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_err,
 168 };
 169
 170 static match_table_t f2fs_tokens = {
 171         {Opt_gc_background, "background_gc=%s"},
 172         {Opt_disable_roll_forward, "disable_roll_forward"},
 173         {Opt_norecovery, "norecovery"},
 174         {Opt_discard, "discard"},
 175         {Opt_nodiscard, "nodiscard"},
 176         {Opt_noheap, "no_heap"},
 177         {Opt_heap, "heap"},
 178         {Opt_user_xattr, "user_xattr"},
 179         {Opt_nouser_xattr, "nouser_xattr"},
 180         {Opt_acl, "acl"},
 181         {Opt_noacl, "noacl"},
 182         {Opt_active_logs, "active_logs=%u"},
 183         {Opt_disable_ext_identify, "disable_ext_identify"},
 184         {Opt_inline_xattr, "inline_xattr"},
 185         {Opt_noinline_xattr, "noinline_xattr"},
 186         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 187         {Opt_inline_data, "inline_data"},
 188         {Opt_inline_dentry, "inline_dentry"},
 189         {Opt_noinline_dentry, "noinline_dentry"},
 190         {Opt_flush_merge, "flush_merge"},
 191         {Opt_noflush_merge, "noflush_merge"},
 192         {Opt_barrier, "barrier"},
 193         {Opt_nobarrier, "nobarrier"},
 194         {Opt_fastboot, "fastboot"},
 195         {Opt_extent_cache, "extent_cache"},
 196         {Opt_noextent_cache, "noextent_cache"},
 197         {Opt_noinline_data, "noinline_data"},
 198         {Opt_data_flush, "data_flush"},
 199         {Opt_reserve_root, "reserve_root=%u"},
 200         {Opt_resgid, "resgid=%u"},
 201         {Opt_resuid, "resuid=%u"},
 202         {Opt_mode, "mode=%s"},
 203         {Opt_io_size_bits, "io_bits=%u"},
 204         {Opt_fault_injection, "fault_injection=%u"},
 205         {Opt_fault_type, "fault_type=%u"},
 206         {Opt_lazytime, "lazytime"},
 207         {Opt_nolazytime, "nolazytime"},
 208         {Opt_quota, "quota"},
 209         {Opt_noquota, "noquota"},
 210         {Opt_usrquota, "usrquota"},
 211         {Opt_grpquota, "grpquota"},
 212         {Opt_prjquota, "prjquota"},
 213         {Opt_usrjquota, "usrjquota=%s"},
 214         {Opt_grpjquota, "grpjquota=%s"},
 215         {Opt_prjjquota, "prjjquota=%s"},
 216         {Opt_offusrjquota, "usrjquota="},
 217         {Opt_offgrpjquota, "grpjquota="},
 218         {Opt_offprjjquota, "prjjquota="},
 219         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 220         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 221         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 222         {Opt_alloc, "alloc_mode=%s"},
 223         {Opt_fsync, "fsync_mode=%s"},
 224         {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
 225         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 226         {Opt_inlinecrypt, "inlinecrypt"},
 227         {Opt_checkpoint_disable, "checkpoint=disable"},
 228         {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 229         {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 230         {Opt_checkpoint_enable, "checkpoint=enable"},
 231         {Opt_checkpoint_merge, "checkpoint_merge"},
 232         {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
 233         {Opt_compress_algorithm, "compress_algorithm=%s"},
 234         {Opt_compress_log_size, "compress_log_size=%u"},
 235         {Opt_compress_extension, "compress_extension=%s"},
 236         {Opt_nocompress_extension, "nocompress_extension=%s"},
 237         {Opt_compress_chksum, "compress_chksum"},
 238         {Opt_compress_mode, "compress_mode=%s"},
 239         {Opt_compress_cache, "compress_cache"},
 240         {Opt_atgc, "atgc"},
 241         {Opt_gc_merge, "gc_merge"},
 242         {Opt_nogc_merge, "nogc_merge"},
 243         {Opt_discard_unit, "discard_unit=%s"},
 244         {Opt_memory_mode, "memory=%s"},
 245         {Opt_age_extent_cache, "age_extent_cache"},
 246         {Opt_err, NULL},
 247 };
 248
 249 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 250 {
 251         struct va_format vaf;
 252         va_list args;
 253         int level;
 254
 255         va_start(args, fmt);
 256
 257         level = printk_get_level(fmt);
 258         vaf.fmt = printk_skip_level(fmt);
 259         vaf.va = &args;
 260         printk("%c%cF2FS-fs (%s): %pV\n",
 261                KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 262
 263         va_end(args);
 264 }
 265
 266 #if IS_ENABLED(CONFIG_UNICODE)
 267 static const struct f2fs_sb_encodings {
 268         __u16 magic;
 269         char *name;
 270         unsigned int version;
 271 } f2fs_sb_encoding_map[] = {
 272         {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
 273 };
 274
 275 static const struct f2fs_sb_encodings *
 276 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
 277 {
 278         __u16 magic = le16_to_cpu(sb->s_encoding);
 279         int i;
 280
 281         for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
 282                 if (magic == f2fs_sb_encoding_map[i].magic)
 283                         return &f2fs_sb_encoding_map[i];
 284
 285         return NULL;
 286 }
 287
 288 struct kmem_cache *f2fs_cf_name_slab;
 289 static int __init f2fs_create_casefold_cache(void)
 290 {
 291         f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
 292                                                         F2FS_NAME_LEN);
 293         return f2fs_cf_name_slab ? 0 : -ENOMEM;
 294 }
 295
 296 static void f2fs_destroy_casefold_cache(void)
 297 {
 298         kmem_cache_destroy(f2fs_cf_name_slab);
 299 }
 300 #else
 301 static int __init f2fs_create_casefold_cache(void) { return 0; }
 302 static void f2fs_destroy_casefold_cache(void) { }
 303 #endif
 304
 305 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 306 {
 307         block_t limit = min((sbi->user_block_count >> 3),
 308                         sbi->user_block_count - sbi->reserved_blocks);
 309
 310         /* limit is 12.5% */
 311         if (test_opt(sbi, RESERVE_ROOT) &&
 312                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 313                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 314                 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 315                           F2FS_OPTION(sbi).root_reserved_blocks);
 316         }
 317         if (!test_opt(sbi, RESERVE_ROOT) &&
 318                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 319                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 320                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 321                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 322                 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 323                           from_kuid_munged(&init_user_ns,
 324                                            F2FS_OPTION(sbi).s_resuid),
 325                           from_kgid_munged(&init_user_ns,
 326                                            F2FS_OPTION(sbi).s_resgid));
 327 }
 328
 329 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
 330 {
 331         unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
 332         unsigned int avg_vblocks;
 333         unsigned int wanted_reserved_segments;
 334         block_t avail_user_block_count;
 335
 336         if (!F2FS_IO_ALIGNED(sbi))
 337                 return 0;
 338
 339         /* average valid block count in section in worst case */
 340         avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
 341
 342         /*
 343          * we need enough free space when migrating one section in worst case
 344          */
 345         wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
 346                                                 reserved_segments(sbi);
 347         wanted_reserved_segments -= reserved_segments(sbi);
 348
 349         avail_user_block_count = sbi->user_block_count -
 350                                 sbi->current_reserved_blocks -
 351                                 F2FS_OPTION(sbi).root_reserved_blocks;
 352
 353         if (wanted_reserved_segments * sbi->blocks_per_seg >
 354                                         avail_user_block_count) {
 355                 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
 356                         wanted_reserved_segments,
 357                         avail_user_block_count >> sbi->log_blocks_per_seg);
 358                 return -ENOSPC;
 359         }
 360
 361         SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
 362
 363         f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
 364                          wanted_reserved_segments);
 365
 366         return 0;
 367 }
 368
 369 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
 370 {
 371         if (!F2FS_OPTION(sbi).unusable_cap_perc)
 372                 return;
 373
 374         if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
 375                 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
 376         else
 377                 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
 378                                         F2FS_OPTION(sbi).unusable_cap_perc;
 379
 380         f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
 381                         F2FS_OPTION(sbi).unusable_cap,
 382                         F2FS_OPTION(sbi).unusable_cap_perc);
 383 }
 384
 385 static void init_once(void *foo)
 386 {
 387         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 388
 389         inode_init_once(&fi->vfs_inode);
 390 }
 391
 392 #ifdef CONFIG_QUOTA
 393 static const char * const quotatypes[] = INITQFNAMES;
 394 #define QTYPE2NAME(t) (quotatypes[t])
 395 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 396                                                         substring_t *args)
 397 {
 398         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 399         char *qname;
 400         int ret = -EINVAL;
 401
 402         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 403                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 404                 return -EINVAL;
 405         }
 406         if (f2fs_sb_has_quota_ino(sbi)) {
 407                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 408                 return 0;
 409         }
 410
 411         qname = match_strdup(args);
 412         if (!qname) {
 413                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 414                 return -ENOMEM;
 415         }
 416         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 417                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 418                         ret = 0;
 419                 else
 420                         f2fs_err(sbi, "%s quota file already specified",
 421                                  QTYPE2NAME(qtype));
 422                 goto errout;
 423         }
 424         if (strchr(qname, '/')) {
 425                 f2fs_err(sbi, "quotafile must be on filesystem root");
 426                 goto errout;
 427         }
 428         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 429         set_opt(sbi, QUOTA);
 430         return 0;
 431 errout:
 432         kfree(qname);
 433         return ret;
 434 }
 435
 436 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 437 {
 438         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 439
 440         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 441                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 442                 return -EINVAL;
 443         }
 444         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 445         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 446         return 0;
 447 }
 448
 449 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 450 {
 451         /*
 452          * We do the test below only for project quotas. 'usrquota' and
 453          * 'grpquota' mount options are allowed even without quota feature
 454          * to support legacy quotas in quota files.
 455          */
 456         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 457                 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 458                 return -1;
 459         }
 460         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 461                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 462                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 463                 if (test_opt(sbi, USRQUOTA) &&
 464                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 465                         clear_opt(sbi, USRQUOTA);
 466
 467                 if (test_opt(sbi, GRPQUOTA) &&
 468                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 469                         clear_opt(sbi, GRPQUOTA);
 470
 471                 if (test_opt(sbi, PRJQUOTA) &&
 472                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 473                         clear_opt(sbi, PRJQUOTA);
 474
 475                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 476                                 test_opt(sbi, PRJQUOTA)) {
 477                         f2fs_err(sbi, "old and new quota format mixing");
 478                         return -1;
 479                 }
 480
 481                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 482                         f2fs_err(sbi, "journaled quota format not specified");
 483                         return -1;
 484                 }
 485         }
 486
 487         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 488                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 489                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 490         }
 491         return 0;
 492 }
 493 #endif
 494
 495 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
 496                                           const char *opt,
 497                                           const substring_t *arg,
 498                                           bool is_remount)
 499 {
 500         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 501         struct fs_parameter param = {
 502                 .type = fs_value_is_string,
 503                 .string = arg->from ? arg->from : "",
 504         };
 505         struct fscrypt_dummy_policy *policy =
 506                 &F2FS_OPTION(sbi).dummy_enc_policy;
 507         int err;
 508
 509         if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
 510                 f2fs_warn(sbi, "test_dummy_encryption option not supported");
 511                 return -EINVAL;
 512         }
 513
 514         if (!f2fs_sb_has_encrypt(sbi)) {
 515                 f2fs_err(sbi, "Encrypt feature is off");
 516                 return -EINVAL;
 517         }
 518
 519         /*
 520          * This mount option is just for testing, and it's not worthwhile to
 521          * implement the extra complexity (e.g. RCU protection) that would be
 522          * needed to allow it to be set or changed during remount.  We do allow
 523          * it to be specified during remount, but only if there is no change.
 524          */
 525         if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
 526                 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
 527                 return -EINVAL;
 528         }
 529
 530         err = fscrypt_parse_test_dummy_encryption(&param, policy);
 531         if (err) {
 532                 if (err == -EEXIST)
 533                         f2fs_warn(sbi,
 534                                   "Can't change test_dummy_encryption on remount");
 535                 else if (err == -EINVAL)
 536                         f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
 537                                   opt);
 538                 else
 539                         f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
 540                                   opt, err);
 541                 return -EINVAL;
 542         }
 543         err = fscrypt_add_test_dummy_key(sb, policy);
 544         if (err) {
 545                 f2fs_warn(sbi, "Error adding test dummy encryption key [%d]",
 546                           err);
 547                 return err;
 548         }
 549         f2fs_warn(sbi, "Test dummy encryption mode enabled");
 550         return 0;
 551 }
 552
 553 #ifdef CONFIG_F2FS_FS_COMPRESSION
 554 /*
 555  * 1. The same extension name cannot not appear in both compress and non-compress extension
 556  * at the same time.
 557  * 2. If the compress extension specifies all files, the types specified by the non-compress
 558  * extension will be treated as special cases and will not be compressed.
 559  * 3. Don't allow the non-compress extension specifies all files.
 560  */
 561 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
 562 {
 563         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 564         unsigned char (*noext)[F2FS_EXTENSION_LEN];
 565         int ext_cnt, noext_cnt, index = 0, no_index = 0;
 566
 567         ext = F2FS_OPTION(sbi).extensions;
 568         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
 569         noext = F2FS_OPTION(sbi).noextensions;
 570         noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
 571
 572         if (!noext_cnt)
 573                 return 0;
 574
 575         for (no_index = 0; no_index < noext_cnt; no_index++) {
 576                 if (!strcasecmp("*", noext[no_index])) {
 577                         f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
 578                         return -EINVAL;
 579                 }
 580                 for (index = 0; index < ext_cnt; index++) {
 581                         if (!strcasecmp(ext[index], noext[no_index])) {
 582                                 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
 583                                                 ext[index]);
 584                                 return -EINVAL;
 585                         }
 586                 }
 587         }
 588         return 0;
 589 }
 590
 591 #ifdef CONFIG_F2FS_FS_LZ4
 592 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
 593 {
 594 #ifdef CONFIG_F2FS_FS_LZ4HC
 595         unsigned int level;
 596 #endif
 597
 598         if (strlen(str) == 3) {
 599                 F2FS_OPTION(sbi).compress_level = 0;
 600                 return 0;
 601         }
 602
 603 #ifdef CONFIG_F2FS_FS_LZ4HC
 604         str += 3;
 605
 606         if (str[0] != ':') {
 607                 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
 608                 return -EINVAL;
 609         }
 610         if (kstrtouint(str + 1, 10, &level))
 611                 return -EINVAL;
 612
 613         if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) {
 614                 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
 615                 return -EINVAL;
 616         }
 617
 618         F2FS_OPTION(sbi).compress_level = level;
 619         return 0;
 620 #else
 621         f2fs_info(sbi, "kernel doesn't support lz4hc compression");
 622         return -EINVAL;
 623 #endif
 624 }
 625 #endif
 626
 627 #ifdef CONFIG_F2FS_FS_ZSTD
 628 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
 629 {
 630         unsigned int level;
 631         int len = 4;
 632
 633         if (strlen(str) == len) {
 634                 F2FS_OPTION(sbi).compress_level = 0;
 635                 return 0;
 636         }
 637
 638         str += len;
 639
 640         if (str[0] != ':') {
 641                 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
 642                 return -EINVAL;
 643         }
 644         if (kstrtouint(str + 1, 10, &level))
 645                 return -EINVAL;
 646
 647         if (!level || level > zstd_max_clevel()) {
 648                 f2fs_info(sbi, "invalid zstd compress level: %d", level);
 649                 return -EINVAL;
 650         }
 651
 652         F2FS_OPTION(sbi).compress_level = level;
 653         return 0;
 654 }
 655 #endif
 656 #endif
 657
 658 static int parse_options(struct super_block *sb, char *options, bool is_remount)
 659 {
 660         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 661         substring_t args[MAX_OPT_ARGS];
 662 #ifdef CONFIG_F2FS_FS_COMPRESSION
 663         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 664         unsigned char (*noext)[F2FS_EXTENSION_LEN];
 665         int ext_cnt, noext_cnt;
 666 #endif
 667         char *p, *name;
 668         int arg = 0;
 669         kuid_t uid;
 670         kgid_t gid;
 671         int ret;
 672
 673         if (!options)
 674                 goto default_check;
 675
 676         while ((p = strsep(&options, ",")) != NULL) {
 677                 int token;
 678
 679                 if (!*p)
 680                         continue;
 681                 /*
 682                  * Initialize args struct so we know whether arg was
 683                  * found; some options take optional arguments.
