_OPTION(sbi).compress_level = level;
 613         return 0;
 614 #else
 615         if (strlen(str) == 3) {
 616                 F2FS_OPTION(sbi).compress_level = 0;
 617                 return 0;
 618         }
 619         f2fs_info(sbi, "kernel doesn't support lz4hc compression");
 620         return -EINVAL;
 621 #endif
 622 }
 623 #endif
 624
 625 #ifdef CONFIG_F2FS_FS_ZSTD
 626 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
 627 {
 628         unsigned int level;
 629         int len = 4;
 630
 631         if (strlen(str) == len) {
 632                 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
 633                 return 0;
 634         }
 635
 636         str += len;
 637
 638         if (str[0] != ':') {
 639                 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
 640                 return -EINVAL;
 641         }
 642         if (kstrtouint(str + 1, 10, &level))
 643                 return -EINVAL;
 644
 645         if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
 646                 f2fs_info(sbi, "invalid zstd compress level: %d", level);
 647                 return -EINVAL;
 648         }
 649
 650         F2FS_OPTION(sbi).compress_level = level;
 651         return 0;
 652 }
 653 #endif
 654 #endif
 655
 656 static int parse_options(struct super_block *sb, char *options, bool is_remount)
 657 {
 658         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 659         substring_t args[MAX_OPT_ARGS];
 660 #ifdef CONFIG_F2FS_FS_COMPRESSION
 661         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 662         unsigned char (*noext)[F2FS_EXTENSION_LEN];
 663         int ext_cnt, noext_cnt;
 664 #endif
 665         char *p, *name;
 666         int arg = 0;
 667         kuid_t uid;
 668         kgid_t gid;
 669         int ret;
 670
 671         if (!options)
 672                 goto default_check;
 673
 674         while ((p = strsep(&options, ",")) != NULL) {
 675                 int token;
 676
 677                 if (!*p)
 678                         continue;
 679                 /*
 680                  * Initialize args struct so we know whether arg was
 681                  * found; some options take optional arguments.
 682                  */
 683                 args[0].to = args[0].from = NULL;
 684                 token = match_token(p, f2fs_tokens, args);
 685
 686                 switch (token) {
 687                 case Opt_gc_background:
 688                         name = match_strdup(&args[0]);
 689
 690                         if (!name)
 691                                 return -ENOMEM;
 692                         if (!strcmp(name, "on")) {
 693                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
 694                         } else if (!strcmp(name, "off")) {
 695                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
 696                         } else if (!strcmp(name, "sync")) {
 697                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
 698                         } else {
 699                                 kfree(name);
 700                                 return -EINVAL;
 701                         }
 702                         kfree(name);
 703                         break;
 704                 case Opt_disable_roll_forward:
 705                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 706                         break;
 707                 case Opt_norecovery:
 708                         /* this option mounts f2fs with ro */
 709                         set_opt(sbi, NORECOVERY);
 710                         if (!f2fs_readonly(sb))
 711                                 return -EINVAL;
 712                         break;
 713                 case Opt_discard:
 714                         if (!f2fs_hw_support_discard(sbi)) {
 715                                 f2fs_warn(sbi, "device does not support discard");
 716                                 break;
 717                         }
 718                         set_opt(sbi, DISCARD);
 719                         break;
 720                 case Opt_nodiscard:
 721                         if (f2fs_hw_should_discard(sbi)) {
 722                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 723                                 return -EINVAL;
 724                         }
 725                         clear_opt(sbi, DISCARD);
 726                         break;
 727                 case Opt_noheap:
 728                         set_opt(sbi, NOHEAP);
 729                         break;
 730                 case Opt_heap:
 731                         clear_opt(sbi, NOHEAP);
 732                         break;
 733 #ifdef CONFIG_F2FS_FS_XATTR
 734                 case Opt_user_xattr:
 735                         set_opt(sbi, XATTR_USER);
 736                         break;
 737                 case Opt_nouser_xattr:
 738                         clear_opt(sbi, XATTR_USER);
 739                         break;
 740                 case Opt_inline_xattr:
 741                         set_opt(sbi, INLINE_XATTR);
 742                         break;
 743                 case Opt_noinline_xattr:
 744                         clear_opt(sbi, INLINE_XATTR);
 745                         break;
 746                 case Opt_inline_xattr_size:
 747                         if (args->from && match_int(args, &arg))
 748                                 return -EINVAL;
 749                         set_opt(sbi, INLINE_XATTR_SIZE);
 750                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 751                         break;
 752 #else
 753                 case Opt_user_xattr:
 754                         f2fs_info(sbi, "user_xattr options not supported");
 755                         break;
 756                 case Opt_nouser_xattr:
 757                         f2fs_info(sbi, "nouser_xattr options not supported");
 758                         break;
 759                 case Opt_inline_xattr:
 760                         f2fs_info(sbi, "inline_xattr options not supported");
 761                         break;
 762                 case Opt_noinline_xattr:
 763                         f2fs_info(sbi, "noinline_xattr options not supported");
 764                         break;
 765 #endif
 766 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 767                 case Opt_acl:
 768                         set_opt(sbi, POSIX_ACL);
 769                         break;
 770                 case Opt_noacl:
 771                         clear_opt(sbi, POSIX_ACL);
 772                         break;
 773 #else
 774                 case Opt_acl:
 775                         f2fs_info(sbi, "acl options not supported");
 776                         break;
 777                 case Opt_noacl:
 778                         f2fs_info(sbi, "noacl options not supported");
 779                         break;
 780 #endif
 781                 case Opt_active_logs:
 782                         if (args->from && match_int(args, &arg))
 783                                 return -EINVAL;
 784                         if (arg != 2 && arg != 4 &&
 785                                 arg != NR_CURSEG_PERSIST_TYPE)
 786                                 return -EINVAL;
 787                         F2FS_OPTION(sbi).active_logs = arg;
 788                         break;
 789                 case Opt_disable_ext_identify:
 790                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 791                         break;
 792                 case Opt_inline_data:
 793                         set_opt(sbi, INLINE_DATA);
 794                         break;
 795                 case Opt_inline_dentry:
 796                         set_opt(sbi, INLINE_DENTRY);
 797                         break;
 798                 case Opt_noinline_dentry:
 799                         clear_opt(sbi, INLINE_DENTRY);
 800                         break;
 801                 case Opt_flush_merge:
 802                         set_opt(sbi, FLUSH_MERGE);
 803                         break;
 804                 case Opt_noflush_merge:
 805                         clear_opt(sbi, FLUSH_MERGE);
 806                         break;
 807                 case Opt_nobarrier:
 808                         set_opt(sbi, NOBARRIER);
 809                         break;
 810                 case Opt_barrier:
 811                         clear_opt(sbi, NOBARRIER);
 812                         break;
 813                 case Opt_fastboot:
 814                         set_opt(sbi, FASTBOOT);
 815                         break;
 816                 case Opt_extent_cache:
 817                         set_opt(sbi, READ_EXTENT_CACHE);
 818                         break;
 819                 case Opt_noextent_cache:
 820                         clear_opt(sbi, READ_EXTENT_CACHE);
 821                         break;
 822                 case Opt_noinline_data:
 823                         clear_opt(sbi, INLINE_DATA);
 824                         break;
 825                 case Opt_data_flush:
 826                         set_opt(sbi, DATA_FLUSH);
 827                         break;
 828                 case Opt_reserve_root:
 829                         if (args->from && match_int(args, &arg))
 830                                 return -EINVAL;
 831                         if (test_opt(sbi, RESERVE_ROOT)) {
 832                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 833                                           F2FS_OPTION(sbi).root_reserved_blocks);
 834                         } else {
 835                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 836                                 set_opt(sbi, RESERVE_ROOT);
 837                         }
 838                         break;
 839                 case Opt_resuid:
 840                         if (args->from && match_int(args, &arg))
 841                                 return -EINVAL;
 842                         uid = make_kuid(current_user_ns(), arg);
 843                         if (!uid_valid(uid)) {
 844                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 845                                 return -EINVAL;
 846                         }
 847                         F2FS_OPTION(sbi).s_resuid = uid;
 848                         break;
 849                 case Opt_resgid:
 850                         if (args->from && match_int(args, &arg))
 851                                 return -EINVAL;
 852                         gid = make_kgid(current_user_ns(), arg);
 853                         if (!gid_valid(gid)) {
 854                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 855                                 return -EINVAL;
 856                         }
 857                         F2FS_OPTION(sbi).s_resgid = gid;
 858                         break;
 859                 case Opt_mode:
 860                         name = match_strdup(&args[0]);
 861
 862                         if (!name)
 863                                 return -ENOMEM;
 864                         if (!strcmp(name, "adaptive")) {
 865                                 if (f2fs_sb_has_blkzoned(sbi)) {
 866                                         f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
 867                                         kfree(name);
 868                                         return -EINVAL;
 869                                 }
 870                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
 871                         } else if (!strcmp(name, "lfs")) {
 872                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
 873                         } else if (!strcmp(name, "fragment:segment")) {
 874                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
 875                         } else if (!strcmp(name, "fragment:block")) {
 876                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
 877                         } else {
 878                                 kfree(name);
 879                                 return -EINVAL;
 880                         }
 881                         kfree(name);
 882                         break;
 883                 case Opt_io_size_bits:
 884                         if (args->from && match_int(args, &arg))
 885                                 return -EINVAL;
 886                         if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
 887                                 f2fs_warn(sbi, "Not support %ld, larger than %d",
 888                                         BIT(arg), BIO_MAX_VECS);
 889                                 return -EINVAL;
 890                         }
 891                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 892                         break;
 893 #ifdef CONFIG_F2FS_FAULT_INJECTION
 894                 case Opt_fault_injection:
 895                         if (args->from && match_int(args, &arg))
 896                                 return -EINVAL;
 897                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 898                         set_opt(sbi, FAULT_INJECTION);
 899                         break;
 900
 901                 case Opt_fault_type:
 902                         if (args->from && match_int(args, &arg))
 903                                 return -EINVAL;
 904                         f2fs_build_fault_attr(sbi, 0, arg);
 905                         set_opt(sbi, FAULT_INJECTION);
 906                         break;
 907 #else
 908                 case Opt_fault_injection:
 909                         f2fs_info(sbi, "fault_injection options not supported");
 910                         break;
 911
 912                 case Opt_fault_type:
 913                         f2fs_info(sbi, "fault_type options not supported");
 914                         break;
 915 #endif
 916                 case Opt_lazytime:
 917                         sb->s_flags |= SB_LAZYTIME;
 918                         break;
 919                 case Opt_nolazytime:
 920                         sb->s_flags &= ~SB_LAZYTIME;
 921                         break;
 922 #ifdef CONFIG_QUOTA
 923                 case Opt_quota:
 924                 case Opt_usrquota:
 925                         set_opt(sbi, USRQUOTA);
 926                         break;
 927                 case Opt_grpquota:
 928                         set_opt(sbi, GRPQUOTA);
 929                         break;
 930                 case Opt_prjquota:
 931                         set_opt(sbi, PRJQUOTA);
 932                         break;
 933                 case Opt_usrjquota:
 934                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 935                         if (ret)
 936                                 return ret;
 937                         break;
 938                 case Opt_grpjquota:
 939                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 940                         if (ret)
 941                                 return ret;
 942                         break;
 943                 case Opt_prjjquota:
 944                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 945                         if (ret)
 946                                 return ret;
 947                         break;
 948                 case Opt_offusrjquota:
 949                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 950                         if (ret)
 951                                 return ret;
 952                         break;
 953                 case Opt_offgrpjquota:
 954                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 955                         if (ret)
 956                                 return ret;
 957                         break;
 958                 case Opt_offprjjquota:
 959                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 960                         if (ret)
 961                                 return ret;
 962                         break;
 963                 case Opt_jqfmt_vfsold:
 964                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 965                         break;
 966                 case Opt_jqfmt_vfsv0:
 967                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 968                         break;
 969                 case Opt_jqfmt_vfsv1:
 970                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 971                         break;
 972                 case Opt_noquota:
 973                         clear_opt(sbi, QUOTA);
 974                         clear_opt(sbi, USRQUOTA);
 975                         clear_opt(sbi, GRPQUOTA);
 976                         clear_opt(sbi, PRJQUOTA);
 977                         break;
 978 #else
 979                 case Opt_quota:
 980                 case Opt_usrquota:
 981                 case Opt_grpquota:
 982                 case Opt_prjquota:
 983                 case Opt_usrjquota:
 984                 case Opt_grpjquota:
 985                 case Opt_prjjquota:
 986                 case Opt_offusrjquota:
 987                 case Opt_offgrpjquota:
 988                 case Opt_offprjjquota:
 989                 case Opt_jqfmt_vfsold:
 990                 case Opt_jqfmt_vfsv0:
 991                 case Opt_jqfmt_vfsv1:
 992                 case Opt_noquota:
 993                         f2fs_info(sbi, "quota operations not supported");
 994                         break;
 995 #endif
 996                 case Opt_alloc:
 997                         name = match_strdup(&args[0]);
 998                         if (!name)
 999                                 return -ENOMEM;
1000
1001                         if (!strcmp(name, "default")) {
1002                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1003                         } else if (!strcmp(name, "reuse")) {
1004                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1005                         } else {
1006                                 kfree(name);
1007                                 return -EINVAL;
1008                         }
1009                         kfree(name);
1010                         break;
1011                 case Opt_fsync:
1012                         name = match_strdup(&args[0]);
1013                         if (!name)
1014                                 return -ENOMEM;
1015                         if (!strcmp(name, "posix")) {
1016                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1017                         } else if (!strcmp(name, "strict")) {
1018                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1019                         } else if (!strcmp(name, "nobarrier")) {
1020                                 F2FS_OPTION(sbi).fsync_mode =
1021                                                         FSYNC_MODE_NOBARRIER;
1022                         } else {
1023                                 kfree(name);
1024                                 return -EINVAL;
1025                         }
1026                         kfree(name);
1027                         break;
1028                 case Opt_test_dummy_encryption:
1029                         ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1030                                                              is_remount);
1031                         if (ret)
1032                                 return ret;
1033                         break;
1034                 case Opt_inlinecrypt:
1035 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1036                         sb->s_flags |= SB_INLINECRYPT;
1037 #else
1038                         f2fs_info(sbi, "inline encryption not supported");
1039 #endif
1040                         break;
1041                 case Opt_checkpoint_disable_cap_perc:
1042                         if (args->from && match_int(args, &arg))
1043                                 return -EINVAL;
1044                         if (arg < 0 || arg > 100)
1045                                 return -EINVAL;
1046                         F2FS_OPTION(sbi).unusable_cap_perc = arg;
1047                         set_opt(sbi, DISABLE_CHECKPOINT);
1048                         break;
1049                 case Opt_checkpoint_disable_cap:
1050                         if (args->from && match_int(args, &arg))
1051                                 return -EINVAL;
1052                         F2FS_OPTION(sbi).unusable_cap = arg;
1053                         set_opt(sbi, DISABLE_CHECKPOINT);
1054                         break;
1055                 case Opt_checkpoint_disable:
1056                         set_opt(sbi, DISABLE_CHECKPOINT);
1057                         break;
1058                 case Opt_checkpoint_enable:
1059                         clear_opt(sbi, DISABLE_CHECKPOINT);
1060                         break;
1061                 case Opt_checkpoint_merge:
1062                         set_opt(sbi, MERGE_CHECKPOINT);
1063                         break;
1064                 case Opt_nocheckpoint_merge:
1065                         clear_opt(sbi, MERGE_CHECKPOINT);
1066                         break;
1067 #ifdef CONFIG_F2FS_FS_COMPRESSION
1068                 case Opt_compress_algorithm:
1069                         if (!f2fs_sb_has_compression(sbi)) {
1070                                 f2fs_info(sbi, "Image doesn't support compression");
1071                                 break;
1072                         }
1073                         name = match_strdup(&args[0]);
1074                         if (!name)
1075                                 return -ENOMEM;
1076                         if (!strcmp(name, "lzo")) {
1077 #ifdef CONFIG_F2FS_FS_LZO
1078                                 F2FS_OPTION(sbi).compress_level = 0;
1079                                 F2FS_OPTION(sbi).compress_algorithm =
1080                                                                 COMPRESS_LZO;
1081 #else
1082                                 f2fs_info(sbi, "kernel doesn't support lzo compression");
1083 #endif
1084                         } else if (!strncmp(name, "lz4", 3)) {
1085 #ifdef CONFIG_F2FS_FS_LZ4
1086                                 ret = f2fs_set_lz4hc_level(sbi, name);
1087                                 if (ret) {
1088                                         kfree(name);
1089                                         return -EINVAL;
1090                                 }
1091                                 F2FS_OPTION(sbi).compress_algorithm =
1092                                                                 COMPRESS_LZ4;
1093 #else
1094                                 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1095 #endif
1096                         } else if (!strncmp(name, "zstd", 4)) {
1097 #ifdef CONFIG_F2FS_FS_ZSTD
1098                                 ret = f2fs_set_zstd_level(sbi, name);
1099                                 if (ret) {
1100                                         kfree(name);
1101                                         return -EINVAL;
1102                                 }
1103                                 F2FS_OPTION(sbi).compress_algorithm =
1104                                                                 COMPRESS_ZSTD;
1105 #else
1106                                 f2fs_info(sbi, "kernel doesn't support zstd compression");
1107 #endif
1108                         } else if (!strcmp(name, "lzo-rle")) {
1109 #ifdef CONFIG_F2FS_FS_LZORLE
1110                                 F2FS_OPTION(sbi).compress_level = 0;
1111                                 F2FS_OPTION(sbi).compress_algorithm =
