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/kthread.h>
  15 #include <linux/parser.h>
  16 #include <linux/mount.h>
  17 #include <linux/seq_file.h>
  18 #include <linux/proc_fs.h>
  19 #include <linux/random.h>
  20 #include <linux/exportfs.h>
  21 #include <linux/blkdev.h>
  22 #include <linux/quotaops.h>
  23 #include <linux/f2fs_fs.h>
  24 #include <linux/sysfs.h>
  25 #include <linux/quota.h>
  26 #include <linux/unicode.h>
  27 #include <linux/part_stat.h>
  28 #include <linux/zstd.h>
  29 #include <linux/lz4.h>
  30
  31 #include "f2fs.h"
  32 #include "node.h"
  33 #include "segment.h"
  34 #include "xattr.h"
  35 #include "gc.h"
  36 #include "iostat.h"
  37
  38 #define CREATE_TRACE_POINTS
  39 #include <trace/events/f2fs.h>
  40
  41 static struct kmem_cache *f2fs_inode_cachep;
  42
  43 #ifdef CONFIG_F2FS_FAULT_INJECTION
  44
  45 const char *f2fs_fault_name[FAULT_MAX] = {
  46         [FAULT_KMALLOC]                 = "kmalloc",
  47         [FAULT_KVMALLOC]                = "kvmalloc",
  48         [FAULT_PAGE_ALLOC]              = "page alloc",
  49         [FAULT_PAGE_GET]                = "page get",
  50         [FAULT_ALLOC_NID]               = "alloc nid",
  51         [FAULT_ORPHAN]                  = "orphan",
  52         [FAULT_BLOCK]                   = "no more block",
  53         [FAULT_DIR_DEPTH]               = "too big dir depth",
  54         [FAULT_EVICT_INODE]             = "evict_inode fail",
  55         [FAULT_TRUNCATE]                = "truncate fail",
  56         [FAULT_READ_IO]                 = "read IO error",
  57         [FAULT_CHECKPOINT]              = "checkpoint error",
  58         [FAULT_DISCARD]                 = "discard error",
  59         [FAULT_WRITE_IO]                = "write IO error",
  60         [FAULT_SLAB_ALLOC]              = "slab alloc",
  61         [FAULT_DQUOT_INIT]              = "dquot initialize",
  62         [FAULT_LOCK_OP]                 = "lock_op",
  63         [FAULT_BLKADDR_VALIDITY]        = "invalid blkaddr",
  64         [FAULT_BLKADDR_CONSISTENCE]     = "inconsistent blkaddr",
  65         [FAULT_NO_SEGMENT]              = "no free segment",
  66 };
  67
  68 int f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned long rate,
  69                                                         unsigned long type)
  70 {
  71         struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  72
  73         if (rate) {
  74                 if (rate > INT_MAX)
  75                         return -EINVAL;
  76                 atomic_set(&ffi->inject_ops, 0);
  77                 ffi->inject_rate = (int)rate;
  78         }
  79
  80         if (type) {
  81                 if (type >= BIT(FAULT_MAX))
  82                         return -EINVAL;
  83                 ffi->inject_type = (unsigned int)type;
  84         }
  85
  86         if (!rate && !type)
  87                 memset(ffi, 0, sizeof(struct f2fs_fault_info));
  88         else
  89                 f2fs_info(sbi,
  90                         "build fault injection attr: rate: %lu, type: 0x%lx",
  91                                                                 rate, type);
  92         return 0;
  93 }
  94 #endif
  95
  96 /* f2fs-wide shrinker description */
  97 static struct shrinker *f2fs_shrinker_info;
  98
  99 static int __init f2fs_init_shrinker(void)
 100 {
 101         f2fs_shrinker_info = shrinker_alloc(0, "f2fs-shrinker");
 102         if (!f2fs_shrinker_info)
 103                 return -ENOMEM;
 104
 105         f2fs_shrinker_info->count_objects = f2fs_shrink_count;
 106         f2fs_shrinker_info->scan_objects = f2fs_shrink_scan;
 107
 108         shrinker_register(f2fs_shrinker_info);
 109
 110         return 0;
 111 }
 112
 113 static void f2fs_exit_shrinker(void)
 114 {
 115         shrinker_free(f2fs_shrinker_info);
 116 }
 117
 118 enum {
 119         Opt_gc_background,
 120         Opt_disable_roll_forward,
 121         Opt_norecovery,
 122         Opt_discard,
 123         Opt_nodiscard,
 124         Opt_noheap,
 125         Opt_heap,
 126         Opt_user_xattr,
 127         Opt_nouser_xattr,
 128         Opt_acl,
 129         Opt_noacl,
 130         Opt_active_logs,
 131         Opt_disable_ext_identify,
 132         Opt_inline_xattr,
 133         Opt_noinline_xattr,
 134         Opt_inline_xattr_size,
 135         Opt_inline_data,
 136         Opt_inline_dentry,
 137         Opt_noinline_dentry,
 138         Opt_flush_merge,
 139         Opt_noflush_merge,
 140         Opt_barrier,
 141         Opt_nobarrier,
 142         Opt_fastboot,
 143         Opt_extent_cache,
 144         Opt_noextent_cache,
 145         Opt_noinline_data,
 146         Opt_data_flush,
 147         Opt_reserve_root,
 148         Opt_resgid,
 149         Opt_resuid,
 150         Opt_mode,
 151         Opt_fault_injection,
 152         Opt_fault_type,
 153         Opt_lazytime,
 154         Opt_nolazytime,
 155         Opt_quota,
 156         Opt_noquota,
 157         Opt_usrquota,
 158         Opt_grpquota,
 159         Opt_prjquota,
 160         Opt_usrjquota,
 161         Opt_grpjquota,
 162         Opt_prjjquota,
 163         Opt_offusrjquota,
 164         Opt_offgrpjquota,
 165         Opt_offprjjquota,
 166         Opt_jqfmt_vfsold,
 167         Opt_jqfmt_vfsv0,
 168         Opt_jqfmt_vfsv1,
 169         Opt_alloc,
 170         Opt_fsync,
 171         Opt_test_dummy_encryption,
 172         Opt_inlinecrypt,
 173         Opt_checkpoint_disable,
 174         Opt_checkpoint_disable_cap,
 175         Opt_checkpoint_disable_cap_perc,
 176         Opt_checkpoint_enable,
 177         Opt_checkpoint_merge,
 178         Opt_nocheckpoint_merge,
 179         Opt_compress_algorithm,
 180         Opt_compress_log_size,
 181         Opt_compress_extension,
 182         Opt_nocompress_extension,
 183         Opt_compress_chksum,
 184         Opt_compress_mode,
 185         Opt_compress_cache,
 186         Opt_atgc,
 187         Opt_gc_merge,
 188         Opt_nogc_merge,
 189         Opt_discard_unit,
 190         Opt_memory_mode,
 191         Opt_age_extent_cache,
 192         Opt_errors,
 193         Opt_err,
 194 };
 195
 196 static match_table_t f2fs_tokens = {
 197         {Opt_gc_background, "background_gc=%s"},
 198         {Opt_disable_roll_forward, "disable_roll_forward"},
 199         {Opt_norecovery, "norecovery"},
 200         {Opt_discard, "discard"},
 201         {Opt_nodiscard, "nodiscard"},
 202         {Opt_noheap, "no_heap"},
 203         {Opt_heap, "heap"},
 204         {Opt_user_xattr, "user_xattr"},
 205         {Opt_nouser_xattr, "nouser_xattr"},
 206         {Opt_acl, "acl"},
 207         {Opt_noacl, "noacl"},
 208         {Opt_active_logs, "active_logs=%u"},
 209         {Opt_disable_ext_identify, "disable_ext_identify"},
 210         {Opt_inline_xattr, "inline_xattr"},
 211         {Opt_noinline_xattr, "noinline_xattr"},
 212         {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 213         {Opt_inline_data, "inline_data"},
 214         {Opt_inline_dentry, "inline_dentry"},
 215         {Opt_noinline_dentry, "noinline_dentry"},
 216         {Opt_flush_merge, "flush_merge"},
 217         {Opt_noflush_merge, "noflush_merge"},
 218         {Opt_barrier, "barrier"},
 219         {Opt_nobarrier, "nobarrier"},
 220         {Opt_fastboot, "fastboot"},
 221         {Opt_extent_cache, "extent_cache"},
 222         {Opt_noextent_cache, "noextent_cache"},
 223         {Opt_noinline_data, "noinline_data"},
 224         {Opt_data_flush, "data_flush"},
 225         {Opt_reserve_root, "reserve_root=%u"},
 226         {Opt_resgid, "resgid=%u"},
 227         {Opt_resuid, "resuid=%u"},
 228         {Opt_mode, "mode=%s"},
 229         {Opt_fault_injection, "fault_injection=%u"},
 230         {Opt_fault_type, "fault_type=%u"},
 231         {Opt_lazytime, "lazytime"},
 232         {Opt_nolazytime, "nolazytime"},
 233         {Opt_quota, "quota"},
 234         {Opt_noquota, "noquota"},
 235         {Opt_usrquota, "usrquota"},
 236         {Opt_grpquota, "grpquota"},
 237         {Opt_prjquota, "prjquota"},
 238         {Opt_usrjquota, "usrjquota=%s"},
 239         {Opt_grpjquota, "grpjquota=%s"},
 240         {Opt_prjjquota, "prjjquota=%s"},
 241         {Opt_offusrjquota, "usrjquota="},
 242         {Opt_offgrpjquota, "grpjquota="},
 243         {Opt_offprjjquota, "prjjquota="},
 244         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 245         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 246         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 247         {Opt_alloc, "alloc_mode=%s"},
 248         {Opt_fsync, "fsync_mode=%s"},
 249         {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
 250         {Opt_test_dummy_encryption, "test_dummy_encryption"},
 251         {Opt_inlinecrypt, "inlinecrypt"},
 252         {Opt_checkpoint_disable, "checkpoint=disable"},
 253         {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 254         {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 255         {Opt_checkpoint_enable, "checkpoint=enable"},
 256         {Opt_checkpoint_merge, "checkpoint_merge"},
 257         {Opt_nocheckpoint_merge, "nocheckpoint_merge"},
 258         {Opt_compress_algorithm, "compress_algorithm=%s"},
 259         {Opt_compress_log_size, "compress_log_size=%u"},
 260         {Opt_compress_extension, "compress_extension=%s"},
 261         {Opt_nocompress_extension, "nocompress_extension=%s"},
 262         {Opt_compress_chksum, "compress_chksum"},
 263         {Opt_compress_mode, "compress_mode=%s"},
 264         {Opt_compress_cache, "compress_cache"},
 265         {Opt_atgc, "atgc"},
 266         {Opt_gc_merge, "gc_merge"},
 267         {Opt_nogc_merge, "nogc_merge"},
 268         {Opt_discard_unit, "discard_unit=%s"},
 269         {Opt_memory_mode, "memory=%s"},
 270         {Opt_age_extent_cache, "age_extent_cache"},
 271         {Opt_errors, "errors=%s"},
 272         {Opt_err, NULL},
 273 };
 274
 275 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate,
 276                                                 const char *fmt, ...)