 684                  */
 685                 args[0].to = args[0].from = NULL;
 686                 token = match_token(p, f2fs_tokens, args);
 687
 688                 switch (token) {
 689                 case Opt_gc_background:
 690                         name = match_strdup(&args[0]);
 691
 692                         if (!name)
 693                                 return -ENOMEM;
 694                         if (!strcmp(name, "on")) {
 695                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
 696                         } else if (!strcmp(name, "off")) {
 697                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
 698                         } else if (!strcmp(name, "sync")) {
 699                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
 700                         } else {
 701                                 kfree(name);
 702                                 return -EINVAL;
 703                         }
 704                         kfree(name);
 705                         break;
 706                 case Opt_disable_roll_forward:
 707                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 708                         break;
 709                 case Opt_norecovery:
 710                         /* this option mounts f2fs with ro */
 711                         set_opt(sbi, NORECOVERY);
 712                         if (!f2fs_readonly(sb))
 713                                 return -EINVAL;
 714                         break;
 715                 case Opt_discard:
 716                         if (!f2fs_hw_support_discard(sbi)) {
 717                                 f2fs_warn(sbi, "device does not support discard");
 718                                 break;
 719                         }
 720                         set_opt(sbi, DISCARD);
 721                         break;
 722                 case Opt_nodiscard:
 723                         if (f2fs_hw_should_discard(sbi)) {
 724                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 725                                 return -EINVAL;
 726                         }
 727                         clear_opt(sbi, DISCARD);
 728                         break;
 729                 case Opt_noheap:
 730                         set_opt(sbi, NOHEAP);
 731                         break;
 732                 case Opt_heap:
 733                         clear_opt(sbi, NOHEAP);
 734                         break;
 735 #ifdef CONFIG_F2FS_FS_XATTR
 736                 case Opt_user_xattr:
 737                         set_opt(sbi, XATTR_USER);
 738                         break;
 739                 case Opt_nouser_xattr:
 740                         clear_opt(sbi, XATTR_USER);
 741                         break;
 742                 case Opt_inline_xattr:
 743                         set_opt(sbi, INLINE_XATTR);
 744                         break;
 745                 case Opt_noinline_xattr:
 746                         clear_opt(sbi, INLINE_XATTR);
 747                         break;
 748                 case Opt_inline_xattr_size:
 749                         if (args->from && match_int(args, &arg))
 750                                 return -EINVAL;
 751                         set_opt(sbi, INLINE_XATTR_SIZE);
 752                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 753                         break;
 754 #else
 755                 case Opt_user_xattr:
 756                         f2fs_info(sbi, "user_xattr options not supported");
 757                         break;
 758                 case Opt_nouser_xattr:
 759                         f2fs_info(sbi, "nouser_xattr options not supported");
 760                         break;
 761                 case Opt_inline_xattr:
 762                         f2fs_info(sbi, "inline_xattr options not supported");
 763                         break;
 764                 case Opt_noinline_xattr:
 765                         f2fs_info(sbi, "noinline_xattr options not supported");
 766                         break;
 767 #endif
 768 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 769                 case Opt_acl:
 770                         set_opt(sbi, POSIX_ACL);
 771                         break;
 772                 case Opt_noacl:
 773                         clear_opt(sbi, POSIX_ACL);
 774                         break;
 775 #else
 776                 case Opt_acl:
 777                         f2fs_info(sbi, "acl options not supported");
 778                         break;
 779                 case Opt_noacl:
 780                         f2fs_info(sbi, "noacl options not supported");
 781                         break;
 782 #endif
 783                 case Opt_active_logs:
 784                         if (args->from && match_int(args, &arg))
 785                                 return -EINVAL;
 786                         if (arg != 2 && arg != 4 &&
 787                                 arg != NR_CURSEG_PERSIST_TYPE)
 788                                 return -EINVAL;
 789                         F2FS_OPTION(sbi).active_logs = arg;
 790                         break;
 791                 case Opt_disable_ext_identify:
 792                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 793                         break;
 794                 case Opt_inline_data:
 795                         set_opt(sbi, INLINE_DATA);
 796                         break;
 797                 case Opt_inline_dentry:
 798                         set_opt(sbi, INLINE_DENTRY);
 799                         break;
 800                 case Opt_noinline_dentry:
 801                         clear_opt(sbi, INLINE_DENTRY);
 802                         break;
 803                 case Opt_flush_merge:
 804                         set_opt(sbi, FLUSH_MERGE);
 805                         break;
 806                 case Opt_noflush_merge:
 807                         clear_opt(sbi, FLUSH_MERGE);
 808                         break;
 809                 case Opt_nobarrier:
 810                         set_opt(sbi, NOBARRIER);
 811                         break;
 812                 case Opt_barrier:
 813                         clear_opt(sbi, NOBARRIER);
 814                         break;
 815                 case Opt_fastboot:
 816                         set_opt(sbi, FASTBOOT);
 817                         break;
 818                 case Opt_extent_cache:
 819                         set_opt(sbi, READ_EXTENT_CACHE);
 820                         break;
 821                 case Opt_noextent_cache:
 822                         clear_opt(sbi, READ_EXTENT_CACHE);
 823                         break;
 824                 case Opt_noinline_data:
 825                         clear_opt(sbi, INLINE_DATA);
 826                         break;
 827                 case Opt_data_flush:
 828                         set_opt(sbi, DATA_FLUSH);
 829                         break;
 830                 case Opt_reserve_root:
 831                         if (args->from && match_int(args, &arg))
 832                                 return -EINVAL;
 833                         if (test_opt(sbi, RESERVE_ROOT)) {
 834                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 835                                           F2FS_OPTION(sbi).root_reserved_blocks);
 836                         } else {
 837                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 838                                 set_opt(sbi, RESERVE_ROOT);
 839                         }
 840                         break;
 841                 case Opt_resuid:
 842                         if (args->from && match_int(args, &arg))
 843                                 return -EINVAL;
 844                         uid = make_kuid(current_user_ns(), arg);
 845                         if (!uid_valid(uid)) {
 846                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 847                                 return -EINVAL;
 848                         }
 849                         F2FS_OPTION(sbi).s_resuid = uid;
 850                         break;
 851                 case Opt_resgid:
 852                         if (args->from && match_int(args, &arg))
 853                                 return -EINVAL;
 854                         gid = make_kgid(current_user_ns(), arg);
 855                         if (!gid_valid(gid)) {
 856                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 857                                 return -EINVAL;
 858                         }
 859                         F2FS_OPTION(sbi).s_resgid = gid;
 860                         break;
 861                 case Opt_mode:
 862                         name = match_strdup(&args[0]);
 863
 864                         if (!name)
 865                                 return -ENOMEM;
 866                         if (!strcmp(name, "adaptive")) {
 867                                 if (f2fs_sb_has_blkzoned(sbi)) {
 868                                         f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
 869                                         kfree(name);
 870                                         return -EINVAL;
 871                                 }
 872                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
 873                         } else if (!strcmp(name, "lfs")) {
 874                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
 875                         } else if (!strcmp(name, "fragment:segment")) {
 876                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
 877                         } else if (!strcmp(name, "fragment:block")) {
 878                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
 879                         } else {
 880                                 kfree(name);
 881                                 return -EINVAL;
 882                         }
 883                         kfree(name);
 884                         break;
 885                 case Opt_io_size_bits:
 886                         if (args->from && match_int(args, &arg))
 887                                 return -EINVAL;
 888                         if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) {
 889                                 f2fs_warn(sbi, "Not support %d, larger than %d",
 890                                           1 << arg, BIO_MAX_VECS);
 891                                 return -EINVAL;
 892                         }
 893                         F2FS_OPTION(sbi).write_io_size_bits = arg;
 894                         break;
 895 #ifdef CONFIG_F2FS_FAULT_INJECTION
 896                 case Opt_fault_injection:
 897                         if (args->from && match_int(args, &arg))
 898                                 return -EINVAL;
 899                         f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 900                         set_opt(sbi, FAULT_INJECTION);
 901                         break;
 902
 903                 case Opt_fault_type:
 904                         if (args->from && match_int(args, &arg))
 905                                 return -EINVAL;
 906                         f2fs_build_fault_attr(sbi, 0, arg);
 907                         set_opt(sbi, FAULT_INJECTION);
 908                         break;
 909 #else
 910                 case Opt_fault_injection:
 911                         f2fs_info(sbi, "fault_injection options not supported");
 912                         break;
 913
 914                 case Opt_fault_type:
 915                         f2fs_info(sbi, "fault_type options not supported");
 916                         break;
 917 #endif
 918                 case Opt_lazytime:
 919                         sb->s_flags |= SB_LAZYTIME;
 920                         break;
 921                 case Opt_nolazytime:
 922                         sb->s_flags &= ~SB_LAZYTIME;
 923                         break;
 924 #ifdef CONFIG_QUOTA
 925                 case Opt_quota:
 926                 case Opt_usrquota:
 927                         set_opt(sbi, USRQUOTA);
 928                         break;
 929                 case Opt_grpquota:
 930                         set_opt(sbi, GRPQUOTA);
 931                         break;
 932                 case Opt_prjquota:
 933                         set_opt(sbi, PRJQUOTA);
 934                         break;
 935                 case Opt_usrjquota:
 936                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 937                         if (ret)
 938                                 return ret;
 939                         break;
 940                 case Opt_grpjquota:
 941                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 942                         if (ret)
 943                                 return ret;
 944                         break;
 945                 case Opt_prjjquota:
 946                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 947                         if (ret)
 948                                 return ret;
 949                         break;
 950                 case Opt_offusrjquota:
 951                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 952                         if (ret)
 953                                 return ret;
 954                         break;
 955                 case Opt_offgrpjquota:
 956                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 957                         if (ret)
 958                                 return ret;
 959                         break;
 960                 case Opt_offprjjquota:
 961                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 962                         if (ret)
 963                                 return ret;
 964                         break;
 965                 case Opt_jqfmt_vfsold:
 966                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 967                         break;
 968                 case Opt_jqfmt_vfsv0:
 969                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 970                         break;
 971                 case Opt_jqfmt_vfsv1:
 972                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 973                         break;
 974                 case Opt_noquota:
 975                         clear_opt(sbi, QUOTA);
 976                         clear_opt(sbi, USRQUOTA);
 977                         clear_opt(sbi, GRPQUOTA);
 978                         clear_opt(sbi, PRJQUOTA);
 979                         break;
 980 #else
 981                 case Opt_quota:
 982                 case Opt_usrquota:
 983                 case Opt_grpquota:
 984                 case Opt_prjquota:
 985                 case Opt_usrjquota:
 986                 case Opt_grpjquota:
 987                 case Opt_prjjquota:
 988                 case Opt_offusrjquota:
 989                 case Opt_offgrpjquota:
 990                 case Opt_offprjjquota:
 991                 case Opt_jqfmt_vfsold:
 992                 case Opt_jqfmt_vfsv0:
 993                 case Opt_jqfmt_vfsv1:
 994                 case Opt_noquota:
 995                         f2fs_info(sbi, "quota operations not supported");
 996                         break;
 997 #endif
 998                 case Opt_alloc:
 999                         name = match_strdup(&args[0]);
1000                         if (!name)
1001                                 return -ENOMEM;
1002
1003                         if (!strcmp(name, "default")) {
1004                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1005                         } else if (!strcmp(name, "reuse")) {
1006                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1007                         } else {
1008                                 kfree(name);
1009                                 return -EINVAL;
1010                         }
1011                         kfree(name);
1012                         break;
1013                 case Opt_fsync:
1014                         name = match_strdup(&args[0]);
1015                         if (!name)
1016                                 return -ENOMEM;
1017                         if (!strcmp(name, "posix")) {
1018                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1019                         } else if (!strcmp(name, "strict")) {
1020                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1021                         } else if (!strcmp(name, "nobarrier")) {
1022                                 F2FS_OPTION(sbi).fsync_mode =
1023                                                         FSYNC_MODE_NOBARRIER;
1024                         } else {
1025                                 kfree(name);
1026                                 return -EINVAL;
1027                         }
1028                         kfree(name);
1029                         break;
1030                 case Opt_test_dummy_encryption:
1031                         ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1032                                                              is_remount);
1033                         if (ret)
1034                                 return ret;
1035                         break;
1036                 case Opt_inlinecrypt:
1037 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1038                         sb->s_flags |= SB_INLINECRYPT;
1039 #else
1040                         f2fs_info(sbi, "inline encryption not supported");
1041 #endif
1042                         break;
1043                 case Opt_checkpoint_disable_cap_perc:
1044                         if (args->from && match_int(args, &arg))
1045                                 return -EINVAL;
1046                         if (arg < 0 || arg > 100)
1047                                 return -EINVAL;
1048                         F2FS_OPTION(sbi).unusable_cap_perc = arg;
1049                         set_opt(sbi, DISABLE_CHECKPOINT);
1050                         break;
1051                 case Opt_checkpoint_disable_cap:
1052                         if (args->from && match_int(args, &arg))
1053                                 return -EINVAL;
1054                         F2FS_OPTION(sbi).unusable_cap = arg;
1055                         set_opt(sbi, DISABLE_CHECKPOINT);
1056                         break;
1057                 case Opt_checkpoint_disable:
1058                         set_opt(sbi, DISABLE_CHECKPOINT);
1059                         break;
1060                 case Opt_checkpoint_enable:
1061                         clear_opt(sbi, DISABLE_CHECKPOINT);
1062                         break;
1063                 case Opt_checkpoint_merge:
1064                         set_opt(sbi, MERGE_CHECKPOINT);
1065                         break;
1066                 case Opt_nocheckpoint_merge:
1067                         clear_opt(sbi, MERGE_CHECKPOINT);
1068                         break;
1069 #ifdef CONFIG_F2FS_FS_COMPRESSION
1070                 case Opt_compress_algorithm:
1071                         if (!f2fs_sb_has_compression(sbi)) {
1072                                 f2fs_info(sbi, "Image doesn't support compression");
1073                                 break;
1074                         }
1075                         name = match_strdup(&args[0]);
1076                         if (!name)
1077                                 return -ENOMEM;
1078                         if (!strcmp(name, "lzo")) {
1079 #ifdef CONFIG_F2FS_FS_LZO
1080                                 F2FS_OPTION(sbi).compress_level = 0;
1081                                 F2FS_OPTION(sbi).compress_algorithm =
1082                                                                 COMPRESS_LZO;
1083 #else
1084                                 f2fs_info(sbi, "kernel doesn't support lzo compression");
1085 #endif
1086                         } else if (!strncmp(name, "lz4", 3)) {
1087 #ifdef CONFIG_F2FS_FS_LZ4
1088                                 ret = f2fs_set_lz4hc_level(sbi, name);
1089                                 if (ret) {
1090                                         kfree(name);
1091                                         return -EINVAL;
1092                                 }
1093                                 F2FS_OPTION(sbi).compress_algorithm =
1094                                                                 COMPRESS_LZ4;
1095 #else
1096                                 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1097 #endif
1098                         } else if (!strncmp(name, "zstd", 4)) {
1099 #ifdef CONFIG_F2FS_FS_ZSTD
1100                                 ret = f2fs_set_zstd_level(sbi, name);
1101                                 if (ret) {
1102                                         kfree(name);
1103                                         return -EINVAL;
1104                                 }
1105                                 F2FS_OPTION(sbi).compress_algorithm =
1106                                                                 COMPRESS_ZSTD;
1107 #else
1108                                 f2fs_info(sbi, "kernel doesn't support zstd compression");
1109 #endif
1110                         } else if (!strcmp(name, "lzo-rle")) {
1111 #ifdef CONFIG_F2FS_FS_LZORLE
1112                                 F2FS_OPTION(sbi).compress_level = 0;
1113                                 F2FS_OPTION(sbi).compress_algorithm =
1114                                                                 COMPRESS_LZORLE;
1115 #else
1116                                 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1117 #endif
1118                         } else {
1119                                 kfree(name);
1120                                 return -EINVAL;
1121                         }
1122                         kfree(name);
1123                         break;
1124                 case Opt_compress_log_size:
1125                         if (!f2fs_sb_has_compression(sbi)) {
1126                                 f2fs_info(sbi, "Image doesn't support compression");
1127                                 break;
1128                         }
1129                         if (args->from && match_int(args, &arg))
1130                                 return -EINVAL;
1131                         if (arg < MIN_COMPRESS_LOG_SIZE ||
1132                                 arg > MAX_COMPRESS_LOG_SIZE) {
1133                                 f2fs_err(sbi,
1134                                         "Compress cluster log size is out of range");
1135                                 return -EINVAL;
1136                         }
1137                         F2FS_OPTION(sbi).compress_log_size = arg;
1138                         break;
1139                 case Opt_compress_extension:
1140                         if (!f2fs_sb_has_compression(sbi)) {
1141                                 f2fs_info(sbi, "Image doesn't support compression");
1142                                 break;
1143                         }
1144                         name = match_strdup(&args[0]);
1145                         if (!name)
1146                                 return -ENOMEM;
1147
1148                         ext = F2FS_OPTION(sbi).extensions;
1149                         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1150
1151                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
1152                                 ext_cnt >= COMPRESS_EXT_NUM) {
1153                                 f2fs_err(sbi,
1154                                         "invalid extension length/number");
1155                                 kfree(name);
1156                                 return -EINVAL;
1157                         }
1158
1159                         strcpy(ext[ext_cnt], name);
1160                         F2FS_OPTION(sbi).compress_ext_cnt++;
1161                         kfree(name);
1162                         break;
1163                 case Opt_nocompress_extension:
1164                         if (!f2fs_sb_has_compression(sbi)) {
1165                                 f2fs_info(sbi, "Image doesn't support compression");
1166                                 break;
1167                         }
1168                         name = match_strdup(&args[0]);
1169                         if (!name)
1170                                 return -ENOMEM;
1171
1172                         noext = F2FS_OPTION(sbi).noextensions;
1173                         noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1174
1175                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
1176                                 noext_cnt >= COMPRESS_EXT_NUM) {
1177                                 f2fs_err(sbi,
1178                                         "invalid extension length/number");
1179                                 kfree(name);
1180                                 return -EINVAL;
1181                         }
1182
1183                         strcpy(noext[noext_cnt], name);
1184                         F2FS_OPTION(sbi).nocompress_ext_cnt++;
1185                         kfree(name);
1186                         break;
1187                 case Opt_compress_chksum:
1188                         F2FS_OPTION(sbi).compress_chksum = true;
1189                         break;
1190                 case Opt_compress_mode:
1191                         name = match_strdup(&args[0]);
1192                         if (!name)
1193                                 return -ENOMEM;
1194                         if (!strcmp(name, "fs")) {
1195                                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1196                         } else if (!strcmp(name, "user")) {
1197                                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1198                         } else {
1199                                 kfree(name);
1200                                 return -EINVAL;
1201                         }
1202                         kfree(name);
1203                         break;
1204                 case Opt_compress_cache:
1205                         set_opt(sbi, COMPRESS_CACHE);
1206                         break;
1207 #else
1208                 case Opt_compress_algorithm:
1209                 case Opt_compress_log_size:
1210                 case Opt_compress_extension:
1211                 case Opt_nocompress_extension:
1212                 case Opt_compress_chksum:
1213                 case Opt_compress_mode:
1214                 case Opt_compress_cache:
1215                         f2fs_info(sbi, "compression options not supported");
1216                         break;
1217 #endif
1218                 case Opt_atgc:
1219                         set_opt(sbi, ATGC);
1220                         break;
1221                 case Opt_gc_merge:
1222                         set_opt(sbi, GC_MERGE);
1223                         break;
1224                 case Opt_nogc_merge:
1225                         clear_opt(sbi, GC_MERGE);
1226                         break;
1227                 case Opt_discard_unit:
1228                         name = match_strdup(&args[0]);
1229                         if (!name)
1230                                 return -ENOMEM;
1231                         if (!strcmp(name, "block")) {
1232                                 F2FS_OPTION(sbi).discard_unit =
1233                                                 DISCARD_UNIT_BLOCK;
1234                         } else if (!strcmp(name, "segment")) {
1235                                 F2FS_OPTION(sbi).discard_unit =
1236                                                 DISCARD_UNIT_SEGMENT;
1237                         } else if (!strcmp(name, "section")) {
1238                                 F2FS_OPTION(sbi).discard_unit =
1239                                                 DISCARD_UNIT_SECTION;
1240                         } else {
1241                                 kfree(name);
1242                                 return -EINVAL;
1243                         }
1244                         kfree(name);
1245                         break;
1246                 case Opt_memory_mode:
1247                         name = match_strdup(&args[0]);
1248                         if (!name)
1249                                 return -ENOMEM;
1250                         if (!strcmp(name, "normal")) {
1251                                 F2FS_OPTION(sbi).memory_mode =
1252                                                 MEMORY_MODE_NORMAL;
1253                         } else if (!strcmp(name, "low")) {
1254                                 F2FS_OPTION(sbi).memory_mode =
1255                                                 MEMORY_MODE_LOW;
1256                         } else {
1257                                 kfree(name);
1258                                 return -EINVAL;
1259                         }
1260                         kfree(name);
1261                         break;
1262                 case Opt_age_extent_cache:
1263                         set_opt(sbi, AGE_EXTENT_CACHE);
1264                         break;
1265                 default:
1266                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1267                                  p);
1268                         return -EINVAL;
1269                 }
1270         }
1271 default_check:
1272 #ifdef CONFIG_QUOTA
1273         if (f2fs_check_quota_options(sbi))
1274                 return -EINVAL;
1275 #else
1276         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1277                 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1278                 return -EINVAL;
1279         }
1280         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1281                 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1282                 return -EINVAL;
1283         }
1284 #endif
1285 #if !IS_ENABLED(CONFIG_UNICODE)
1286         if (f2fs_sb_has_casefold(sbi)) {
1287                 f2fs_err(sbi,
1288                         "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1289                 return -EINVAL;
1290         }
1291 #endif
1292         /*
1293          * The BLKZONED feature indicates that the drive was formatted with
1294          * zone alignment optimization. This is optional for host-aware
1295          * devices, but mandatory for host-managed zoned block devices.