1112                                                                 COMPRESS_LZORLE;
1113 #else
1114                                 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1115 #endif
1116                         } else {
1117                                 kfree(name);
1118                                 return -EINVAL;
1119                         }
1120                         kfree(name);
1121                         break;
1122                 case Opt_compress_log_size:
1123                         if (!f2fs_sb_has_compression(sbi)) {
1124                                 f2fs_info(sbi, "Image doesn't support compression");
1125                                 break;
1126                         }
1127                         if (args->from && match_int(args, &arg))
1128                                 return -EINVAL;
1129                         if (arg < MIN_COMPRESS_LOG_SIZE ||
1130                                 arg > MAX_COMPRESS_LOG_SIZE) {
1131                                 f2fs_err(sbi,
1132                                         "Compress cluster log size is out of range");
1133                                 return -EINVAL;
1134                         }
1135                         F2FS_OPTION(sbi).compress_log_size = arg;
1136                         break;
1137                 case Opt_compress_extension:
1138                         if (!f2fs_sb_has_compression(sbi)) {
1139                                 f2fs_info(sbi, "Image doesn't support compression");
1140                                 break;
1141                         }
1142                         name = match_strdup(&args[0]);
1143                         if (!name)
1144                                 return -ENOMEM;
1145
1146                         ext = F2FS_OPTION(sbi).extensions;
1147                         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1148
1149                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
1150                                 ext_cnt >= COMPRESS_EXT_NUM) {
1151                                 f2fs_err(sbi,
1152                                         "invalid extension length/number");
1153                                 kfree(name);
1154                                 return -EINVAL;
1155                         }
1156
1157                         strcpy(ext[ext_cnt], name);
1158                         F2FS_OPTION(sbi).compress_ext_cnt++;
1159                         kfree(name);
1160                         break;
1161                 case Opt_nocompress_extension:
1162                         if (!f2fs_sb_has_compression(sbi)) {
1163                                 f2fs_info(sbi, "Image doesn't support compression");
1164                                 break;
1165                         }
1166                         name = match_strdup(&args[0]);
1167                         if (!name)
1168                                 return -ENOMEM;
1169
1170                         noext = F2FS_OPTION(sbi).noextensions;
1171                         noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1172
1173                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
1174                                 noext_cnt >= COMPRESS_EXT_NUM) {
1175                                 f2fs_err(sbi,
1176                                         "invalid extension length/number");
1177                                 kfree(name);
1178                                 return -EINVAL;
1179                         }
1180
1181                         strcpy(noext[noext_cnt], name);
1182                         F2FS_OPTION(sbi).nocompress_ext_cnt++;
1183                         kfree(name);
1184                         break;
1185                 case Opt_compress_chksum:
1186                         if (!f2fs_sb_has_compression(sbi)) {
1187                                 f2fs_info(sbi, "Image doesn't support compression");
1188                                 break;
1189                         }
1190                         F2FS_OPTION(sbi).compress_chksum = true;
1191                         break;
1192                 case Opt_compress_mode:
1193                         if (!f2fs_sb_has_compression(sbi)) {
1194                                 f2fs_info(sbi, "Image doesn't support compression");
1195                                 break;
1196                         }
1197                         name = match_strdup(&args[0]);
1198                         if (!name)
1199                                 return -ENOMEM;
1200                         if (!strcmp(name, "fs")) {
1201                                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1202                         } else if (!strcmp(name, "user")) {
1203                                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1204                         } else {
1205                                 kfree(name);
1206                                 return -EINVAL;
1207                         }
1208                         kfree(name);
1209                         break;
1210                 case Opt_compress_cache:
1211                         if (!f2fs_sb_has_compression(sbi)) {
1212                                 f2fs_info(sbi, "Image doesn't support compression");
1213                                 break;
1214                         }
1215                         set_opt(sbi, COMPRESS_CACHE);
1216                         break;
1217 #else
1218                 case Opt_compress_algorithm:
1219                 case Opt_compress_log_size:
1220                 case Opt_compress_extension:
1221                 case Opt_nocompress_extension:
1222                 case Opt_compress_chksum:
1223                 case Opt_compress_mode:
1224                 case Opt_compress_cache:
1225                         f2fs_info(sbi, "compression options not supported");
1226                         break;
1227 #endif
1228                 case Opt_atgc:
1229                         set_opt(sbi, ATGC);
1230                         break;
1231                 case Opt_gc_merge:
1232                         set_opt(sbi, GC_MERGE);
1233                         break;
1234                 case Opt_nogc_merge:
1235                         clear_opt(sbi, GC_MERGE);
1236                         break;
1237                 case Opt_discard_unit:
1238                         name = match_strdup(&args[0]);
1239                         if (!name)
1240                                 return -ENOMEM;
1241                         if (!strcmp(name, "block")) {
1242                                 F2FS_OPTION(sbi).discard_unit =
1243                                                 DISCARD_UNIT_BLOCK;
1244                         } else if (!strcmp(name, "segment")) {
1245                                 F2FS_OPTION(sbi).discard_unit =
1246                                                 DISCARD_UNIT_SEGMENT;
1247                         } else if (!strcmp(name, "section")) {
1248                                 F2FS_OPTION(sbi).discard_unit =
1249                                                 DISCARD_UNIT_SECTION;
1250                         } else {
1251                                 kfree(name);
1252                                 return -EINVAL;
1253                         }
1254                         kfree(name);
1255                         break;
1256                 case Opt_memory_mode:
1257                         name = match_strdup(&args[0]);
1258                         if (!name)
1259                                 return -ENOMEM;
1260                         if (!strcmp(name, "normal")) {
1261                                 F2FS_OPTION(sbi).memory_mode =
1262                                                 MEMORY_MODE_NORMAL;
1263                         } else if (!strcmp(name, "low")) {
1264                                 F2FS_OPTION(sbi).memory_mode =
1265                                                 MEMORY_MODE_LOW;
1266                         } else {
1267                                 kfree(name);
1268                                 return -EINVAL;
1269                         }
1270                         kfree(name);
1271                         break;
1272                 case Opt_age_extent_cache:
1273                         set_opt(sbi, AGE_EXTENT_CACHE);
1274                         break;
1275                 case Opt_errors:
1276                         name = match_strdup(&args[0]);
1277                         if (!name)
1278                                 return -ENOMEM;
1279                         if (!strcmp(name, "remount-ro")) {
1280                                 F2FS_OPTION(sbi).errors =
1281                                                 MOUNT_ERRORS_READONLY;
1282                         } else if (!strcmp(name, "continue")) {
1283                                 F2FS_OPTION(sbi).errors =
1284                                                 MOUNT_ERRORS_CONTINUE;
1285                         } else if (!strcmp(name, "panic")) {
1286                                 F2FS_OPTION(sbi).errors =
1287                                                 MOUNT_ERRORS_PANIC;
1288                         } else {
1289                                 kfree(name);
1290                                 return -EINVAL;
1291                         }
1292                         kfree(name);
1293                         break;
1294                 default:
1295                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1296                                  p);
1297                         return -EINVAL;
1298                 }
1299         }
1300 default_check:
1301 #ifdef CONFIG_QUOTA
1302         if (f2fs_check_quota_options(sbi))
1303                 return -EINVAL;
1304 #else
1305         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1306                 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1307                 return -EINVAL;
1308         }
1309         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1310                 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1311                 return -EINVAL;
1312         }
1313 #endif
1314 #if !IS_ENABLED(CONFIG_UNICODE)
1315         if (f2fs_sb_has_casefold(sbi)) {
1316                 f2fs_err(sbi,
1317                         "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1318                 return -EINVAL;
1319         }
1320 #endif
1321         /*
1322          * The BLKZONED feature indicates that the drive was formatted with
1323          * zone alignment optimization. This is optional for host-aware
1324          * devices, but mandatory for host-managed zoned block devices.
1325          */
1326         if (f2fs_sb_has_blkzoned(sbi)) {
1327 #ifdef CONFIG_BLK_DEV_ZONED
1328                 if (F2FS_OPTION(sbi).discard_unit !=
1329                                                 DISCARD_UNIT_SECTION) {
1330                         f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1331                         F2FS_OPTION(sbi).discard_unit =
1332                                         DISCARD_UNIT_SECTION;
1333                 }
1334 #else
1335                 f2fs_err(sbi, "Zoned block device support is not enabled");
1336                 return -EINVAL;
1337 #endif
1338         }
1339
1340 #ifdef CONFIG_F2FS_FS_COMPRESSION
1341         if (f2fs_test_compress_extension(sbi)) {
1342                 f2fs_err(sbi, "invalid compress or nocompress extension");
1343                 return -EINVAL;
1344         }
1345 #endif
1346
1347         if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1348                 f2fs_err(sbi, "Should set mode=lfs with %luKB-sized IO",
1349                          F2FS_IO_SIZE_KB(sbi));
1350                 return -EINVAL;
1351         }
1352
1353         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1354                 int min_size, max_size;
1355
1356                 if (!f2fs_sb_has_extra_attr(sbi) ||
1357                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1358                         f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1359                         return -EINVAL;
1360                 }
1361                 if (!test_opt(sbi, INLINE_XATTR)) {
1362                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1363                         return -EINVAL;
1364                 }
1365
1366                 min_size = MIN_INLINE_XATTR_SIZE;
1367                 max_size = MAX_INLINE_XATTR_SIZE;
1368
1369                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1370                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1371                         f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1372                                  min_size, max_size);
1373                         return -EINVAL;
1374                 }
1375         }
1376
1377         if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1378                 f2fs_err(sbi, "LFS is not compatible with checkpoint=disable");
1379                 return -EINVAL;
1380         }
1381
1382         if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1383                 f2fs_err(sbi, "LFS is not compatible with ATGC");
1384                 return -EINVAL;
1385         }
1386
1387         if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1388                 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1389                 return -EINVAL;
1390         }
1391
1392         if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1393                 f2fs_err(sbi, "Allow to mount readonly mode only");
1394                 return -EROFS;
1395         }
1396         return 0;
1397 }
1398
1399 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1400 {
1401         struct f2fs_inode_info *fi;
1402
1403         if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1404                 return NULL;
1405
1406         fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1407         if (!fi)
1408                 return NULL;
1409
1410         init_once((void *) fi);
1411
1412         /* Initialize f2fs-specific inode info */
1413         atomic_set(&fi->dirty_pages, 0);
1414         atomic_set(&fi->i_compr_blocks, 0);
1415         init_f2fs_rwsem(&fi->i_sem);
1416         spin_lock_init(&fi->i_size_lock);
1417         INIT_LIST_HEAD(&fi->dirty_list);
1418         INIT_LIST_HEAD(&fi->gdirty_list);
1419         init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1420         init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1421         init_f2fs_rwsem(&fi->i_xattr_sem);
1422
1423         /* Will be used by directory only */
1424         fi->i_dir_level = F2FS_SB(sb)->dir_level;
1425
1426         return &fi->vfs_inode;
1427 }
1428
1429 static int f2fs_drop_inode(struct inode *inode)
1430 {
1431         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1432         int ret;
1433
1434         /*
1435          * during filesystem shutdown, if checkpoint is disabled,
1436          * drop useless meta/node dirty pages.
1437          */
1438         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1439                 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1440                         inode->i_ino == F2FS_META_INO(sbi)) {
1441                         trace_f2fs_drop_inode(inode, 1);
1442                         return 1;
1443                 }
1444         }
1445
1446         /*
1447          * This is to avoid a deadlock condition like below.
1448          * writeback_single_inode(inode)
1449          *  - f2fs_write_data_page
1450          *    - f2fs_gc -> iput -> evict
1451          *       - inode_wait_for_writeback(inode)
1452          */
1453         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1454                 if (!inode->i_nlink && !is_bad_inode(inode)) {
1455                         /* to avoid evict_inode call simultaneously */
1456                         atomic_inc(&inode->i_count);
1457                         spin_unlock(&inode->i_lock);
1458
1459                         /* should remain fi->extent_tree for writepage */
1460                         f2fs_destroy_extent_node(inode);
1461
1462                         sb_start_intwrite(inode->i_sb);
1463                         f2fs_i_size_write(inode, 0);
1464
1465                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1466                                         inode, NULL, 0, DATA);
1467                         truncate_inode_pages_final(inode->i_mapping);
1468
1469                         if (F2FS_HAS_BLOCKS(inode))
1470                                 f2fs_truncate(inode);
1471
1472                         sb_end_intwrite(inode->i_sb);
1473
1474                         spin_lock(&inode->i_lock);
1475                         atomic_dec(&inode->i_count);
1476                 }
1477                 trace_f2fs_drop_inode(inode, 0);
1478                 return 0;
1479         }
1480         ret = generic_drop_inode(inode);
1481         if (!ret)
1482                 ret = fscrypt_drop_inode(inode);
1483         trace_f2fs_drop_inode(inode, ret);
1484         return ret;
1485 }
1486
1487 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1488 {
1489         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1490         int ret = 0;
1491
1492         spin_lock(&sbi->inode_lock[DIRTY_META]);
1493         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1494                 ret = 1;
1495         } else {
1496                 set_inode_flag(inode, FI_DIRTY_INODE);
1497                 stat_inc_dirty_inode(sbi, DIRTY_META);
1498         }
1499         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1500                 list_add_tail(&F2FS_I(inode)->gdirty_list,
1501                                 &sbi->inode_list[DIRTY_META]);
1502                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1503         }
1504         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1505         return ret;
1506 }
1507
1508 void f2fs_inode_synced(struct inode *inode)
1509 {
1510         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1511
1512         spin_lock(&sbi->inode_lock[DIRTY_META]);
1513         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1514                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1515                 return;
1516         }
1517         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1518                 list_del_init(&F2FS_I(inode)->gdirty_list);
1519                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1520         }
1521         clear_inode_flag(inode, FI_DIRTY_INODE);
1522         clear_inode_flag(inode, FI_AUTO_RECOVER);
1523         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1524         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1525 }
1526
1527 /*
1528  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1529  *
1530  * We should call set_dirty_inode to write the dirty inode through write_inode.
1531  */
1532 static void f2fs_dirty_inode(struct inode *inode, int flags)
1533 {
1534         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1535
1536         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1537                         inode->i_ino == F2FS_META_INO(sbi))
1538                 return;
1539
1540         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1541                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1542
1543         f2fs_inode_dirtied(inode, false);
1544 }
1545
1546 static void f2fs_free_inode(struct inode *inode)
1547 {
1548         fscrypt_free_inode(inode);
1549         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1550 }
1551
1552 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1553 {
1554         percpu_counter_destroy(&sbi->total_valid_inode_count);
1555         percpu_counter_destroy(&sbi->rf_node_block_count);
1556         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1557 }
1558
1559 static void destroy_device_list(struct f2fs_sb_info *sbi)
1560 {
1561         int i;
1562
1563         for (i = 0; i < sbi->s_ndevs; i++) {
1564                 blkdev_put(FDEV(i).bdev, sbi->sb->s_type);
1565 #ifdef CONFIG_BLK_DEV_ZONED
1566                 kvfree(FDEV(i).blkz_seq);
1567 #endif
1568         }
1569         kvfree(sbi->devs);
1570 }
1571
1572 static void f2fs_put_super(struct super_block *sb)
1573 {
1574         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1575         int i;
1576         int err = 0;
1577         bool done;
1578
1579         /* unregister procfs/sysfs entries in advance to avoid race case */
1580         f2fs_unregister_sysfs(sbi);
1581
1582         f2fs_quota_off_umount(sb);
1583
1584         /* prevent remaining shrinker jobs */
1585         mutex_lock(&sbi->umount_mutex);
1586
1587         /*
1588          * flush all issued checkpoints and stop checkpoint issue thread.
1589          * after then, all checkpoints should be done by each process context.
1590          */
1591         f2fs_stop_ckpt_thread(sbi);
1592
1593         /*
1594          * We don't need to do checkpoint when superblock is clean.
1595          * But, the previous checkpoint was not done by umount, it needs to do
1596          * clean checkpoint again.
1597          */
1598         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1599                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1600                 struct cp_control cpc = {
1601                         .reason = CP_UMOUNT,
1602                 };
1603                 err = f2fs_write_checkpoint(sbi, &cpc);
1604         }
1605
1606         /* be sure to wait for any on-going discard commands */
1607         done = f2fs_issue_discard_timeout(sbi);
1608         if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1609                 struct cp_control cpc = {
1610                         .reason = CP_UMOUNT | CP_TRIMMED,
1611                 };
1612                 err = f2fs_write_checkpoint(sbi, &cpc);
1613         }
1614
1615         /*
1616          * normally superblock is clean, so we need to release this.
1617          * In addition, EIO will skip do checkpoint, we need this as well.
1618          */
1619         f2fs_release_ino_entry(sbi, true);
1620
1621         f2fs_leave_shrinker(sbi);
1622         mutex_unlock(&sbi->umount_mutex);
1623
1624         /* our cp_error case, we can wait for any writeback page */
1625         f2fs_flush_merged_writes(sbi);
1626
1627         f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1628
1629         if (err) {
1630                 truncate_inode_pages_final(NODE_MAPPING(sbi));
1631                 truncate_inode_pages_final(META_MAPPING(sbi));
1632         }
1633
1634         for (i = 0; i < NR_COUNT_TYPE; i++) {
1635                 if (!get_pages(sbi, i))
1636                         continue;
1637                 f2fs_err(sbi, "detect filesystem reference count leak during "
1638                         "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1639                 f2fs_bug_on(sbi, 1);
1640         }
1641
1642         f2fs_bug_on(sbi, sbi->fsync_node_num);
1643
1644         f2fs_destroy_compress_inode(sbi);
1645
1646         iput(sbi->node_inode);
1647         sbi->node_inode = NULL;
1648
1649         iput(sbi->meta_inode);
1650         sbi->meta_inode = NULL;
1651
1652         /*
1653          * iput() can update stat information, if f2fs_write_checkpoint()
1654          * above failed with error.