 277 {
 278         struct va_format vaf;
 279         va_list args;
 280         int level;
 281
 282         va_start(args, fmt);
 283
 284         level = printk_get_level(fmt);
 285         vaf.fmt = printk_skip_level(fmt);
 286         vaf.va = &args;
 287         if (limit_rate)
 288                 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n",
 289                         KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 290         else
 291                 printk("%c%cF2FS-fs (%s): %pV\n",
 292                         KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 293
 294         va_end(args);
 295 }
 296
 297 #if IS_ENABLED(CONFIG_UNICODE)
 298 static const struct f2fs_sb_encodings {
 299         __u16 magic;
 300         char *name;
 301         unsigned int version;
 302 } f2fs_sb_encoding_map[] = {
 303         {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
 304 };
 305
 306 static const struct f2fs_sb_encodings *
 307 f2fs_sb_read_encoding(const struct f2fs_super_block *sb)
 308 {
 309         __u16 magic = le16_to_cpu(sb->s_encoding);
 310         int i;
 311
 312         for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
 313                 if (magic == f2fs_sb_encoding_map[i].magic)
 314                         return &f2fs_sb_encoding_map[i];
 315
 316         return NULL;
 317 }
 318
 319 struct kmem_cache *f2fs_cf_name_slab;
 320 static int __init f2fs_create_casefold_cache(void)
 321 {
 322         f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name",
 323                                                    F2FS_NAME_LEN);
 324         return f2fs_cf_name_slab ? 0 : -ENOMEM;
 325 }
 326
 327 static void f2fs_destroy_casefold_cache(void)
 328 {
 329         kmem_cache_destroy(f2fs_cf_name_slab);
 330 }
 331 #else
 332 static int __init f2fs_create_casefold_cache(void) { return 0; }
 333 static void f2fs_destroy_casefold_cache(void) { }
 334 #endif
 335
 336 static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 337 {
 338         block_t limit = min((sbi->user_block_count >> 3),
 339                         sbi->user_block_count - sbi->reserved_blocks);
 340
 341         /* limit is 12.5% */
 342         if (test_opt(sbi, RESERVE_ROOT) &&
 343                         F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 344                 F2FS_OPTION(sbi).root_reserved_blocks = limit;
 345                 f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 346                           F2FS_OPTION(sbi).root_reserved_blocks);
 347         }
 348         if (!test_opt(sbi, RESERVE_ROOT) &&
 349                 (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 350                                 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 351                 !gid_eq(F2FS_OPTION(sbi).s_resgid,
 352                                 make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 353                 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 354                           from_kuid_munged(&init_user_ns,
 355                                            F2FS_OPTION(sbi).s_resuid),
 356                           from_kgid_munged(&init_user_ns,
 357                                            F2FS_OPTION(sbi).s_resgid));
 358 }
 359
 360 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
 361 {
 362         if (!F2FS_OPTION(sbi).unusable_cap_perc)
 363                 return;
 364
 365         if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
 366                 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
 367         else
 368                 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
 369                                         F2FS_OPTION(sbi).unusable_cap_perc;
 370
 371         f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
 372                         F2FS_OPTION(sbi).unusable_cap,
 373                         F2FS_OPTION(sbi).unusable_cap_perc);
 374 }
 375
 376 static void init_once(void *foo)
 377 {
 378         struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 379
 380         inode_init_once(&fi->vfs_inode);
 381 }
 382
 383 #ifdef CONFIG_QUOTA
 384 static const char * const quotatypes[] = INITQFNAMES;
 385 #define QTYPE2NAME(t) (quotatypes[t])
 386 static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 387                                                         substring_t *args)
 388 {
 389         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 390         char *qname;
 391         int ret = -EINVAL;
 392
 393         if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 394                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 395                 return -EINVAL;
 396         }
 397         if (f2fs_sb_has_quota_ino(sbi)) {
 398                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 399                 return 0;
 400         }
 401
 402         qname = match_strdup(args);
 403         if (!qname) {
 404                 f2fs_err(sbi, "Not enough memory for storing quotafile name");
 405                 return -ENOMEM;
 406         }
 407         if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 408                 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 409                         ret = 0;
 410                 else
 411                         f2fs_err(sbi, "%s quota file already specified",
 412                                  QTYPE2NAME(qtype));
 413                 goto errout;
 414         }
 415         if (strchr(qname, '/')) {
 416                 f2fs_err(sbi, "quotafile must be on filesystem root");
 417                 goto errout;
 418         }
 419         F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 420         set_opt(sbi, QUOTA);
 421         return 0;
 422 errout:
 423         kfree(qname);
 424         return ret;
 425 }
 426
 427 static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 428 {
 429         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 430
 431         if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 432                 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 433                 return -EINVAL;
 434         }
 435         kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 436         F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 437         return 0;
 438 }
 439
 440 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 441 {
 442         /*
 443          * We do the test below only for project quotas. 'usrquota' and
 444          * 'grpquota' mount options are allowed even without quota feature
 445          * to support legacy quotas in quota files.
 446          */
 447         if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 448                 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 449                 return -1;
 450         }
 451         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 452                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 453                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 454                 if (test_opt(sbi, USRQUOTA) &&
 455                                 F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 456                         clear_opt(sbi, USRQUOTA);
 457
 458                 if (test_opt(sbi, GRPQUOTA) &&
 459                                 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 460                         clear_opt(sbi, GRPQUOTA);
 461
 462                 if (test_opt(sbi, PRJQUOTA) &&
 463                                 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 464                         clear_opt(sbi, PRJQUOTA);
 465
 466                 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 467                                 test_opt(sbi, PRJQUOTA)) {
 468                         f2fs_err(sbi, "old and new quota format mixing");
 469                         return -1;
 470                 }
 471
 472                 if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 473                         f2fs_err(sbi, "journaled quota format not specified");
 474                         return -1;
 475                 }
 476         }
 477
 478         if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 479                 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 480                 F2FS_OPTION(sbi).s_jquota_fmt = 0;
 481         }
 482         return 0;
 483 }
 484 #endif
 485
 486 static int f2fs_set_test_dummy_encryption(struct super_block *sb,
 487                                           const char *opt,
 488                                           const substring_t *arg,
 489                                           bool is_remount)
 490 {
 491         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 492         struct fs_parameter param = {
 493                 .type = fs_value_is_string,
 494                 .string = arg->from ? arg->from : "",
 495         };
 496         struct fscrypt_dummy_policy *policy =
 497                 &F2FS_OPTION(sbi).dummy_enc_policy;
 498         int err;
 499
 500         if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
 501                 f2fs_warn(sbi, "test_dummy_encryption option not supported");
 502                 return -EINVAL;
 503         }
 504
 505         if (!f2fs_sb_has_encrypt(sbi)) {
 506                 f2fs_err(sbi, "Encrypt feature is off");
 507                 return -EINVAL;
 508         }
 509
 510         /*
 511          * This mount option is just for testing, and it's not worthwhile to
 512          * implement the extra complexity (e.g. RCU protection) that would be
 513          * needed to allow it to be set or changed during remount.  We do allow
 514          * it to be specified during remount, but only if there is no change.
 515          */
 516         if (is_remount && !fscrypt_is_dummy_policy_set(policy)) {
 517                 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
 518                 return -EINVAL;
 519         }
 520
 521         err = fscrypt_parse_test_dummy_encryption(&param, policy);
 522         if (err) {
 523                 if (err == -EEXIST)
 524                         f2fs_warn(sbi,
 525                                   "Can't change test_dummy_encryption on remount");
 526                 else if (err == -EINVAL)
 527                         f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
 528                                   opt);
 529                 else
 530                         f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
 531                                   opt, err);
 532                 return -EINVAL;
 533         }
 534         f2fs_warn(sbi, "Test dummy encryption mode enabled");
 535         return 0;
 536 }
 537
 538 #ifdef CONFIG_F2FS_FS_COMPRESSION
 539 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi,
 540                                         const char *new_ext, bool is_ext)
 541 {
 542         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 543         int ext_cnt;
 544         int i;
 545
 546         if (is_ext) {
 547                 ext = F2FS_OPTION(sbi).extensions;
 548                 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
 549         } else {
 550                 ext = F2FS_OPTION(sbi).noextensions;
 551                 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
 552         }
 553
 554         for (i = 0; i < ext_cnt; i++) {
 555                 if (!strcasecmp(new_ext, ext[i]))
 556                         return true;
 557         }
 558
 559         return false;
 560 }
 561
 562 /*
 563  * 1. The same extension name cannot not appear in both compress and non-compress extension
 564  * at the same time.