1296          */
1297 #ifndef CONFIG_BLK_DEV_ZONED
1298         if (f2fs_sb_has_blkzoned(sbi)) {
1299                 f2fs_err(sbi, "Zoned block device support is not enabled");
1300                 return -EINVAL;
1301         }
1302 #endif
1303         if (f2fs_sb_has_blkzoned(sbi)) {
1304                 if (F2FS_OPTION(sbi).discard_unit !=
1305                                                 DISCARD_UNIT_SECTION) {
1306                         f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1307                         F2FS_OPTION(sbi).discard_unit =
1308                                         DISCARD_UNIT_SECTION;
1309                 }
1310         }
1311
1312 #ifdef CONFIG_F2FS_FS_COMPRESSION
1313         if (f2fs_test_compress_extension(sbi)) {
1314                 f2fs_err(sbi, "invalid compress or nocompress extension");
1315                 return -EINVAL;
1316         }
1317 #endif
1318
1319         if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1320                 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1321                          F2FS_IO_SIZE_KB(sbi));
1322                 return -EINVAL;
1323         }
1324
1325         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1326                 int min_size, max_size;
1327
1328                 if (!f2fs_sb_has_extra_attr(sbi) ||
1329                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1330                         f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1331                         return -EINVAL;
1332                 }
1333                 if (!test_opt(sbi, INLINE_XATTR)) {
1334                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1335                         return -EINVAL;
1336                 }
1337
1338                 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1339                 max_size = MAX_INLINE_XATTR_SIZE;
1340
1341                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1342                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1343                         f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1344                                  min_size, max_size);
1345                         return -EINVAL;
1346                 }
1347         }
1348
1349         if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1350                 f2fs_err(sbi, "LFS not compatible with checkpoint=disable");
1351                 return -EINVAL;
1352         }
1353
1354         if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1355                 f2fs_err(sbi, "LFS not compatible with ATGC");
1356                 return -EINVAL;
1357         }
1358
1359         if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1360                 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1361                 return -EINVAL;
1362         }
1363
1364         if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1365                 f2fs_err(sbi, "Allow to mount readonly mode only");
1366                 return -EROFS;
1367         }
1368         return 0;
1369 }
1370
1371 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1372 {
1373         struct f2fs_inode_info *fi;
1374
1375         if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) {
1376                 f2fs_show_injection_info(F2FS_SB(sb), FAULT_SLAB_ALLOC);
1377                 return NULL;
1378         }
1379
1380         fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1381         if (!fi)
1382                 return NULL;
1383
1384         init_once((void *) fi);
1385
1386         /* Initialize f2fs-specific inode info */
1387         atomic_set(&fi->dirty_pages, 0);
1388         atomic_set(&fi->i_compr_blocks, 0);
1389         init_f2fs_rwsem(&fi->i_sem);
1390         spin_lock_init(&fi->i_size_lock);
1391         INIT_LIST_HEAD(&fi->dirty_list);
1392         INIT_LIST_HEAD(&fi->gdirty_list);
1393         init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1394         init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1395         init_f2fs_rwsem(&fi->i_xattr_sem);
1396
1397         /* Will be used by directory only */
1398         fi->i_dir_level = F2FS_SB(sb)->dir_level;
1399
1400         return &fi->vfs_inode;
1401 }
1402
1403 static int f2fs_drop_inode(struct inode *inode)
1404 {
1405         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1406         int ret;
1407
1408         /*
1409          * during filesystem shutdown, if checkpoint is disabled,
1410          * drop useless meta/node dirty pages.
1411          */
1412         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1413                 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1414                         inode->i_ino == F2FS_META_INO(sbi)) {
1415                         trace_f2fs_drop_inode(inode, 1);
1416                         return 1;
1417                 }
1418         }
1419
1420         /*
1421          * This is to avoid a deadlock condition like below.
1422          * writeback_single_inode(inode)
1423          *  - f2fs_write_data_page
1424          *    - f2fs_gc -> iput -> evict
1425          *       - inode_wait_for_writeback(inode)
1426          */
1427         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1428                 if (!inode->i_nlink && !is_bad_inode(inode)) {
1429                         /* to avoid evict_inode call simultaneously */
1430                         atomic_inc(&inode->i_count);
1431                         spin_unlock(&inode->i_lock);
1432
1433                         f2fs_abort_atomic_write(inode, true);
1434
1435                         /* should remain fi->extent_tree for writepage */
1436                         f2fs_destroy_extent_node(inode);
1437
1438                         sb_start_intwrite(inode->i_sb);
1439                         f2fs_i_size_write(inode, 0);
1440
1441                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1442                                         inode, NULL, 0, DATA);
1443                         truncate_inode_pages_final(inode->i_mapping);
1444
1445                         if (F2FS_HAS_BLOCKS(inode))
1446                                 f2fs_truncate(inode);
1447
1448                         sb_end_intwrite(inode->i_sb);
1449
1450                         spin_lock(&inode->i_lock);
1451                         atomic_dec(&inode->i_count);
1452                 }
1453                 trace_f2fs_drop_inode(inode, 0);
1454                 return 0;
1455         }
1456         ret = generic_drop_inode(inode);
1457         if (!ret)
1458                 ret = fscrypt_drop_inode(inode);
1459         trace_f2fs_drop_inode(inode, ret);
1460         return ret;
1461 }
1462
1463 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1464 {
1465         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1466         int ret = 0;
1467
1468         spin_lock(&sbi->inode_lock[DIRTY_META]);
1469         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1470                 ret = 1;
1471         } else {
1472                 set_inode_flag(inode, FI_DIRTY_INODE);
1473                 stat_inc_dirty_inode(sbi, DIRTY_META);
1474         }
1475         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1476                 list_add_tail(&F2FS_I(inode)->gdirty_list,
1477                                 &sbi->inode_list[DIRTY_META]);
1478                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1479         }
1480         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1481         return ret;
1482 }
1483
1484 void f2fs_inode_synced(struct inode *inode)
1485 {
1486         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1487
1488         spin_lock(&sbi->inode_lock[DIRTY_META]);
1489         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1490                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1491                 return;
1492         }
1493         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1494                 list_del_init(&F2FS_I(inode)->gdirty_list);
1495                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1496         }
1497         clear_inode_flag(inode, FI_DIRTY_INODE);
1498         clear_inode_flag(inode, FI_AUTO_RECOVER);
1499         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1500         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1501 }
1502
1503 /*
1504  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1505  *
1506  * We should call set_dirty_inode to write the dirty inode through write_inode.
1507  */
1508 static void f2fs_dirty_inode(struct inode *inode, int flags)
1509 {
1510         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1511
1512         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1513                         inode->i_ino == F2FS_META_INO(sbi))
1514                 return;
1515
1516         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1517                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1518
1519         f2fs_inode_dirtied(inode, false);
1520 }
1521
1522 static void f2fs_free_inode(struct inode *inode)
1523 {
1524         fscrypt_free_inode(inode);
1525         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1526 }
1527
1528 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1529 {
1530         percpu_counter_destroy(&sbi->total_valid_inode_count);
1531         percpu_counter_destroy(&sbi->rf_node_block_count);
1532         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1533 }
1534
1535 static void destroy_device_list(struct f2fs_sb_info *sbi)
1536 {
1537         int i;
1538
1539         for (i = 0; i < sbi->s_ndevs; i++) {
1540                 blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1541 #ifdef CONFIG_BLK_DEV_ZONED
1542                 kvfree(FDEV(i).blkz_seq);
1543 #endif
1544         }
1545         kvfree(sbi->devs);
1546 }
1547
1548 static void f2fs_put_super(struct super_block *sb)
1549 {
1550         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1551         int i;
1552         bool dropped;
1553
1554         /* unregister procfs/sysfs entries in advance to avoid race case */
1555         f2fs_unregister_sysfs(sbi);
1556
1557         f2fs_quota_off_umount(sb);
1558
1559         /* prevent remaining shrinker jobs */
1560         mutex_lock(&sbi->umount_mutex);
1561
1562         /*
1563          * flush all issued checkpoints and stop checkpoint issue thread.
1564          * after then, all checkpoints should be done by each process context.
1565          */
1566         f2fs_stop_ckpt_thread(sbi);
1567
1568         /*
1569          * We don't need to do checkpoint when superblock is clean.
1570          * But, the previous checkpoint was not done by umount, it needs to do
1571          * clean checkpoint again.
1572          */
1573         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1574                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1575                 struct cp_control cpc = {
1576                         .reason = CP_UMOUNT,
1577                 };
1578                 f2fs_write_checkpoint(sbi, &cpc);
1579         }
1580
1581         /* be sure to wait for any on-going discard commands */
1582         dropped = f2fs_issue_discard_timeout(sbi);
1583
1584         if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && !dropped) {
1585                 struct cp_control cpc = {
1586                         .reason = CP_UMOUNT | CP_TRIMMED,
1587                 };
1588                 f2fs_write_checkpoint(sbi, &cpc);
1589         }
1590
1591         /*
1592          * normally superblock is clean, so we need to release this.
1593          * In addition, EIO will skip do checkpoint, we need this as well.
1594          */
1595         f2fs_release_ino_entry(sbi, true);
1596
1597         f2fs_leave_shrinker(sbi);
1598         mutex_unlock(&sbi->umount_mutex);
1599
1600         /* our cp_error case, we can wait for any writeback page */
1601         f2fs_flush_merged_writes(sbi);
1602
1603         f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1604
1605         f2fs_bug_on(sbi, sbi->fsync_node_num);
1606
1607         f2fs_destroy_compress_inode(sbi);
1608
1609         iput(sbi->node_inode);
1610         sbi->node_inode = NULL;
1611
1612         iput(sbi->meta_inode);
1613         sbi->meta_inode = NULL;
1614
1615         /*
1616          * iput() can update stat information, if f2fs_write_checkpoint()
1617          * above failed with error.
1618          */
1619         f2fs_destroy_stats(sbi);
1620
1621         /* destroy f2fs internal modules */
1622         f2fs_destroy_node_manager(sbi);
1623         f2fs_destroy_segment_manager(sbi);
1624
1625         f2fs_destroy_post_read_wq(sbi);
1626
1627         kvfree(sbi->ckpt);
1628
1629         sb->s_fs_info = NULL;
1630         if (sbi->s_chksum_driver)
1631                 crypto_free_shash(sbi->s_chksum_driver);
1632         kfree(sbi->raw_super);
1633
1634         destroy_device_list(sbi);
1635         f2fs_destroy_page_array_cache(sbi);
1636         f2fs_destroy_xattr_caches(sbi);
1637         mempool_destroy(sbi->write_io_dummy);
1638 #ifdef CONFIG_QUOTA
1639         for (i = 0; i < MAXQUOTAS; i++)
1640                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1641 #endif
1642         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1643         destroy_percpu_info(sbi);
1644         f2fs_destroy_iostat(sbi);
1645         for (i = 0; i < NR_PAGE_TYPE; i++)
1646                 kvfree(sbi->write_io[i]);
1647 #if IS_ENABLED(CONFIG_UNICODE)
1648         utf8_unload(sb->s_encoding);
1649 #endif
1650         kfree(sbi);
1651 }
1652
1653 int f2fs_sync_fs(struct super_block *sb, int sync)
1654 {
1655         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1656         int err = 0;
1657
1658         if (unlikely(f2fs_cp_error(sbi)))
1659                 return 0;
1660         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1661                 return 0;
1662
1663         trace_f2fs_sync_fs(sb, sync);
1664
1665         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1666                 return -EAGAIN;
1667
1668         if (sync)
1669                 err = f2fs_issue_checkpoint(sbi);
1670
1671         return err;
1672 }
1673
1674 static int f2fs_freeze(struct super_block *sb)
1675 {
1676         if (f2fs_readonly(sb))
1677                 return 0;
1678
1679         /* IO error happened before */
1680         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1681                 return -EIO;
1682
1683         /* must be clean, since sync_filesystem() was already called */
1684         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1685                 return -EINVAL;
1686
1687         /* Let's flush checkpoints and stop the thread. */
1688         f2fs_flush_ckpt_thread(F2FS_SB(sb));
1689
1690         /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1691         set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1692         return 0;
1693 }
1694
1695 static int f2fs_unfreeze(struct super_block *sb)
1696 {
1697         clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1698         return 0;
1699 }
1700
1701 #ifdef CONFIG_QUOTA
1702 static int f2fs_statfs_project(struct super_block *sb,
1703                                 kprojid_t projid, struct kstatfs *buf)
1704 {
1705         struct kqid qid;
1706         struct dquot *dquot;
1707         u64 limit;
1708         u64 curblock;
1709
1710         qid = make_kqid_projid(projid);
1711         dquot = dqget(sb, qid);
1712         if (IS_ERR(dquot))
1713                 return PTR_ERR(dquot);
1714         spin_lock(&dquot->dq_dqb_lock);
1715
1716         limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1717                                         dquot->dq_dqb.dqb_bhardlimit);
1718         if (limit)
1719                 limit >>= sb->s_blocksize_bits;
1720
1721         if (limit && buf->f_blocks > limit) {
1722                 curblock = (dquot->dq_dqb.dqb_curspace +
1723                             dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1724                 buf->f_blocks = limit;
1725                 buf->f_bfree = buf->f_bavail =
1726                         (buf->f_blocks > curblock) ?