1655          */
1656         f2fs_destroy_stats(sbi);
1657
1658         /* destroy f2fs internal modules */
1659         f2fs_destroy_node_manager(sbi);
1660         f2fs_destroy_segment_manager(sbi);
1661
1662         /* flush s_error_work before sbi destroy */
1663         flush_work(&sbi->s_error_work);
1664
1665         f2fs_destroy_post_read_wq(sbi);
1666
1667         kvfree(sbi->ckpt);
1668
1669         sb->s_fs_info = NULL;
1670         if (sbi->s_chksum_driver)
1671                 crypto_free_shash(sbi->s_chksum_driver);
1672         kfree(sbi->raw_super);
1673
1674         destroy_device_list(sbi);
1675         f2fs_destroy_page_array_cache(sbi);
1676         f2fs_destroy_xattr_caches(sbi);
1677         mempool_destroy(sbi->write_io_dummy);
1678 #ifdef CONFIG_QUOTA
1679         for (i = 0; i < MAXQUOTAS; i++)
1680                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1681 #endif
1682         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1683         destroy_percpu_info(sbi);
1684         f2fs_destroy_iostat(sbi);
1685         for (i = 0; i < NR_PAGE_TYPE; i++)
1686                 kvfree(sbi->write_io[i]);
1687 #if IS_ENABLED(CONFIG_UNICODE)
1688         utf8_unload(sb->s_encoding);
1689 #endif
1690         kfree(sbi);
1691 }
1692
1693 int f2fs_sync_fs(struct super_block *sb, int sync)
1694 {
1695         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1696         int err = 0;
1697
1698         if (unlikely(f2fs_cp_error(sbi)))
1699                 return 0;
1700         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1701                 return 0;
1702
1703         trace_f2fs_sync_fs(sb, sync);
1704
1705         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1706                 return -EAGAIN;
1707
1708         if (sync)
1709                 err = f2fs_issue_checkpoint(sbi);
1710
1711         return err;
1712 }
1713
1714 static int f2fs_freeze(struct super_block *sb)
1715 {
1716         if (f2fs_readonly(sb))
1717                 return 0;
1718
1719         /* IO error happened before */
1720         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1721                 return -EIO;
1722
1723         /* must be clean, since sync_filesystem() was already called */
1724         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1725                 return -EINVAL;
1726
1727         /* Let's flush checkpoints and stop the thread. */
1728         f2fs_flush_ckpt_thread(F2FS_SB(sb));
1729
1730         /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1731         set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1732         return 0;
1733 }
1734
1735 static int f2fs_unfreeze(struct super_block *sb)
1736 {
1737         clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1738         return 0;
1739 }
1740
1741 #ifdef CONFIG_QUOTA
1742 static int f2fs_statfs_project(struct super_block *sb,
1743                                 kprojid_t projid, struct kstatfs *buf)
1744 {
1745         struct kqid qid;
1746         struct dquot *dquot;
1747         u64 limit;
1748         u64 curblock;
1749
1750         qid = make_kqid_projid(projid);
1751         dquot = dqget(sb, qid);
1752         if (IS_ERR(dquot))
1753                 return PTR_ERR(dquot);
1754         spin_lock(&dquot->dq_dqb_lock);
1755
1756         limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1757                                         dquot->dq_dqb.dqb_bhardlimit);
1758         if (limit)
1759                 limit >>= sb->s_blocksize_bits;
1760
1761         if (limit && buf->f_blocks > limit) {
1762                 curblock = (dquot->dq_dqb.dqb_curspace +
1763                             dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1764                 buf->f_blocks = limit;
1765                 buf->f_bfree = buf->f_bavail =
1766                         (buf->f_blocks > curblock) ?
1767                          (buf->f_blocks - curblock) : 0;
1768         }
1769
1770         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1771                                         dquot->dq_dqb.dqb_ihardlimit);
1772
1773         if (limit && buf->f_files > limit) {
1774                 buf->f_files = limit;
1775                 buf->f_ffree =
1776                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1777                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1778         }
1779
1780         spin_unlock(&dquot->dq_dqb_lock);
1781         dqput(dquot);
1782         return 0;
1783 }
1784 #endif
1785
1786 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1787 {
1788         struct super_block *sb = dentry->d_sb;
1789         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1790         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1791         block_t total_count, user_block_count, start_count;
1792         u64 avail_node_count;
1793         unsigned int total_valid_node_count;
1794
1795         total_count = le64_to_cpu(sbi->raw_super->block_count);
1796         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1797         buf->f_type = F2FS_SUPER_MAGIC;
1798         buf->f_bsize = sbi->blocksize;
1799
1800         buf->f_blocks = total_count - start_count;
1801
1802         spin_lock(&sbi->stat_lock);
1803
1804         user_block_count = sbi->user_block_count;
1805         total_valid_node_count = valid_node_count(sbi);
1806         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1807         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1808                                                 sbi->current_reserved_blocks;
1809
1810         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1811                 buf->f_bfree = 0;
1812         else
1813                 buf->f_bfree -= sbi->unusable_block_count;
1814         spin_unlock(&sbi->stat_lock);
1815
1816         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1817                 buf->f_bavail = buf->f_bfree -
1818                                 F2FS_OPTION(sbi).root_reserved_blocks;
1819         else
1820                 buf->f_bavail = 0;
1821
1822         if (avail_node_count > user_block_count) {
1823                 buf->f_files = user_block_count;
1824                 buf->f_ffree = buf->f_bavail;
1825         } else {
1826                 buf->f_files = avail_node_count;
1827                 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1828                                         buf->f_bavail);
1829         }
1830
1831         buf->f_namelen = F2FS_NAME_LEN;
1832         buf->f_fsid    = u64_to_fsid(id);
1833
1834 #ifdef CONFIG_QUOTA
1835         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1836                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1837                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1838         }
1839 #endif
1840         return 0;
1841 }
1842
1843 static inline void f2fs_show_quota_options(struct seq_file *seq,
1844                                            struct super_block *sb)
1845 {
1846 #ifdef CONFIG_QUOTA
1847         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1848
1849         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1850                 char *fmtname = "";
1851
1852                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1853                 case QFMT_VFS_OLD:
1854                         fmtname = "vfsold";
1855                         break;
1856                 case QFMT_VFS_V0:
1857                         fmtname = "vfsv0";
1858                         break;
1859                 case QFMT_VFS_V1:
1860                         fmtname = "vfsv1";
1861                         break;
1862                 }
1863                 seq_printf(seq, ",jqfmt=%s", fmtname);
1864         }
1865
1866         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1867                 seq_show_option(seq, "usrjquota",
1868                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1869
1870         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1871                 seq_show_option(seq, "grpjquota",
1872                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1873
1874         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1875                 seq_show_option(seq, "prjjquota",
1876                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1877 #endif
1878 }
1879
1880 #ifdef CONFIG_F2FS_FS_COMPRESSION
1881 static inline void f2fs_show_compress_options(struct seq_file *seq,
1882                                                         struct super_block *sb)
1883 {
1884         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1885         char *algtype = "";
1886         int i;
1887
1888         if (!f2fs_sb_has_compression(sbi))
1889                 return;
1890
1891         switch (F2FS_OPTION(sbi).compress_algorithm) {
1892         case COMPRESS_LZO:
1893                 algtype = "lzo";
1894                 break;
1895         case COMPRESS_LZ4:
1896                 algtype = "lz4";
1897                 break;
1898         case COMPRESS_ZSTD:
1899                 algtype = "zstd";
1900                 break;
1901         case COMPRESS_LZORLE:
1902                 algtype = "lzo-rle";
1903                 break;
1904         }
1905         seq_printf(seq, ",compress_algorithm=%s", algtype);
1906
1907         if (F2FS_OPTION(sbi).compress_level)
1908                 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1909
1910         seq_printf(seq, ",compress_log_size=%u",
1911                         F2FS_OPTION(sbi).compress_log_size);
1912
1913         for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1914                 seq_printf(seq, ",compress_extension=%s",
1915                         F2FS_OPTION(sbi).extensions[i]);
1916         }
1917
1918         for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1919                 seq_printf(seq, ",nocompress_extension=%s",
1920                         F2FS_OPTION(sbi).noextensions[i]);
1921         }
1922
1923         if (F2FS_OPTION(sbi).compress_chksum)
1924                 seq_puts(seq, ",compress_chksum");
1925
1926         if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1927                 seq_printf(seq, ",compress_mode=%s", "fs");
1928         else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1929                 seq_printf(seq, ",compress_mode=%s", "user");
1930
1931         if (test_opt(sbi, COMPRESS_CACHE))
1932                 seq_puts(seq, ",compress_cache");
1933 }
1934 #endif
1935
1936 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1937 {
1938         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1939
1940         if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1941                 seq_printf(seq, ",background_gc=%s", "sync");
1942         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1943                 seq_printf(seq, ",background_gc=%s", "on");
1944         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1945                 seq_printf(seq, ",background_gc=%s", "off");
1946
1947         if (test_opt(sbi, GC_MERGE))
1948                 seq_puts(seq, ",gc_merge");
1949         else
1950                 seq_puts(seq, ",nogc_merge");
1951
1952         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1953                 seq_puts(seq, ",disable_roll_forward");
1954         if (test_opt(sbi, NORECOVERY))
1955                 seq_puts(seq, ",norecovery");
1956         if (test_opt(sbi, DISCARD)) {
1957                 seq_puts(seq, ",discard");
1958                 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1959                         seq_printf(seq, ",discard_unit=%s", "block");
1960                 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1961                         seq_printf(seq, ",discard_unit=%s", "segment");
1962                 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1963                         seq_printf(seq, ",discard_unit=%s", "section");
1964         } else {
1965                 seq_puts(seq, ",nodiscard");
1966         }
1967         if (test_opt(sbi, NOHEAP))
1968                 seq_puts(seq, ",no_heap");
1969         else
1970                 seq_puts(seq, ",heap");
1971 #ifdef CONFIG_F2FS_FS_XATTR
1972         if (test_opt(sbi, XATTR_USER))
1973                 seq_puts(seq, ",user_xattr");
1974         else
1975                 seq_puts(seq, ",nouser_xattr");
1976         if (test_opt(sbi, INLINE_XATTR))
1977                 seq_puts(seq, ",inline_xattr");
1978         else
1979                 seq_puts(seq, ",noinline_xattr");
1980         if (test_opt(sbi, INLINE_XATTR_SIZE))
1981                 seq_printf(seq, ",inline_xattr_size=%u",
1982                                         F2FS_OPTION(sbi).inline_xattr_size);
1983 #endif
1984 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1985         if (test_opt(sbi, POSIX_ACL))
1986                 seq_puts(seq, ",acl");
1987         else
1988                 seq_puts(seq, ",noacl");
1989 #endif
1990         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1991                 seq_puts(seq, ",disable_ext_identify");
1992         if (test_opt(sbi, INLINE_DATA))
1993                 seq_puts(seq, ",inline_data");
1994         else
1995                 seq_puts(seq, ",noinline_data");
1996         if (test_opt(sbi, INLINE_DENTRY))
1997                 seq_puts(seq, ",inline_dentry");
1998         else
1999                 seq_puts(seq, ",noinline_dentry");
2000         if (test_opt(sbi, FLUSH_MERGE))
2001                 seq_puts(seq, ",flush_merge");
2002         else
2003                 seq_puts(seq, ",noflush_merge");
2004         if (test_opt(sbi, NOBARRIER))
2005                 seq_puts(seq, ",nobarrier");
2006         else
2007                 seq_puts(seq, ",barrier");
2008         if (test_opt(sbi, FASTBOOT))
2009                 seq_puts(seq, ",fastboot");
2010         if (test_opt(sbi, READ_EXTENT_CACHE))
2011                 seq_puts(seq, ",extent_cache");
2012         else
2013                 seq_puts(seq, ",noextent_cache");
2014         if (test_opt(sbi, AGE_EXTENT_CACHE))
2015                 seq_puts(seq, ",age_extent_cache");
2016         if (test_opt(sbi, DATA_FLUSH))
2017                 seq_puts(seq, ",data_flush");
2018
2019         seq_puts(seq, ",mode=");
2020         if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2021                 seq_puts(seq, "adaptive");
2022         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2023                 seq_puts(seq, "lfs");
2024         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2025                 seq_puts(seq, "fragment:segment");
2026         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2027                 seq_puts(seq, "fragment:block");
2028         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2029         if (test_opt(sbi, RESERVE_ROOT))
2030                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2031                                 F2FS_OPTION(sbi).root_reserved_blocks,
2032                                 from_kuid_munged(&init_user_ns,
2033                                         F2FS_OPTION(sbi).s_resuid),
2034                                 from_kgid_munged(&init_user_ns,
2035                                         F2FS_OPTION(sbi).s_resgid));
2036         if (F2FS_IO_SIZE_BITS(sbi))
2037                 seq_printf(seq, ",io_bits=%u",
2038                                 F2FS_OPTION(sbi).write_io_size_bits);
2039 #ifdef CONFIG_F2FS_FAULT_INJECTION
2040         if (test_opt(sbi, FAULT_INJECTION)) {
2041                 seq_printf(seq, ",fault_injection=%u",
2042                                 F2FS_OPTION(sbi).fault_info.inject_rate);
2043                 seq_printf(seq, ",fault_type=%u",
2044                                 F2FS_OPTION(sbi).fault_info.inject_type);
2045         }
2046 #endif
2047 #ifdef CONFIG_QUOTA
2048         if (test_opt(sbi, QUOTA))
2049                 seq_puts(seq, ",quota");
2050         if (test_opt(sbi, USRQUOTA))
2051                 seq_puts(seq, ",usrquota");
2052         if (test_opt(sbi, GRPQUOTA))
2053                 seq_puts(seq, ",grpquota");
2054         if (test_opt(sbi, PRJQUOTA))
2055                 seq_puts(seq, ",prjquota");
2056 #endif
2057         f2fs_show_quota_options(seq, sbi->sb);
2058
2059         fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2060
2061         if (sbi->sb->s_flags & SB_INLINECRYPT)
2062                 seq_puts(seq, ",inlinecrypt");
2063
2064         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2065                 seq_printf(seq, ",alloc_mode=%s", "default");
2066         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2067                 seq_printf(seq, ",alloc_mode=%s", "reuse");
2068
2069         if (test_opt(sbi, DISABLE_CHECKPOINT))
2070                 seq_printf(seq, ",checkpoint=disable:%u",
2071                                 F2FS_OPTION(sbi).unusable_cap);
2072         if (test_opt(sbi, MERGE_CHECKPOINT))
2073                 seq_puts(seq, ",checkpoint_merge");
2074         else
2075                 seq_puts(seq, ",nocheckpoint_merge");
2076         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2077                 seq_printf(seq, ",fsync_mode=%s", "posix");
2078         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2079                 seq_printf(seq, ",fsync_mode=%s", "strict");
2080         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2081                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2082
2083 #ifdef CONFIG_F2FS_FS_COMPRESSION
2084         f2fs_show_compress_options(seq, sbi->sb);
2085 #endif
2086
2087         if (test_opt(sbi, ATGC))
2088                 seq_puts(seq, ",atgc");
2089
2090         if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2091                 seq_printf(seq, ",memory=%s", "normal");
2092         else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2093                 seq_printf(seq, ",memory=%s", "low");
2094
2095         if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2096                 seq_printf(seq, ",errors=%s", "remount-ro");
2097         else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2098                 seq_printf(seq, ",errors=%s", "continue");
2099         else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2100                 seq_printf(seq, ",errors=%s", "panic");
2101
2102         return 0;
2103 }
2104
2105 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2106 {
2107         /* init some FS parameters */
2108         if (!remount) {
2109                 set_opt(sbi, READ_EXTENT_CACHE);
2110                 clear_opt(sbi, DISABLE_CHECKPOINT);
2111
2112                 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2113                         set_opt(sbi, DISCARD);
2114
2115                 if (f2fs_sb_has_blkzoned(sbi))
2116                         F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2117                 else
2118                         F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2119         }
2120
2121         if (f2fs_sb_has_readonly(sbi))
2122                 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2123         else
2124                 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2125
2126         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2127         if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2128                                                         SMALL_VOLUME_SEGMENTS)
2129                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2130         else
2131                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2132         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2133         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2134         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2135         if (f2fs_sb_has_compression(sbi)) {
2136                 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2137                 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2138                 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2139                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2140         }
2141         F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2142         F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2143         F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2144
2145         sbi->sb->s_flags &= ~SB_INLINECRYPT;
2146
2147         set_opt(sbi, INLINE_XATTR);
2148         set_opt(sbi, INLINE_DATA);
2149         set_opt(sbi, INLINE_DENTRY);
2150         set_opt(sbi, NOHEAP);
2151         set_opt(sbi, MERGE_CHECKPOINT);
2152         F2FS_OPTION(sbi).unusable_cap = 0;
2153         sbi->sb->s_flags |= SB_LAZYTIME;
2154         if (!f2fs_is_readonly(sbi))
2155                 set_opt(sbi, FLUSH_MERGE);
2156         if (f2fs_sb_has_blkzoned(sbi))
2157                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2158         else
2159                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2160
2161 #ifdef CONFIG_F2FS_FS_XATTR
2162         set_opt(sbi, XATTR_USER);
2163 #endif
2164 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2165         set_opt(sbi, POSIX_ACL);
2166 #endif
2167
2168         f2fs_build_fault_attr(sbi, 0, 0);
2169 }
2170
2171 #ifdef CONFIG_QUOTA
2172 static int f2fs_enable_quotas(struct super_block *sb);
2173 #endif
2174
2175 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2176 {
2177         unsigned int s_flags = sbi->sb->s_flags;
2178         struct cp_control cpc;
2179         unsigned int gc_mode = sbi->gc_mode;
2180         int err = 0;
2181         int ret;
2182         block_t unusable;
2183
2184         if (s_flags & SB_RDONLY) {
2185                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2186                 return -EINVAL;
2187         }
2188         sbi->sb->s_flags |= SB_ACTIVE;
2189
2190         /* check if we need more GC first */
2191         unusable = f2fs_get_unusable_blocks(sbi);
2192         if (!f2fs_disable_cp_again(sbi, unusable))
2193                 goto skip_gc;
2194
2195         f2fs_update_time(sbi, DISABLE_TIME);
2196
2197         sbi->gc_mode = GC_URGENT_HIGH;
2198
2199         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2200                 struct f2fs_gc_control gc_control = {
2201                         .victim_segno = NULL_SEGNO,
2202                         .init_gc_type = FG_GC,
2203                         .should_migrate_blocks = false,
2204                         .err_gc_skipped = true,
2205                         .nr_free_secs = 1 };
2206
2207                 f2fs_down_write(&sbi->gc_lock);
2208                 err = f2fs_gc(sbi, &gc_control);
2209                 if (err == -ENODATA) {
2210                         err = 0;
2211                         break;
2212                 }
2213                 if (err && err != -EAGAIN)
2214                         break;
2215         }
2216
2217         ret = sync_filesystem(sbi->sb);
2218         if (ret || err) {
2219                 err = ret ? ret : err;
2220                 goto restore_flag;
2221         }
2222
2223         unusable = f2fs_get_unusable_blocks(sbi);
2224         if (f2fs_disable_cp_again(sbi, unusable)) {
2225                 err = -EAGAIN;
2226                 goto restore_flag;
2227         }
2228
2229 skip_gc:
2230         f2fs_down_write(&sbi->gc_lock);
2231         cpc.reason = CP_PAUSE;
2232         set_sbi_flag(sbi, SBI_CP_DISABLED);
2233         err = f2fs_write_checkpoint(sbi, &cpc);
2234         if (err)
2235                 goto out_unlock;
2236
2237         spin_lock(&sbi->stat_lock);
2238         sbi->unusable_block_count = unusable;
2239         spin_unlock(&sbi->stat_lock);
2240
2241 out_unlock:
2242         f2fs_up_write(&sbi->gc_lock);
2243 restore_flag:
2244         sbi->gc_mode = gc_mode;
2245         sbi->sb->s_flags = s_flags;     /* Restore SB_RDONLY status */
2246         return err;
2247 }
2248
2249 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2250 {
2251         int retry = DEFAULT_RETRY_IO_COUNT;
2252
2253         /* we should flush all the data to keep data consistency */
2254         do {
2255                 sync_inodes_sb(sbi->sb);
2256                 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2257         } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2258
2259         if (unlikely(retry < 0))
2260                 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2261
2262         f2fs_down_write(&sbi->gc_lock);
2263         f2fs_dirty_to_prefree(sbi);
2264
2265         clear_sbi_flag(sbi, SBI_CP_DISABLED);
2266         set_sbi_flag(sbi, SBI_IS_DIRTY);
2267         f2fs_up_write(&sbi->gc_lock);
2268
2269         f2fs_sync_fs(sbi->sb, 1);
2270
2271         /* Let's ensure there's no pending checkpoint anymore */
2272         f2fs_flush_ckpt_thread(sbi);
2273 }
2274
2275 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2276 {
2277         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2278         struct f2fs_mount_info org_mount_opt;
2279         unsigned long old_sb_flags;
2280         int err;
2281         bool need_restart_gc = false, need_stop_gc = false;
2282         bool need_restart_ckpt = false, need_stop_ckpt = false;
2283         bool need_restart_flush = false, need_stop_flush = false;
2284         bool need_restart_discard = false, need_stop_discard = false;
2285         bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2286         bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2287         bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2288         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2289         bool no_atgc = !test_opt(sbi, ATGC);
2290         bool no_discard = !test_opt(sbi, DISCARD);
2291         bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2292         bool block_unit_discard = f2fs_block_unit_discard(sbi);
2293 #ifdef CONFIG_QUOTA
2294         int i, j;
2295 #endif
2296
2297         /*
2298          * Save the old mount options in case we
2299          * need to restore them.