 565  * 2. If the compress extension specifies all files, the types specified by the non-compress
 566  * extension will be treated as special cases and will not be compressed.
 567  * 3. Don't allow the non-compress extension specifies all files.
 568  */
 569 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi)
 570 {
 571         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 572         unsigned char (*noext)[F2FS_EXTENSION_LEN];
 573         int ext_cnt, noext_cnt, index = 0, no_index = 0;
 574
 575         ext = F2FS_OPTION(sbi).extensions;
 576         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
 577         noext = F2FS_OPTION(sbi).noextensions;
 578         noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
 579
 580         if (!noext_cnt)
 581                 return 0;
 582
 583         for (no_index = 0; no_index < noext_cnt; no_index++) {
 584                 if (!strcasecmp("*", noext[no_index])) {
 585                         f2fs_info(sbi, "Don't allow the nocompress extension specifies all files");
 586                         return -EINVAL;
 587                 }
 588                 for (index = 0; index < ext_cnt; index++) {
 589                         if (!strcasecmp(ext[index], noext[no_index])) {
 590                                 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension",
 591                                                 ext[index]);
 592                                 return -EINVAL;
 593                         }
 594                 }
 595         }
 596         return 0;
 597 }
 598
 599 #ifdef CONFIG_F2FS_FS_LZ4
 600 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str)
 601 {
 602 #ifdef CONFIG_F2FS_FS_LZ4HC
 603         unsigned int level;
 604
 605         if (strlen(str) == 3) {
 606                 F2FS_OPTION(sbi).compress_level = 0;
 607                 return 0;
 608         }
 609
 610         str += 3;
 611
 612         if (str[0] != ':') {
 613                 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
 614                 return -EINVAL;
 615         }
 616         if (kstrtouint(str + 1, 10, &level))
 617                 return -EINVAL;
 618
 619         if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) {
 620                 f2fs_info(sbi, "invalid lz4hc compress level: %d", level);
 621                 return -EINVAL;
 622         }
 623
 624         F2FS_OPTION(sbi).compress_level = level;
 625         return 0;
 626 #else
 627         if (strlen(str) == 3) {
 628                 F2FS_OPTION(sbi).compress_level = 0;
 629                 return 0;
 630         }
 631         f2fs_info(sbi, "kernel doesn't support lz4hc compression");
 632         return -EINVAL;
 633 #endif
 634 }
 635 #endif
 636
 637 #ifdef CONFIG_F2FS_FS_ZSTD
 638 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str)
 639 {
 640         int level;
 641         int len = 4;
 642
 643         if (strlen(str) == len) {
 644                 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL;
 645                 return 0;
 646         }
 647
 648         str += len;
 649
 650         if (str[0] != ':') {
 651                 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>");
 652                 return -EINVAL;
 653         }
 654         if (kstrtoint(str + 1, 10, &level))
 655                 return -EINVAL;
 656
 657         /* f2fs does not support negative compress level now */
 658         if (level < 0) {
 659                 f2fs_info(sbi, "do not support negative compress level: %d", level);
 660                 return -ERANGE;
 661         }
 662
 663         if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) {
 664                 f2fs_info(sbi, "invalid zstd compress level: %d", level);
 665                 return -EINVAL;
 666         }
 667
 668         F2FS_OPTION(sbi).compress_level = level;
 669         return 0;
 670 }
 671 #endif
 672 #endif
 673
 674 static int parse_options(struct super_block *sb, char *options, bool is_remount)
 675 {
 676         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 677         substring_t args[MAX_OPT_ARGS];
 678 #ifdef CONFIG_F2FS_FS_COMPRESSION
 679         unsigned char (*ext)[F2FS_EXTENSION_LEN];
 680         unsigned char (*noext)[F2FS_EXTENSION_LEN];
 681         int ext_cnt, noext_cnt;
 682 #endif
 683         char *p, *name;
 684         int arg = 0;
 685         kuid_t uid;
 686         kgid_t gid;
 687         int ret;
 688
 689         if (!options)
 690                 goto default_check;
 691
 692         while ((p = strsep(&options, ",")) != NULL) {
 693                 int token;
 694
 695                 if (!*p)
 696                         continue;
 697                 /*
 698                  * Initialize args struct so we know whether arg was
 699                  * found; some options take optional arguments.
 700                  */
 701                 args[0].to = args[0].from = NULL;
 702                 token = match_token(p, f2fs_tokens, args);
 703
 704                 switch (token) {
 705                 case Opt_gc_background:
 706                         name = match_strdup(&args[0]);
 707
 708                         if (!name)
 709                                 return -ENOMEM;
 710                         if (!strcmp(name, "on")) {
 711                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
 712                         } else if (!strcmp(name, "off")) {
 713                                 if (f2fs_sb_has_blkzoned(sbi)) {
 714                                         f2fs_warn(sbi, "zoned devices need bggc");
 715                                         kfree(name);
 716                                         return -EINVAL;
 717                                 }
 718                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
 719                         } else if (!strcmp(name, "sync")) {
 720                                 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
 721                         } else {
 722                                 kfree(name);
 723                                 return -EINVAL;
 724                         }
 725                         kfree(name);
 726                         break;
 727                 case Opt_disable_roll_forward:
 728                         set_opt(sbi, DISABLE_ROLL_FORWARD);
 729                         break;
 730                 case Opt_norecovery:
 731                         /* this option mounts f2fs with ro */
 732                         set_opt(sbi, NORECOVERY);
 733                         if (!f2fs_readonly(sb))
 734                                 return -EINVAL;
 735                         break;
 736                 case Opt_discard:
 737                         if (!f2fs_hw_support_discard(sbi)) {
 738                                 f2fs_warn(sbi, "device does not support discard");
 739                                 break;
 740                         }
 741                         set_opt(sbi, DISCARD);
 742                         break;
 743                 case Opt_nodiscard:
 744                         if (f2fs_hw_should_discard(sbi)) {
 745                                 f2fs_warn(sbi, "discard is required for zoned block devices");
 746                                 return -EINVAL;
 747                         }
 748                         clear_opt(sbi, DISCARD);
 749                         break;
 750                 case Opt_noheap:
 751                 case Opt_heap:
 752                         f2fs_warn(sbi, "heap/no_heap options were deprecated");
 753                         break;
 754 #ifdef CONFIG_F2FS_FS_XATTR
 755                 case Opt_user_xattr:
 756                         set_opt(sbi, XATTR_USER);
 757                         break;
 758                 case Opt_nouser_xattr:
 759                         clear_opt(sbi, XATTR_USER);
 760                         break;
 761                 case Opt_inline_xattr:
 762                         set_opt(sbi, INLINE_XATTR);
 763                         break;
 764                 case Opt_noinline_xattr:
 765                         clear_opt(sbi, INLINE_XATTR);
 766                         break;
 767                 case Opt_inline_xattr_size:
 768                         if (args->from && match_int(args, &arg))
 769                                 return -EINVAL;
 770                         set_opt(sbi, INLINE_XATTR_SIZE);
 771                         F2FS_OPTION(sbi).inline_xattr_size = arg;
 772                         break;
 773 #else
 774                 case Opt_user_xattr:
 775                         f2fs_info(sbi, "user_xattr options not supported");
 776                         break;
 777                 case Opt_nouser_xattr:
 778                         f2fs_info(sbi, "nouser_xattr options not supported");
 779                         break;
 780                 case Opt_inline_xattr:
 781                         f2fs_info(sbi, "inline_xattr options not supported");
 782                         break;
 783                 case Opt_noinline_xattr:
 784                         f2fs_info(sbi, "noinline_xattr options not supported");
 785                         break;
 786 #endif
 787 #ifdef CONFIG_F2FS_FS_POSIX_ACL
 788                 case Opt_acl:
 789                         set_opt(sbi, POSIX_ACL);
 790                         break;
 791                 case Opt_noacl:
 792                         clear_opt(sbi, POSIX_ACL);
 793                         break;
 794 #else
 795                 case Opt_acl:
 796                         f2fs_info(sbi, "acl options not supported");
 797                         break;
 798                 case Opt_noacl:
 799                         f2fs_info(sbi, "noacl options not supported");
 800                         break;
 801 #endif
 802                 case Opt_active_logs:
 803                         if (args->from && match_int(args, &arg))
 804                                 return -EINVAL;
 805                         if (arg != 2 && arg != 4 &&
 806                                 arg != NR_CURSEG_PERSIST_TYPE)
 807                                 return -EINVAL;
 808                         F2FS_OPTION(sbi).active_logs = arg;
 809                         break;
 810                 case Opt_disable_ext_identify:
 811                         set_opt(sbi, DISABLE_EXT_IDENTIFY);
 812                         break;
 813                 case Opt_inline_data:
 814                         set_opt(sbi, INLINE_DATA);
 815                         break;
 816                 case Opt_inline_dentry:
 817                         set_opt(sbi, INLINE_DENTRY);
 818                         break;
 819                 case Opt_noinline_dentry:
 820                         clear_opt(sbi, INLINE_DENTRY);
 821                         break;
 822                 case Opt_flush_merge:
 823                         set_opt(sbi, FLUSH_MERGE);
 824                         break;
 825                 case Opt_noflush_merge:
 826                         clear_opt(sbi, FLUSH_MERGE);
 827                         break;
 828                 case Opt_nobarrier:
 829                         set_opt(sbi, NOBARRIER);
 830                         break;
 831                 case Opt_barrier:
 832                         clear_opt(sbi, NOBARRIER);
 833                         break;
 834                 case Opt_fastboot:
 835                         set_opt(sbi, FASTBOOT);
 836                         break;
 837                 case Opt_extent_cache:
 838                         set_opt(sbi, READ_EXTENT_CACHE);
 839                         break;
 840                 case Opt_noextent_cache:
 841                         if (F2FS_HAS_FEATURE(sbi, F2FS_FEATURE_DEVICE_ALIAS)) {
 842                                 f2fs_err(sbi, "device aliasing requires extent cache");
 843                                 return -EINVAL;
 844                         }
 845                         clear_opt(sbi, READ_EXTENT_CACHE);
 846                         break;
 847                 case Opt_noinline_data:
 848                         clear_opt(sbi, INLINE_DATA);
 849                         break;
 850                 case Opt_data_flush:
 851                         set_opt(sbi, DATA_FLUSH);
 852                         break;
 853                 case Opt_reserve_root:
 854                         if (args->from && match_int(args, &arg))
 855                                 return -EINVAL;
 856                         if (test_opt(sbi, RESERVE_ROOT)) {