1727                          (buf->f_blocks - curblock) : 0;
1728         }
1729
1730         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1731                                         dquot->dq_dqb.dqb_ihardlimit);
1732
1733         if (limit && buf->f_files > limit) {
1734                 buf->f_files = limit;
1735                 buf->f_ffree =
1736                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1737                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1738         }
1739
1740         spin_unlock(&dquot->dq_dqb_lock);
1741         dqput(dquot);
1742         return 0;
1743 }
1744 #endif
1745
1746 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1747 {
1748         struct super_block *sb = dentry->d_sb;
1749         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1750         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1751         block_t total_count, user_block_count, start_count;
1752         u64 avail_node_count;
1753         unsigned int total_valid_node_count;
1754
1755         total_count = le64_to_cpu(sbi->raw_super->block_count);
1756         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1757         buf->f_type = F2FS_SUPER_MAGIC;
1758         buf->f_bsize = sbi->blocksize;
1759
1760         buf->f_blocks = total_count - start_count;
1761
1762         spin_lock(&sbi->stat_lock);
1763
1764         user_block_count = sbi->user_block_count;
1765         total_valid_node_count = valid_node_count(sbi);
1766         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1767         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1768                                                 sbi->current_reserved_blocks;
1769
1770         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1771                 buf->f_bfree = 0;
1772         else
1773                 buf->f_bfree -= sbi->unusable_block_count;
1774         spin_unlock(&sbi->stat_lock);
1775
1776         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1777                 buf->f_bavail = buf->f_bfree -
1778                                 F2FS_OPTION(sbi).root_reserved_blocks;
1779         else
1780                 buf->f_bavail = 0;
1781
1782         if (avail_node_count > user_block_count) {
1783                 buf->f_files = user_block_count;
1784                 buf->f_ffree = buf->f_bavail;
1785         } else {
1786                 buf->f_files = avail_node_count;
1787                 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1788                                         buf->f_bavail);
1789         }
1790
1791         buf->f_namelen = F2FS_NAME_LEN;
1792         buf->f_fsid    = u64_to_fsid(id);
1793
1794 #ifdef CONFIG_QUOTA
1795         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1796                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1797                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1798         }
1799 #endif
1800         return 0;
1801 }
1802
1803 static inline void f2fs_show_quota_options(struct seq_file *seq,
1804                                            struct super_block *sb)
1805 {
1806 #ifdef CONFIG_QUOTA
1807         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1808
1809         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1810                 char *fmtname = "";
1811
1812                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1813                 case QFMT_VFS_OLD:
1814                         fmtname = "vfsold";
1815                         break;
1816                 case QFMT_VFS_V0:
1817                         fmtname = "vfsv0";
1818                         break;
1819                 case QFMT_VFS_V1:
1820                         fmtname = "vfsv1";
1821                         break;
1822                 }
1823                 seq_printf(seq, ",jqfmt=%s", fmtname);
1824         }
1825
1826         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1827                 seq_show_option(seq, "usrjquota",
1828                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1829
1830         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1831                 seq_show_option(seq, "grpjquota",
1832                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1833
1834         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1835                 seq_show_option(seq, "prjjquota",
1836                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1837 #endif
1838 }
1839
1840 #ifdef CONFIG_F2FS_FS_COMPRESSION
1841 static inline void f2fs_show_compress_options(struct seq_file *seq,
1842                                                         struct super_block *sb)
1843 {
1844         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1845         char *algtype = "";
1846         int i;
1847
1848         if (!f2fs_sb_has_compression(sbi))
1849                 return;
1850
1851         switch (F2FS_OPTION(sbi).compress_algorithm) {
1852         case COMPRESS_LZO:
1853                 algtype = "lzo";
1854                 break;
1855         case COMPRESS_LZ4:
1856                 algtype = "lz4";
1857                 break;
1858         case COMPRESS_ZSTD:
1859                 algtype = "zstd";
1860                 break;
1861         case COMPRESS_LZORLE:
1862                 algtype = "lzo-rle";
1863                 break;
1864         }
1865         seq_printf(seq, ",compress_algorithm=%s", algtype);
1866
1867         if (F2FS_OPTION(sbi).compress_level)
1868                 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1869
1870         seq_printf(seq, ",compress_log_size=%u",
1871                         F2FS_OPTION(sbi).compress_log_size);
1872
1873         for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1874                 seq_printf(seq, ",compress_extension=%s",
1875                         F2FS_OPTION(sbi).extensions[i]);
1876         }
1877
1878         for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1879                 seq_printf(seq, ",nocompress_extension=%s",
1880                         F2FS_OPTION(sbi).noextensions[i]);
1881         }
1882
1883         if (F2FS_OPTION(sbi).compress_chksum)
1884                 seq_puts(seq, ",compress_chksum");
1885
1886         if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1887                 seq_printf(seq, ",compress_mode=%s", "fs");
1888         else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1889                 seq_printf(seq, ",compress_mode=%s", "user");
1890
1891         if (test_opt(sbi, COMPRESS_CACHE))
1892                 seq_puts(seq, ",compress_cache");
1893 }
1894 #endif
1895
1896 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1897 {
1898         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1899
1900         if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1901                 seq_printf(seq, ",background_gc=%s", "sync");
1902         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1903                 seq_printf(seq, ",background_gc=%s", "on");
1904         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1905                 seq_printf(seq, ",background_gc=%s", "off");
1906
1907         if (test_opt(sbi, GC_MERGE))
1908                 seq_puts(seq, ",gc_merge");
1909
1910         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1911                 seq_puts(seq, ",disable_roll_forward");
1912         if (test_opt(sbi, NORECOVERY))
1913                 seq_puts(seq, ",norecovery");
1914         if (test_opt(sbi, DISCARD))
1915                 seq_puts(seq, ",discard");
1916         else
1917                 seq_puts(seq, ",nodiscard");
1918         if (test_opt(sbi, NOHEAP))
1919                 seq_puts(seq, ",no_heap");
1920         else
1921                 seq_puts(seq, ",heap");
1922 #ifdef CONFIG_F2FS_FS_XATTR
1923         if (test_opt(sbi, XATTR_USER))
1924                 seq_puts(seq, ",user_xattr");
1925         else
1926                 seq_puts(seq, ",nouser_xattr");
1927         if (test_opt(sbi, INLINE_XATTR))
1928                 seq_puts(seq, ",inline_xattr");
1929         else
1930                 seq_puts(seq, ",noinline_xattr");
1931         if (test_opt(sbi, INLINE_XATTR_SIZE))
1932                 seq_printf(seq, ",inline_xattr_size=%u",
1933                                         F2FS_OPTION(sbi).inline_xattr_size);
1934 #endif
1935 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1936         if (test_opt(sbi, POSIX_ACL))
1937                 seq_puts(seq, ",acl");
1938         else
1939                 seq_puts(seq, ",noacl");
1940 #endif
1941         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1942                 seq_puts(seq, ",disable_ext_identify");
1943         if (test_opt(sbi, INLINE_DATA))
1944                 seq_puts(seq, ",inline_data");
1945         else
1946                 seq_puts(seq, ",noinline_data");
1947         if (test_opt(sbi, INLINE_DENTRY))
1948                 seq_puts(seq, ",inline_dentry");
1949         else
1950                 seq_puts(seq, ",noinline_dentry");
1951         if (test_opt(sbi, FLUSH_MERGE))
1952                 seq_puts(seq, ",flush_merge");
1953         else
1954                 seq_puts(seq, ",noflush_merge");
1955         if (test_opt(sbi, NOBARRIER))
1956                 seq_puts(seq, ",nobarrier");
1957         else
1958                 seq_puts(seq, ",barrier");
1959         if (test_opt(sbi, FASTBOOT))
1960                 seq_puts(seq, ",fastboot");
1961         if (test_opt(sbi, READ_EXTENT_CACHE))
1962                 seq_puts(seq, ",extent_cache");
1963         else
1964                 seq_puts(seq, ",noextent_cache");
1965         if (test_opt(sbi, AGE_EXTENT_CACHE))
1966                 seq_puts(seq, ",age_extent_cache");
1967         if (test_opt(sbi, DATA_FLUSH))
1968                 seq_puts(seq, ",data_flush");
1969
1970         seq_puts(seq, ",mode=");
1971         if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1972                 seq_puts(seq, "adaptive");
1973         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1974                 seq_puts(seq, "lfs");
1975         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
1976                 seq_puts(seq, "fragment:segment");
1977         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
1978                 seq_puts(seq, "fragment:block");
1979         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1980         if (test_opt(sbi, RESERVE_ROOT))
1981                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1982                                 F2FS_OPTION(sbi).root_reserved_blocks,
1983                                 from_kuid_munged(&init_user_ns,
1984                                         F2FS_OPTION(sbi).s_resuid),
1985                                 from_kgid_munged(&init_user_ns,
1986                                         F2FS_OPTION(sbi).s_resgid));
1987         if (F2FS_IO_SIZE_BITS(sbi))
1988                 seq_printf(seq, ",io_bits=%u",
1989                                 F2FS_OPTION(sbi).write_io_size_bits);
1990 #ifdef CONFIG_F2FS_FAULT_INJECTION
1991         if (test_opt(sbi, FAULT_INJECTION)) {
1992                 seq_printf(seq, ",fault_injection=%u",
1993                                 F2FS_OPTION(sbi).fault_info.inject_rate);
1994                 seq_printf(seq, ",fault_type=%u",
1995                                 F2FS_OPTION(sbi).fault_info.inject_type);
1996         }
1997 #endif
1998 #ifdef CONFIG_QUOTA
1999         if (test_opt(sbi, QUOTA))
2000                 seq_puts(seq, ",quota");
2001         if (test_opt(sbi, USRQUOTA))
2002                 seq_puts(seq, ",usrquota");
2003         if (test_opt(sbi, GRPQUOTA))
2004                 seq_puts(seq, ",grpquota");
2005         if (test_opt(sbi, PRJQUOTA))
2006                 seq_puts(seq, ",prjquota");
2007 #endif
2008         f2fs_show_quota_options(seq, sbi->sb);
2009
2010         fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2011
2012         if (sbi->sb->s_flags & SB_INLINECRYPT)
2013                 seq_puts(seq, ",inlinecrypt");
2014
2015         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2016                 seq_printf(seq, ",alloc_mode=%s", "default");
2017         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2018                 seq_printf(seq, ",alloc_mode=%s", "reuse");
2019
2020         if (test_opt(sbi, DISABLE_CHECKPOINT))
2021                 seq_printf(seq, ",checkpoint=disable:%u",
2022                                 F2FS_OPTION(sbi).unusable_cap);
2023         if (test_opt(sbi, MERGE_CHECKPOINT))
2024                 seq_puts(seq, ",checkpoint_merge");
2025         else
2026                 seq_puts(seq, ",nocheckpoint_merge");
2027         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2028                 seq_printf(seq, ",fsync_mode=%s", "posix");
2029         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2030                 seq_printf(seq, ",fsync_mode=%s", "strict");
2031         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2032                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2033
2034 #ifdef CONFIG_F2FS_FS_COMPRESSION
2035         f2fs_show_compress_options(seq, sbi->sb);
2036 #endif
2037
2038         if (test_opt(sbi, ATGC))
2039                 seq_puts(seq, ",atgc");
2040
2041         if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
2042                 seq_printf(seq, ",discard_unit=%s", "block");
2043         else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
2044                 seq_printf(seq, ",discard_unit=%s", "segment");
2045         else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
2046                 seq_printf(seq, ",discard_unit=%s", "section");
2047
2048         if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2049                 seq_printf(seq, ",memory=%s", "normal");
2050         else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2051                 seq_printf(seq, ",memory=%s", "low");
2052
2053         return 0;
2054 }
2055
2056 static void default_options(struct f2fs_sb_info *sbi)
2057 {
2058         /* init some FS parameters */
2059         if (f2fs_sb_has_readonly(sbi))
2060                 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2061         else
2062                 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2063
2064         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2065         if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2066                                                         SMALL_VOLUME_SEGMENTS)
2067                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2068         else
2069                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2070         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2071         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2072         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2073         F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2074         F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2075         F2FS_OPTION(sbi).compress_ext_cnt = 0;
2076         F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2077         F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2078         F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2079
2080         sbi->sb->s_flags &= ~SB_INLINECRYPT;
2081
2082         set_opt(sbi, INLINE_XATTR);
2083         set_opt(sbi, INLINE_DATA);
2084         set_opt(sbi, INLINE_DENTRY);
2085         set_opt(sbi, READ_EXTENT_CACHE);
2086         set_opt(sbi, NOHEAP);
2087         clear_opt(sbi, DISABLE_CHECKPOINT);
2088         set_opt(sbi, MERGE_CHECKPOINT);
2089         F2FS_OPTION(sbi).unusable_cap = 0;
2090         sbi->sb->s_flags |= SB_LAZYTIME;
2091         if (!f2fs_is_readonly(sbi))
2092                 set_opt(sbi, FLUSH_MERGE);
2093         if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2094                 set_opt(sbi, DISCARD);
2095         if (f2fs_sb_has_blkzoned(sbi)) {
2096                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2097                 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2098         } else {
2099                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2100                 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2101         }
2102
2103 #ifdef CONFIG_F2FS_FS_XATTR
2104         set_opt(sbi, XATTR_USER);
2105 #endif
2106 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2107         set_opt(sbi, POSIX_ACL);
2108 #endif
2109
2110         f2fs_build_fault_attr(sbi, 0, 0);
2111 }
2112
2113 #ifdef CONFIG_QUOTA
2114 static int f2fs_enable_quotas(struct super_block *sb);
2115 #endif
2116
2117 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2118 {
2119         unsigned int s_flags = sbi->sb->s_flags;
2120         struct cp_control cpc;
2121         unsigned int gc_mode = sbi->gc_mode;
2122         int err = 0;
2123         int ret;
2124         block_t unusable;
2125
2126         if (s_flags & SB_RDONLY) {
2127                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2128                 return -EINVAL;
2129         }
2130         sbi->sb->s_flags |= SB_ACTIVE;
2131
2132         /* check if we need more GC first */
2133         unusable = f2fs_get_unusable_blocks(sbi);
2134         if (!f2fs_disable_cp_again(sbi, unusable))
2135                 goto skip_gc;
2136
2137         f2fs_update_time(sbi, DISABLE_TIME);
2138
2139         sbi->gc_mode = GC_URGENT_HIGH;
2140
2141         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2142                 struct f2fs_gc_control gc_control = {
2143                         .victim_segno = NULL_SEGNO,
2144                         .init_gc_type = FG_GC,
2145                         .should_migrate_blocks = false,
2146                         .err_gc_skipped = true,
2147                         .nr_free_secs = 1 };
2148
2149                 f2fs_down_write(&sbi->gc_lock);
2150                 err = f2fs_gc(sbi, &gc_control);
2151                 if (err == -ENODATA) {
2152                         err = 0;
2153                         break;
2154                 }
2155                 if (err && err != -EAGAIN)
2156                         break;
2157         }
2158
2159         ret = sync_filesystem(sbi->sb);
2160         if (ret || err) {
2161                 err = ret ? ret : err;
2162                 goto restore_flag;
2163         }
2164
2165         unusable = f2fs_get_unusable_blocks(sbi);
2166         if (f2fs_disable_cp_again(sbi, unusable)) {
2167                 err = -EAGAIN;
2168                 goto restore_flag;
2169         }
2170
2171 skip_gc:
2172         f2fs_down_write(&sbi->gc_lock);
2173         cpc.reason = CP_PAUSE;
2174         set_sbi_flag(sbi, SBI_CP_DISABLED);
2175         err = f2fs_write_checkpoint(sbi, &cpc);
2176         if (err)
2177                 goto out_unlock;
2178
2179         spin_lock(&sbi->stat_lock);
2180         sbi->unusable_block_count = unusable;
2181         spin_unlock(&sbi->stat_lock);
2182
2183 out_unlock:
2184         f2fs_up_write(&sbi->gc_lock);
2185 restore_flag:
2186         sbi->gc_mode = gc_mode;
2187         sbi->sb->s_flags = s_flags;     /* Restore SB_RDONLY status */
2188         return err;
2189 }
2190
2191 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2192 {
2193         int retry = DEFAULT_RETRY_IO_COUNT;
2194
2195         /* we should flush all the data to keep data consistency */
2196         do {
2197                 sync_inodes_sb(sbi->sb);
2198                 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2199         } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2200
2201         if (unlikely(retry < 0))
2202                 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2203
2204         f2fs_down_write(&sbi->gc_lock);
2205         f2fs_dirty_to_prefree(sbi);
2206
2207         clear_sbi_flag(sbi, SBI_CP_DISABLED);
2208         set_sbi_flag(sbi, SBI_IS_DIRTY);
2209         f2fs_up_write(&sbi->gc_lock);
2210
2211         f2fs_sync_fs(sbi->sb, 1);
2212
2213         /* Let's ensure there's no pending checkpoint anymore */
2214         f2fs_flush_ckpt_thread(sbi);
2215 }
2216
2217 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2218 {
2219         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2220         struct f2fs_mount_info org_mount_opt;
2221         unsigned long old_sb_flags;
2222         int err;
2223         bool need_restart_gc = false, need_stop_gc = false;
2224         bool need_restart_ckpt = false, need_stop_ckpt = false;
2225         bool need_restart_flush = false, need_stop_flush = false;
2226         bool need_restart_discard = false, need_stop_discard = false;
2227         bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2228         bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2229         bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2230         bool no_io_align = !F2FS_IO_ALIGNED(sbi);
2231         bool no_atgc = !test_opt(sbi, ATGC);
2232         bool no_discard = !test_opt(sbi, DISCARD);
2233         bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2234         bool block_unit_discard = f2fs_block_unit_discard(sbi);
2235 #ifdef CONFIG_QUOTA
2236         int i, j;
2237 #endif
2238
2239         /*
2240          * Save the old mount options in case we
2241          * need to restore them.
2242          */
2243         org_mount_opt = sbi->mount_opt;
2244         old_sb_flags = sb->s_flags;
2245
2246 #ifdef CONFIG_QUOTA
2247         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2248         for (i = 0; i < MAXQUOTAS; i++) {
2249                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2250                         org_mount_opt.s_qf_names[i] =
2251                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2252                                 GFP_KERNEL);
2253                         if (!org_mount_opt.s_qf_names[i]) {
2254                                 for (j = 0; j < i; j++)
2255                                         kfree(org_mount_opt.s_qf_names[j]);
2256                                 return -ENOMEM;
2257                         }
2258                 } else {
2259                         org_mount_opt.s_qf_names[i] = NULL;
2260                 }
2261         }
2262 #endif
2263
2264         /* recover superblocks we couldn't write due to previous RO mount */
2265         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2266                 err = f2fs_commit_super(sbi, false);
2267                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2268                           err);
2269                 if (!err)
2270                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2271         }
2272
2273         default_options(sbi);
2274
2275         /* parse mount options */
2276         err = parse_options(sb, data, true);
2277         if (err)
2278                 goto restore_opts;
2279
2280         /*
2281          * Previous and new state of filesystem is RO,
2282          * so skip checking GC and FLUSH_MERGE conditions.
2283          */
2284         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2285                 goto skip;
2286
2287         if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) {
2288                 err = -EROFS;
2289                 goto restore_opts;
2290         }
2291
2292 #ifdef CONFIG_QUOTA
2293         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2294                 err = dquot_suspend(sb, -1);
2295                 if (err < 0)
2296                         goto restore_opts;
2297         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2298                 /* dquot_resume needs RW */
2299                 sb->s_flags &= ~SB_RDONLY;
2300                 if (sb_any_quota_suspended(sb)) {
2301                         dquot_resume(sb, -1);
2302                 } else if (f2fs_sb_has_quota_ino(sbi)) {
2303                         err = f2fs_enable_quotas(sb);
2304                         if (err)
2305                                 goto restore_opts;
2306                 }
2307         }
2308 #endif
2309         /* disallow enable atgc dynamically */
2310         if (no_atgc == !!test_opt(sbi, ATGC)) {
2311                 err = -EINVAL;
2312                 f2fs_warn(sbi, "switch atgc option is not allowed");
2313                 goto restore_opts;
2314         }
2315
2316         /* disallow enable/disable extent_cache dynamically */
2317         if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2318                 err = -EINVAL;
2319                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2320                 goto restore_opts;
2321         }
2322         /* disallow enable/disable age extent_cache dynamically */
2323         if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2324                 err = -EINVAL;
2325                 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2326                 goto restore_opts;
2327         }
2328
2329         if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
2330                 err = -EINVAL;
2331                 f2fs_warn(sbi, "switch io_bits option is not allowed");
2332                 goto restore_opts;
2333         }
2334
2335         if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2336                 err = -EINVAL;
2337                 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2338                 goto restore_opts;
2339         }
2340
2341         if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2342                 err = -EINVAL;
2343                 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2344                 goto restore_opts;
2345         }
2346
2347         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2348                 err = -EINVAL;
2349                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2350                 goto restore_opts;
2351         }
2352
2353         /*
2354          * We stop the GC thread if FS is mounted as RO
2355          * or if background_gc = off is passed in mount
2356          * option. Also sync the filesystem.
2357          */
2358         if ((*flags & SB_RDONLY) ||
2359                         (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2360                         !test_opt(sbi, GC_MERGE))) {
2361                 if (sbi->gc_thread) {
2362                         f2fs_stop_gc_thread(sbi);
2363                         need_restart_gc = true;
2364                 }
2365         } else if (!sbi->gc_thread) {
2366                 err = f2fs_start_gc_thread(sbi);
2367                 if (err)
2368                         goto restore_opts;
2369                 need_stop_gc = true;
2370         }
2371
2372         if (*flags & SB_RDONLY) {
2373                 sync_inodes_sb(sb);
2374
2375                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2376                 set_sbi_flag(sbi, SBI_IS_CLOSE);
2377                 f2fs_sync_fs(sb, 1);
2378                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2379         }
2380
2381         if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2382                         !test_opt(sbi, MERGE_CHECKPOINT)) {
2383                 f2fs_stop_ckpt_thread(sbi);
2384                 need_restart_ckpt = true;
2385         } else {
2386                 /* Flush if the prevous checkpoint, if exists. */
2387                 f2fs_flush_ckpt_thread(sbi);
2388
2389                 err = f2fs_start_ckpt_thread(sbi);
2390                 if (err) {
2391                         f2fs_err(sbi,
2392                             "Failed to start F2FS issue_checkpoint_thread (%d)",
2393                             err);
2394                         goto restore_gc;
2395                 }
2396                 need_stop_ckpt = true;
2397         }
2398
2399         /*
2400          * We stop issue flush thread if FS is mounted as RO
2401          * or if flush_merge is not passed in mount option.