2300          */
2301         org_mount_opt = sbi->mount_opt;
2302         old_sb_flags = sb->s_flags;
2303
2304 #ifdef CONFIG_QUOTA
2305         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2306         for (i = 0; i < MAXQUOTAS; i++) {
2307                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2308                         org_mount_opt.s_qf_names[i] =
2309                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2310                                 GFP_KERNEL);
2311                         if (!org_mount_opt.s_qf_names[i]) {
2312                                 for (j = 0; j < i; j++)
2313                                         kfree(org_mount_opt.s_qf_names[j]);
2314                                 return -ENOMEM;
2315                         }
2316                 } else {
2317                         org_mount_opt.s_qf_names[i] = NULL;
2318                 }
2319         }
2320 #endif
2321
2322         /* recover superblocks we couldn't write due to previous RO mount */
2323         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2324                 err = f2fs_commit_super(sbi, false);
2325                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2326                           err);
2327                 if (!err)
2328                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2329         }
2330
2331         default_options(sbi, true);
2332
2333         /* parse mount options */
2334         err = parse_options(sb, data, true);
2335         if (err)
2336                 goto restore_opts;
2337
2338         /* flush outstanding errors before changing fs state */
2339         flush_work(&sbi->s_error_work);
2340
2341         /*
2342          * Previous and new state of filesystem is RO,
2343          * so skip checking GC and FLUSH_MERGE conditions.
2344          */
2345         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2346                 goto skip;
2347
2348         if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2349                 err = -EROFS;
2350                 goto restore_opts;
2351         }
2352
2353 #ifdef CONFIG_QUOTA
2354         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2355                 err = dquot_suspend(sb, -1);
2356                 if (err < 0)
2357                         goto restore_opts;
2358         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2359                 /* dquot_resume needs RW */
2360                 sb->s_flags &= ~SB_RDONLY;
2361                 if (sb_any_quota_suspended(sb)) {
2362                         dquot_resume(sb, -1);
2363                 } else if (f2fs_sb_has_quota_ino(sbi)) {
2364                         err = f2fs_enable_quotas(sb);
2365                         if (err)
2366                                 goto restore_opts;
2367                 }
2368         }
2369 #endif
2370         if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2371                 err = -EINVAL;
2372                 f2fs_warn(sbi, "LFS is not compatible with IPU");
2373                 goto restore_opts;
2374         }
2375
2376         /* disallow enable atgc dynamically */
2377         if (no_atgc == !!test_opt(sbi, ATGC)) {
2378                 err = -EINVAL;
2379                 f2fs_warn(sbi, "switch atgc option is not allowed");
2380                 goto restore_opts;
2381         }
2382
2383         /* disallow enable/disable extent_cache dynamically */
2384         if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2385                 err = -EINVAL;
2386                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2387                 goto restore_opts;
2388         }
2389         /* disallow enable/disable age extent_cache dynamically */
2390         if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2391                 err = -EINVAL;
2392                 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2393                 goto restore_opts;
2394         }
2395
2396         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2397                 err = -EINVAL;
2398                 f2fs_warn(sbi, "switch io_bits option is not allowed");
2399                 goto restore_opts;
2400         }
2401
2402         if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2403                 err = -EINVAL;
2404                 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2405                 goto restore_opts;
2406         }
2407
2408         if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2409                 err = -EINVAL;
2410                 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2411                 goto restore_opts;
2412         }
2413
2414         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2415                 err = -EINVAL;
2416                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2417                 goto restore_opts;
2418         }
2419
2420         /*
2421          * We stop the GC thread if FS is mounted as RO
2422          * or if background_gc = off is passed in mount
2423          * option. Also sync the filesystem.
2424          */
2425         if ((*flags & SB_RDONLY) ||
2426                         (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2427                         !test_opt(sbi, GC_MERGE))) {
2428                 if (sbi->gc_thread) {
2429                         f2fs_stop_gc_thread(sbi);
2430                         need_restart_gc = true;
2431                 }
2432         } else if (!sbi->gc_thread) {
2433                 err = f2fs_start_gc_thread(sbi);
2434                 if (err)
2435                         goto restore_opts;
2436                 need_stop_gc = true;
2437         }
2438
2439         if (*flags & SB_RDONLY) {
2440                 sync_inodes_sb(sb);
2441
2442                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2443                 set_sbi_flag(sbi, SBI_IS_CLOSE);
2444                 f2fs_sync_fs(sb, 1);
2445                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2446         }
2447
2448         if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2449                         !test_opt(sbi, MERGE_CHECKPOINT)) {
2450                 f2fs_stop_ckpt_thread(sbi);
2451                 need_restart_ckpt = true;
2452         } else {
2453                 /* Flush if the prevous checkpoint, if exists. */
2454                 f2fs_flush_ckpt_thread(sbi);
2455
2456                 err = f2fs_start_ckpt_thread(sbi);
2457                 if (err) {
2458                         f2fs_err(sbi,
2459                             "Failed to start F2FS issue_checkpoint_thread (%d)",
2460                             err);
2461                         goto restore_gc;
2462                 }
2463                 need_stop_ckpt = true;
2464         }
2465
2466         /*
2467          * We stop issue flush thread if FS is mounted as RO
2468          * or if flush_merge is not passed in mount option.
2469          */
2470         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2471                 clear_opt(sbi, FLUSH_MERGE);
2472                 f2fs_destroy_flush_cmd_control(sbi, false);
2473                 need_restart_flush = true;
2474         } else {
2475                 err = f2fs_create_flush_cmd_control(sbi);
2476                 if (err)
2477                         goto restore_ckpt;
2478                 need_stop_flush = true;
2479         }
2480
2481         if (no_discard == !!test_opt(sbi, DISCARD)) {
2482                 if (test_opt(sbi, DISCARD)) {
2483                         err = f2fs_start_discard_thread(sbi);
2484                         if (err)
2485                                 goto restore_flush;
2486                         need_stop_discard = true;
2487                 } else {
2488                         f2fs_stop_discard_thread(sbi);
2489                         f2fs_issue_discard_timeout(sbi);
2490                         need_restart_discard = true;
2491                 }
2492         }
2493
2494         if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2495                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2496                         err = f2fs_disable_checkpoint(sbi);
2497                         if (err)
2498                                 goto restore_discard;
2499                 } else {
2500                         f2fs_enable_checkpoint(sbi);
2501                 }
2502         }
2503
2504 skip:
2505 #ifdef CONFIG_QUOTA
2506         /* Release old quota file names */
2507         for (i = 0; i < MAXQUOTAS; i++)
2508                 kfree(org_mount_opt.s_qf_names[i]);
2509 #endif
2510         /* Update the POSIXACL Flag */
2511         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2512                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2513
2514         limit_reserve_root(sbi);
2515         adjust_unusable_cap_perc(sbi);
2516         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2517         return 0;
2518 restore_discard:
2519         if (need_restart_discard) {
2520                 if (f2fs_start_discard_thread(sbi))
2521                         f2fs_warn(sbi, "discard has been stopped");
2522         } else if (need_stop_discard) {
2523                 f2fs_stop_discard_thread(sbi);
2524         }
2525 restore_flush:
2526         if (need_restart_flush) {
2527                 if (f2fs_create_flush_cmd_control(sbi))
2528                         f2fs_warn(sbi, "background flush thread has stopped");
2529         } else if (need_stop_flush) {
2530                 clear_opt(sbi, FLUSH_MERGE);
2531                 f2fs_destroy_flush_cmd_control(sbi, false);
2532         }
2533 restore_ckpt:
2534         if (need_restart_ckpt) {
2535                 if (f2fs_start_ckpt_thread(sbi))
2536                         f2fs_warn(sbi, "background ckpt thread has stopped");
2537         } else if (need_stop_ckpt) {
2538                 f2fs_stop_ckpt_thread(sbi);
2539         }
2540 restore_gc:
2541         if (need_restart_gc) {
2542                 if (f2fs_start_gc_thread(sbi))
2543                         f2fs_warn(sbi, "background gc thread has stopped");
2544         } else if (need_stop_gc) {
2545                 f2fs_stop_gc_thread(sbi);
2546         }
2547 restore_opts:
2548 #ifdef CONFIG_QUOTA
2549         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2550         for (i = 0; i < MAXQUOTAS; i++) {
2551                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2552                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2553         }
2554 #endif
2555         sbi->mount_opt = org_mount_opt;
2556         sb->s_flags = old_sb_flags;
2557         return err;
2558 }
2559
2560 #ifdef CONFIG_QUOTA
2561 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2562 {
2563         /* need to recovery orphan */
2564         if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2565                 return true;
2566         /* need to recovery data */
2567         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2568                 return false;
2569         if (test_opt(sbi, NORECOVERY))
2570                 return false;
2571         return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2572 }
2573
2574 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2575 {
2576         bool readonly = f2fs_readonly(sbi->sb);
2577
2578         if (!f2fs_need_recovery(sbi))
2579                 return false;
2580
2581         /* it doesn't need to check f2fs_sb_has_readonly() */
2582         if (f2fs_hw_is_readonly(sbi))
2583                 return false;
2584
2585         if (readonly) {
2586                 sbi->sb->s_flags &= ~SB_RDONLY;
2587                 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2588         }
2589
2590         /*
2591          * Turn on quotas which were not enabled for read-only mounts if
2592          * filesystem has quota feature, so that they are updated correctly.
2593          */
2594         return f2fs_enable_quota_files(sbi, readonly);
2595 }
2596
2597 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2598                                                 bool quota_enabled)
2599 {
2600         if (quota_enabled)
2601                 f2fs_quota_off_umount(sbi->sb);
2602
2603         if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2604                 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2605                 sbi->sb->s_flags |= SB_RDONLY;
2606         }
2607 }
2608
2609 /* Read data from quotafile */
2610 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2611                                size_t len, loff_t off)
2612 {
2613         struct inode *inode = sb_dqopt(sb)->files[type];
2614         struct address_space *mapping = inode->i_mapping;
2615         block_t blkidx = F2FS_BYTES_TO_BLK(off);
2616         int offset = off & (sb->s_blocksize - 1);
2617         int tocopy;
2618         size_t toread;
2619         loff_t i_size = i_size_read(inode);
2620         struct page *page;
2621
2622         if (off > i_size)
2623                 return 0;
2624
2625         if (off + len > i_size)
2626                 len = i_size - off;
2627         toread = len;
2628         while (toread > 0) {
2629                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2630 repeat:
2631                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2632                 if (IS_ERR(page)) {
2633                         if (PTR_ERR(page) == -ENOMEM) {
2634                                 memalloc_retry_wait(GFP_NOFS);
2635                                 goto repeat;
2636                         }
2637                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2638                         return PTR_ERR(page);
2639                 }
2640
2641                 lock_page(page);
2642
2643                 if (unlikely(page->mapping != mapping)) {
2644                         f2fs_put_page(page, 1);
2645                         goto repeat;
2646                 }
2647                 if (unlikely(!PageUptodate(page))) {
2648                         f2fs_put_page(page, 1);
2649                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2650                         return -EIO;
2651                 }
2652
2653                 memcpy_from_page(data, page, offset, tocopy);
2654                 f2fs_put_page(page, 1);
2655
2656                 offset = 0;
2657                 toread -= tocopy;
2658                 data += tocopy;
2659                 blkidx++;
2660         }
2661         return len;
2662 }
2663
2664 /* Write to quotafile */
2665 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2666                                 const char *data, size_t len, loff_t off)
2667 {
2668         struct inode *inode = sb_dqopt(sb)->files[type];
2669         struct address_space *mapping = inode->i_mapping;
2670         const struct address_space_operations *a_ops = mapping->a_ops;
2671         int offset = off & (sb->s_blocksize - 1);
2672         size_t towrite = len;
2673         struct page *page;
2674         void *fsdata = NULL;
2675         int err = 0;
2676         int tocopy;
2677
2678         while (towrite > 0) {
2679                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2680                                                                 towrite);
2681 retry:
2682                 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2683                                                         &page, &fsdata);
2684                 if (unlikely(err)) {
2685                         if (err == -ENOMEM) {
2686                                 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2687                                 goto retry;
2688                         }
2689                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2690                         break;
2691                 }
2692
2693                 memcpy_to_page(page, offset, data, tocopy);
2694
2695                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2696                                                 page, fsdata);
2697                 offset = 0;
2698                 towrite -= tocopy;
2699                 off += tocopy;
2700                 data += tocopy;
2701                 cond_resched();
2702         }
2703
2704         if (len == towrite)
2705                 return err;
2706         inode->i_mtime = inode->i_ctime = current_time(inode);
2707         f2fs_mark_inode_dirty_sync(inode, false);
2708         return len - towrite;
2709 }
2710
2711 int f2fs_dquot_initialize(struct inode *inode)
2712 {
2713         if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2714                 return -ESRCH;
2715
2716         return dquot_initialize(inode);
2717 }
2718
2719 static struct dquot **f2fs_get_dquots(struct inode *inode)
2720 {
2721         return F2FS_I(inode)->i_dquot;
2722 }
2723
2724 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2725 {
2726         return &F2FS_I(inode)->i_reserved_quota;
2727 }
2728
2729 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2730 {
2731         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2732                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2733                 return 0;
2734         }
2735
2736         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2737                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
2738 }
2739
2740 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2741 {
2742         int enabled = 0;
2743         int i, err;
2744
2745         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2746                 err = f2fs_enable_quotas(sbi->sb);
2747                 if (err) {
2748                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2749                         return 0;
2750                 }
2751                 return 1;
2752         }
2753
2754         for (i = 0; i < MAXQUOTAS; i++) {
2755                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2756                         err = f2fs_quota_on_mount(sbi, i);
2757                         if (!err) {
2758                                 enabled = 1;
2759                                 continue;
2760                         }
2761                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2762                                  err, i);
2763                 }
2764         }
2765         return enabled;
2766 }
2767
2768 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2769                              unsigned int flags)
2770 {
2771         struct inode *qf_inode;
2772         unsigned long qf_inum;
2773         unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2774         int err;
2775
2776         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2777
2778         qf_inum = f2fs_qf_ino(sb, type);
2779         if (!qf_inum)
2780                 return -EPERM;
2781
2782         qf_inode = f2fs_iget(sb, qf_inum);
2783         if (IS_ERR(qf_inode)) {
2784                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2785                 return PTR_ERR(qf_inode);
2786         }
2787
2788         /* Don't account quota for quota files to avoid recursion */
2789         inode_lock(qf_inode);
2790         qf_inode->i_flags |= S_NOQUOTA;
2791
2792         if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2793                 F2FS_I(qf_inode)->i_flags |= qf_flag;
2794                 f2fs_set_inode_flags(qf_inode);
2795         }
2796         inode_unlock(qf_inode);
2797
2798         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2799         iput(qf_inode);
2800         return err;
2801 }
2802
2803 static int f2fs_enable_quotas(struct super_block *sb)
2804 {
2805         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2806         int type, err = 0;
2807         unsigned long qf_inum;
2808         bool quota_mopt[MAXQUOTAS] = {
2809                 test_opt(sbi, USRQUOTA),
2810                 test_opt(sbi, GRPQUOTA),
2811                 test_opt(sbi, PRJQUOTA),
2812         };
2813
2814         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2815                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2816                 return 0;
2817         }
2818
2819         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2820
2821         for (type = 0; type < MAXQUOTAS; type++) {
2822                 qf_inum = f2fs_qf_ino(sb, type);
2823                 if (qf_inum) {
2824                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2825                                 DQUOT_USAGE_ENABLED |
2826                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2827                         if (err) {
2828                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2829                                          type, err);
2830                                 for (type--; type >= 0; type--)
2831                                         dquot_quota_off(sb, type);
2832                                 set_sbi_flag(F2FS_SB(sb),
2833                                                 SBI_QUOTA_NEED_REPAIR);
2834                                 return err;
2835                         }
2836                 }
2837         }
2838         return 0;
2839 }
2840
2841 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2842 {
2843         struct quota_info *dqopt = sb_dqopt(sbi->sb);
2844         struct address_space *mapping = dqopt->files[type]->i_mapping;
2845         int ret = 0;
2846
2847         ret = dquot_writeback_dquots(sbi->sb, type);
2848         if (ret)
2849                 goto out;
2850
2851         ret = filemap_fdatawrite(mapping);
2852         if (ret)
2853                 goto out;
2854
2855         /* if we are using journalled quota */
2856         if (is_journalled_quota(sbi))
2857                 goto out;
2858
2859         ret = filemap_fdatawait(mapping);
2860
2861         truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2862 out:
2863         if (ret)
2864                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2865         return ret;
2866 }
2867
2868 int f2fs_quota_sync(struct super_block *sb, int type)
2869 {
2870         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2871         struct quota_info *dqopt = sb_dqopt(sb);
2872         int cnt;
2873         int ret = 0;
2874
2875         /*
2876          * Now when everything is written we can discard the pagecache so
2877          * that userspace sees the changes.