 857                                 f2fs_info(sbi, "Preserve previous reserve_root=%u",
 858                                           F2FS_OPTION(sbi).root_reserved_blocks);
 859                         } else {
 860                                 F2FS_OPTION(sbi).root_reserved_blocks = arg;
 861                                 set_opt(sbi, RESERVE_ROOT);
 862                         }
 863                         break;
 864                 case Opt_resuid:
 865                         if (args->from && match_int(args, &arg))
 866                                 return -EINVAL;
 867                         uid = make_kuid(current_user_ns(), arg);
 868                         if (!uid_valid(uid)) {
 869                                 f2fs_err(sbi, "Invalid uid value %d", arg);
 870                                 return -EINVAL;
 871                         }
 872                         F2FS_OPTION(sbi).s_resuid = uid;
 873                         break;
 874                 case Opt_resgid:
 875                         if (args->from && match_int(args, &arg))
 876                                 return -EINVAL;
 877                         gid = make_kgid(current_user_ns(), arg);
 878                         if (!gid_valid(gid)) {
 879                                 f2fs_err(sbi, "Invalid gid value %d", arg);
 880                                 return -EINVAL;
 881                         }
 882                         F2FS_OPTION(sbi).s_resgid = gid;
 883                         break;
 884                 case Opt_mode:
 885                         name = match_strdup(&args[0]);
 886
 887                         if (!name)
 888                                 return -ENOMEM;
 889                         if (!strcmp(name, "adaptive")) {
 890                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
 891                         } else if (!strcmp(name, "lfs")) {
 892                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
 893                         } else if (!strcmp(name, "fragment:segment")) {
 894                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
 895                         } else if (!strcmp(name, "fragment:block")) {
 896                                 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
 897                         } else {
 898                                 kfree(name);
 899                                 return -EINVAL;
 900                         }
 901                         kfree(name);
 902                         break;
 903 #ifdef CONFIG_F2FS_FAULT_INJECTION
 904                 case Opt_fault_injection:
 905                         if (args->from && match_int(args, &arg))
 906                                 return -EINVAL;
 907                         if (f2fs_build_fault_attr(sbi, arg,
 908                                         F2FS_ALL_FAULT_TYPE))
 909                                 return -EINVAL;
 910                         set_opt(sbi, FAULT_INJECTION);
 911                         break;
 912
 913                 case Opt_fault_type:
 914                         if (args->from && match_int(args, &arg))
 915                                 return -EINVAL;
 916                         if (f2fs_build_fault_attr(sbi, 0, arg))
 917                                 return -EINVAL;
 918                         set_opt(sbi, FAULT_INJECTION);
 919                         break;
 920 #else
 921                 case Opt_fault_injection:
 922                         f2fs_info(sbi, "fault_injection options not supported");
 923                         break;
 924
 925                 case Opt_fault_type:
 926                         f2fs_info(sbi, "fault_type options not supported");
 927                         break;
 928 #endif
 929                 case Opt_lazytime:
 930                         sb->s_flags |= SB_LAZYTIME;
 931                         break;
 932                 case Opt_nolazytime:
 933                         sb->s_flags &= ~SB_LAZYTIME;
 934                         break;
 935 #ifdef CONFIG_QUOTA
 936                 case Opt_quota:
 937                 case Opt_usrquota:
 938                         set_opt(sbi, USRQUOTA);
 939                         break;
 940                 case Opt_grpquota:
 941                         set_opt(sbi, GRPQUOTA);
 942                         break;
 943                 case Opt_prjquota:
 944                         set_opt(sbi, PRJQUOTA);
 945                         break;
 946                 case Opt_usrjquota:
 947                         ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 948                         if (ret)
 949                                 return ret;
 950                         break;
 951                 case Opt_grpjquota:
 952                         ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 953                         if (ret)
 954                                 return ret;
 955                         break;
 956                 case Opt_prjjquota:
 957                         ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 958                         if (ret)
 959                                 return ret;
 960                         break;
 961                 case Opt_offusrjquota:
 962                         ret = f2fs_clear_qf_name(sb, USRQUOTA);
 963                         if (ret)
 964                                 return ret;
 965                         break;
 966                 case Opt_offgrpjquota:
 967                         ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 968                         if (ret)
 969                                 return ret;
 970                         break;
 971                 case Opt_offprjjquota:
 972                         ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 973                         if (ret)
 974                                 return ret;
 975                         break;
 976                 case Opt_jqfmt_vfsold:
 977                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 978                         break;
 979                 case Opt_jqfmt_vfsv0:
 980                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 981                         break;
 982                 case Opt_jqfmt_vfsv1:
 983                         F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 984                         break;
 985                 case Opt_noquota:
 986                         clear_opt(sbi, QUOTA);
 987                         clear_opt(sbi, USRQUOTA);
 988                         clear_opt(sbi, GRPQUOTA);
 989                         clear_opt(sbi, PRJQUOTA);
 990                         break;
 991 #else
 992                 case Opt_quota:
 993                 case Opt_usrquota:
 994                 case Opt_grpquota:
 995                 case Opt_prjquota:
 996                 case Opt_usrjquota:
 997                 case Opt_grpjquota:
 998                 case Opt_prjjquota:
 999                 case Opt_offusrjquota:
1000                 case Opt_offgrpjquota:
1001                 case Opt_offprjjquota:
1002                 case Opt_jqfmt_vfsold:
1003                 case Opt_jqfmt_vfsv0:
1004                 case Opt_jqfmt_vfsv1:
1005                 case Opt_noquota:
1006                         f2fs_info(sbi, "quota operations not supported");
1007                         break;
1008 #endif
1009                 case Opt_alloc:
1010                         name = match_strdup(&args[0]);
1011                         if (!name)
1012                                 return -ENOMEM;
1013
1014                         if (!strcmp(name, "default")) {
1015                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1016                         } else if (!strcmp(name, "reuse")) {
1017                                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
1018                         } else {
1019                                 kfree(name);
1020                                 return -EINVAL;
1021                         }
1022                         kfree(name);
1023                         break;
1024                 case Opt_fsync:
1025                         name = match_strdup(&args[0]);
1026                         if (!name)
1027                                 return -ENOMEM;
1028                         if (!strcmp(name, "posix")) {
1029                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1030                         } else if (!strcmp(name, "strict")) {
1031                                 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
1032                         } else if (!strcmp(name, "nobarrier")) {
1033                                 F2FS_OPTION(sbi).fsync_mode =
1034                                                         FSYNC_MODE_NOBARRIER;
1035                         } else {
1036                                 kfree(name);
1037                                 return -EINVAL;
1038                         }
1039                         kfree(name);
1040                         break;
1041                 case Opt_test_dummy_encryption:
1042                         ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
1043                                                              is_remount);
1044                         if (ret)
1045                                 return ret;
1046                         break;
1047                 case Opt_inlinecrypt:
1048 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
1049                         sb->s_flags |= SB_INLINECRYPT;
1050 #else
1051                         f2fs_info(sbi, "inline encryption not supported");
1052 #endif
1053                         break;
1054                 case Opt_checkpoint_disable_cap_perc:
1055                         if (args->from && match_int(args, &arg))
1056                                 return -EINVAL;
1057                         if (arg < 0 || arg > 100)
1058                                 return -EINVAL;
1059                         F2FS_OPTION(sbi).unusable_cap_perc = arg;
1060                         set_opt(sbi, DISABLE_CHECKPOINT);
1061                         break;
1062                 case Opt_checkpoint_disable_cap:
1063                         if (args->from && match_int(args, &arg))
1064                                 return -EINVAL;
1065                         F2FS_OPTION(sbi).unusable_cap = arg;
1066                         set_opt(sbi, DISABLE_CHECKPOINT);
1067                         break;
1068                 case Opt_checkpoint_disable:
1069                         set_opt(sbi, DISABLE_CHECKPOINT);
1070                         break;
1071                 case Opt_checkpoint_enable:
1072                         clear_opt(sbi, DISABLE_CHECKPOINT);
1073                         break;
1074                 case Opt_checkpoint_merge:
1075                         set_opt(sbi, MERGE_CHECKPOINT);
1076                         break;
1077                 case Opt_nocheckpoint_merge:
1078                         clear_opt(sbi, MERGE_CHECKPOINT);
1079                         break;
1080 #ifdef CONFIG_F2FS_FS_COMPRESSION
1081                 case Opt_compress_algorithm:
1082                         if (!f2fs_sb_has_compression(sbi)) {
1083                                 f2fs_info(sbi, "Image doesn't support compression");
1084                                 break;
1085                         }
1086                         name = match_strdup(&args[0]);
1087                         if (!name)
1088                                 return -ENOMEM;
1089                         if (!strcmp(name, "lzo")) {
1090 #ifdef CONFIG_F2FS_FS_LZO
1091                                 F2FS_OPTION(sbi).compress_level = 0;
1092                                 F2FS_OPTION(sbi).compress_algorithm =
1093                                                                 COMPRESS_LZO;
1094 #else
1095                                 f2fs_info(sbi, "kernel doesn't support lzo compression");
1096 #endif
1097                         } else if (!strncmp(name, "lz4", 3)) {
1098 #ifdef CONFIG_F2FS_FS_LZ4
1099                                 ret = f2fs_set_lz4hc_level(sbi, name);
1100                                 if (ret) {
1101                                         kfree(name);
1102                                         return -EINVAL;
1103                                 }
1104                                 F2FS_OPTION(sbi).compress_algorithm =
1105                                                                 COMPRESS_LZ4;