2402          */
2403         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2404                 clear_opt(sbi, FLUSH_MERGE);
2405                 f2fs_destroy_flush_cmd_control(sbi, false);
2406                 need_restart_flush = true;
2407         } else {
2408                 err = f2fs_create_flush_cmd_control(sbi);
2409                 if (err)
2410                         goto restore_ckpt;
2411                 need_stop_flush = true;
2412         }
2413
2414         if (no_discard == !!test_opt(sbi, DISCARD)) {
2415                 if (test_opt(sbi, DISCARD)) {
2416                         err = f2fs_start_discard_thread(sbi);
2417                         if (err)
2418                                 goto restore_flush;
2419                         need_stop_discard = true;
2420                 } else {
2421                         f2fs_stop_discard_thread(sbi);
2422                         f2fs_issue_discard_timeout(sbi);
2423                         need_restart_discard = true;
2424                 }
2425         }
2426
2427         if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2428                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2429                         err = f2fs_disable_checkpoint(sbi);
2430                         if (err)
2431                                 goto restore_discard;
2432                 } else {
2433                         f2fs_enable_checkpoint(sbi);
2434                 }
2435         }
2436
2437 skip:
2438 #ifdef CONFIG_QUOTA
2439         /* Release old quota file names */
2440         for (i = 0; i < MAXQUOTAS; i++)
2441                 kfree(org_mount_opt.s_qf_names[i]);
2442 #endif
2443         /* Update the POSIXACL Flag */
2444         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2445                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2446
2447         limit_reserve_root(sbi);
2448         adjust_unusable_cap_perc(sbi);
2449         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2450         return 0;
2451 restore_discard:
2452         if (need_restart_discard) {
2453                 if (f2fs_start_discard_thread(sbi))
2454                         f2fs_warn(sbi, "discard has been stopped");
2455         } else if (need_stop_discard) {
2456                 f2fs_stop_discard_thread(sbi);
2457         }
2458 restore_flush:
2459         if (need_restart_flush) {
2460                 if (f2fs_create_flush_cmd_control(sbi))
2461                         f2fs_warn(sbi, "background flush thread has stopped");
2462         } else if (need_stop_flush) {
2463                 clear_opt(sbi, FLUSH_MERGE);
2464                 f2fs_destroy_flush_cmd_control(sbi, false);
2465         }
2466 restore_ckpt:
2467         if (need_restart_ckpt) {
2468                 if (f2fs_start_ckpt_thread(sbi))
2469                         f2fs_warn(sbi, "background ckpt thread has stopped");
2470         } else if (need_stop_ckpt) {
2471                 f2fs_stop_ckpt_thread(sbi);
2472         }
2473 restore_gc:
2474         if (need_restart_gc) {
2475                 if (f2fs_start_gc_thread(sbi))
2476                         f2fs_warn(sbi, "background gc thread has stopped");
2477         } else if (need_stop_gc) {
2478                 f2fs_stop_gc_thread(sbi);
2479         }
2480 restore_opts:
2481 #ifdef CONFIG_QUOTA
2482         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2483         for (i = 0; i < MAXQUOTAS; i++) {
2484                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2485                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2486         }
2487 #endif
2488         sbi->mount_opt = org_mount_opt;
2489         sb->s_flags = old_sb_flags;
2490         return err;
2491 }
2492
2493 #ifdef CONFIG_QUOTA
2494 /* Read data from quotafile */
2495 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2496                                size_t len, loff_t off)
2497 {
2498         struct inode *inode = sb_dqopt(sb)->files[type];
2499         struct address_space *mapping = inode->i_mapping;
2500         block_t blkidx = F2FS_BYTES_TO_BLK(off);
2501         int offset = off & (sb->s_blocksize - 1);
2502         int tocopy;
2503         size_t toread;
2504         loff_t i_size = i_size_read(inode);
2505         struct page *page;
2506
2507         if (off > i_size)
2508                 return 0;
2509
2510         if (off + len > i_size)
2511                 len = i_size - off;
2512         toread = len;
2513         while (toread > 0) {
2514                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2515 repeat:
2516                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2517                 if (IS_ERR(page)) {
2518                         if (PTR_ERR(page) == -ENOMEM) {
2519                                 memalloc_retry_wait(GFP_NOFS);
2520                                 goto repeat;
2521                         }
2522                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2523                         return PTR_ERR(page);
2524                 }
2525
2526                 lock_page(page);
2527
2528                 if (unlikely(page->mapping != mapping)) {
2529                         f2fs_put_page(page, 1);
2530                         goto repeat;
2531                 }
2532                 if (unlikely(!PageUptodate(page))) {
2533                         f2fs_put_page(page, 1);
2534                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2535                         return -EIO;
2536                 }
2537
2538                 memcpy_from_page(data, page, offset, tocopy);
2539                 f2fs_put_page(page, 1);
2540
2541                 offset = 0;
2542                 toread -= tocopy;
2543                 data += tocopy;
2544                 blkidx++;
2545         }
2546         return len;
2547 }
2548
2549 /* Write to quotafile */
2550 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2551                                 const char *data, size_t len, loff_t off)
2552 {
2553         struct inode *inode = sb_dqopt(sb)->files[type];
2554         struct address_space *mapping = inode->i_mapping;
2555         const struct address_space_operations *a_ops = mapping->a_ops;
2556         int offset = off & (sb->s_blocksize - 1);
2557         size_t towrite = len;
2558         struct page *page;
2559         void *fsdata = NULL;
2560         int err = 0;
2561         int tocopy;
2562
2563         while (towrite > 0) {
2564                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2565                                                                 towrite);
2566 retry:
2567                 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2568                                                         &page, &fsdata);
2569                 if (unlikely(err)) {
2570                         if (err == -ENOMEM) {
2571                                 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2572                                 goto retry;
2573                         }
2574                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2575                         break;
2576                 }
2577
2578                 memcpy_to_page(page, offset, data, tocopy);
2579
2580                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2581                                                 page, fsdata);
2582                 offset = 0;
2583                 towrite -= tocopy;
2584                 off += tocopy;
2585                 data += tocopy;
2586                 cond_resched();
2587         }
2588
2589         if (len == towrite)
2590                 return err;
2591         inode->i_mtime = inode->i_ctime = current_time(inode);
2592         f2fs_mark_inode_dirty_sync(inode, false);
2593         return len - towrite;
2594 }
2595
2596 int f2fs_dquot_initialize(struct inode *inode)
2597 {
2598         if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) {
2599                 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT);
2600                 return -ESRCH;
2601         }
2602
2603         return dquot_initialize(inode);
2604 }
2605
2606 static struct dquot **f2fs_get_dquots(struct inode *inode)
2607 {
2608         return F2FS_I(inode)->i_dquot;
2609 }
2610
2611 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2612 {
2613         return &F2FS_I(inode)->i_reserved_quota;
2614 }
2615
2616 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2617 {
2618         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2619                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2620                 return 0;
2621         }
2622
2623         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2624                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
2625 }
2626
2627 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2628 {
2629         int enabled = 0;
2630         int i, err;
2631
2632         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2633                 err = f2fs_enable_quotas(sbi->sb);
2634                 if (err) {
2635                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2636                         return 0;
2637                 }
2638                 return 1;
2639         }
2640
2641         for (i = 0; i < MAXQUOTAS; i++) {
2642                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2643                         err = f2fs_quota_on_mount(sbi, i);
2644                         if (!err) {
2645                                 enabled = 1;
2646                                 continue;
2647                         }
2648                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2649                                  err, i);
2650                 }
2651         }
2652         return enabled;
2653 }
2654
2655 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2656                              unsigned int flags)
2657 {
2658         struct inode *qf_inode;
2659         unsigned long qf_inum;
2660         int err;
2661
2662         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2663
2664         qf_inum = f2fs_qf_ino(sb, type);
2665         if (!qf_inum)
2666                 return -EPERM;
2667
2668         qf_inode = f2fs_iget(sb, qf_inum);
2669         if (IS_ERR(qf_inode)) {
2670                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2671                 return PTR_ERR(qf_inode);
2672         }
2673
2674         /* Don't account quota for quota files to avoid recursion */
2675         qf_inode->i_flags |= S_NOQUOTA;
2676         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2677         iput(qf_inode);
2678         return err;
2679 }
2680
2681 static int f2fs_enable_quotas(struct super_block *sb)
2682 {
2683         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2684         int type, err = 0;
2685         unsigned long qf_inum;
2686         bool quota_mopt[MAXQUOTAS] = {
2687                 test_opt(sbi, USRQUOTA),
2688                 test_opt(sbi, GRPQUOTA),
2689                 test_opt(sbi, PRJQUOTA),
2690         };
2691
2692         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2693                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2694                 return 0;
2695         }
2696
2697         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2698
2699         for (type = 0; type < MAXQUOTAS; type++) {
2700                 qf_inum = f2fs_qf_ino(sb, type);
2701                 if (qf_inum) {
2702                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2703                                 DQUOT_USAGE_ENABLED |
2704                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2705                         if (err) {
2706                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2707                                          type, err);
2708                                 for (type--; type >= 0; type--)
2709                                         dquot_quota_off(sb, type);
2710                                 set_sbi_flag(F2FS_SB(sb),
2711                                                 SBI_QUOTA_NEED_REPAIR);
2712                                 return err;
2713                         }
2714                 }
2715         }
2716         return 0;
2717 }
2718
2719 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2720 {
2721         struct quota_info *dqopt = sb_dqopt(sbi->sb);
2722         struct address_space *mapping = dqopt->files[type]->i_mapping;
2723         int ret = 0;
2724
2725         ret = dquot_writeback_dquots(sbi->sb, type);
2726         if (ret)
2727                 goto out;
2728
2729         ret = filemap_fdatawrite(mapping);
2730         if (ret)
2731                 goto out;
2732
2733         /* if we are using journalled quota */
2734         if (is_journalled_quota(sbi))
2735                 goto out;
2736
2737         ret = filemap_fdatawait(mapping);
2738
2739         truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2740 out:
2741         if (ret)
2742                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2743         return ret;
2744 }
2745
2746 int f2fs_quota_sync(struct super_block *sb, int type)
2747 {
2748         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2749         struct quota_info *dqopt = sb_dqopt(sb);
2750         int cnt;
2751         int ret = 0;
2752
2753         /*
2754          * Now when everything is written we can discard the pagecache so
2755          * that userspace sees the changes.
2756          */
2757         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2758
2759                 if (type != -1 && cnt != type)
2760                         continue;
2761
2762                 if (!sb_has_quota_active(sb, cnt))
2763                         continue;
2764
2765                 if (!f2fs_sb_has_quota_ino(sbi))
2766                         inode_lock(dqopt->files[cnt]);
2767
2768                 /*
2769                  * do_quotactl
2770                  *  f2fs_quota_sync
2771                  *  f2fs_down_read(quota_sem)
2772                  *  dquot_writeback_dquots()
2773                  *  f2fs_dquot_commit
2774                  *                            block_operation
2775                  *                            f2fs_down_read(quota_sem)
2776                  */
2777                 f2fs_lock_op(sbi);
2778                 f2fs_down_read(&sbi->quota_sem);
2779
2780                 ret = f2fs_quota_sync_file(sbi, cnt);
2781
2782                 f2fs_up_read(&sbi->quota_sem);
2783                 f2fs_unlock_op(sbi);
2784
2785                 if (!f2fs_sb_has_quota_ino(sbi))
2786                         inode_unlock(dqopt->files[cnt]);
2787
2788                 if (ret)
2789                         break;
2790         }
2791         return ret;
2792 }
2793
2794 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2795                                                         const struct path *path)
2796 {
2797         struct inode *inode;
2798         int err;
2799
2800         /* if quota sysfile exists, deny enabling quota with specific file */
2801         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2802                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2803                 return -EBUSY;
2804         }
2805
2806         err = f2fs_quota_sync(sb, type);
2807         if (err)
2808                 return err;
2809
2810         err = dquot_quota_on(sb, type, format_id, path);
2811         if (err)
2812                 return err;
2813
2814         inode = d_inode(path->dentry);
2815
2816         inode_lock(inode);
2817         F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2818         f2fs_set_inode_flags(inode);
2819         inode_unlock(inode);
2820         f2fs_mark_inode_dirty_sync(inode, false);
2821
2822         return 0;
2823 }
2824
2825 static int __f2fs_quota_off(struct super_block *sb, int type)
2826 {
2827         struct inode *inode = sb_dqopt(sb)->files[type];
2828         int err;
2829
2830         if (!inode || !igrab(inode))
2831                 return dquot_quota_off(sb, type);
2832
2833         err = f2fs_quota_sync(sb, type);
2834         if (err)
2835                 goto out_put;
2836
2837         err = dquot_quota_off(sb, type);
2838         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2839                 goto out_put;
2840
2841         inode_lock(inode);
2842         F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2843         f2fs_set_inode_flags(inode);
2844         inode_unlock(inode);
2845         f2fs_mark_inode_dirty_sync(inode, false);
2846 out_put:
2847         iput(inode);
2848         return err;
2849 }
2850
2851 static int f2fs_quota_off(struct super_block *sb, int type)
2852 {
2853         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2854         int err;
2855
2856         err = __f2fs_quota_off(sb, type);
2857
2858         /*
2859          * quotactl can shutdown journalled quota, result in inconsistence
2860          * between quota record and fs data by following updates, tag the
2861          * flag to let fsck be aware of it.
2862          */
2863         if (is_journalled_quota(sbi))
2864                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2865         return err;
2866 }
2867
2868 void f2fs_quota_off_umount(struct super_block *sb)
2869 {
2870         int type;
2871         int err;
2872
2873         for (type = 0; type < MAXQUOTAS; type++) {
2874                 err = __f2fs_quota_off(sb, type);
2875                 if (err) {
2876                         int ret = dquot_quota_off(sb, type);
2877
2878                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2879                                  type, err, ret);
2880                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2881                 }
2882         }
2883         /*
2884          * In case of checkpoint=disable, we must flush quota blocks.
2885          * This can cause NULL exception for node_inode in end_io, since
2886          * put_super already dropped it.
2887          */
2888         sync_filesystem(sb);
2889 }
2890
2891 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2892 {
2893         struct quota_info *dqopt = sb_dqopt(sb);
2894         int type;
2895
2896         for (type = 0; type < MAXQUOTAS; type++) {
2897                 if (!dqopt->files[type])
2898                         continue;
2899                 f2fs_inode_synced(dqopt->files[type]);
2900         }
2901 }
2902
2903 static int f2fs_dquot_commit(struct dquot *dquot)
2904 {
2905         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2906         int ret;
2907
2908         f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2909         ret = dquot_commit(dquot);
2910         if (ret < 0)
2911                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2912         f2fs_up_read(&sbi->quota_sem);
2913         return ret;
2914 }
2915
2916 static int f2fs_dquot_acquire(struct dquot *dquot)
2917 {
2918         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2919         int ret;
2920
2921         f2fs_down_read(&sbi->quota_sem);
2922         ret = dquot_acquire(dquot);
2923         if (ret < 0)
2924                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2925         f2fs_up_read(&sbi->quota_sem);
2926         return ret;
2927 }
2928
2929 static int f2fs_dquot_release(struct dquot *dquot)
2930 {
2931         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2932         int ret = dquot_release(dquot);
2933
2934         if (ret < 0)
2935                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2936         return ret;
2937 }
2938
2939 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2940 {
2941         struct super_block *sb = dquot->dq_sb;
2942         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2943         int ret = dquot_mark_dquot_dirty(dquot);
2944
2945         /* if we are using journalled quota */
2946         if (is_journalled_quota(sbi))
2947                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2948
2949         return ret;
2950 }
2951
2952 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2953 {
2954         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2955         int ret = dquot_commit_info(sb, type);
2956
2957         if (ret < 0)
2958                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2959         return ret;
2960 }
2961
2962 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2963 {
2964         *projid = F2FS_I(inode)->i_projid;
2965         return 0;
2966 }
2967
2968 static const struct dquot_operations f2fs_quota_operations = {
2969         .get_reserved_space = f2fs_get_reserved_space,
2970         .write_dquot    = f2fs_dquot_commit,
2971         .acquire_dquot  = f2fs_dquot_acquire,
2972         .release_dquot  = f2fs_dquot_release,
2973         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2974         .write_info     = f2fs_dquot_commit_info,
2975         .alloc_dquot    = dquot_alloc,
2976         .destroy_dquot  = dquot_destroy,
2977         .get_projid     = f2fs_get_projid,
2978         .get_next_id    = dquot_get_next_id,
2979 };
2980
2981 static const struct quotactl_ops f2fs_quotactl_ops = {
2982         .quota_on       = f2fs_quota_on,
2983         .quota_off      = f2fs_quota_off,
2984         .quota_sync     = f2fs_quota_sync,
2985         .get_state      = dquot_get_state,
2986         .set_info       = dquot_set_dqinfo,
2987         .get_dqblk      = dquot_get_dqblk,
2988         .set_dqblk      = dquot_set_dqblk,
2989         .get_nextdqblk  = dquot_get_next_dqblk,
2990 };
2991 #else
2992 int f2fs_dquot_initialize(struct inode *inode)
2993 {
2994         return 0;
2995 }
2996
2997 int f2fs_quota_sync(struct super_block *sb, int type)
2998 {
2999         return 0;
3000 }
3001
3002 void f2fs_quota_off_umount(struct super_block *sb)
3003 {
3004 }
3005 #endif
3006
3007 static const struct super_operations f2fs_sops = {
3008         .alloc_inode    = f2fs_alloc_inode,
3009         .free_inode     = f2fs_free_inode,
3010         .drop_inode     = f2fs_drop_inode,
3011         .write_inode    = f2fs_write_inode,
3012         .dirty_inode    = f2fs_dirty_inode,
3013         .show_options   = f2fs_show_options,
3014 #ifdef CONFIG_QUOTA
3015         .quota_read     = f2fs_quota_read,
3016         .quota_write    = f2fs_quota_write,
3017         .get_dquots     = f2fs_get_dquots,
3018 #endif
3019         .evict_inode    = f2fs_evict_inode,
3020         .put_super      = f2fs_put_super,
3021         .sync_fs        = f2fs_sync_fs,
3022         .freeze_fs      = f2fs_freeze,
3023         .unfreeze_fs    = f2fs_unfreeze,
3024         .statfs         = f2fs_statfs,
3025         .remount_fs     = f2fs_remount,
3026 };
3027
3028 #ifdef CONFIG_FS_ENCRYPTION
3029 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3030 {
3031         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3032                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3033                                 ctx, len, NULL);
3034 }
3035
3036 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3037                                                         void *fs_data)
3038 {
3039         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3040
3041         /*
3042          * Encrypting the root directory is not allowed because fsck
3043          * expects lost+found directory to exist and remain unencrypted
3044          * if LOST_FOUND feature is enabled.