2878          */
2879         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2880
2881                 if (type != -1 && cnt != type)
2882                         continue;
2883
2884                 if (!sb_has_quota_active(sb, cnt))
2885                         continue;
2886
2887                 if (!f2fs_sb_has_quota_ino(sbi))
2888                         inode_lock(dqopt->files[cnt]);
2889
2890                 /*
2891                  * do_quotactl
2892                  *  f2fs_quota_sync
2893                  *  f2fs_down_read(quota_sem)
2894                  *  dquot_writeback_dquots()
2895                  *  f2fs_dquot_commit
2896                  *                            block_operation
2897                  *                            f2fs_down_read(quota_sem)
2898                  */
2899                 f2fs_lock_op(sbi);
2900                 f2fs_down_read(&sbi->quota_sem);
2901
2902                 ret = f2fs_quota_sync_file(sbi, cnt);
2903
2904                 f2fs_up_read(&sbi->quota_sem);
2905                 f2fs_unlock_op(sbi);
2906
2907                 if (!f2fs_sb_has_quota_ino(sbi))
2908                         inode_unlock(dqopt->files[cnt]);
2909
2910                 if (ret)
2911                         break;
2912         }
2913         return ret;
2914 }
2915
2916 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2917                                                         const struct path *path)
2918 {
2919         struct inode *inode;
2920         int err;
2921
2922         /* if quota sysfile exists, deny enabling quota with specific file */
2923         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2924                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2925                 return -EBUSY;
2926         }
2927
2928         if (path->dentry->d_sb != sb)
2929                 return -EXDEV;
2930
2931         err = f2fs_quota_sync(sb, type);
2932         if (err)
2933                 return err;
2934
2935         inode = d_inode(path->dentry);
2936
2937         err = filemap_fdatawrite(inode->i_mapping);
2938         if (err)
2939                 return err;
2940
2941         err = filemap_fdatawait(inode->i_mapping);
2942         if (err)
2943                 return err;
2944
2945         err = dquot_quota_on(sb, type, format_id, path);
2946         if (err)
2947                 return err;
2948
2949         inode_lock(inode);
2950         F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2951         f2fs_set_inode_flags(inode);
2952         inode_unlock(inode);
2953         f2fs_mark_inode_dirty_sync(inode, false);
2954
2955         return 0;
2956 }
2957
2958 static int __f2fs_quota_off(struct super_block *sb, int type)
2959 {
2960         struct inode *inode = sb_dqopt(sb)->files[type];
2961         int err;
2962
2963         if (!inode || !igrab(inode))
2964                 return dquot_quota_off(sb, type);
2965
2966         err = f2fs_quota_sync(sb, type);
2967         if (err)
2968                 goto out_put;
2969
2970         err = dquot_quota_off(sb, type);
2971         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2972                 goto out_put;
2973
2974         inode_lock(inode);
2975         F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
2976         f2fs_set_inode_flags(inode);
2977         inode_unlock(inode);
2978         f2fs_mark_inode_dirty_sync(inode, false);
2979 out_put:
2980         iput(inode);
2981         return err;
2982 }
2983
2984 static int f2fs_quota_off(struct super_block *sb, int type)
2985 {
2986         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2987         int err;
2988
2989         err = __f2fs_quota_off(sb, type);
2990
2991         /*
2992          * quotactl can shutdown journalled quota, result in inconsistence
2993          * between quota record and fs data by following updates, tag the
2994          * flag to let fsck be aware of it.
2995          */
2996         if (is_journalled_quota(sbi))
2997                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2998         return err;
2999 }
3000
3001 void f2fs_quota_off_umount(struct super_block *sb)
3002 {
3003         int type;
3004         int err;
3005
3006         for (type = 0; type < MAXQUOTAS; type++) {
3007                 err = __f2fs_quota_off(sb, type);
3008                 if (err) {
3009                         int ret = dquot_quota_off(sb, type);
3010
3011                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3012                                  type, err, ret);
3013                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3014                 }
3015         }
3016         /*
3017          * In case of checkpoint=disable, we must flush quota blocks.
3018          * This can cause NULL exception for node_inode in end_io, since
3019          * put_super already dropped it.
3020          */
3021         sync_filesystem(sb);
3022 }
3023
3024 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3025 {
3026         struct quota_info *dqopt = sb_dqopt(sb);
3027         int type;
3028
3029         for (type = 0; type < MAXQUOTAS; type++) {
3030                 if (!dqopt->files[type])
3031                         continue;
3032                 f2fs_inode_synced(dqopt->files[type]);
3033         }
3034 }
3035
3036 static int f2fs_dquot_commit(struct dquot *dquot)
3037 {
3038         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3039         int ret;
3040
3041         f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3042         ret = dquot_commit(dquot);
3043         if (ret < 0)
3044                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3045         f2fs_up_read(&sbi->quota_sem);
3046         return ret;
3047 }
3048
3049 static int f2fs_dquot_acquire(struct dquot *dquot)
3050 {
3051         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3052         int ret;
3053
3054         f2fs_down_read(&sbi->quota_sem);
3055         ret = dquot_acquire(dquot);
3056         if (ret < 0)
3057                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3058         f2fs_up_read(&sbi->quota_sem);
3059         return ret;
3060 }
3061
3062 static int f2fs_dquot_release(struct dquot *dquot)
3063 {
3064         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3065         int ret = dquot_release(dquot);
3066
3067         if (ret < 0)
3068                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3069         return ret;
3070 }
3071
3072 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3073 {
3074         struct super_block *sb = dquot->dq_sb;
3075         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3076         int ret = dquot_mark_dquot_dirty(dquot);
3077
3078         /* if we are using journalled quota */
3079         if (is_journalled_quota(sbi))
3080                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3081
3082         return ret;
3083 }
3084
3085 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3086 {
3087         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3088         int ret = dquot_commit_info(sb, type);
3089
3090         if (ret < 0)
3091                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3092         return ret;
3093 }
3094
3095 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3096 {
3097         *projid = F2FS_I(inode)->i_projid;
3098         return 0;
3099 }
3100
3101 static const struct dquot_operations f2fs_quota_operations = {
3102         .get_reserved_space = f2fs_get_reserved_space,
3103         .write_dquot    = f2fs_dquot_commit,
3104         .acquire_dquot  = f2fs_dquot_acquire,
3105         .release_dquot  = f2fs_dquot_release,
3106         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
3107         .write_info     = f2fs_dquot_commit_info,
3108         .alloc_dquot    = dquot_alloc,
3109         .destroy_dquot  = dquot_destroy,
3110         .get_projid     = f2fs_get_projid,
3111         .get_next_id    = dquot_get_next_id,
3112 };
3113
3114 static const struct quotactl_ops f2fs_quotactl_ops = {
3115         .quota_on       = f2fs_quota_on,
3116         .quota_off      = f2fs_quota_off,
3117         .quota_sync     = f2fs_quota_sync,
3118         .get_state      = dquot_get_state,
3119         .set_info       = dquot_set_dqinfo,
3120         .get_dqblk      = dquot_get_dqblk,
3121         .set_dqblk      = dquot_set_dqblk,
3122         .get_nextdqblk  = dquot_get_next_dqblk,
3123 };
3124 #else
3125 int f2fs_dquot_initialize(struct inode *inode)
3126 {
3127         return 0;
3128 }
3129
3130 int f2fs_quota_sync(struct super_block *sb, int type)
3131 {
3132         return 0;
3133 }
3134
3135 void f2fs_quota_off_umount(struct super_block *sb)
3136 {
3137 }
3138 #endif
3139
3140 static const struct super_operations f2fs_sops = {
3141         .alloc_inode    = f2fs_alloc_inode,
3142         .free_inode     = f2fs_free_inode,
3143         .drop_inode     = f2fs_drop_inode,
3144         .write_inode    = f2fs_write_inode,
3145         .dirty_inode    = f2fs_dirty_inode,
3146         .show_options   = f2fs_show_options,
3147 #ifdef CONFIG_QUOTA
3148         .quota_read     = f2fs_quota_read,
3149         .quota_write    = f2fs_quota_write,
3150         .get_dquots     = f2fs_get_dquots,
3151 #endif
3152         .evict_inode    = f2fs_evict_inode,
3153         .put_super      = f2fs_put_super,
3154         .sync_fs        = f2fs_sync_fs,
3155         .freeze_fs      = f2fs_freeze,
3156         .unfreeze_fs    = f2fs_unfreeze,
3157         .statfs         = f2fs_statfs,
3158         .remount_fs     = f2fs_remount,
3159 };
3160
3161 #ifdef CONFIG_FS_ENCRYPTION
3162 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3163 {
3164         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3165                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3166                                 ctx, len, NULL);
3167 }
3168
3169 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3170                                                         void *fs_data)
3171 {
3172         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3173
3174         /*
3175          * Encrypting the root directory is not allowed because fsck
3176          * expects lost+found directory to exist and remain unencrypted
3177          * if LOST_FOUND feature is enabled.
3178          *
3179          */
3180         if (f2fs_sb_has_lost_found(sbi) &&
3181                         inode->i_ino == F2FS_ROOT_INO(sbi))
3182                 return -EPERM;
3183
3184         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3185                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3186                                 ctx, len, fs_data, XATTR_CREATE);
3187 }
3188
3189 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3190 {
3191         return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3192 }
3193
3194 static bool f2fs_has_stable_inodes(struct super_block *sb)
3195 {
3196         return true;
3197 }
3198
3199 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3200                                        int *ino_bits_ret, int *lblk_bits_ret)
3201 {
3202         *ino_bits_ret = 8 * sizeof(nid_t);
3203         *lblk_bits_ret = 8 * sizeof(block_t);
3204 }
3205
3206 static struct block_device **f2fs_get_devices(struct super_block *sb,
3207                                               unsigned int *num_devs)
3208 {
3209         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3210         struct block_device **devs;
3211         int i;
3212
3213         if (!f2fs_is_multi_device(sbi))
3214                 return NULL;
3215
3216         devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3217         if (!devs)
3218                 return ERR_PTR(-ENOMEM);
3219
3220         for (i = 0; i < sbi->s_ndevs; i++)
3221                 devs[i] = FDEV(i).bdev;
3222         *num_devs = sbi->s_ndevs;
3223         return devs;
3224 }
3225
3226 static const struct fscrypt_operations f2fs_cryptops = {
3227         .key_prefix             = "f2fs:",
3228         .get_context            = f2fs_get_context,
3229         .set_context            = f2fs_set_context,
3230         .get_dummy_policy       = f2fs_get_dummy_policy,
3231         .empty_dir              = f2fs_empty_dir,
3232         .has_stable_inodes      = f2fs_has_stable_inodes,
3233         .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
3234         .get_devices            = f2fs_get_devices,
3235 };
3236 #endif
3237
3238 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3239                 u64 ino, u32 generation)
3240 {
3241         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3242         struct inode *inode;
3243
3244         if (f2fs_check_nid_range(sbi, ino))
3245                 return ERR_PTR(-ESTALE);
3246
3247         /*
3248          * f2fs_iget isn't quite right if the inode is currently unallocated!
3249          * However f2fs_iget currently does appropriate checks to handle stale
3250          * inodes so everything is OK.
3251          */
3252         inode = f2fs_iget(sb, ino);
3253         if (IS_ERR(inode))
3254                 return ERR_CAST(inode);
3255         if (unlikely(generation && inode->i_generation != generation)) {
3256                 /* we didn't find the right inode.. */
3257                 iput(inode);
3258                 return ERR_PTR(-ESTALE);
3259         }
3260         return inode;
3261 }
3262
3263 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3264                 int fh_len, int fh_type)
3265 {
3266         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3267                                     f2fs_nfs_get_inode);
3268 }
3269
3270 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3271                 int fh_len, int fh_type)
3272 {
3273         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3274                                     f2fs_nfs_get_inode);
3275 }
3276
3277 static const struct export_operations f2fs_export_ops = {
3278         .fh_to_dentry = f2fs_fh_to_dentry,
3279         .fh_to_parent = f2fs_fh_to_parent,
3280         .get_parent = f2fs_get_parent,
3281 };
3282
3283 loff_t max_file_blocks(struct inode *inode)
3284 {
3285         loff_t result = 0;
3286         loff_t leaf_count;
3287
3288         /*
3289          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3290          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3291          * space in inode.i_addr, it will be more safe to reassign
3292          * result as zero.