1106 #else
1107                                 f2fs_info(sbi, "kernel doesn't support lz4 compression");
1108 #endif
1109                         } else if (!strncmp(name, "zstd", 4)) {
1110 #ifdef CONFIG_F2FS_FS_ZSTD
1111                                 ret = f2fs_set_zstd_level(sbi, name);
1112                                 if (ret) {
1113                                         kfree(name);
1114                                         return -EINVAL;
1115                                 }
1116                                 F2FS_OPTION(sbi).compress_algorithm =
1117                                                                 COMPRESS_ZSTD;
1118 #else
1119                                 f2fs_info(sbi, "kernel doesn't support zstd compression");
1120 #endif
1121                         } else if (!strcmp(name, "lzo-rle")) {
1122 #ifdef CONFIG_F2FS_FS_LZORLE
1123                                 F2FS_OPTION(sbi).compress_level = 0;
1124                                 F2FS_OPTION(sbi).compress_algorithm =
1125                                                                 COMPRESS_LZORLE;
1126 #else
1127                                 f2fs_info(sbi, "kernel doesn't support lzorle compression");
1128 #endif
1129                         } else {
1130                                 kfree(name);
1131                                 return -EINVAL;
1132                         }
1133                         kfree(name);
1134                         break;
1135                 case Opt_compress_log_size:
1136                         if (!f2fs_sb_has_compression(sbi)) {
1137                                 f2fs_info(sbi, "Image doesn't support compression");
1138                                 break;
1139                         }
1140                         if (args->from && match_int(args, &arg))
1141                                 return -EINVAL;
1142                         if (arg < MIN_COMPRESS_LOG_SIZE ||
1143                                 arg > MAX_COMPRESS_LOG_SIZE) {
1144                                 f2fs_err(sbi,
1145                                         "Compress cluster log size is out of range");
1146                                 return -EINVAL;
1147                         }
1148                         F2FS_OPTION(sbi).compress_log_size = arg;
1149                         break;
1150                 case Opt_compress_extension:
1151                         if (!f2fs_sb_has_compression(sbi)) {
1152                                 f2fs_info(sbi, "Image doesn't support compression");
1153                                 break;
1154                         }
1155                         name = match_strdup(&args[0]);
1156                         if (!name)
1157                                 return -ENOMEM;
1158
1159                         ext = F2FS_OPTION(sbi).extensions;
1160                         ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
1161
1162                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
1163                                 ext_cnt >= COMPRESS_EXT_NUM) {
1164                                 f2fs_err(sbi,
1165                                         "invalid extension length/number");
1166                                 kfree(name);
1167                                 return -EINVAL;
1168                         }
1169
1170                         if (is_compress_extension_exist(sbi, name, true)) {
1171                                 kfree(name);
1172                                 break;
1173                         }
1174
1175                         ret = strscpy(ext[ext_cnt], name);
1176                         if (ret < 0) {
1177                                 kfree(name);
1178                                 return ret;
1179                         }
1180                         F2FS_OPTION(sbi).compress_ext_cnt++;
1181                         kfree(name);
1182                         break;
1183                 case Opt_nocompress_extension:
1184                         if (!f2fs_sb_has_compression(sbi)) {
1185                                 f2fs_info(sbi, "Image doesn't support compression");
1186                                 break;
1187                         }
1188                         name = match_strdup(&args[0]);
1189                         if (!name)
1190                                 return -ENOMEM;
1191
1192                         noext = F2FS_OPTION(sbi).noextensions;
1193                         noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt;
1194
1195                         if (strlen(name) >= F2FS_EXTENSION_LEN ||
1196                                 noext_cnt >= COMPRESS_EXT_NUM) {
1197                                 f2fs_err(sbi,
1198                                         "invalid extension length/number");
1199                                 kfree(name);
1200                                 return -EINVAL;
1201                         }
1202
1203                         if (is_compress_extension_exist(sbi, name, false)) {
1204                                 kfree(name);
1205                                 break;
1206                         }
1207
1208                         ret = strscpy(noext[noext_cnt], name);
1209                         if (ret < 0) {
1210                                 kfree(name);
1211                                 return ret;
1212                         }
1213                         F2FS_OPTION(sbi).nocompress_ext_cnt++;
1214                         kfree(name);
1215                         break;
1216                 case Opt_compress_chksum:
1217                         if (!f2fs_sb_has_compression(sbi)) {
1218                                 f2fs_info(sbi, "Image doesn't support compression");
1219                                 break;
1220                         }
1221                         F2FS_OPTION(sbi).compress_chksum = true;
1222                         break;
1223                 case Opt_compress_mode:
1224                         if (!f2fs_sb_has_compression(sbi)) {
1225                                 f2fs_info(sbi, "Image doesn't support compression");
1226                                 break;
1227                         }
1228                         name = match_strdup(&args[0]);
1229                         if (!name)
1230                                 return -ENOMEM;
1231                         if (!strcmp(name, "fs")) {
1232                                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
1233                         } else if (!strcmp(name, "user")) {
1234                                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER;
1235                         } else {
1236                                 kfree(name);
1237                                 return -EINVAL;
1238                         }
1239                         kfree(name);
1240                         break;
1241                 case Opt_compress_cache:
1242                         if (!f2fs_sb_has_compression(sbi)) {
1243                                 f2fs_info(sbi, "Image doesn't support compression");
1244                                 break;
1245                         }
1246                         set_opt(sbi, COMPRESS_CACHE);
1247                         break;
1248 #else
1249                 case Opt_compress_algorithm:
1250                 case Opt_compress_log_size:
1251                 case Opt_compress_extension:
1252                 case Opt_nocompress_extension:
1253                 case Opt_compress_chksum:
1254                 case Opt_compress_mode:
1255                 case Opt_compress_cache:
1256                         f2fs_info(sbi, "compression options not supported");
1257                         break;
1258 #endif
1259                 case Opt_atgc:
1260                         set_opt(sbi, ATGC);
1261                         break;
1262                 case Opt_gc_merge:
1263                         set_opt(sbi, GC_MERGE);
1264                         break;
1265                 case Opt_nogc_merge:
1266                         clear_opt(sbi, GC_MERGE);
1267                         break;
1268                 case Opt_discard_unit:
1269                         name = match_strdup(&args[0]);
1270                         if (!name)
1271                                 return -ENOMEM;
1272                         if (!strcmp(name, "block")) {
1273                                 F2FS_OPTION(sbi).discard_unit =
1274                                                 DISCARD_UNIT_BLOCK;
1275                         } else if (!strcmp(name, "segment")) {
1276                                 F2FS_OPTION(sbi).discard_unit =
1277                                                 DISCARD_UNIT_SEGMENT;
1278                         } else if (!strcmp(name, "section")) {
1279                                 F2FS_OPTION(sbi).discard_unit =
1280                                                 DISCARD_UNIT_SECTION;
1281                         } else {
1282                                 kfree(name);
1283                                 return -EINVAL;
1284                         }
1285                         kfree(name);
1286                         break;
1287                 case Opt_memory_mode:
1288                         name = match_strdup(&args[0]);
1289                         if (!name)
1290                                 return -ENOMEM;
1291                         if (!strcmp(name, "normal")) {
1292                                 F2FS_OPTION(sbi).memory_mode =
1293                                                 MEMORY_MODE_NORMAL;
1294                         } else if (!strcmp(name, "low")) {
1295                                 F2FS_OPTION(sbi).memory_mode =
1296                                                 MEMORY_MODE_LOW;
1297                         } else {
1298                                 kfree(name);
1299                                 return -EINVAL;
1300                         }
1301                         kfree(name);
1302                         break;
1303                 case Opt_age_extent_cache:
1304                         set_opt(sbi, AGE_EXTENT_CACHE);
1305                         break;
1306                 case Opt_errors:
1307                         name = match_strdup(&args[0]);
1308                         if (!name)
1309                                 return -ENOMEM;
1310                         if (!strcmp(name, "remount-ro")) {
1311                                 F2FS_OPTION(sbi).errors =
1312                                                 MOUNT_ERRORS_READONLY;
1313                         } else if (!strcmp(name, "continue")) {
1314                                 F2FS_OPTION(sbi).errors =
1315                                                 MOUNT_ERRORS_CONTINUE;
1316                         } else if (!strcmp(name, "panic")) {
1317                                 F2FS_OPTION(sbi).errors =
1318                                                 MOUNT_ERRORS_PANIC;
1319                         } else {
1320                                 kfree(name);
1321                                 return -EINVAL;
1322                         }
1323                         kfree(name);
1324                         break;
1325                 default:
1326                         f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1327                                  p);
1328                         return -EINVAL;
1329                 }
1330         }
1331 default_check:
1332 #ifdef CONFIG_QUOTA
1333         if (f2fs_check_quota_options(sbi))
1334                 return -EINVAL;
1335 #else
1336         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1337                 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1338                 return -EINVAL;
1339         }
1340         if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1341                 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1342                 return -EINVAL;
1343         }
1344 #endif
1345
1346         if (!IS_ENABLED(CONFIG_UNICODE) && f2fs_sb_has_casefold(sbi)) {
1347                 f2fs_err(sbi,
1348                         "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1349                 return -EINVAL;
1350         }
1351
1352         /*
1353          * The BLKZONED feature indicates that the drive was formatted with
1354          * zone alignment optimization. This is optional for host-aware
1355          * devices, but mandatory for host-managed zoned block devices.