3045          *
3046          */
3047         if (f2fs_sb_has_lost_found(sbi) &&
3048                         inode->i_ino == F2FS_ROOT_INO(sbi))
3049                 return -EPERM;
3050
3051         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3052                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3053                                 ctx, len, fs_data, XATTR_CREATE);
3054 }
3055
3056 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3057 {
3058         return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3059 }
3060
3061 static bool f2fs_has_stable_inodes(struct super_block *sb)
3062 {
3063         return true;
3064 }
3065
3066 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
3067                                        int *ino_bits_ret, int *lblk_bits_ret)
3068 {
3069         *ino_bits_ret = 8 * sizeof(nid_t);
3070         *lblk_bits_ret = 8 * sizeof(block_t);
3071 }
3072
3073 static struct block_device **f2fs_get_devices(struct super_block *sb,
3074                                               unsigned int *num_devs)
3075 {
3076         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3077         struct block_device **devs;
3078         int i;
3079
3080         if (!f2fs_is_multi_device(sbi))
3081                 return NULL;
3082
3083         devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3084         if (!devs)
3085                 return ERR_PTR(-ENOMEM);
3086
3087         for (i = 0; i < sbi->s_ndevs; i++)
3088                 devs[i] = FDEV(i).bdev;
3089         *num_devs = sbi->s_ndevs;
3090         return devs;
3091 }
3092
3093 static const struct fscrypt_operations f2fs_cryptops = {
3094         .key_prefix             = "f2fs:",
3095         .get_context            = f2fs_get_context,
3096         .set_context            = f2fs_set_context,
3097         .get_dummy_policy       = f2fs_get_dummy_policy,
3098         .empty_dir              = f2fs_empty_dir,
3099         .has_stable_inodes      = f2fs_has_stable_inodes,
3100         .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
3101         .get_devices            = f2fs_get_devices,
3102 };
3103 #endif
3104
3105 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3106                 u64 ino, u32 generation)
3107 {
3108         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3109         struct inode *inode;
3110
3111         if (f2fs_check_nid_range(sbi, ino))
3112                 return ERR_PTR(-ESTALE);
3113
3114         /*
3115          * f2fs_iget isn't quite right if the inode is currently unallocated!
3116          * However f2fs_iget currently does appropriate checks to handle stale
3117          * inodes so everything is OK.
3118          */
3119         inode = f2fs_iget(sb, ino);
3120         if (IS_ERR(inode))
3121                 return ERR_CAST(inode);
3122         if (unlikely(generation && inode->i_generation != generation)) {
3123                 /* we didn't find the right inode.. */
3124                 iput(inode);
3125                 return ERR_PTR(-ESTALE);
3126         }
3127         return inode;
3128 }
3129
3130 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3131                 int fh_len, int fh_type)
3132 {
3133         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3134                                     f2fs_nfs_get_inode);
3135 }
3136
3137 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3138                 int fh_len, int fh_type)
3139 {
3140         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3141                                     f2fs_nfs_get_inode);
3142 }
3143
3144 static const struct export_operations f2fs_export_ops = {
3145         .fh_to_dentry = f2fs_fh_to_dentry,
3146         .fh_to_parent = f2fs_fh_to_parent,
3147         .get_parent = f2fs_get_parent,
3148 };
3149
3150 loff_t max_file_blocks(struct inode *inode)
3151 {
3152         loff_t result = 0;
3153         loff_t leaf_count;
3154
3155         /*
3156          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3157          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3158          * space in inode.i_addr, it will be more safe to reassign
3159          * result as zero.
3160          */
3161
3162         if (inode && f2fs_compressed_file(inode))
3163                 leaf_count = ADDRS_PER_BLOCK(inode);
3164         else
3165                 leaf_count = DEF_ADDRS_PER_BLOCK;
3166
3167         /* two direct node blocks */
3168         result += (leaf_count * 2);
3169
3170         /* two indirect node blocks */
3171         leaf_count *= NIDS_PER_BLOCK;
3172         result += (leaf_count * 2);
3173
3174         /* one double indirect node block */
3175         leaf_count *= NIDS_PER_BLOCK;
3176         result += leaf_count;
3177
3178         return result;
3179 }
3180
3181 static int __f2fs_commit_super(struct buffer_head *bh,
3182                         struct f2fs_super_block *super)
3183 {
3184         lock_buffer(bh);
3185         if (super)
3186                 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
3187         set_buffer_dirty(bh);
3188         unlock_buffer(bh);
3189
3190         /* it's rare case, we can do fua all the time */
3191         return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
3192 }
3193
3194 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3195                                         struct buffer_head *bh)
3196 {
3197         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3198                                         (bh->b_data + F2FS_SUPER_OFFSET);
3199         struct super_block *sb = sbi->sb;
3200         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3201         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3202         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3203         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3204         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3205         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3206         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3207         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3208         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3209         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3210         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3211         u32 segment_count = le32_to_cpu(raw_super->segment_count);
3212         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3213         u64 main_end_blkaddr = main_blkaddr +
3214                                 (segment_count_main << log_blocks_per_seg);
3215         u64 seg_end_blkaddr = segment0_blkaddr +
3216                                 (segment_count << log_blocks_per_seg);
3217
3218         if (segment0_blkaddr != cp_blkaddr) {
3219                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3220                           segment0_blkaddr, cp_blkaddr);
3221                 return true;
3222         }
3223
3224         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3225                                                         sit_blkaddr) {
3226                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3227                           cp_blkaddr, sit_blkaddr,
3228                           segment_count_ckpt << log_blocks_per_seg);
3229                 return true;
3230         }
3231
3232         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3233                                                         nat_blkaddr) {
3234                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3235                           sit_blkaddr, nat_blkaddr,
3236                           segment_count_sit << log_blocks_per_seg);
3237                 return true;
3238         }
3239
3240         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3241                                                         ssa_blkaddr) {
3242                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3243                           nat_blkaddr, ssa_blkaddr,
3244                           segment_count_nat << log_blocks_per_seg);
3245                 return true;
3246         }
3247
3248         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3249                                                         main_blkaddr) {
3250                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3251                           ssa_blkaddr, main_blkaddr,
3252                           segment_count_ssa << log_blocks_per_seg);
3253                 return true;
3254         }
3255
3256         if (main_end_blkaddr > seg_end_blkaddr) {
3257                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3258                           main_blkaddr, seg_end_blkaddr,
3259                           segment_count_main << log_blocks_per_seg);
3260                 return true;
3261         } else if (main_end_blkaddr < seg_end_blkaddr) {
3262                 int err = 0;
3263                 char *res;
3264
3265                 /* fix in-memory information all the time */
3266                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3267                                 segment0_blkaddr) >> log_blocks_per_seg);
3268
3269                 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
3270                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3271                         res = "internally";
3272                 } else {
3273                         err = __f2fs_commit_super(bh, NULL);
3274                         res = err ? "failed" : "done";
3275                 }
3276                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3277                           res, main_blkaddr, seg_end_blkaddr,
3278                           segment_count_main << log_blocks_per_seg);
3279                 if (err)
3280                         return true;
3281         }
3282         return false;
3283 }
3284
3285 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3286                                 struct buffer_head *bh)
3287 {
3288         block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3289         block_t total_sections, blocks_per_seg;
3290         struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
3291                                         (bh->b_data + F2FS_SUPER_OFFSET);
3292         size_t crc_offset = 0;
3293         __u32 crc = 0;
3294
3295         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3296                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3297                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3298                 return -EINVAL;
3299         }
3300
3301         /* Check checksum_offset and crc in superblock */
3302         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3303                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3304                 if (crc_offset !=
3305                         offsetof(struct f2fs_super_block, crc)) {
3306                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3307                                   crc_offset);
3308                         return -EFSCORRUPTED;
3309                 }
3310                 crc = le32_to_cpu(raw_super->crc);
3311                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3312                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3313                         return -EFSCORRUPTED;
3314                 }
3315         }
3316
3317         /* Currently, support only 4KB block size */
3318         if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3319                 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3320                           le32_to_cpu(raw_super->log_blocksize),
3321                           F2FS_BLKSIZE_BITS);
3322                 return -EFSCORRUPTED;
3323         }
3324
3325         /* check log blocks per segment */
3326         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3327                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3328                           le32_to_cpu(raw_super->log_blocks_per_seg));
3329                 return -EFSCORRUPTED;
3330         }
3331
3332         /* Currently, support 512/1024/2048/4096 bytes sector size */
3333         if (le32_to_cpu(raw_super->log_sectorsize) >
3334                                 F2FS_MAX_LOG_SECTOR_SIZE ||
3335                 le32_to_cpu(raw_super->log_sectorsize) <
3336                                 F2FS_MIN_LOG_SECTOR_SIZE) {
3337                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3338                           le32_to_cpu(raw_super->log_sectorsize));
3339                 return -EFSCORRUPTED;
3340         }
3341         if (le32_to_cpu(raw_super->log_sectors_per_block) +
3342                 le32_to_cpu(raw_super->log_sectorsize) !=
3343                         F2FS_MAX_LOG_SECTOR_SIZE) {
3344                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3345                           le32_to_cpu(raw_super->log_sectors_per_block),
3346                           le32_to_cpu(raw_super->log_sectorsize));
3347                 return -EFSCORRUPTED;
3348         }
3349
3350         segment_count = le32_to_cpu(raw_super->segment_count);
3351         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3352         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3353         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3354         total_sections = le32_to_cpu(raw_super->section_count);
3355
3356         /* blocks_per_seg should be 512, given the above check */
3357         blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
3358
3359         if (segment_count > F2FS_MAX_SEGMENT ||
3360                                 segment_count < F2FS_MIN_SEGMENTS) {
3361                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3362                 return -EFSCORRUPTED;
3363         }
3364
3365         if (total_sections > segment_count_main || total_sections < 1 ||
3366                         segs_per_sec > segment_count || !segs_per_sec) {
3367                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3368                           segment_count, total_sections, segs_per_sec);
3369                 return -EFSCORRUPTED;
3370         }
3371
3372         if (segment_count_main != total_sections * segs_per_sec) {
3373                 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3374                           segment_count_main, total_sections, segs_per_sec);
3375                 return -EFSCORRUPTED;
3376         }
3377
3378         if ((segment_count / segs_per_sec) < total_sections) {
3379                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3380                           segment_count, segs_per_sec, total_sections);
3381                 return -EFSCORRUPTED;
3382         }
3383
3384         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3385                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3386                           segment_count, le64_to_cpu(raw_super->block_count));
3387                 return -EFSCORRUPTED;
3388         }
3389
3390         if (RDEV(0).path[0]) {
3391                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3392                 int i = 1;
3393
3394                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3395                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3396                         i++;
3397                 }
3398                 if (segment_count != dev_seg_count) {
3399                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3400                                         segment_count, dev_seg_count);
3401                         return -EFSCORRUPTED;
3402                 }
3403         } else {
3404                 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3405                                         !bdev_is_zoned(sbi->sb->s_bdev)) {
3406                         f2fs_info(sbi, "Zoned block device path is missing");
3407                         return -EFSCORRUPTED;
3408                 }
3409         }
3410
3411         if (secs_per_zone > total_sections || !secs_per_zone) {
3412                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3413                           secs_per_zone, total_sections);
3414                 return -EFSCORRUPTED;
3415         }
3416         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3417                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3418                         (le32_to_cpu(raw_super->extension_count) +
3419                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3420                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3421                           le32_to_cpu(raw_super->extension_count),
3422                           raw_super->hot_ext_count,
3423                           F2FS_MAX_EXTENSION);
3424                 return -EFSCORRUPTED;
3425         }
3426
3427         if (le32_to_cpu(raw_super->cp_payload) >=
3428                                 (blocks_per_seg - F2FS_CP_PACKS -
3429                                 NR_CURSEG_PERSIST_TYPE)) {
3430                 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3431                           le32_to_cpu(raw_super->cp_payload),
3432                           blocks_per_seg - F2FS_CP_PACKS -
3433                           NR_CURSEG_PERSIST_TYPE);
3434                 return -EFSCORRUPTED;
3435         }
3436
3437         /* check reserved ino info */
3438         if (le32_to_cpu(raw_super->node_ino) != 1 ||
3439                 le32_to_cpu(raw_super->meta_ino) != 2 ||
3440                 le32_to_cpu(raw_super->root_ino) != 3) {
3441                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3442                           le32_to_cpu(raw_super->node_ino),
3443                           le32_to_cpu(raw_super->meta_ino),
3444                           le32_to_cpu(raw_super->root_ino));
3445                 return -EFSCORRUPTED;
3446         }
3447
3448         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3449         if (sanity_check_area_boundary(sbi, bh))
3450                 return -EFSCORRUPTED;
3451
3452         return 0;
3453 }
3454
3455 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3456 {
3457         unsigned int total, fsmeta;
3458         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3459         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3460         unsigned int ovp_segments, reserved_segments;
3461         unsigned int main_segs, blocks_per_seg;
3462         unsigned int sit_segs, nat_segs;
3463         unsigned int sit_bitmap_size, nat_bitmap_size;
3464         unsigned int log_blocks_per_seg;
3465         unsigned int segment_count_main;
3466         unsigned int cp_pack_start_sum, cp_payload;
3467         block_t user_block_count, valid_user_blocks;
3468         block_t avail_node_count, valid_node_count;
3469         unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3470         int i, j;
3471
3472         total = le32_to_cpu(raw_super->segment_count);
3473         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3474         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3475         fsmeta += sit_segs;
3476         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3477         fsmeta += nat_segs;
3478         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3479         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3480
3481         if (unlikely(fsmeta >= total))
3482                 return 1;
3483
3484         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3485         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3486
3487         if (!f2fs_sb_has_readonly(sbi) &&
3488                         unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3489                         ovp_segments == 0 || reserved_segments == 0)) {
3490                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3491                 return 1;
3492         }
3493         user_block_count = le64_to_cpu(ckpt->user_block_count);
3494         segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3495                         (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3496         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3497         if (!user_block_count || user_block_count >=
3498                         segment_count_main << log_blocks_per_seg) {
3499                 f2fs_err(sbi, "Wrong user_block_count: %u",
3500                          user_block_count);
3501                 return 1;
3502         }
3503
3504         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3505         if (valid_user_blocks > user_block_count) {
3506                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3507                          valid_user_blocks, user_block_count);
3508                 return 1;
3509         }
3510
3511         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3512         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3513         if (valid_node_count > avail_node_count) {
3514                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3515                          valid_node_count, avail_node_count);
3516                 return 1;
3517         }
3518
3519         main_segs = le32_to_cpu(raw_super->segment_count_main);
3520         blocks_per_seg = sbi->blocks_per_seg;
3521
3522         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3523                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3524                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3525                         return 1;
3526
3527                 if (f2fs_sb_has_readonly(sbi))
3528                         goto check_data;
3529
3530                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3531                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3532                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
3533                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3534                                          i, j,
3535                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3536                                 return 1;
3537                         }
3538                 }
3539         }
3540 check_data:
3541         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3542                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3543                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3544                         return 1;
3545
3546                 if (f2fs_sb_has_readonly(sbi))
3547                         goto skip_cross;
3548
3549                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3550                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3551                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3552                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3553                                          i, j,
3554                                          le32_to_cpu(ckpt->cur_data_segno[i]));
3555                                 return 1;
3556                         }
3557                 }
3558         }
3559         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3560                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3561                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3562                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3563                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3564                                          i, j,
3565                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3566                                 return 1;
3567                         }
3568                 }
3569         }
3570 skip_cross:
3571         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3572         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3573
3574         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3575                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3576                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3577                          sit_bitmap_size, nat_bitmap_size);
3578                 return 1;
3579         }
3580
3581         cp_pack_start_sum = __start_sum_addr(sbi);
3582         cp_payload = __cp_payload(sbi);
3583         if (cp_pack_start_sum < cp_payload + 1 ||
3584                 cp_pack_start_sum > blocks_per_seg - 1 -
3585                         NR_CURSEG_PERSIST_TYPE) {
3586                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3587                          cp_pack_start_sum);
3588                 return 1;
3589         }
3590
3591         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3592                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3593                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3594                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3595                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3596                           le32_to_cpu(ckpt->checksum_offset));
3597                 return 1;
3598         }
3599
3600         nat_blocks = nat_segs << log_blocks_per_seg;
3601         nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3602         nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3603         if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3604                 (cp_payload + F2FS_CP_PACKS +
3605                 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3606                 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3607                           cp_payload, nat_bits_blocks);
3608                 return 1;
3609         }
3610
3611         if (unlikely(f2fs_cp_error(sbi))) {
3612                 f2fs_err(sbi, "A bug case: need to run fsck");
3613                 return 1;
3614         }
3615         return 0;
3616 }
3617
3618 static void init_sb_info(struct f2fs_sb_info *sbi)
3619 {
3620         struct f2fs_super_block *raw_super = sbi->raw_super;
3621         int i;
3622
3623         sbi->log_sectors_per_block =
3624                 le32_to_cpu(raw_super->log_sectors_per_block);
3625         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3626         sbi->blocksize = 1 << sbi->log_blocksize;
3627         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3628         sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3629         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3630         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3631         sbi->total_sections = le32_to_cpu(raw_super->section_count);
3632         sbi->total_node_count =
3633                 (le32_to_cpu(raw_super->segment_count_nat) / 2)
3634                         * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3635         F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3636         F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3637         F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3638         sbi->cur_victim_sec = NULL_SECNO;
3639         sbi->gc_mode = GC_NORMAL;
3640         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3641         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3642         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3643         sbi->migration_granularity = sbi->segs_per_sec;
3644         sbi->seq_file_ra_mul = MIN_RA_MUL;
3645         sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3646         sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3647         spin_lock_init(&sbi->gc_remaining_trials_lock);
3648         atomic64_set(&sbi->current_atomic_write, 0);
3649
3650         sbi->dir_level = DEF_DIR_LEVEL;
3651         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3652         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3653         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3654         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3655         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3656         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3657                                 DEF_UMOUNT_DISCARD_TIMEOUT;
3658         clear_sbi_flag(sbi, SBI_NEED_FSCK);
3659
3660         for (i = 0; i < NR_COUNT_TYPE; i++)
3661                 atomic_set(&sbi->nr_pages[i], 0);
3662
3663         for (i = 0; i < META; i++)
3664                 atomic_set(&sbi->wb_sync_req[i], 0);
3665
3666         INIT_LIST_HEAD(&sbi->s_list);
3667         mutex_init(&sbi->umount_mutex);
3668         init_f2fs_rwsem(&sbi->io_order_lock);
3669         spin_lock_init(&sbi->cp_lock);
3670
3671         sbi->dirty_device = 0;
3672         spin_lock_init(&sbi->dev_lock);
3673
3674         init_f2fs_rwsem(&sbi->sb_lock);
3675         init_f2fs_rwsem(&sbi->pin_sem);
3676 }
3677
3678 static int init_percpu_info(struct f2fs_sb_info *sbi)
3679 {
3680         int err;
3681
3682         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3683         if (err)
3684                 return err;
3685
3686         err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3687         if (err)
3688                 goto err_valid_block;
3689
3690         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3691                                                                 GFP_KERNEL);
3692         if (err)
3693                 goto err_node_block;
3694         return 0;
3695
3696 err_node_block:
3697         percpu_counter_destroy(&sbi->rf_node_block_count);
3698 err_valid_block:
3699         percpu_counter_destroy(&sbi->alloc_valid_block_count);
3700         return err;
3701 }
3702
3703 #ifdef CONFIG_BLK_DEV_ZONED
3704
3705 struct f2fs_report_zones_args {
3706         struct f2fs_sb_info *sbi;
3707         struct f2fs_dev_info *dev;
3708 };
3709
3710 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3711                               void *data)
3712 {
3713         struct f2fs_report_zones_args *rz_args = data;
3714         block_t unusable_blocks = (zone->len - zone->capacity) >>
3715                                         F2FS_LOG_SECTORS_PER_BLOCK;
3716
3717         if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3718                 return 0;
3719
3720         set_bit(idx, rz_args->dev->blkz_seq);
3721         if (!rz_args->sbi->unusable_blocks_per_sec) {
3722                 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3723                 return 0;
3724         }
3725         if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3726                 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3727                 return -EINVAL;
3728         }
3729         return 0;
3730 }
3731
3732 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3733 {
3734         struct block_device *bdev = FDEV(devi).bdev;
3735         sector_t nr_sectors = bdev_nr_sectors(bdev);
3736         struct f2fs_report_zones_args rep_zone_arg;
3737         u64 zone_sectors;
3738         int ret;
3739
3740         if (!f2fs_sb_has_blkzoned(sbi))
3741                 return 0;
3742
3743         zone_sectors = bdev_zone_sectors(bdev);
3744         if (!is_power_of_2(zone_sectors)) {
3745                 f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n");
3746                 return -EINVAL;
3747         }
3748
3749         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3750                                 SECTOR_TO_BLOCK(zone_sectors))
3751                 return -EINVAL;
3752         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3753         if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3754                                 __ilog2_u32(sbi->blocks_per_blkz))
3755                 return -EINVAL;
3756         sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3757         FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3758                                         sbi->log_blocks_per_blkz;
3759         if (nr_sectors & (zone_sectors - 1))
3760                 FDEV(devi).nr_blkz++;
3761
3762         FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3763                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
3764                                         * sizeof(unsigned long),
3765                                         GFP_KERNEL);
3766         if (!FDEV(devi).blkz_seq)
3767                 return -ENOMEM;
3768
3769         rep_zone_arg.sbi = sbi;
3770         rep_zone_arg.dev = &FDEV(devi);
3771
3772         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3773                                   &rep_zone_arg);
3774         if (ret < 0)
3775                 return ret;
3776         return 0;
3777 }
3778 #endif
3779
3780 /*
3781  * Read f2fs raw super block.