3293          */
3294
3295         if (inode && f2fs_compressed_file(inode))
3296                 leaf_count = ADDRS_PER_BLOCK(inode);
3297         else
3298                 leaf_count = DEF_ADDRS_PER_BLOCK;
3299
3300         /* two direct node blocks */
3301         result += (leaf_count * 2);
3302
3303         /* two indirect node blocks */
3304         leaf_count *= NIDS_PER_BLOCK;
3305         result += (leaf_count * 2);
3306
3307         /* one double indirect node block */
3308         leaf_count *= NIDS_PER_BLOCK;
3309         result += leaf_count;
3310
3311         return result;
3312 }
3313
3314 static int __f2fs_commit_super(struct buffer_head *bh,
3315                         struct f2fs_super_block *super)
3316 {
3317         lock_buffer(bh);
3318         if (super)
3319                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3320         set_buffer_dirty(bh);
3321         unlock_buffer(bh);
3322
3323         /* it's rare case, we can do fua all the time */
3324         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3325 }
3326
3327 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3328                                         struct buffer_head *bh)
3329 {
3330         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3331                                         (bh->b_data + F2FS_SUPER_OFFSET);
3332         struct super_block *sb = sbi->sb;
3333         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3334         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3335         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3336         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3337         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3338         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3339         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3340         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3341         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3342         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3343         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3344         u32 segment_count = le32_to_cpu(raw_super->segment_count);
3345         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3346         u64 main_end_blkaddr = main_blkaddr +
3347                                 (segment_count_main << log_blocks_per_seg);
3348         u64 seg_end_blkaddr = segment0_blkaddr +
3349                                 (segment_count << log_blocks_per_seg);
3350
3351         if (segment0_blkaddr != cp_blkaddr) {
3352                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3353                           segment0_blkaddr, cp_blkaddr);
3354                 return true;
3355         }
3356
3357         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3358                                                         sit_blkaddr) {
3359                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3360                           cp_blkaddr, sit_blkaddr,
3361                           segment_count_ckpt << log_blocks_per_seg);
3362                 return true;
3363         }
3364
3365         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3366                                                         nat_blkaddr) {
3367                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3368                           sit_blkaddr, nat_blkaddr,
3369                           segment_count_sit << log_blocks_per_seg);
3370                 return true;
3371         }
3372
3373         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3374                                                         ssa_blkaddr) {
3375                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3376                           nat_blkaddr, ssa_blkaddr,
3377                           segment_count_nat << log_blocks_per_seg);
3378                 return true;
3379         }
3380
3381         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3382                                                         main_blkaddr) {
3383                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3384                           ssa_blkaddr, main_blkaddr,
3385                           segment_count_ssa << log_blocks_per_seg);
3386                 return true;
3387         }
3388
3389         if (main_end_blkaddr > seg_end_blkaddr) {
3390                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3391                           main_blkaddr, seg_end_blkaddr,
3392                           segment_count_main << log_blocks_per_seg);
3393                 return true;
3394         } else if (main_end_blkaddr < seg_end_blkaddr) {
3395                 int err = 0;
3396                 char *res;
3397
3398                 /* fix in-memory information all the time */
3399                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3400                                 segment0_blkaddr) >> log_blocks_per_seg);
3401
3402                 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3403                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3404                         res = "internally";
3405                 } else {
3406                         err = __f2fs_commit_super(bh, NULL);
3407                         res = err ? "failed" : "done";
3408                 }
3409                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3410                           res, main_blkaddr, seg_end_blkaddr,
3411                           segment_count_main << log_blocks_per_seg);
3412                 if (err)
3413                         return true;
3414         }
3415         return false;
3416 }
3417
3418 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3419                                 struct buffer_head *bh)
3420 {
3421         block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3422         block_t total_sections, blocks_per_seg;
3423         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3424                                         (bh->b_data + F2FS_SUPER_OFFSET);
3425         size_t crc_offset = 0;
3426         __u32 crc = 0;
3427
3428         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3429                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3430                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3431                 return -EINVAL;
3432         }
3433
3434         /* Check checksum_offset and crc in superblock */
3435         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3436                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3437                 if (crc_offset !=
3438                         offsetof(struct f2fs_super_block, crc)) {
3439                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3440                                   crc_offset);
3441                         return -EFSCORRUPTED;
3442                 }
3443                 crc = le32_to_cpu(raw_super->crc);
3444                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3445                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3446                         return -EFSCORRUPTED;
3447                 }
3448         }
3449
3450         /* Currently, support only 4KB block size */
3451         if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3452                 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3453                           le32_to_cpu(raw_super->log_blocksize),
3454                           F2FS_BLKSIZE_BITS);
3455                 return -EFSCORRUPTED;
3456         }
3457
3458         /* check log blocks per segment */
3459         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3460                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3461                           le32_to_cpu(raw_super->log_blocks_per_seg));
3462                 return -EFSCORRUPTED;
3463         }
3464
3465         /* Currently, support 512/1024/2048/4096 bytes sector size */
3466         if (le32_to_cpu(raw_super->log_sectorsize) >
3467                                 F2FS_MAX_LOG_SECTOR_SIZE ||
3468                 le32_to_cpu(raw_super->log_sectorsize) <
3469                                 F2FS_MIN_LOG_SECTOR_SIZE) {
3470                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3471                           le32_to_cpu(raw_super->log_sectorsize));
3472                 return -EFSCORRUPTED;
3473         }
3474         if (le32_to_cpu(raw_super->log_sectors_per_block) +
3475                 le32_to_cpu(raw_super->log_sectorsize) !=
3476                         F2FS_MAX_LOG_SECTOR_SIZE) {
3477                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3478                           le32_to_cpu(raw_super->log_sectors_per_block),
3479                           le32_to_cpu(raw_super->log_sectorsize));
3480                 return -EFSCORRUPTED;
3481         }
3482
3483         segment_count = le32_to_cpu(raw_super->segment_count);
3484         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3485         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3486         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3487         total_sections = le32_to_cpu(raw_super->section_count);
3488
3489         /* blocks_per_seg should be 512, given the above check */
3490         blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3491
3492         if (segment_count > F2FS_MAX_SEGMENT ||
3493                                 segment_count < F2FS_MIN_SEGMENTS) {
3494                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3495                 return -EFSCORRUPTED;
3496         }
3497
3498         if (total_sections > segment_count_main || total_sections < 1 ||
3499                         segs_per_sec > segment_count || !segs_per_sec) {
3500                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3501                           segment_count, total_sections, segs_per_sec);
3502                 return -EFSCORRUPTED;
3503         }
3504
3505         if (segment_count_main != total_sections * segs_per_sec) {
3506                 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3507                           segment_count_main, total_sections, segs_per_sec);
3508                 return -EFSCORRUPTED;
3509         }
3510
3511         if ((segment_count / segs_per_sec) < total_sections) {
3512                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3513                           segment_count, segs_per_sec, total_sections);
3514                 return -EFSCORRUPTED;
3515         }
3516
3517         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3518                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3519                           segment_count, le64_to_cpu(raw_super->block_count));
3520                 return -EFSCORRUPTED;
3521         }
3522
3523         if (RDEV(0).path[0]) {
3524                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3525                 int i = 1;
3526
3527                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3528                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3529                         i++;
3530                 }
3531                 if (segment_count != dev_seg_count) {
3532                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3533                                         segment_count, dev_seg_count);
3534                         return -EFSCORRUPTED;
3535                 }
3536         } else {
3537                 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3538                                         !bdev_is_zoned(sbi->sb->s_bdev)) {
3539                         f2fs_info(sbi, "Zoned block device path is missing");
3540                         return -EFSCORRUPTED;
3541                 }
3542         }
3543
3544         if (secs_per_zone > total_sections || !secs_per_zone) {
3545                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3546                           secs_per_zone, total_sections);
3547                 return -EFSCORRUPTED;
3548         }
3549         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3550                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3551                         (le32_to_cpu(raw_super->extension_count) +
3552                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3553                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3554                           le32_to_cpu(raw_super->extension_count),
3555                           raw_super->hot_ext_count,
3556                           F2FS_MAX_EXTENSION);
3557                 return -EFSCORRUPTED;
3558         }
3559
3560         if (le32_to_cpu(raw_super->cp_payload) >=
3561                                 (blocks_per_seg - F2FS_CP_PACKS -
3562                                 NR_CURSEG_PERSIST_TYPE)) {
3563                 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3564                           le32_to_cpu(raw_super->cp_payload),
3565                           blocks_per_seg - F2FS_CP_PACKS -
3566                           NR_CURSEG_PERSIST_TYPE);
3567                 return -EFSCORRUPTED;
3568         }
3569
3570         /* check reserved ino info */
3571         if (le32_to_cpu(raw_super->node_ino) != 1 ||
3572                 le32_to_cpu(raw_super->meta_ino) != 2 ||
3573                 le32_to_cpu(raw_super->root_ino) != 3) {
3574                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3575                           le32_to_cpu(raw_super->node_ino),
3576                           le32_to_cpu(raw_super->meta_ino),
3577                           le32_to_cpu(raw_super->root_ino));
3578                 return -EFSCORRUPTED;
3579         }
3580
3581         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3582         if (sanity_check_area_boundary(sbi, bh))
3583                 return -EFSCORRUPTED;
3584
3585         return 0;
3586 }
3587
3588 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3589 {
3590         unsigned int total, fsmeta;
3591         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3592         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3593         unsigned int ovp_segments, reserved_segments;
3594         unsigned int main_segs, blocks_per_seg;
3595         unsigned int sit_segs, nat_segs;
3596         unsigned int sit_bitmap_size, nat_bitmap_size;
3597         unsigned int log_blocks_per_seg;
3598         unsigned int segment_count_main;
3599         unsigned int cp_pack_start_sum, cp_payload;
3600         block_t user_block_count, valid_user_blocks;
3601         block_t avail_node_count, valid_node_count;
3602         unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3603         int i, j;
3604
3605         total = le32_to_cpu(raw_super->segment_count);
3606         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3607         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3608         fsmeta += sit_segs;
3609         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3610         fsmeta += nat_segs;
3611         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3612         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3613
3614         if (unlikely(fsmeta >= total))
3615                 return 1;
3616
3617         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3618         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3619
3620         if (!f2fs_sb_has_readonly(sbi) &&
3621                         unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3622                         ovp_segments == 0 || reserved_segments == 0)) {
3623                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3624                 return 1;
3625         }
3626         user_block_count = le64_to_cpu(ckpt->user_block_count);
3627         segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3628                         (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3629         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3630         if (!user_block_count || user_block_count >=
3631                         segment_count_main << log_blocks_per_seg) {
3632                 f2fs_err(sbi, "Wrong user_block_count: %u",
3633                          user_block_count);
3634                 return 1;
3635         }
3636
3637         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3638         if (valid_user_blocks > user_block_count) {
3639                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3640                          valid_user_blocks, user_block_count);
3641                 return 1;
3642         }
3643
3644         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3645         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3646         if (valid_node_count > avail_node_count) {
3647                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3648                          valid_node_count, avail_node_count);
3649                 return 1;
3650         }
3651
3652         main_segs = le32_to_cpu(raw_super->segment_count_main);
3653         blocks_per_seg = sbi->blocks_per_seg;
3654
3655         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3656                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3657                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3658                         return 1;
3659
3660                 if (f2fs_sb_has_readonly(sbi))
3661                         goto check_data;
3662
3663                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3664                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3665                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
3666                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3667                                          i, j,
3668                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3669                                 return 1;
3670                         }
3671                 }
3672         }
3673 check_data:
3674         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3675                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3676                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3677                         return 1;
3678
3679                 if (f2fs_sb_has_readonly(sbi))
3680                         goto skip_cross;
3681
3682                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3683                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3684                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3685                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3686                                          i, j,
3687                                          le32_to_cpu(ckpt->cur_data_segno[i]));
3688                                 return 1;
3689                         }
3690                 }
3691         }
3692         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3693                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3694                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3695                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3696                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3697                                          i, j,
3698                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3699                                 return 1;
3700                         }
3701                 }
3702         }
3703 skip_cross:
3704         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3705         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3706
3707         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3708                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3709                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3710                          sit_bitmap_size, nat_bitmap_size);
3711                 return 1;
3712         }
3713
3714         cp_pack_start_sum = __start_sum_addr(sbi);
3715         cp_payload = __cp_payload(sbi);
3716         if (cp_pack_start_sum < cp_payload + 1 ||
3717                 cp_pack_start_sum > blocks_per_seg - 1 -
3718                         NR_CURSEG_PERSIST_TYPE) {
3719                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3720                          cp_pack_start_sum);
3721                 return 1;
3722         }
3723
3724         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3725                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3726                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3727                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3728                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3729                           le32_to_cpu(ckpt->checksum_offset));
3730                 return 1;
3731         }
3732
3733         nat_blocks = nat_segs << log_blocks_per_seg;
3734         nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3735         nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3736         if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3737                 (cp_payload + F2FS_CP_PACKS +
3738                 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3739                 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3740                           cp_payload, nat_bits_blocks);
3741                 return 1;
3742         }
3743
3744         if (unlikely(f2fs_cp_error(sbi))) {
3745                 f2fs_err(sbi, "A bug case: need to run fsck");
3746                 return 1;
3747         }
3748         return 0;
3749 }
3750
3751 static void init_sb_info(struct f2fs_sb_info *sbi)
3752 {
3753         struct f2fs_super_block *raw_super = sbi->raw_super;
3754         int i;
3755
3756         sbi->log_sectors_per_block =
3757                 le32_to_cpu(raw_super->log_sectors_per_block);
3758         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3759         sbi->blocksize = BIT(sbi->log_blocksize);
3760         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3761         sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3762         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3763         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3764         sbi->total_sections = le32_to_cpu(raw_super->section_count);
3765         sbi->total_node_count =
3766                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3767                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3768         F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3769         F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3770         F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3771         sbi->cur_victim_sec = NULL_SECNO;
3772         sbi->gc_mode = GC_NORMAL;
3773         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3774         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3775         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3776         sbi->migration_granularity = sbi->segs_per_sec;
3777         sbi->seq_file_ra_mul = MIN_RA_MUL;
3778         sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3779         sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3780         spin_lock_init(&sbi->gc_remaining_trials_lock);
3781         atomic64_set(&sbi->current_atomic_write, 0);
3782
3783         sbi->dir_level = DEF_DIR_LEVEL;
3784         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3785         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3786         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3787         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3788         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3789         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3790                                 DEF_UMOUNT_DISCARD_TIMEOUT;
3791         clear_sbi_flag(sbi, SBI_NEED_FSCK);
3792
3793         for (i = 0; i < NR_COUNT_TYPE; i++)
3794                 atomic_set(&sbi->nr_pages[i], 0);
3795
3796         for (i = 0; i < META; i++)
3797                 atomic_set(&sbi->wb_sync_req[i], 0);
3798
3799         INIT_LIST_HEAD(&sbi->s_list);
3800         mutex_init(&sbi->umount_mutex);
3801         init_f2fs_rwsem(&sbi->io_order_lock);
3802         spin_lock_init(&sbi->cp_lock);
3803
3804         sbi->dirty_device = 0;
3805         spin_lock_init(&sbi->dev_lock);
3806
3807         init_f2fs_rwsem(&sbi->sb_lock);
3808         init_f2fs_rwsem(&sbi->pin_sem);
3809 }
3810
3811 static int init_percpu_info(struct f2fs_sb_info *sbi)
3812 {
3813         int err;
3814
3815         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3816         if (err)
3817                 return err;
3818
3819         err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3820         if (err)
3821                 goto err_valid_block;
3822
3823         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3824                                                                 GFP_KERNEL);
3825         if (err)
3826                 goto err_node_block;
3827         return 0;
3828
3829 err_node_block:
3830         percpu_counter_destroy(&sbi->rf_node_block_count);
3831 err_valid_block:
3832         percpu_counter_destroy(&sbi->alloc_valid_block_count);
3833         return err;
3834 }
3835
3836 #ifdef CONFIG_BLK_DEV_ZONED
3837
3838 struct f2fs_report_zones_args {
3839         struct f2fs_sb_info *sbi;
3840         struct f2fs_dev_info *dev;
3841 };
3842
3843 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3844                               void *data)
3845 {
3846         struct f2fs_report_zones_args *rz_args = data;
3847         block_t unusable_blocks = (zone->len - zone->capacity) >>
3848                                         F2FS_LOG_SECTORS_PER_BLOCK;
3849
3850         if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3851                 return 0;
3852
3853         set_bit(idx, rz_args->dev->blkz_seq);
3854         if (!rz_args->sbi->unusable_blocks_per_sec) {
3855                 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3856                 return 0;
3857         }
3858         if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3859                 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3860                 return -EINVAL;
3861         }
3862         return 0;
3863 }
3864
3865 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3866 {
3867         struct block_device *bdev = FDEV(devi).bdev;
3868         sector_t nr_sectors = bdev_nr_sectors(bdev);
3869         struct f2fs_report_zones_args rep_zone_arg;
3870         u64 zone_sectors;
3871         int ret;
3872
3873         if (!f2fs_sb_has_blkzoned(sbi))
3874                 return 0;
3875
3876         zone_sectors = bdev_zone_sectors(bdev);
3877         if (!is_power_of_2(zone_sectors)) {
3878                 f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n");
3879                 return -EINVAL;
3880         }
3881
3882         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3883                                 SECTOR_TO_BLOCK(zone_sectors))
3884                 return -EINVAL;
3885         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3886         FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3887                                         sbi->blocks_per_blkz);
3888         if (nr_sectors & (zone_sectors - 1))
3889                 FDEV(devi).nr_blkz++;
3890
3891         FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3892                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
3893                                         * sizeof(unsigned long),
3894                                         GFP_KERNEL);
3895         if (!FDEV(devi).blkz_seq)
3896                 return -ENOMEM;
3897
3898         rep_zone_arg.sbi = sbi;
3899         rep_zone_arg.dev = &FDEV(devi);
3900
3901         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3902                                   &rep_zone_arg);
3903         if (ret < 0)
3904                 return ret;
3905         return 0;
3906 }
3907 #endif
3908
3909 /*
3910  * Read f2fs raw super block.
3911  * Because we have two copies of super block, so read both of them
3912  * to get the first valid one. If any one of them is broken, we pass
3913  * them recovery flag back to the caller.