1356          */
1357         if (f2fs_sb_has_blkzoned(sbi)) {
1358 #ifdef CONFIG_BLK_DEV_ZONED
1359                 if (F2FS_OPTION(sbi).discard_unit !=
1360                                                 DISCARD_UNIT_SECTION) {
1361                         f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default");
1362                         F2FS_OPTION(sbi).discard_unit =
1363                                         DISCARD_UNIT_SECTION;
1364                 }
1365
1366                 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) {
1367                         f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature");
1368                         return -EINVAL;
1369                 }
1370 #else
1371                 f2fs_err(sbi, "Zoned block device support is not enabled");
1372                 return -EINVAL;
1373 #endif
1374         }
1375
1376 #ifdef CONFIG_F2FS_FS_COMPRESSION
1377         if (f2fs_test_compress_extension(sbi)) {
1378                 f2fs_err(sbi, "invalid compress or nocompress extension");
1379                 return -EINVAL;
1380         }
1381 #endif
1382
1383         if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1384                 int min_size, max_size;
1385
1386                 if (!f2fs_sb_has_extra_attr(sbi) ||
1387                         !f2fs_sb_has_flexible_inline_xattr(sbi)) {
1388                         f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1389                         return -EINVAL;
1390                 }
1391                 if (!test_opt(sbi, INLINE_XATTR)) {
1392                         f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1393                         return -EINVAL;
1394                 }
1395
1396                 min_size = MIN_INLINE_XATTR_SIZE;
1397                 max_size = MAX_INLINE_XATTR_SIZE;
1398
1399                 if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1400                                 F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1401                         f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1402                                  min_size, max_size);
1403                         return -EINVAL;
1404                 }
1405         }
1406
1407         if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) {
1408                 f2fs_err(sbi, "LFS is not compatible with ATGC");
1409                 return -EINVAL;
1410         }
1411
1412         if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) {
1413                 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode");
1414                 return -EINVAL;
1415         }
1416
1417         if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) {
1418                 f2fs_err(sbi, "Allow to mount readonly mode only");
1419                 return -EROFS;
1420         }
1421         return 0;
1422 }
1423
1424 static struct inode *f2fs_alloc_inode(struct super_block *sb)
1425 {
1426         struct f2fs_inode_info *fi;
1427
1428         if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC))
1429                 return NULL;
1430
1431         fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO);
1432         if (!fi)
1433                 return NULL;
1434
1435         init_once((void *) fi);
1436
1437         /* Initialize f2fs-specific inode info */
1438         atomic_set(&fi->dirty_pages, 0);
1439         atomic_set(&fi->i_compr_blocks, 0);
1440         init_f2fs_rwsem(&fi->i_sem);
1441         spin_lock_init(&fi->i_size_lock);
1442         INIT_LIST_HEAD(&fi->dirty_list);
1443         INIT_LIST_HEAD(&fi->gdirty_list);
1444         init_f2fs_rwsem(&fi->i_gc_rwsem[READ]);
1445         init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]);
1446         init_f2fs_rwsem(&fi->i_xattr_sem);
1447
1448         /* Will be used by directory only */
1449         fi->i_dir_level = F2FS_SB(sb)->dir_level;
1450
1451         return &fi->vfs_inode;
1452 }
1453
1454 static int f2fs_drop_inode(struct inode *inode)
1455 {
1456         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1457         int ret;
1458
1459         /*
1460          * during filesystem shutdown, if checkpoint is disabled,
1461          * drop useless meta/node dirty pages.
1462          */
1463         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1464                 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1465                         inode->i_ino == F2FS_META_INO(sbi)) {
1466                         trace_f2fs_drop_inode(inode, 1);
1467                         return 1;
1468                 }
1469         }
1470
1471         /*
1472          * This is to avoid a deadlock condition like below.
1473          * writeback_single_inode(inode)
1474          *  - f2fs_write_data_page
1475          *    - f2fs_gc -> iput -> evict
1476          *       - inode_wait_for_writeback(inode)
1477          */
1478         if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1479                 if (!inode->i_nlink && !is_bad_inode(inode)) {
1480                         /* to avoid evict_inode call simultaneously */
1481                         atomic_inc(&inode->i_count);
1482                         spin_unlock(&inode->i_lock);
1483
1484                         /* should remain fi->extent_tree for writepage */
1485                         f2fs_destroy_extent_node(inode);
1486
1487                         sb_start_intwrite(inode->i_sb);
1488                         f2fs_i_size_write(inode, 0);
1489
1490                         f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1491                                         inode, NULL, 0, DATA);
1492                         truncate_inode_pages_final(inode->i_mapping);
1493
1494                         if (F2FS_HAS_BLOCKS(inode))
1495                                 f2fs_truncate(inode);
1496
1497                         sb_end_intwrite(inode->i_sb);
1498
1499                         spin_lock(&inode->i_lock);
1500                         atomic_dec(&inode->i_count);
1501                 }
1502                 trace_f2fs_drop_inode(inode, 0);
1503                 return 0;
1504         }
1505         ret = generic_drop_inode(inode);
1506         if (!ret)
1507                 ret = fscrypt_drop_inode(inode);
1508         trace_f2fs_drop_inode(inode, ret);
1509         return ret;
1510 }
1511
1512 int f2fs_inode_dirtied(struct inode *inode, bool sync)
1513 {
1514         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1515         int ret = 0;
1516
1517         spin_lock(&sbi->inode_lock[DIRTY_META]);
1518         if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1519                 ret = 1;
1520         } else {
1521                 set_inode_flag(inode, FI_DIRTY_INODE);
1522                 stat_inc_dirty_inode(sbi, DIRTY_META);
1523         }
1524         if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1525                 list_add_tail(&F2FS_I(inode)->gdirty_list,
1526                                 &sbi->inode_list[DIRTY_META]);
1527                 inc_page_count(sbi, F2FS_DIRTY_IMETA);
1528         }
1529         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1530         return ret;
1531 }
1532
1533 void f2fs_inode_synced(struct inode *inode)
1534 {
1535         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1536
1537         spin_lock(&sbi->inode_lock[DIRTY_META]);
1538         if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1539                 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1540                 return;
1541         }
1542         if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1543                 list_del_init(&F2FS_I(inode)->gdirty_list);
1544                 dec_page_count(sbi, F2FS_DIRTY_IMETA);
1545         }
1546         clear_inode_flag(inode, FI_DIRTY_INODE);
1547         clear_inode_flag(inode, FI_AUTO_RECOVER);
1548         stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1549         spin_unlock(&sbi->inode_lock[DIRTY_META]);
1550 }
1551
1552 /*
1553  * f2fs_dirty_inode() is called from __mark_inode_dirty()
1554  *
1555  * We should call set_dirty_inode to write the dirty inode through write_inode.
1556  */
1557 static void f2fs_dirty_inode(struct inode *inode, int flags)
1558 {
1559         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1560
1561         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1562                         inode->i_ino == F2FS_META_INO(sbi))
1563                 return;
1564
1565         if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1566                 clear_inode_flag(inode, FI_AUTO_RECOVER);
1567
1568         f2fs_inode_dirtied(inode, false);
1569 }
1570
1571 static void f2fs_free_inode(struct inode *inode)
1572 {
1573         fscrypt_free_inode(inode);
1574         kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1575 }
1576
1577 static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1578 {
1579         percpu_counter_destroy(&sbi->total_valid_inode_count);
1580         percpu_counter_destroy(&sbi->rf_node_block_count);
1581         percpu_counter_destroy(&sbi->alloc_valid_block_count);
1582 }
1583
1584 static void destroy_device_list(struct f2fs_sb_info *sbi)
1585 {
1586         int i;
1587
1588         for (i = 0; i < sbi->s_ndevs; i++) {
1589                 if (i > 0)
1590                         bdev_fput(FDEV(i).bdev_file);
1591 #ifdef CONFIG_BLK_DEV_ZONED
1592                 kvfree(FDEV(i).blkz_seq);
1593 #endif
1594         }
1595         kvfree(sbi->devs);
1596 }
1597
1598 static void f2fs_put_super(struct super_block *sb)
1599 {
1600         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1601         int i;
1602         int err = 0;
1603         bool done;
1604
1605         /* unregister procfs/sysfs entries in advance to avoid race case */
1606         f2fs_unregister_sysfs(sbi);
1607
1608         f2fs_quota_off_umount(sb);
1609
1610         /* prevent remaining shrinker jobs */
1611         mutex_lock(&sbi->umount_mutex);
1612
1613         /*
1614          * flush all issued checkpoints and stop checkpoint issue thread.