3782  * Because we have two copies of super block, so read both of them
3783  * to get the first valid one. If any one of them is broken, we pass
3784  * them recovery flag back to the caller.
3785  */
3786 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3787                         struct f2fs_super_block **raw_super,
3788                         int *valid_super_block, int *recovery)
3789 {
3790         struct super_block *sb = sbi->sb;
3791         int block;
3792         struct buffer_head *bh;
3793         struct f2fs_super_block *super;
3794         int err = 0;
3795
3796         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3797         if (!super)
3798                 return -ENOMEM;
3799
3800         for (block = 0; block < 2; block++) {
3801                 bh = sb_bread(sb, block);
3802                 if (!bh) {
3803                         f2fs_err(sbi, "Unable to read %dth superblock",
3804                                  block + 1);
3805                         err = -EIO;
3806                         *recovery = 1;
3807                         continue;
3808                 }
3809
3810                 /* sanity checking of raw super */
3811                 err = sanity_check_raw_super(sbi, bh);
3812                 if (err) {
3813                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3814                                  block + 1);
3815                         brelse(bh);
3816                         *recovery = 1;
3817                         continue;
3818                 }
3819
3820                 if (!*raw_super) {
3821                         memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3822                                                         sizeof(*super));
3823                         *valid_super_block = block;
3824                         *raw_super = super;
3825                 }
3826                 brelse(bh);
3827         }
3828
3829         /* No valid superblock */
3830         if (!*raw_super)
3831                 kfree(super);
3832         else
3833                 err = 0;
3834
3835         return err;
3836 }
3837
3838 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3839 {
3840         struct buffer_head *bh;
3841         __u32 crc = 0;
3842         int err;
3843
3844         if ((recover && f2fs_readonly(sbi->sb)) ||
3845                                 bdev_read_only(sbi->sb->s_bdev)) {
3846                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3847                 return -EROFS;
3848         }
3849
3850         /* we should update superblock crc here */
3851         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3852                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3853                                 offsetof(struct f2fs_super_block, crc));
3854                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3855         }
3856
3857         /* write back-up superblock first */
3858         bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3859         if (!bh)
3860                 return -EIO;
3861         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3862         brelse(bh);
3863
3864         /* if we are in recovery path, skip writing valid superblock */
3865         if (recover || err)
3866                 return err;
3867
3868         /* write current valid superblock */
3869         bh = sb_bread(sbi->sb, sbi->valid_super_block);
3870         if (!bh)
3871                 return -EIO;
3872         err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3873         brelse(bh);
3874         return err;
3875 }
3876
3877 void f2fs_handle_stop(struct f2fs_sb_info *sbi, unsigned char reason)
3878 {
3879         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3880         int err;
3881
3882         f2fs_down_write(&sbi->sb_lock);
3883
3884         if (raw_super->s_stop_reason[reason] < ((1 << BITS_PER_BYTE) - 1))
3885                 raw_super->s_stop_reason[reason]++;
3886
3887         err = f2fs_commit_super(sbi, false);
3888         if (err)
3889                 f2fs_err(sbi, "f2fs_commit_super fails to record reason:%u err:%d",
3890                                                                 reason, err);
3891         f2fs_up_write(&sbi->sb_lock);
3892 }
3893
3894 static void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
3895 {
3896         spin_lock(&sbi->error_lock);
3897         if (!test_bit(flag, (unsigned long *)sbi->errors)) {
3898                 set_bit(flag, (unsigned long *)sbi->errors);
3899                 sbi->error_dirty = true;
3900         }
3901         spin_unlock(&sbi->error_lock);
3902 }
3903
3904 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
3905 {
3906         bool need_update = false;
3907
3908         spin_lock(&sbi->error_lock);
3909         if (sbi->error_dirty) {
3910                 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
3911                                                         MAX_F2FS_ERRORS);
3912                 sbi->error_dirty = false;
3913                 need_update = true;
3914         }
3915         spin_unlock(&sbi->error_lock);
3916
3917         return need_update;
3918 }
3919
3920 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
3921 {
3922         int err;
3923
3924         f2fs_save_errors(sbi, error);
3925
3926         f2fs_down_write(&sbi->sb_lock);
3927
3928         if (!f2fs_update_errors(sbi))
3929                 goto out_unlock;
3930
3931         err = f2fs_commit_super(sbi, false);
3932         if (err)
3933                 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d",
3934                                                                 error, err);
3935 out_unlock:
3936         f2fs_up_write(&sbi->sb_lock);
3937 }
3938
3939 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3940 {
3941         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3942         unsigned int max_devices = MAX_DEVICES;
3943         unsigned int logical_blksize;
3944         int i;
3945
3946         /* Initialize single device information */
3947         if (!RDEV(0).path[0]) {
3948                 if (!bdev_is_zoned(sbi->sb->s_bdev))
3949                         return 0;
3950                 max_devices = 1;
3951         }
3952
3953         /*
3954          * Initialize multiple devices information, or single
3955          * zoned block device information.
3956          */
3957         sbi->devs = f2fs_kzalloc(sbi,
3958                                  array_size(max_devices,
3959                                             sizeof(struct f2fs_dev_info)),
3960                                  GFP_KERNEL);
3961         if (!sbi->devs)
3962                 return -ENOMEM;
3963
3964         logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
3965         sbi->aligned_blksize = true;
3966
3967         for (i = 0; i < max_devices; i++) {
3968
3969                 if (i > 0 && !RDEV(i).path[0])
3970                         break;
3971
3972                 if (max_devices == 1) {
3973                         /* Single zoned block device mount */
3974                         FDEV(0).bdev =
3975                                 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3976                                         sbi->sb->s_mode, sbi->sb->s_type);
3977                 } else {
3978                         /* Multi-device mount */
3979                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3980                         FDEV(i).total_segments =
3981                                 le32_to_cpu(RDEV(i).total_segments);
3982                         if (i == 0) {
3983                                 FDEV(i).start_blk = 0;
3984                                 FDEV(i).end_blk = FDEV(i).start_blk +
3985                                     (FDEV(i).total_segments <<
3986                                     sbi->log_blocks_per_seg) - 1 +
3987                                     le32_to_cpu(raw_super->segment0_blkaddr);
3988                         } else {
3989                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3990                                 FDEV(i).end_blk = FDEV(i).start_blk +
3991                                         (FDEV(i).total_segments <<
3992                                         sbi->log_blocks_per_seg) - 1;
3993                         }
3994                         FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3995                                         sbi->sb->s_mode, sbi->sb->s_type);
3996                 }
3997                 if (IS_ERR(FDEV(i).bdev))
3998                         return PTR_ERR(FDEV(i).bdev);
3999
4000                 /* to release errored devices */
4001                 sbi->s_ndevs = i + 1;
4002
4003                 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4004                         sbi->aligned_blksize = false;
4005
4006 #ifdef CONFIG_BLK_DEV_ZONED
4007                 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
4008                                 !f2fs_sb_has_blkzoned(sbi)) {
4009                         f2fs_err(sbi, "Zoned block device feature not enabled");
4010                         return -EINVAL;
4011                 }
4012                 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
4013                         if (init_blkz_info(sbi, i)) {
4014                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4015                                 return -EINVAL;
4016                         }
4017                         if (max_devices == 1)
4018                                 break;
4019                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
4020                                   i, FDEV(i).path,
4021                                   FDEV(i).total_segments,
4022                                   FDEV(i).start_blk, FDEV(i).end_blk,
4023                                   bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
4024                                   "Host-aware" : "Host-managed");
4025                         continue;
4026                 }
4027 #endif
4028                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4029                           i, FDEV(i).path,
4030                           FDEV(i).total_segments,
4031                           FDEV(i).start_blk, FDEV(i).end_blk);
4032         }
4033         f2fs_info(sbi,
4034                   "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
4035         return 0;
4036 }
4037
4038 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4039 {
4040 #if IS_ENABLED(CONFIG_UNICODE)
4041         if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4042                 const struct f2fs_sb_encodings *encoding_info;
4043                 struct unicode_map *encoding;
4044                 __u16 encoding_flags;
4045
4046                 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4047                 if (!encoding_info) {
4048                         f2fs_err(sbi,
4049                                  "Encoding requested by superblock is unknown");
4050                         return -EINVAL;
4051                 }
4052
4053                 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4054                 encoding = utf8_load(encoding_info->version);
4055                 if (IS_ERR(encoding)) {
4056                         f2fs_err(sbi,
4057                                  "can't mount with superblock charset: %s-%u.%u.%u "
4058                                  "not supported by the kernel. flags: 0x%x.",
4059                                  encoding_info->name,
4060                                  unicode_major(encoding_info->version),
4061                                  unicode_minor(encoding_info->version),
4062                                  unicode_rev(encoding_info->version),
4063                                  encoding_flags);
4064                         return PTR_ERR(encoding);
4065                 }
4066                 f2fs_info(sbi, "Using encoding defined by superblock: "
4067                          "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4068                          unicode_major(encoding_info->version),
4069                          unicode_minor(encoding_info->version),
4070                          unicode_rev(encoding_info->version),
4071                          encoding_flags);
4072
4073                 sbi->sb->s_encoding = encoding;
4074                 sbi->sb->s_encoding_flags = encoding_flags;
4075         }
4076 #else
4077         if (f2fs_sb_has_casefold(sbi)) {
4078                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4079                 return -EINVAL;
4080         }
4081 #endif
4082         return 0;
4083 }
4084
4085 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4086 {
4087         /* adjust parameters according to the volume size */
4088         if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4089                 if (f2fs_block_unit_discard(sbi))
4090                         SM_I(sbi)->dcc_info->discard_granularity =
4091                                                 MIN_DISCARD_GRANULARITY;
4092                 SM_I(sbi)->ipu_policy = 1 << F2FS_IPU_FORCE |
4093                                         1 << F2FS_IPU_HONOR_OPU_WRITE;
4094         }
4095
4096         sbi->readdir_ra = true;
4097 }
4098
4099 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4100 {
4101         struct f2fs_sb_info *sbi;
4102         struct f2fs_super_block *raw_super;
4103         struct inode *root;
4104         int err;
4105         bool skip_recovery = false, need_fsck = false;
4106         char *options = NULL;
4107         int recovery, i, valid_super_block;
4108         struct curseg_info *seg_i;
4109         int retry_cnt = 1;
4110
4111 try_onemore:
4112         err = -EINVAL;
4113         raw_super = NULL;
4114         valid_super_block = -1;
4115         recovery = 0;
4116
4117         /* allocate memory for f2fs-specific super block info */
4118         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4119         if (!sbi)
4120                 return -ENOMEM;
4121
4122         sbi->sb = sb;
4123
4124         /* initialize locks within allocated memory */
4125         init_f2fs_rwsem(&sbi->gc_lock);
4126         mutex_init(&sbi->writepages);
4127         init_f2fs_rwsem(&sbi->cp_global_sem);
4128         init_f2fs_rwsem(&sbi->node_write);
4129         init_f2fs_rwsem(&sbi->node_change);
4130         spin_lock_init(&sbi->stat_lock);
4131         init_f2fs_rwsem(&sbi->cp_rwsem);
4132         init_f2fs_rwsem(&sbi->quota_sem);
4133         init_waitqueue_head(&sbi->cp_wait);
4134         spin_lock_init(&sbi->error_lock);
4135
4136         for (i = 0; i < NR_INODE_TYPE; i++) {
4137                 INIT_LIST_HEAD(&sbi->inode_list[i]);
4138                 spin_lock_init(&sbi->inode_lock[i]);
4139         }
4140         mutex_init(&sbi->flush_lock);
4141
4142         /* Load the checksum driver */
4143         sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
4144         if (IS_ERR(sbi->s_chksum_driver)) {
4145                 f2fs_err(sbi, "Cannot load crc32 driver.");
4146                 err = PTR_ERR(sbi->s_chksum_driver);
4147                 sbi->s_chksum_driver = NULL;
4148                 goto free_sbi;
4149         }
4150
4151         /* set a block size */
4152         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4153                 f2fs_err(sbi, "unable to set blocksize");
4154                 goto free_sbi;
4155         }
4156
4157         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4158                                                                 &recovery);
4159         if (err)
4160                 goto free_sbi;
4161
4162         sb->s_fs_info = sbi;
4163         sbi->raw_super = raw_super;
4164
4165         memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4166
4167         /* precompute checksum seed for metadata */
4168         if (f2fs_sb_has_inode_chksum(sbi))
4169                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4170                                                 sizeof(raw_super->uuid));
4171
4172         default_options(sbi);
4173         /* parse mount options */
4174         options = kstrdup((const char *)data, GFP_KERNEL);
4175         if (data && !options) {
4176                 err = -ENOMEM;
4177                 goto free_sb_buf;
4178         }
4179
4180         err = parse_options(sb, options, false);
4181         if (err)
4182                 goto free_options;
4183
4184         sb->s_maxbytes = max_file_blocks(NULL) <<
4185                                 le32_to_cpu(raw_super->log_blocksize);
4186         sb->s_max_links = F2FS_LINK_MAX;
4187
4188         err = f2fs_setup_casefold(sbi);
4189         if (err)
4190                 goto free_options;
4191
4192 #ifdef CONFIG_QUOTA
4193         sb->dq_op = &f2fs_quota_operations;
4194         sb->s_qcop = &f2fs_quotactl_ops;
4195         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4196
4197         if (f2fs_sb_has_quota_ino(sbi)) {
4198                 for (i = 0; i < MAXQUOTAS; i++) {
4199                         if (f2fs_qf_ino(sbi->sb, i))
4200                                 sbi->nquota_files++;
4201                 }
4202         }
4203 #endif
4204
4205         sb->s_op = &f2fs_sops;
4206 #ifdef CONFIG_FS_ENCRYPTION
4207         sb->s_cop = &f2fs_cryptops;
4208 #endif
4209 #ifdef CONFIG_FS_VERITY
4210         sb->s_vop = &f2fs_verityops;
4211 #endif
4212         sb->s_xattr = f2fs_xattr_handlers;
4213         sb->s_export_op = &f2fs_export_ops;
4214         sb->s_magic = F2FS_SUPER_MAGIC;
4215         sb->s_time_gran = 1;
4216         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4217                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4218         memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
4219         sb->s_iflags |= SB_I_CGROUPWB;
4220
4221         /* init f2fs-specific super block info */
4222         sbi->valid_super_block = valid_super_block;
4223
4224         /* disallow all the data/node/meta page writes */
4225         set_sbi_flag(sbi, SBI_POR_DOING);
4226
4227         err = f2fs_init_write_merge_io(sbi);
4228         if (err)
4229                 goto free_bio_info;
4230
4231         init_sb_info(sbi);
4232
4233         err = f2fs_init_iostat(sbi);
4234         if (err)
4235                 goto free_bio_info;
4236
4237         err = init_percpu_info(sbi);
4238         if (err)
4239                 goto free_iostat;
4240
4241         if (F2FS_IO_ALIGNED(sbi)) {
4242                 sbi->write_io_dummy =
4243                         mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
4244                 if (!sbi->write_io_dummy) {
4245                         err = -ENOMEM;
4246                         goto free_percpu;
4247                 }
4248         }
4249
4250         /* init per sbi slab cache */
4251         err = f2fs_init_xattr_caches(sbi);
4252         if (err)
4253                 goto free_io_dummy;
4254         err = f2fs_init_page_array_cache(sbi);
4255         if (err)
4256                 goto free_xattr_cache;
4257
4258         /* get an inode for meta space */
4259         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4260         if (IS_ERR(sbi->meta_inode)) {
4261                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4262                 err = PTR_ERR(sbi->meta_inode);
4263                 goto free_page_array_cache;
4264         }
4265
4266         err = f2fs_get_valid_checkpoint(sbi);
4267         if (err) {
4268                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4269                 goto free_meta_inode;
4270         }
4271
4272         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4273                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4274         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4275                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4276                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4277         }
4278
4279         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4280                 set_sbi_flag(sbi, SBI_NEED_FSCK);
4281
4282         /* Initialize device list */
4283         err = f2fs_scan_devices(sbi);
4284         if (err) {
4285                 f2fs_err(sbi, "Failed to find devices");
4286                 goto free_devices;
4287         }
4288
4289         err = f2fs_init_post_read_wq(sbi);
4290         if (err) {
4291                 f2fs_err(sbi, "Failed to initialize post read workqueue");
4292                 goto free_devices;
4293         }
4294
4295         sbi->total_valid_node_count =
4296                                 le32_to_cpu(sbi->ckpt->valid_node_count);
4297         percpu_counter_set(&sbi->total_valid_inode_count,
4298                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
4299         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4300         sbi->total_valid_block_count =
4301                                 le64_to_cpu(sbi->ckpt->valid_block_count);
4302         sbi->last_valid_block_count = sbi->total_valid_block_count;
4303         sbi->reserved_blocks = 0;
4304         sbi->current_reserved_blocks = 0;
4305         limit_reserve_root(sbi);
4306         adjust_unusable_cap_perc(sbi);
4307
4308         f2fs_init_extent_cache_info(sbi);
4309
4310         f2fs_init_ino_entry_info(sbi);
4311
4312         f2fs_init_fsync_node_info(sbi);
4313
4314         /* setup checkpoint request control and start checkpoint issue thread */
4315         f2fs_init_ckpt_req_control(sbi);
4316         if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4317                         test_opt(sbi, MERGE_CHECKPOINT)) {
4318                 err = f2fs_start_ckpt_thread(sbi);
4319                 if (err) {
4320                         f2fs_err(sbi,
4321                             "Failed to start F2FS issue_checkpoint_thread (%d)",
4322                             err);
4323                         goto stop_ckpt_thread;
4324                 }
4325         }
4326
4327         /* setup f2fs internal modules */
4328         err = f2fs_build_segment_manager(sbi);
4329         if (err) {
4330                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4331                          err);
4332                 goto free_sm;
4333         }
4334         err = f2fs_build_node_manager(sbi);
4335         if (err) {
4336                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4337                          err);
4338                 goto free_nm;
4339         }
4340
4341         err = adjust_reserved_segment(sbi);
4342         if (err)
4343                 goto free_nm;
4344
4345         /* For write statistics */
4346         sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4347
4348         /* Read accumulated write IO statistics if exists */
4349         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4350         if (__exist_node_summaries(sbi))
4351                 sbi->kbytes_written =
4352                         le64_to_cpu(seg_i->journal->info.kbytes_written);
4353
4354         f2fs_build_gc_manager(sbi);
4355
4356         err = f2fs_build_stats(sbi);
4357         if (err)
4358                 goto free_nm;
4359
4360         /* get an inode for node space */
4361         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4362         if (IS_ERR(sbi->node_inode)) {
4363                 f2fs_err(sbi, "Failed to read node inode");
4364                 err = PTR_ERR(sbi->node_inode);
4365                 goto free_stats;
4366         }
4367
4368         /* read root inode and dentry */
4369         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4370         if (IS_ERR(root)) {
4371                 f2fs_err(sbi, "Failed to read root inode");
4372                 err = PTR_ERR(root);
4373                 goto free_node_inode;
4374         }
4375         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4376                         !root->i_size || !root->i_nlink) {
4377                 iput(root);
4378                 err = -EINVAL;
4379                 goto free_node_inode;
4380         }
4381
4382         sb->s_root = d_make_root(root); /* allocate root dentry */
4383         if (!sb->s_root) {
4384                 err = -ENOMEM;
4385                 goto free_node_inode;
4386         }
4387
4388         err = f2fs_init_compress_inode(sbi);
4389         if (err)
4390                 goto free_root_inode;
4391
4392         err = f2fs_register_sysfs(sbi);
4393         if (err)
4394                 goto free_compress_inode;
4395
4396 #ifdef CONFIG_QUOTA
4397         /* Enable quota usage during mount */
4398         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4399                 err = f2fs_enable_quotas(sb);
4400                 if (err)
4401                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4402         }
4403 #endif
4404         /* if there are any orphan inodes, free them */
4405         err = f2fs_recover_orphan_inodes(sbi);
4406         if (err)
4407                 goto free_meta;
4408
4409         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4410                 goto reset_checkpoint;
4411
4412         /* recover fsynced data */
4413         if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4414                         !test_opt(sbi, NORECOVERY)) {
4415                 /*
4416                  * mount should be failed, when device has readonly mode, and
4417                  * previous checkpoint was not done by clean system shutdown.