3914  */
3915 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3916                         struct f2fs_super_block **raw_super,
3917                         int *valid_super_block, int *recovery)
3918 {
3919         struct super_block *sb = sbi->sb;
3920         int block;
3921         struct buffer_head *bh;
3922         struct f2fs_super_block *super;
3923         int err = 0;
3924
3925         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3926         if (!super)
3927                 return -ENOMEM;
3928
3929         for (block = 0; block < 2; block++) {
3930                 bh = sb_bread(sb, block);
3931                 if (!bh) {
3932                         f2fs_err(sbi, "Unable to read %dth superblock",
3933                                  block + 1);
3934                         err = -EIO;
3935                         *recovery = 1;
3936                         continue;
3937                 }
3938
3939                 /* sanity checking of raw super */
3940                 err = sanity_check_raw_super(sbi, bh);
3941                 if (err) {
3942                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3943                                  block + 1);
3944                         brelse(bh);
3945                         *recovery = 1;
3946                         continue;
3947                 }
3948
3949                 if (!*raw_super) {
3950                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3951                                                         sizeof(*super));
3952                         *valid_super_block = block;
3953                         *raw_super = super;
3954                 }
3955                 brelse(bh);
3956         }
3957
3958         /* No valid superblock */
3959         if (!*raw_super)
3960                 kfree(super);
3961         else
3962                 err = 0;
3963
3964         return err;
3965 }
3966
3967 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3968 {
3969         struct buffer_head *bh;
3970         __u32 crc = 0;
3971         int err;
3972
3973         if ((recover && f2fs_readonly(sbi->sb)) ||
3974                                 f2fs_hw_is_readonly(sbi)) {
3975                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3976                 return -EROFS;
3977         }
3978
3979         /* we should update superblock crc here */
3980         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3981                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3982                                 offsetof(struct f2fs_super_block, crc));
3983                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3984         }
3985
3986         /* write back-up superblock first */
3987         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3988         if (!bh)
3989                 return -EIO;
3990         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3991         brelse(bh);
3992
3993         /* if we are in recovery path, skip writing valid superblock */
3994         if (recover || err)
3995                 return err;
3996
3997         /* write current valid superblock */
3998         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3999         if (!bh)
4000                 return -EIO;
4001         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
4002         brelse(bh);
4003         return err;
4004 }
4005
4006 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4007 {
4008         unsigned long flags;
4009
4010         spin_lock_irqsave(&sbi->error_lock, flags);
4011         if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4012                 sbi->stop_reason[reason]++;
4013         spin_unlock_irqrestore(&sbi->error_lock, flags);
4014 }
4015
4016 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4017 {
4018         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4019         unsigned long flags;
4020         int err;
4021
4022         f2fs_down_write(&sbi->sb_lock);
4023
4024         spin_lock_irqsave(&sbi->error_lock, flags);
4025         if (sbi->error_dirty) {
4026                 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4027                                                         MAX_F2FS_ERRORS);
4028                 sbi->error_dirty = false;
4029         }
4030         memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4031         spin_unlock_irqrestore(&sbi->error_lock, flags);
4032
4033         err = f2fs_commit_super(sbi, false);
4034
4035         f2fs_up_write(&sbi->sb_lock);
4036         if (err)
4037                 f2fs_err(sbi, "f2fs_commit_super fails to record err:%d", err);
4038 }
4039
4040 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4041 {
4042         unsigned long flags;
4043
4044         spin_lock_irqsave(&sbi->error_lock, flags);
4045         if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4046                 set_bit(flag, (unsigned long *)sbi->errors);
4047                 sbi->error_dirty = true;
4048         }
4049         spin_unlock_irqrestore(&sbi->error_lock, flags);
4050 }
4051
4052 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4053 {
4054         unsigned long flags;
4055         bool need_update = false;
4056
4057         spin_lock_irqsave(&sbi->error_lock, flags);
4058         if (sbi->error_dirty) {
4059                 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4060                                                         MAX_F2FS_ERRORS);
4061                 sbi->error_dirty = false;
4062                 need_update = true;
4063         }
4064         spin_unlock_irqrestore(&sbi->error_lock, flags);
4065
4066         return need_update;
4067 }
4068
4069 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4070 {
4071         int err;
4072
4073         f2fs_down_write(&sbi->sb_lock);
4074
4075         if (!f2fs_update_errors(sbi))
4076                 goto out_unlock;
4077
4078         err = f2fs_commit_super(sbi, false);
4079         if (err)
4080                 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
4081                                                                 error, err);
4082 out_unlock:
4083         f2fs_up_write(&sbi->sb_lock);
4084 }
4085
4086 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4087 {
4088         f2fs_save_errors(sbi, error);
4089         f2fs_record_errors(sbi, error);
4090 }
4091
4092 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4093 {
4094         f2fs_save_errors(sbi, error);
4095
4096         if (!sbi->error_dirty)
4097                 return;
4098         if (!test_bit(error, (unsigned long *)sbi->errors))
4099                 return;
4100         schedule_work(&sbi->s_error_work);
4101 }
4102
4103 static bool system_going_down(void)
4104 {
4105         return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4106                 || system_state == SYSTEM_RESTART;
4107 }
4108
4109 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason,
4110                                                         bool irq_context)
4111 {
4112         struct super_block *sb = sbi->sb;
4113         bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4114         bool continue_fs = !shutdown &&
4115                         F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4116
4117         set_ckpt_flags(sbi, CP_ERROR_FLAG);
4118
4119         if (!f2fs_hw_is_readonly(sbi)) {
4120                 save_stop_reason(sbi, reason);
4121
4122                 if (irq_context && !shutdown)
4123                         schedule_work(&sbi->s_error_work);
4124                 else
4125                         f2fs_record_stop_reason(sbi);
4126         }
4127
4128         /*
4129          * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4130          * could panic during 'reboot -f' as the underlying device got already
4131          * disabled.
4132          */
4133         if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4134                                 !shutdown && !system_going_down() &&
4135                                 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4136                 panic("F2FS-fs (device %s): panic forced after error\n",
4137                                                         sb->s_id);
4138
4139         if (shutdown)
4140                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4141
4142         /* continue filesystem operators if errors=continue */
4143         if (continue_fs || f2fs_readonly(sb))
4144                 return;
4145
4146         f2fs_warn(sbi, "Remounting filesystem read-only");
4147         /*
4148          * Make sure updated value of ->s_mount_flags will be visible before
4149          * ->s_flags update
4150          */
4151         smp_wmb();
4152         sb->s_flags |= SB_RDONLY;
4153 }
4154
4155 static void f2fs_record_error_work(struct work_struct *work)
4156 {
4157         struct f2fs_sb_info *sbi = container_of(work,
4158                                         struct f2fs_sb_info, s_error_work);
4159
4160         f2fs_record_stop_reason(sbi);
4161 }
4162
4163 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4164 {
4165         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4166         unsigned int max_devices = MAX_DEVICES;
4167         unsigned int logical_blksize;
4168         blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4169         int i;
4170
4171         /* Initialize single device information */
4172         if (!RDEV(0).path[0]) {
4173                 if (!bdev_is_zoned(sbi->sb->s_bdev))
4174                         return 0;
4175                 max_devices = 1;
4176         }
4177
4178         /*
4179          * Initialize multiple devices information, or single
4180          * zoned block device information.
4181          */
4182         sbi->devs = f2fs_kzalloc(sbi,
4183                                  array_size(max_devices,
4184                                             sizeof(struct f2fs_dev_info)),
4185                                  GFP_KERNEL);
4186         if (!sbi->devs)
4187                 return -ENOMEM;
4188
4189         logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4190         sbi->aligned_blksize = true;
4191
4192         for (i = 0; i < max_devices; i++) {
4193
4194                 if (i > 0 && !RDEV(i).path[0])
4195                         break;
4196
4197                 if (max_devices == 1) {
4198                         /* Single zoned block device mount */
4199                         FDEV(0).bdev =
4200                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, mode,
4201                                                   sbi->sb->s_type, NULL);
4202                 } else {
4203                         /* Multi-device mount */
4204                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4205                         FDEV(i).total_segments =
4206                                 le32_to_cpu(RDEV(i).total_segments);
4207                         if (i == 0) {
4208                                 FDEV(i).start_blk = 0;
4209                                 FDEV(i).end_blk = FDEV(i).start_blk +
4210                                     (FDEV(i).total_segments <<
4211                                     sbi->log_blocks_per_seg) - 1 +
4212                                     le32_to_cpu(raw_super->segment0_blkaddr);
4213                         } else {
4214                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4215                                 FDEV(i).end_blk = FDEV(i).start_blk +
4216                                         (FDEV(i).total_segments <<
4217                                         sbi->log_blocks_per_seg) - 1;
4218                         }
4219                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, mode,
4220                                                           sbi->sb->s_type,
4221                                                           NULL);
4222                 }
4223                 if (IS_ERR(FDEV(i).bdev))
4224                         return PTR_ERR(FDEV(i).bdev);
4225
4226                 /* to release errored devices */
4227                 sbi->s_ndevs = i + 1;
4228
4229                 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4230                         sbi->aligned_blksize = false;
4231
4232 #ifdef CONFIG_BLK_DEV_ZONED
4233                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
4234                                 !f2fs_sb_has_blkzoned(sbi)) {
4235                         f2fs_err(sbi, "Zoned block device feature not enabled");
4236                         return -EINVAL;
4237                 }
4238                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
4239                         if (init_blkz_info(sbi, i)) {
4240                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4241                                 return -EINVAL;
4242                         }
4243                         if (max_devices == 1)
4244                                 break;
4245                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
4246                                   i, FDEV(i).path,
4247                                   FDEV(i).total_segments,
4248                                   FDEV(i).start_blk, FDEV(i).end_blk,
4249                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
4250                                   "Host-aware" : "Host-managed");
4251                         continue;
4252                 }
4253 #endif
4254                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4255                           i, FDEV(i).path,
4256                           FDEV(i).total_segments,
4257                           FDEV(i).start_blk, FDEV(i).end_blk);
4258         }
4259         f2fs_info(sbi,
4260                   "IO Block Size: %8ld KB", F2FS_IO_SIZE_KB(sbi));
4261         return 0;
4262 }
4263
4264 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4265 {
4266 #if IS_ENABLED(CONFIG_UNICODE)
4267         if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4268                 const struct f2fs_sb_encodings *encoding_info;
4269                 struct unicode_map *encoding;
4270                 __u16 encoding_flags;
4271
4272                 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4273                 if (!encoding_info) {
4274                         f2fs_err(sbi,
4275                                  "Encoding requested by superblock is unknown");
4276                         return -EINVAL;
4277                 }
4278
4279                 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4280                 encoding = utf8_load(encoding_info->version);
4281                 if (IS_ERR(encoding)) {
4282                         f2fs_err(sbi,
4283                                  "can't mount with superblock charset: %s-%u.%u.%u "
4284                                  "not supported by the kernel. flags: 0x%x.",
4285                                  encoding_info->name,
4286                                  unicode_major(encoding_info->version),
4287                                  unicode_minor(encoding_info->version),
4288                                  unicode_rev(encoding_info->version),
4289                                  encoding_flags);
4290                         return PTR_ERR(encoding);
4291                 }
4292                 f2fs_info(sbi, "Using encoding defined by superblock: "
4293                          "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4294                          unicode_major(encoding_info->version),
4295                          unicode_minor(encoding_info->version),
4296                          unicode_rev(encoding_info->version),
4297                          encoding_flags);
4298
4299                 sbi->sb->s_encoding = encoding;
4300                 sbi->sb->s_encoding_flags = encoding_flags;
4301         }
4302 #else
4303         if (f2fs_sb_has_casefold(sbi)) {
4304                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4305                 return -EINVAL;
4306         }
4307 #endif
4308         return 0;
4309 }
4310
4311 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4312 {
4313         /* adjust parameters according to the volume size */
4314         if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4315                 if (f2fs_block_unit_discard(sbi))
4316                         SM_I(sbi)->dcc_info->discard_granularity =
4317                                                 MIN_DISCARD_GRANULARITY;
4318                 if (!f2fs_lfs_mode(sbi))
4319                         SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4320                                                 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4321         }
4322
4323         sbi->readdir_ra = true;
4324 }
4325
4326 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4327 {
4328         struct f2fs_sb_info *sbi;
4329         struct f2fs_super_block *raw_super;
4330         struct inode *root;
4331         int err;
4332         bool skip_recovery = false, need_fsck = false;
4333         char *options = NULL;
4334         int recovery, i, valid_super_block;
4335         struct curseg_info *seg_i;
4336         int retry_cnt = 1;
4337 #ifdef CONFIG_QUOTA
4338         bool quota_enabled = false;
4339 #endif
4340
4341 try_onemore:
4342         err = -EINVAL;
4343         raw_super = NULL;
4344         valid_super_block = -1;
4345         recovery = 0;
4346
4347         /* allocate memory for f2fs-specific super block info */
4348         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4349         if (!sbi)
4350                 return -ENOMEM;
4351
4352         sbi->sb = sb;
4353
4354         /* initialize locks within allocated memory */
4355         init_f2fs_rwsem(&sbi->gc_lock);
4356         mutex_init(&sbi->writepages);
4357         init_f2fs_rwsem(&sbi->cp_global_sem);
4358         init_f2fs_rwsem(&sbi->node_write);
4359         init_f2fs_rwsem(&sbi->node_change);
4360         spin_lock_init(&sbi->stat_lock);
4361         init_f2fs_rwsem(&sbi->cp_rwsem);
4362         init_f2fs_rwsem(&sbi->quota_sem);
4363         init_waitqueue_head(&sbi->cp_wait);
4364         spin_lock_init(&sbi->error_lock);
4365
4366         for (i = 0; i < NR_INODE_TYPE; i++) {
4367                 INIT_LIST_HEAD(&sbi->inode_list[i]);
4368                 spin_lock_init(&sbi->inode_lock[i]);
4369         }
4370         mutex_init(&sbi->flush_lock);
4371
4372         /* Load the checksum driver */
4373         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4374         if (IS_ERR(sbi->s_chksum_driver)) {
4375                 f2fs_err(sbi, "Cannot load crc32 driver.");
4376                 err = PTR_ERR(sbi->s_chksum_driver);
4377                 sbi->s_chksum_driver = NULL;
4378                 goto free_sbi;
4379         }
4380
4381         /* set a block size */
4382         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4383                 f2fs_err(sbi, "unable to set blocksize");
4384                 goto free_sbi;
4385         }
4386
4387         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4388                                                                 &recovery);
4389         if (err)
4390                 goto free_sbi;
4391
4392         sb->s_fs_info = sbi;
4393         sbi->raw_super = raw_super;
4394
4395         INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4396         memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4397         memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4398
4399         /* precompute checksum seed for metadata */
4400         if (f2fs_sb_has_inode_chksum(sbi))
4401                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4402                                                 sizeof(raw_super->uuid));
4403
4404         default_options(sbi, false);
4405         /* parse mount options */
4406         options = kstrdup((const char *)data, GFP_KERNEL);
4407         if (data && !options) {
4408                 err = -ENOMEM;
4409                 goto free_sb_buf;
4410         }
4411
4412         err = parse_options(sb, options, false);
4413         if (err)
4414                 goto free_options;
4415
4416         sb->s_maxbytes = max_file_blocks(NULL) <<
4417                                 le32_to_cpu(raw_super->log_blocksize);
4418         sb->s_max_links = F2FS_LINK_MAX;
4419
4420         err = f2fs_setup_casefold(sbi);
4421         if (err)
4422                 goto free_options;
4423
4424 #ifdef CONFIG_QUOTA
4425         sb->dq_op = &f2fs_quota_operations;
4426         sb->s_qcop = &f2fs_quotactl_ops;
4427         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4428
4429         if (f2fs_sb_has_quota_ino(sbi)) {
4430                 for (i = 0; i < MAXQUOTAS; i++) {
4431                         if (f2fs_qf_ino(sbi->sb, i))
4432                                 sbi->nquota_files++;
4433                 }
4434         }
4435 #endif
4436
4437         sb->s_op = &f2fs_sops;
4438 #ifdef CONFIG_FS_ENCRYPTION
4439         sb->s_cop = &f2fs_cryptops;
4440 #endif
4441 #ifdef CONFIG_FS_VERITY
4442         sb->s_vop = &f2fs_verityops;
4443 #endif
4444         sb->s_xattr = f2fs_xattr_handlers;
4445         sb->s_export_op = &f2fs_export_ops;
4446         sb->s_magic = F2FS_SUPER_MAGIC;
4447         sb->s_time_gran = 1;
4448         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4449                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4450         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4451         sb->s_iflags |= SB_I_CGROUPWB;
4452
4453         /* init f2fs-specific super block info */
4454         sbi->valid_super_block = valid_super_block;
4455
4456         /* disallow all the data/node/meta page writes */
4457         set_sbi_flag(sbi, SBI_POR_DOING);
4458
4459         err = f2fs_init_write_merge_io(sbi);
4460         if (err)
4461                 goto free_bio_info;
4462
4463         init_sb_info(sbi);
4464
4465         err = f2fs_init_iostat(sbi);
4466         if (err)
4467                 goto free_bio_info;
4468
4469         err = init_percpu_info(sbi);
4470         if (err)
4471                 goto free_iostat;
4472
4473         if (F2FS_IO_ALIGNED(sbi)) {
4474                 sbi->write_io_dummy =
4475                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4476                 if (!sbi->write_io_dummy) {
4477                         err = -ENOMEM;
4478                         goto free_percpu;
4479                 }
4480         }
4481
4482         /* init per sbi slab cache */
4483         err = f2fs_init_xattr_caches(sbi);
4484         if (err)
4485                 goto free_io_dummy;
4486         err = f2fs_init_page_array_cache(sbi);
4487         if (err)
4488                 goto free_xattr_cache;
4489
4490         /* get an inode for meta space */
4491         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4492         if (IS_ERR(sbi->meta_inode)) {
4493                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4494                 err = PTR_ERR(sbi->meta_inode);
4495                 goto free_page_array_cache;
4496         }
4497
4498         err = f2fs_get_valid_checkpoint(sbi);
4499         if (err) {
4500                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4501                 goto free_meta_inode;
4502         }
4503
4504         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4505                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4506         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4507                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4508                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4509         }
4510
4511         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4512                 set_sbi_flag(sbi, SBI_NEED_FSCK);
4513
4514         /* Initialize device list */
4515         err = f2fs_scan_devices(sbi);
4516         if (err) {
4517                 f2fs_err(sbi, "Failed to find devices");
4518                 goto free_devices;
4519         }
4520
4521         err = f2fs_init_post_read_wq(sbi);
4522         if (err) {
4523                 f2fs_err(sbi, "Failed to initialize post read workqueue");
4524                 goto free_devices;
4525         }
4526
4527         sbi->total_valid_node_count =
4528                                 le32_to_cpu(sbi->ckpt->valid_node_count);
4529         percpu_counter_set(&sbi->total_valid_inode_count,
4530                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
4531         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4532         sbi->total_valid_block_count =
4533                                 le64_to_cpu(sbi->ckpt->valid_block_count);
4534         sbi->last_valid_block_count = sbi->total_valid_block_count;
4535         sbi->reserved_blocks = 0;
4536         sbi->current_reserved_blocks = 0;
4537         limit_reserve_root(sbi);
4538         adjust_unusable_cap_perc(sbi);
4539
4540         f2fs_init_extent_cache_info(sbi);
4541
4542         f2fs_init_ino_entry_info(sbi);
4543
4544         f2fs_init_fsync_node_info(sbi);
4545
4546         /* setup checkpoint request control and start checkpoint issue thread */
4547         f2fs_init_ckpt_req_control(sbi);
4548         if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4549                         test_opt(sbi, MERGE_CHECKPOINT)) {
4550                 err = f2fs_start_ckpt_thread(sbi);
4551                 if (err) {
4552                         f2fs_err(sbi,
4553                             "Failed to start F2FS issue_checkpoint_thread (%d)",
4554                             err);
4555                         goto stop_ckpt_thread;
4556                 }
4557         }
4558
4559         /* setup f2fs internal modules */
4560         err = f2fs_build_segment_manager(sbi);
4561         if (err) {
4562                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4563                          err);
4564                 goto free_sm;
4565         }
4566         err = f2fs_build_node_manager(sbi);
4567         if (err) {
4568                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4569                          err);
4570                 goto free_nm;
4571         }
4572
4573         err = adjust_reserved_segment(sbi);
4574         if (err)
4575                 goto free_nm;
4576
4577         /* For write statistics */
4578         sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4579
4580         /* Read accumulated write IO statistics if exists */
4581         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4582         if (__exist_node_summaries(sbi))
4583                 sbi->kbytes_written =
4584                         le64_to_cpu(seg_i->journal->info.kbytes_written);
4585
4586         f2fs_build_gc_manager(sbi);
4587
4588         err = f2fs_build_stats(sbi);
4589         if (err)
4590                 goto free_nm;
4591
4592         /* get an inode for node space */
4593         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4594         if (IS_ERR(sbi->node_inode)) {
4595                 f2fs_err(sbi, "Failed to read node inode");
4596                 err = PTR_ERR(sbi->node_inode);
4597                 goto free_stats;
4598         }
4599
4600         /* read root inode and dentry */
4601         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4602         if (IS_ERR(root)) {
4603                 f2fs_err(sbi, "Failed to read root inode");
4604                 err = PTR_ERR(root);
4605                 goto free_node_inode;
4606         }
4607         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4608                         !root->i_size || !root->i_nlink) {
4609                 iput(root);
4610                 err = -EINVAL;
4611                 goto free_node_inode;
4612         }
4613
4614         sb->s_root = d_make_root(root); /* allocate root dentry */
4615         if (!sb->s_root) {
4616                 err = -ENOMEM;
4617                 goto free_node_inode;
4618         }
4619
4620         err = f2fs_init_compress_inode(sbi);
4621         if (err)
4622                 goto free_root_inode;
4623
4624         err = f2fs_register_sysfs(sbi);
4625         if (err)
4626                 goto free_compress_inode;
4627
4628 #ifdef CONFIG_QUOTA
4629         /* Enable quota usage during mount */
4630         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4631                 err = f2fs_enable_quotas(sb);
4632                 if (err)
4633                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4634         }
4635
4636         quota_enabled = f2fs_recover_quota_begin(sbi);
4637 #endif
4638         /* if there are any orphan inodes, free them */
4639         err = f2fs_recover_orphan_inodes(sbi);
4640         if (err)
4641                 goto free_meta;
4642
4643         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4644                 goto reset_checkpoint;
4645
4646         /* recover fsynced data */
4647         if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4648                         !test_opt(sbi, NORECOVERY)) {
4649                 /*
4650                  * mount should be failed, when device has readonly mode, and
4651                  * previous checkpoint was not done by clean system shutdown.