1615          * after then, all checkpoints should be done by each process context.
1616          */
1617         f2fs_stop_ckpt_thread(sbi);
1618
1619         /*
1620          * We don't need to do checkpoint when superblock is clean.
1621          * But, the previous checkpoint was not done by umount, it needs to do
1622          * clean checkpoint again.
1623          */
1624         if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1625                         !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1626                 struct cp_control cpc = {
1627                         .reason = CP_UMOUNT,
1628                 };
1629                 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1630                 err = f2fs_write_checkpoint(sbi, &cpc);
1631         }
1632
1633         /* be sure to wait for any on-going discard commands */
1634         done = f2fs_issue_discard_timeout(sbi);
1635         if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) {
1636                 struct cp_control cpc = {
1637                         .reason = CP_UMOUNT | CP_TRIMMED,
1638                 };
1639                 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1640                 err = f2fs_write_checkpoint(sbi, &cpc);
1641         }
1642
1643         /*
1644          * normally superblock is clean, so we need to release this.
1645          * In addition, EIO will skip do checkpoint, we need this as well.
1646          */
1647         f2fs_release_ino_entry(sbi, true);
1648
1649         f2fs_leave_shrinker(sbi);
1650         mutex_unlock(&sbi->umount_mutex);
1651
1652         /* our cp_error case, we can wait for any writeback page */
1653         f2fs_flush_merged_writes(sbi);
1654
1655         f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1656
1657         if (err || f2fs_cp_error(sbi)) {
1658                 truncate_inode_pages_final(NODE_MAPPING(sbi));
1659                 truncate_inode_pages_final(META_MAPPING(sbi));
1660         }
1661
1662         for (i = 0; i < NR_COUNT_TYPE; i++) {
1663                 if (!get_pages(sbi, i))
1664                         continue;
1665                 f2fs_err(sbi, "detect filesystem reference count leak during "
1666                         "umount, type: %d, count: %lld", i, get_pages(sbi, i));
1667                 f2fs_bug_on(sbi, 1);
1668         }
1669
1670         f2fs_bug_on(sbi, sbi->fsync_node_num);
1671
1672         f2fs_destroy_compress_inode(sbi);
1673
1674         iput(sbi->node_inode);
1675         sbi->node_inode = NULL;
1676
1677         iput(sbi->meta_inode);
1678         sbi->meta_inode = NULL;
1679
1680         /*
1681          * iput() can update stat information, if f2fs_write_checkpoint()
1682          * above failed with error.
1683          */
1684         f2fs_destroy_stats(sbi);
1685
1686         /* destroy f2fs internal modules */
1687         f2fs_destroy_node_manager(sbi);
1688         f2fs_destroy_segment_manager(sbi);
1689
1690         /* flush s_error_work before sbi destroy */
1691         flush_work(&sbi->s_error_work);
1692
1693         f2fs_destroy_post_read_wq(sbi);
1694
1695         kvfree(sbi->ckpt);
1696
1697         kfree(sbi->raw_super);
1698
1699         f2fs_destroy_page_array_cache(sbi);
1700         f2fs_destroy_xattr_caches(sbi);
1701 #ifdef CONFIG_QUOTA
1702         for (i = 0; i < MAXQUOTAS; i++)
1703                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1704 #endif
1705         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1706         destroy_percpu_info(sbi);
1707         f2fs_destroy_iostat(sbi);
1708         for (i = 0; i < NR_PAGE_TYPE; i++)
1709                 kvfree(sbi->write_io[i]);
1710 #if IS_ENABLED(CONFIG_UNICODE)
1711         utf8_unload(sb->s_encoding);
1712 #endif
1713 }
1714
1715 int f2fs_sync_fs(struct super_block *sb, int sync)
1716 {
1717         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1718         int err = 0;
1719
1720         if (unlikely(f2fs_cp_error(sbi)))
1721                 return 0;
1722         if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1723                 return 0;
1724
1725         trace_f2fs_sync_fs(sb, sync);
1726
1727         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1728                 return -EAGAIN;
1729
1730         if (sync) {
1731                 stat_inc_cp_call_count(sbi, TOTAL_CALL);
1732                 err = f2fs_issue_checkpoint(sbi);
1733         }
1734
1735         return err;
1736 }
1737
1738 static int f2fs_freeze(struct super_block *sb)
1739 {
1740         if (f2fs_readonly(sb))
1741                 return 0;
1742
1743         /* IO error happened before */
1744         if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1745                 return -EIO;
1746
1747         /* must be clean, since sync_filesystem() was already called */
1748         if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1749                 return -EINVAL;
1750
1751         /* Let's flush checkpoints and stop the thread. */
1752         f2fs_flush_ckpt_thread(F2FS_SB(sb));
1753
1754         /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */
1755         set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1756         return 0;
1757 }
1758
1759 static int f2fs_unfreeze(struct super_block *sb)
1760 {
1761         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1762
1763         /*
1764          * It will update discard_max_bytes of mounted lvm device to zero
1765          * after creating snapshot on this lvm device, let's drop all
1766          * remained discards.
1767          * We don't need to disable real-time discard because discard_max_bytes
1768          * will recover after removal of snapshot.
1769          */
1770         if (test_opt(sbi, DISCARD) && !f2fs_hw_support_discard(sbi))
1771                 f2fs_issue_discard_timeout(sbi);
1772
1773         clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING);
1774         return 0;
1775 }
1776
1777 #ifdef CONFIG_QUOTA
1778 static int f2fs_statfs_project(struct super_block *sb,
1779                                 kprojid_t projid, struct kstatfs *buf)
1780 {
1781         struct kqid qid;
1782         struct dquot *dquot;
1783         u64 limit;
1784         u64 curblock;
1785
1786         qid = make_kqid_projid(projid);
1787         dquot = dqget(sb, qid);
1788         if (IS_ERR(dquot))
1789                 return PTR_ERR(dquot);
1790         spin_lock(&dquot->dq_dqb_lock);
1791
1792         limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1793                                         dquot->dq_dqb.dqb_bhardlimit);
1794         if (limit)
1795                 limit >>= sb->s_blocksize_bits;
1796
1797         if (limit && buf->f_blocks > limit) {
1798                 curblock = (dquot->dq_dqb.dqb_curspace +
1799                             dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1800                 buf->f_blocks = limit;
1801                 buf->f_bfree = buf->f_bavail =
1802                         (buf->f_blocks > curblock) ?