4418                  */
4419                 if (f2fs_hw_is_readonly(sbi)) {
4420                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4421                                 err = f2fs_recover_fsync_data(sbi, true);
4422                                 if (err > 0) {
4423                                         err = -EROFS;
4424                                         f2fs_err(sbi, "Need to recover fsync data, but "
4425                                                 "write access unavailable, please try "
4426                                                 "mount w/ disable_roll_forward or norecovery");
4427                                 }
4428                                 if (err < 0)
4429                                         goto free_meta;
4430                         }
4431                         f2fs_info(sbi, "write access unavailable, skipping recovery");
4432                         goto reset_checkpoint;
4433                 }
4434
4435                 if (need_fsck)
4436                         set_sbi_flag(sbi, SBI_NEED_FSCK);
4437
4438                 if (skip_recovery)
4439                         goto reset_checkpoint;
4440
4441                 err = f2fs_recover_fsync_data(sbi, false);
4442                 if (err < 0) {
4443                         if (err != -ENOMEM)
4444                                 skip_recovery = true;
4445                         need_fsck = true;
4446                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4447                                  err);
4448                         goto free_meta;
4449                 }
4450         } else {
4451                 err = f2fs_recover_fsync_data(sbi, true);
4452
4453                 if (!f2fs_readonly(sb) && err > 0) {
4454                         err = -EINVAL;
4455                         f2fs_err(sbi, "Need to recover fsync data");
4456                         goto free_meta;
4457                 }
4458         }
4459
4460         /*
4461          * If the f2fs is not readonly and fsync data recovery succeeds,
4462          * check zoned block devices' write pointer consistency.
4463          */
4464         if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
4465                 err = f2fs_check_write_pointer(sbi);
4466                 if (err)
4467                         goto free_meta;
4468         }
4469
4470 reset_checkpoint:
4471         f2fs_init_inmem_curseg(sbi);
4472
4473         /* f2fs_recover_fsync_data() cleared this already */
4474         clear_sbi_flag(sbi, SBI_POR_DOING);
4475
4476         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4477                 err = f2fs_disable_checkpoint(sbi);
4478                 if (err)
4479                         goto sync_free_meta;
4480         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4481                 f2fs_enable_checkpoint(sbi);
4482         }
4483
4484         /*
4485          * If filesystem is not mounted as read-only then
4486          * do start the gc_thread.
4487          */
4488         if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4489                 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4490                 /* After POR, we can run background GC thread.*/
4491                 err = f2fs_start_gc_thread(sbi);
4492                 if (err)
4493                         goto sync_free_meta;
4494         }
4495         kvfree(options);
4496
4497         /* recover broken superblock */
4498         if (recovery) {
4499                 err = f2fs_commit_super(sbi, true);
4500                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4501                           sbi->valid_super_block ? 1 : 2, err);
4502         }
4503
4504         f2fs_join_shrinker(sbi);
4505
4506         f2fs_tuning_parameters(sbi);
4507
4508         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4509                     cur_cp_version(F2FS_CKPT(sbi)));
4510         f2fs_update_time(sbi, CP_TIME);
4511         f2fs_update_time(sbi, REQ_TIME);
4512         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4513         return 0;
4514
4515 sync_free_meta:
4516         /* safe to flush all the data */
4517         sync_filesystem(sbi->sb);
4518         retry_cnt = 0;
4519
4520 free_meta:
4521 #ifdef CONFIG_QUOTA
4522         f2fs_truncate_quota_inode_pages(sb);
4523         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4524                 f2fs_quota_off_umount(sbi->sb);
4525 #endif
4526         /*
4527          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4528          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4529          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4530          * falls into an infinite loop in f2fs_sync_meta_pages().
4531          */
4532         truncate_inode_pages_final(META_MAPPING(sbi));
4533         /* evict some inodes being cached by GC */
4534         evict_inodes(sb);
4535         f2fs_unregister_sysfs(sbi);
4536 free_compress_inode:
4537         f2fs_destroy_compress_inode(sbi);
4538 free_root_inode:
4539         dput(sb->s_root);
4540         sb->s_root = NULL;
4541 free_node_inode:
4542         f2fs_release_ino_entry(sbi, true);
4543         truncate_inode_pages_final(NODE_MAPPING(sbi));
4544         iput(sbi->node_inode);
4545         sbi->node_inode = NULL;
4546 free_stats:
4547         f2fs_destroy_stats(sbi);
4548 free_nm:
4549         /* stop discard thread before destroying node manager */
4550         f2fs_stop_discard_thread(sbi);
4551         f2fs_destroy_node_manager(sbi);
4552 free_sm:
4553         f2fs_destroy_segment_manager(sbi);
4554 stop_ckpt_thread:
4555         f2fs_stop_ckpt_thread(sbi);
4556         f2fs_destroy_post_read_wq(sbi);
4557 free_devices:
4558         destroy_device_list(sbi);
4559         kvfree(sbi->ckpt);
4560 free_meta_inode:
4561         make_bad_inode(sbi->meta_inode);
4562         iput(sbi->meta_inode);
4563         sbi->meta_inode = NULL;
4564 free_page_array_cache:
4565         f2fs_destroy_page_array_cache(sbi);
4566 free_xattr_cache:
4567         f2fs_destroy_xattr_caches(sbi);
4568 free_io_dummy:
4569         mempool_destroy(sbi->write_io_dummy);
4570 free_percpu:
4571         destroy_percpu_info(sbi);
4572 free_iostat:
4573         f2fs_destroy_iostat(sbi);
4574 free_bio_info:
4575         for (i = 0; i < NR_PAGE_TYPE; i++)
4576                 kvfree(sbi->write_io[i]);
4577
4578 #if IS_ENABLED(CONFIG_UNICODE)
4579         utf8_unload(sb->s_encoding);
4580         sb->s_encoding = NULL;
4581 #endif
4582 free_options:
4583 #ifdef CONFIG_QUOTA
4584         for (i = 0; i < MAXQUOTAS; i++)
4585                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4586 #endif
4587         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4588         kvfree(options);
4589 free_sb_buf:
4590         kfree(raw_super);
4591 free_sbi:
4592         if (sbi->s_chksum_driver)
4593                 crypto_free_shash(sbi->s_chksum_driver);
4594         kfree(sbi);
4595
4596         /* give only one another chance */
4597         if (retry_cnt > 0 && skip_recovery) {
4598                 retry_cnt--;
4599                 shrink_dcache_sb(sb);
4600                 goto try_onemore;
4601         }
4602         return err;
4603 }
4604
4605 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4606                         const char *dev_name, void *data)
4607 {
4608         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4609 }
4610
4611 static void kill_f2fs_super(struct super_block *sb)
4612 {
4613         if (sb->s_root) {
4614                 struct f2fs_sb_info *sbi = F2FS_SB(sb);
4615
4616                 set_sbi_flag(sbi, SBI_IS_CLOSE);
4617                 f2fs_stop_gc_thread(sbi);
4618                 f2fs_stop_discard_thread(sbi);
4619
4620 #ifdef CONFIG_F2FS_FS_COMPRESSION
4621                 /*
4622                  * latter evict_inode() can bypass checking and invalidating
4623                  * compress inode cache.
4624                  */
4625                 if (test_opt(sbi, COMPRESS_CACHE))
4626                         truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4627 #endif
4628
4629                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4630                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4631                         struct cp_control cpc = {
4632                                 .reason = CP_UMOUNT,
4633                         };
4634                         f2fs_write_checkpoint(sbi, &cpc);
4635                 }
4636
4637                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4638                         sb->s_flags &= ~SB_RDONLY;
4639         }
4640         kill_block_super(sb);
4641 }
4642
4643 static struct file_system_type f2fs_fs_type = {
4644         .owner          = THIS_MODULE,
4645         .name           = "f2fs",
4646         .mount          = f2fs_mount,
4647         .kill_sb        = kill_f2fs_super,
4648         .fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4649 };
4650 MODULE_ALIAS_FS("f2fs");
4651
4652 static int __init init_inodecache(void)
4653 {
4654         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4655                         sizeof(struct f2fs_inode_info), 0,
4656                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4657         return f2fs_inode_cachep ? 0 : -ENOMEM;
4658 }
4659
4660 static void destroy_inodecache(void)
4661 {
4662         /*
4663          * Make sure all delayed rcu free inodes are flushed before we
4664          * destroy cache.
4665          */
4666         rcu_barrier();
4667         kmem_cache_destroy(f2fs_inode_cachep);
4668 }
4669
4670 static int __init init_f2fs_fs(void)
4671 {
4672         int err;
4673
4674         if (PAGE_SIZE != F2FS_BLKSIZE) {
4675                 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4676                                 PAGE_SIZE, F2FS_BLKSIZE);
4677                 return -EINVAL;
4678         }
4679
4680         err = init_inodecache();
4681         if (err)
4682                 goto fail;
4683         err = f2fs_create_node_manager_caches();
4684         if (err)
4685                 goto free_inodecache;
4686         err = f2fs_create_segment_manager_caches();
4687         if (err)
4688                 goto free_node_manager_caches;
4689         err = f2fs_create_checkpoint_caches();
4690         if (err)
4691                 goto free_segment_manager_caches;
4692         err = f2fs_create_recovery_cache();
4693         if (err)
4694                 goto free_checkpoint_caches;
4695         err = f2fs_create_extent_cache();
4696         if (err)
4697                 goto free_recovery_cache;
4698         err = f2fs_create_garbage_collection_cache();
4699         if (err)
4700                 goto free_extent_cache;
4701         err = f2fs_init_sysfs();
4702         if (err)
4703                 goto free_garbage_collection_cache;
4704         err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker");
4705         if (err)
4706                 goto free_sysfs;
4707         err = register_filesystem(&f2fs_fs_type);
4708         if (err)
4709                 goto free_shrinker;
4710         f2fs_create_root_stats();
4711         err = f2fs_init_post_read_processing();
4712         if (err)
4713                 goto free_root_stats;
4714         err = f2fs_init_iostat_processing();
4715         if (err)
4716                 goto free_post_read;
4717         err = f2fs_init_bio_entry_cache();
4718         if (err)
4719                 goto free_iostat;
4720         err = f2fs_init_bioset();
4721         if (err)
4722                 goto free_bio_entry_cache;
4723         err = f2fs_init_compress_mempool();
4724         if (err)
4725                 goto free_bioset;
4726         err = f2fs_init_compress_cache();
4727         if (err)
4728                 goto free_compress_mempool;
4729         err = f2fs_create_casefold_cache();
4730         if (err)
4731                 goto free_compress_cache;
4732         return 0;
4733 free_compress_cache:
4734         f2fs_destroy_compress_cache();
4735 free_compress_mempool:
4736         f2fs_destroy_compress_mempool();
4737 free_bioset:
4738         f2fs_destroy_bioset();
4739 free_bio_entry_cache:
4740         f2fs_destroy_bio_entry_cache();
4741 free_iostat:
4742         f2fs_destroy_iostat_processing();
4743 free_post_read:
4744         f2fs_destroy_post_read_processing();
4745 free_root_stats:
4746         f2fs_destroy_root_stats();
4747         unregister_filesystem(&f2fs_fs_type);
4748 free_shrinker:
4749         unregister_shrinker(&f2fs_shrinker_info);
4750 free_sysfs:
4751         f2fs_exit_sysfs();
4752 free_garbage_collection_cache:
4753         f2fs_destroy_garbage_collection_cache();
4754 free_extent_cache:
4755         f2fs_destroy_extent_cache();
4756 free_recovery_cache:
4757         f2fs_destroy_recovery_cache();
4758 free_checkpoint_caches:
4759         f2fs_destroy_checkpoint_caches();
4760 free_segment_manager_caches:
4761         f2fs_destroy_segment_manager_caches();
4762 free_node_manager_caches:
4763         f2fs_destroy_node_manager_caches();
4764 free_inodecache:
4765         destroy_inodecache();
4766 fail:
4767         return err;
4768 }
4769
4770 static void __exit exit_f2fs_fs(void)
4771 {
4772         f2fs_destroy_casefold_cache();
4773         f2fs_destroy_compress_cache();
4774         f2fs_destroy_compress_mempool();
4775         f2fs_destroy_bioset();
4776         f2fs_destroy_bio_entry_cache();
4777         f2fs_destroy_iostat_processing();
4778         f2fs_destroy_post_read_processing();
4779         f2fs_destroy_root_stats();
4780         unregister_filesystem(&f2fs_fs_type);
4781         unregister_shrinker(&f2fs_shrinker_info);
4782         f2fs_exit_sysfs();
4783         f2fs_destroy_garbage_collection_cache();
4784         f2fs_destroy_extent_cache();
4785         f2fs_destroy_recovery_cache();
4786         f2fs_destroy_checkpoint_caches();
4787         f2fs_destroy_segment_manager_caches();
4788         f2fs_destroy_node_manager_caches();
4789         destroy_inodecache();
4790 }
4791
4792 module_init(init_f2fs_fs)
4793 module_exit(exit_f2fs_fs)
4794
4795 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4796 MODULE_DESCRIPTION("Flash Friendly File System");
4797 MODULE_LICENSE("GPL");
4798 MODULE_SOFTDEP("pre: crc32");
4799