4652                  */
4653                 if (f2fs_hw_is_readonly(sbi)) {
4654                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4655                                 err = f2fs_recover_fsync_data(sbi, true);
4656                                 if (err > 0) {
4657                                         err = -EROFS;
4658                                         f2fs_err(sbi, "Need to recover fsync data, but "
4659                                                 "write access unavailable, please try "
4660                                                 "mount w/ disable_roll_forward or norecovery");
4661                                 }
4662                                 if (err < 0)
4663                                         goto free_meta;
4664                         }
4665                         f2fs_info(sbi, "write access unavailable, skipping recovery");
4666                         goto reset_checkpoint;
4667                 }
4668
4669                 if (need_fsck)
4670                         set_sbi_flag(sbi, SBI_NEED_FSCK);
4671
4672                 if (skip_recovery)
4673                         goto reset_checkpoint;
4674
4675                 err = f2fs_recover_fsync_data(sbi, false);
4676                 if (err < 0) {
4677                         if (err != -ENOMEM)
4678                                 skip_recovery = true;
4679                         need_fsck = true;
4680                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4681                                  err);
4682                         goto free_meta;
4683                 }
4684         } else {
4685                 err = f2fs_recover_fsync_data(sbi, true);
4686
4687                 if (!f2fs_readonly(sb) && err > 0) {
4688                         err = -EINVAL;
4689                         f2fs_err(sbi, "Need to recover fsync data");
4690                         goto free_meta;
4691                 }
4692         }
4693
4694 #ifdef CONFIG_QUOTA
4695         f2fs_recover_quota_end(sbi, quota_enabled);
4696 #endif
4697
4698         /*
4699          * If the f2fs is not readonly and fsync data recovery succeeds,
4700          * check zoned block devices' write pointer consistency.
4701          */
4702         if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4703                 err = f2fs_check_write_pointer(sbi);
4704                 if (err)
4705                         goto free_meta;
4706         }
4707
4708 reset_checkpoint:
4709         f2fs_init_inmem_curseg(sbi);
4710
4711         /* f2fs_recover_fsync_data() cleared this already */
4712         clear_sbi_flag(sbi, SBI_POR_DOING);
4713
4714         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4715                 err = f2fs_disable_checkpoint(sbi);
4716                 if (err)
4717                         goto sync_free_meta;
4718         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4719                 f2fs_enable_checkpoint(sbi);
4720         }
4721
4722         /*
4723          * If filesystem is not mounted as read-only then
4724          * do start the gc_thread.
4725          */
4726         if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4727                 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4728                 /* After POR, we can run background GC thread.*/
4729                 err = f2fs_start_gc_thread(sbi);
4730                 if (err)
4731                         goto sync_free_meta;
4732         }
4733         kvfree(options);
4734
4735         /* recover broken superblock */
4736         if (recovery) {
4737                 err = f2fs_commit_super(sbi, true);
4738                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4739                           sbi->valid_super_block ? 1 : 2, err);
4740         }
4741
4742         f2fs_join_shrinker(sbi);
4743
4744         f2fs_tuning_parameters(sbi);
4745
4746         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4747                     cur_cp_version(F2FS_CKPT(sbi)));
4748         f2fs_update_time(sbi, CP_TIME);
4749         f2fs_update_time(sbi, REQ_TIME);
4750         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4751         return 0;
4752
4753 sync_free_meta:
4754         /* safe to flush all the data */
4755         sync_filesystem(sbi->sb);
4756         retry_cnt = 0;
4757
4758 free_meta:
4759 #ifdef CONFIG_QUOTA
4760         f2fs_truncate_quota_inode_pages(sb);
4761         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4762                 f2fs_quota_off_umount(sbi->sb);
4763 #endif
4764         /*
4765          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4766          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4767          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4768          * falls into an infinite loop in f2fs_sync_meta_pages().
4769          */
4770         truncate_inode_pages_final(META_MAPPING(sbi));
4771         /* evict some inodes being cached by GC */
4772         evict_inodes(sb);
4773         f2fs_unregister_sysfs(sbi);
4774 free_compress_inode:
4775         f2fs_destroy_compress_inode(sbi);
4776 free_root_inode:
4777         dput(sb->s_root);
4778         sb->s_root = NULL;
4779 free_node_inode:
4780         f2fs_release_ino_entry(sbi, true);
4781         truncate_inode_pages_final(NODE_MAPPING(sbi));
4782         iput(sbi->node_inode);
4783         sbi->node_inode = NULL;
4784 free_stats:
4785         f2fs_destroy_stats(sbi);
4786 free_nm:
4787         /* stop discard thread before destroying node manager */
4788         f2fs_stop_discard_thread(sbi);
4789         f2fs_destroy_node_manager(sbi);
4790 free_sm:
4791         f2fs_destroy_segment_manager(sbi);
4792 stop_ckpt_thread:
4793         f2fs_stop_ckpt_thread(sbi);
4794         /* flush s_error_work before sbi destroy */
4795         flush_work(&sbi->s_error_work);
4796         f2fs_destroy_post_read_wq(sbi);
4797 free_devices:
4798         destroy_device_list(sbi);
4799         kvfree(sbi->ckpt);
4800 free_meta_inode:
4801         make_bad_inode(sbi->meta_inode);
4802         iput(sbi->meta_inode);
4803         sbi->meta_inode = NULL;
4804 free_page_array_cache:
4805         f2fs_destroy_page_array_cache(sbi);
4806 free_xattr_cache:
4807         f2fs_destroy_xattr_caches(sbi);
4808 free_io_dummy:
4809         mempool_destroy(sbi->write_io_dummy);
4810 free_percpu:
4811         destroy_percpu_info(sbi);
4812 free_iostat:
4813         f2fs_destroy_iostat(sbi);
4814 free_bio_info:
4815         for (i = 0; i < NR_PAGE_TYPE; i++)
4816                 kvfree(sbi->write_io[i]);
4817
4818 #if IS_ENABLED(CONFIG_UNICODE)
4819         utf8_unload(sb->s_encoding);
4820         sb->s_encoding = NULL;
4821 #endif
4822 free_options:
4823 #ifdef CONFIG_QUOTA
4824         for (i = 0; i < MAXQUOTAS; i++)
4825                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4826 #endif
4827         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4828         kvfree(options);
4829 free_sb_buf:
4830         kfree(raw_super);
4831 free_sbi:
4832         if (sbi->s_chksum_driver)
4833                 crypto_free_shash(sbi->s_chksum_driver);
4834         kfree(sbi);
4835
4836         /* give only one another chance */
4837         if (retry_cnt > 0 && skip_recovery) {
4838                 retry_cnt--;
4839                 shrink_dcache_sb(sb);
4840                 goto try_onemore;
4841         }
4842         return err;
4843 }
4844
4845 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4846                         const char *dev_name, void *data)
4847 {
4848         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4849 }
4850
4851 static void kill_f2fs_super(struct super_block *sb)
4852 {
4853         if (sb->s_root) {
4854                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4855
4856                 set_sbi_flag(sbi, SBI_IS_CLOSE);
4857                 f2fs_stop_gc_thread(sbi);
4858                 f2fs_stop_discard_thread(sbi);
4859
4860 #ifdef CONFIG_F2FS_FS_COMPRESSION
4861                 /*
4862                  * latter evict_inode() can bypass checking and invalidating
4863                  * compress inode cache.
4864                  */
4865                 if (test_opt(sbi, COMPRESS_CACHE))
4866                         truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4867 #endif
4868
4869                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4870                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4871                         struct cp_control cpc = {
4872                                 .reason = CP_UMOUNT,
4873                         };
4874                         f2fs_write_checkpoint(sbi, &cpc);
4875                 }
4876
4877                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4878                         sb->s_flags &= ~SB_RDONLY;
4879         }
4880         kill_block_super(sb);
4881 }
4882
4883 static struct file_system_type f2fs_fs_type = {
4884         .owner          = THIS_MODULE,
4885         .name           = "f2fs",
4886         .mount          = f2fs_mount,
4887         .kill_sb        = kill_f2fs_super,
4888         .fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4889 };
4890 MODULE_ALIAS_FS("f2fs");
4891
4892 static int __init init_inodecache(void)
4893 {
4894         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4895                         sizeof(struct f2fs_inode_info), 0,
4896                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4897         return f2fs_inode_cachep ? 0 : -ENOMEM;
4898 }
4899
4900 static void destroy_inodecache(void)
4901 {
4902         /*
4903          * Make sure all delayed rcu free inodes are flushed before we
4904          * destroy cache.
4905          */
4906         rcu_barrier();
4907         kmem_cache_destroy(f2fs_inode_cachep);
4908 }
4909
4910 static int __init init_f2fs_fs(void)
4911 {
4912         int err;
4913
4914         if (PAGE_SIZE != F2FS_BLKSIZE) {
4915                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4916                                 PAGE_SIZE, F2FS_BLKSIZE);
4917                 return -EINVAL;
4918         }
4919
4920         err = init_inodecache();
4921         if (err)
4922                 goto fail;
4923         err = f2fs_create_node_manager_caches();
4924         if (err)
4925                 goto free_inodecache;
4926         err = f2fs_create_segment_manager_caches();
4927         if (err)
4928                 goto free_node_manager_caches;
4929         err = f2fs_create_checkpoint_caches();
4930         if (err)
4931                 goto free_segment_manager_caches;
4932         err = f2fs_create_recovery_cache();
4933         if (err)
4934                 goto free_checkpoint_caches;
4935         err = f2fs_create_extent_cache();
4936         if (err)
4937                 goto free_recovery_cache;
4938         err = f2fs_create_garbage_collection_cache();
4939         if (err)
4940                 goto free_extent_cache;
4941         err = f2fs_init_sysfs();
4942         if (err)
4943                 goto free_garbage_collection_cache;
4944         err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4945         if (err)
4946                 goto free_sysfs;
4947         err = register_filesystem(&f2fs_fs_type);
4948         if (err)
4949                 goto free_shrinker;
4950         f2fs_create_root_stats();
4951         err = f2fs_init_post_read_processing();
4952         if (err)
4953                 goto free_root_stats;
4954         err = f2fs_init_iostat_processing();
4955         if (err)
4956                 goto free_post_read;
4957         err = f2fs_init_bio_entry_cache();
4958         if (err)
4959                 goto free_iostat;
4960         err = f2fs_init_bioset();
4961         if (err)
4962                 goto free_bio_entry_cache;
4963         err = f2fs_init_compress_mempool();
4964         if (err)
4965                 goto free_bioset;
4966         err = f2fs_init_compress_cache();
4967         if (err)
4968                 goto free_compress_mempool;
4969         err = f2fs_create_casefold_cache();
4970         if (err)
4971                 goto free_compress_cache;
4972         return 0;
4973 free_compress_cache:
4974         f2fs_destroy_compress_cache();
4975 free_compress_mempool:
4976         f2fs_destroy_compress_mempool();
4977 free_bioset:
4978         f2fs_destroy_bioset();
4979 free_bio_entry_cache:
4980         f2fs_destroy_bio_entry_cache();
4981 free_iostat:
4982         f2fs_destroy_iostat_processing();
4983 free_post_read:
4984         f2fs_destroy_post_read_processing();
4985 free_root_stats:
4986         f2fs_destroy_root_stats();
4987         unregister_filesystem(&f2fs_fs_type);
4988 free_shrinker:
4989         unregister_shrinker(&f2fs_shrinker_info);
4990 free_sysfs:
4991         f2fs_exit_sysfs();
4992 free_garbage_collection_cache:
4993         f2fs_destroy_garbage_collection_cache();
4994 free_extent_cache:
4995         f2fs_destroy_extent_cache();
4996 free_recovery_cache:
4997         f2fs_destroy_recovery_cache();
4998 free_checkpoint_caches:
4999         f2fs_destroy_checkpoint_caches();
5000 free_segment_manager_caches:
5001         f2fs_destroy_segment_manager_caches();
5002 free_node_manager_caches:
5003         f2fs_destroy_node_manager_caches();
5004 free_inodecache:
5005         destroy_inodecache();
5006 fail:
5007         return err;
5008 }
5009
5010 static void __exit exit_f2fs_fs(void)
5011 {
5012         f2fs_destroy_casefold_cache();
5013         f2fs_destroy_compress_cache();
5014         f2fs_destroy_compress_mempool();
5015         f2fs_destroy_bioset();
5016         f2fs_destroy_bio_entry_cache();
5017         f2fs_destroy_iostat_processing();
5018         f2fs_destroy_post_read_processing();
5019         f2fs_destroy_root_stats();
5020         unregister_filesystem(&f2fs_fs_type);
5021         unregister_shrinker(&f2fs_shrinker_info);
5022         f2fs_exit_sysfs();
5023         f2fs_destroy_garbage_collection_cache();
5024         f2fs_destroy_extent_cache();
5025         f2fs_destroy_recovery_cache();
5026         f2fs_destroy_checkpoint_caches();
5027         f2fs_destroy_segment_manager_caches();
5028         f2fs_destroy_node_manager_caches();
5029         destroy_inodecache();
5030 }
5031
5032 module_init(init_f2fs_fs)
5033 module_exit(exit_f2fs_fs)
5034
5035 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5036 MODULE_DESCRIPTION("Flash Friendly File System");
5037 MODULE_LICENSE("GPL");
5038 MODULE_SOFTDEP("pre: crc32");
5039