1803                          (buf->f_blocks - curblock) : 0;
1804         }
1805
1806         limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1807                                         dquot->dq_dqb.dqb_ihardlimit);
1808
1809         if (limit && buf->f_files > limit) {
1810                 buf->f_files = limit;
1811                 buf->f_ffree =
1812                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1813                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1814         }
1815
1816         spin_unlock(&dquot->dq_dqb_lock);
1817         dqput(dquot);
1818         return 0;
1819 }
1820 #endif
1821
1822 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1823 {
1824         struct super_block *sb = dentry->d_sb;
1825         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1826         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1827         block_t total_count, user_block_count, start_count;
1828         u64 avail_node_count;
1829         unsigned int total_valid_node_count;
1830
1831         total_count = le64_to_cpu(sbi->raw_super->block_count);
1832         start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1833         buf->f_type = F2FS_SUPER_MAGIC;
1834         buf->f_bsize = sbi->blocksize;
1835
1836         buf->f_blocks = total_count - start_count;
1837
1838         spin_lock(&sbi->stat_lock);
1839
1840         user_block_count = sbi->user_block_count;
1841         total_valid_node_count = valid_node_count(sbi);
1842         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1843         buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1844                                                 sbi->current_reserved_blocks;
1845
1846         if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1847                 buf->f_bfree = 0;
1848         else
1849                 buf->f_bfree -= sbi->unusable_block_count;
1850         spin_unlock(&sbi->stat_lock);
1851
1852         if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1853                 buf->f_bavail = buf->f_bfree -
1854                                 F2FS_OPTION(sbi).root_reserved_blocks;
1855         else
1856                 buf->f_bavail = 0;
1857
1858         if (avail_node_count > user_block_count) {
1859                 buf->f_files = user_block_count;
1860                 buf->f_ffree = buf->f_bavail;
1861         } else {
1862                 buf->f_files = avail_node_count;
1863                 buf->f_ffree = min(avail_node_count - total_valid_node_count,
1864                                         buf->f_bavail);
1865         }
1866
1867         buf->f_namelen = F2FS_NAME_LEN;
1868         buf->f_fsid    = u64_to_fsid(id);
1869
1870 #ifdef CONFIG_QUOTA
1871         if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1872                         sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1873                 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1874         }
1875 #endif
1876         return 0;
1877 }
1878
1879 static inline void f2fs_show_quota_options(struct seq_file *seq,
1880                                            struct super_block *sb)
1881 {
1882 #ifdef CONFIG_QUOTA
1883         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1884
1885         if (F2FS_OPTION(sbi).s_jquota_fmt) {
1886                 char *fmtname = "";
1887
1888                 switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1889                 case QFMT_VFS_OLD:
1890                         fmtname = "vfsold";
1891                         break;
1892                 case QFMT_VFS_V0:
1893                         fmtname = "vfsv0";
1894                         break;
1895                 case QFMT_VFS_V1:
1896                         fmtname = "vfsv1";
1897                         break;
1898                 }
1899                 seq_printf(seq, ",jqfmt=%s", fmtname);
1900         }
1901
1902         if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1903                 seq_show_option(seq, "usrjquota",
1904                         F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1905
1906         if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1907                 seq_show_option(seq, "grpjquota",
1908                         F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1909
1910         if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1911                 seq_show_option(seq, "prjjquota",
1912                         F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1913 #endif
1914 }
1915
1916 #ifdef CONFIG_F2FS_FS_COMPRESSION
1917 static inline void f2fs_show_compress_options(struct seq_file *seq,
1918                                                         struct super_block *sb)
1919 {
1920         struct f2fs_sb_info *sbi = F2FS_SB(sb);
1921         char *algtype = "";
1922         int i;
1923
1924         if (!f2fs_sb_has_compression(sbi))
1925                 return;
1926
1927         switch (F2FS_OPTION(sbi).compress_algorithm) {
1928         case COMPRESS_LZO:
1929                 algtype = "lzo";
1930                 break;
1931         case COMPRESS_LZ4:
1932                 algtype = "lz4";
1933                 break;
1934         case COMPRESS_ZSTD:
1935                 algtype = "zstd";
1936                 break;
1937         case COMPRESS_LZORLE:
1938                 algtype = "lzo-rle";
1939                 break;
1940         }
1941         seq_printf(seq, ",compress_algorithm=%s", algtype);
1942
1943         if (F2FS_OPTION(sbi).compress_level)
1944                 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level);
1945
1946         seq_printf(seq, ",compress_log_size=%u",
1947                         F2FS_OPTION(sbi).compress_log_size);
1948
1949         for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1950                 seq_printf(seq, ",compress_extension=%s",
1951                         F2FS_OPTION(sbi).extensions[i]);
1952         }
1953
1954         for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) {
1955                 seq_printf(seq, ",nocompress_extension=%s",
1956                         F2FS_OPTION(sbi).noextensions[i]);
1957         }
1958
1959         if (F2FS_OPTION(sbi).compress_chksum)
1960                 seq_puts(seq, ",compress_chksum");
1961
1962         if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS)
1963                 seq_printf(seq, ",compress_mode=%s", "fs");
1964         else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER)
1965                 seq_printf(seq, ",compress_mode=%s", "user");
1966
1967         if (test_opt(sbi, COMPRESS_CACHE))
1968                 seq_puts(seq, ",compress_cache");
1969 }
1970 #endif
1971
1972 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1973 {
1974         struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1975
1976         if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1977                 seq_printf(seq, ",background_gc=%s", "sync");
1978         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1979                 seq_printf(seq, ",background_gc=%s", "on");
1980         else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1981                 seq_printf(seq, ",background_gc=%s", "off");
1982
1983         if (test_opt(sbi, GC_MERGE))
1984                 seq_puts(seq, ",gc_merge");
1985         else
1986                 seq_puts(seq, ",nogc_merge");
1987
1988         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1989                 seq_puts(seq, ",disable_roll_forward");
1990         if (test_opt(sbi, NORECOVERY))
1991                 seq_puts(seq, ",norecovery");
1992         if (test_opt(sbi, DISCARD)) {
1993                 seq_puts(seq, ",discard");
1994                 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK)
1995                         seq_printf(seq, ",discard_unit=%s", "block");
1996                 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT)
1997                         seq_printf(seq, ",discard_unit=%s", "segment");
1998                 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION)
1999                         seq_printf(seq, ",discard_unit=%s", "section");
2000         } else {
2001                 seq_puts(seq, ",nodiscard");
2002         }
2003 #ifdef CONFIG_F2FS_FS_XATTR
2004         if (test_opt(sbi, XATTR_USER))
2005                 seq_puts(seq, ",user_xattr");
2006         else
2007                 seq_puts(seq, ",nouser_xattr");
2008         if (test_opt(sbi, INLINE_XATTR))
2009                 seq_puts(seq, ",inline_xattr");
2010         else
2011                 seq_puts(seq, ",noinline_xattr");
2012         if (test_opt(sbi, INLINE_XATTR_SIZE))
2013                 seq_printf(seq, ",inline_xattr_size=%u",
2014                                         F2FS_OPTION(sbi).inline_xattr_size);
2015 #endif
2016 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2017         if (test_opt(sbi, POSIX_ACL))
2018                 seq_puts(seq, ",acl");
2019         else
2020                 seq_puts(seq, ",noacl");
2021 #endif
2022         if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
2023                 seq_puts(seq, ",disable_ext_identify");
2024         if (test_opt(sbi, INLINE_DATA))
2025                 seq_puts(seq, ",inline_data");
2026         else
2027                 seq_puts(seq, ",noinline_data");
2028         if (test_opt(sbi, INLINE_DENTRY))
2029                 seq_puts(seq, ",inline_dentry");
2030         else
2031                 seq_puts(seq, ",noinline_dentry");
2032         if (test_opt(sbi, FLUSH_MERGE))
2033                 seq_puts(seq, ",flush_merge");
2034         else
2035                 seq_puts(seq, ",noflush_merge");
2036         if (test_opt(sbi, NOBARRIER))
2037                 seq_puts(seq, ",nobarrier");
2038         else
2039                 seq_puts(seq, ",barrier");
2040         if (test_opt(sbi, FASTBOOT))
2041                 seq_puts(seq, ",fastboot");
2042         if (test_opt(sbi, READ_EXTENT_CACHE))
2043                 seq_puts(seq, ",extent_cache");
2044         else
2045                 seq_puts(seq, ",noextent_cache");
2046         if (test_opt(sbi, AGE_EXTENT_CACHE))
2047                 seq_puts(seq, ",age_extent_cache");
2048         if (test_opt(sbi, DATA_FLUSH))
2049                 seq_puts(seq, ",data_flush");
2050
2051         seq_puts(seq, ",mode=");
2052         if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
2053                 seq_puts(seq, "adaptive");
2054         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
2055                 seq_puts(seq, "lfs");
2056         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
2057                 seq_puts(seq, "fragment:segment");
2058         else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
2059                 seq_puts(seq, "fragment:block");
2060         seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
2061         if (test_opt(sbi, RESERVE_ROOT))
2062                 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
2063                                 F2FS_OPTION(sbi).root_reserved_blocks,
2064                                 from_kuid_munged(&init_user_ns,
2065                                         F2FS_OPTION(sbi).s_resuid),
2066                                 from_kgid_munged(&init_user_ns,
2067                                         F2FS_OPTION(sbi).s_resgid));
2068 #ifdef CONFIG_F2FS_FAULT_INJECTION
2069         if (test_opt(sbi, FAULT_INJECTION)) {
2070                 seq_printf(seq, ",fault_injection=%u",
2071                                 F2FS_OPTION(sbi).fault_info.inject_rate);
2072                 seq_printf(seq, ",fault_type=%u",
2073                                 F2FS_OPTION(sbi).fault_info.inject_type);
2074         }
2075 #endif
2076 #ifdef CONFIG_QUOTA
2077         if (test_opt(sbi, QUOTA))
2078                 seq_puts(seq, ",quota");
2079         if (test_opt(sbi, USRQUOTA))
2080                 seq_puts(seq, ",usrquota");
2081         if (test_opt(sbi, GRPQUOTA))
2082                 seq_puts(seq, ",grpquota");
2083         if (test_opt(sbi, PRJQUOTA))
2084                 seq_puts(seq, ",prjquota");
2085 #endif
2086         f2fs_show_quota_options(seq, sbi->sb);
2087
2088         fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
2089
2090         if (sbi->sb->s_flags & SB_INLINECRYPT)
2091                 seq_puts(seq, ",inlinecrypt");
2092
2093         if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
2094                 seq_printf(seq, ",alloc_mode=%s", "default");
2095         else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
2096                 seq_printf(seq, ",alloc_mode=%s", "reuse");
2097
2098         if (test_opt(sbi, DISABLE_CHECKPOINT))
2099                 seq_printf(seq, ",checkpoint=disable:%u",
2100                                 F2FS_OPTION(sbi).unusable_cap);
2101         if (test_opt(sbi, MERGE_CHECKPOINT))
2102                 seq_puts(seq, ",checkpoint_merge");
2103         else
2104                 seq_puts(seq, ",nocheckpoint_merge");
2105         if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
2106                 seq_printf(seq, ",fsync_mode=%s", "posix");
2107         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
2108                 seq_printf(seq, ",fsync_mode=%s", "strict");
2109         else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
2110                 seq_printf(seq, ",fsync_mode=%s", "nobarrier");
2111
2112 #ifdef CONFIG_F2FS_FS_COMPRESSION
2113         f2fs_show_compress_options(seq, sbi->sb);
2114 #endif
2115
2116         if (test_opt(sbi, ATGC))
2117                 seq_puts(seq, ",atgc");
2118
2119         if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL)
2120                 seq_printf(seq, ",memory=%s", "normal");
2121         else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW)
2122                 seq_printf(seq, ",memory=%s", "low");
2123
2124         if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
2125                 seq_printf(seq, ",errors=%s", "remount-ro");
2126         else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE)
2127                 seq_printf(seq, ",errors=%s", "continue");
2128         else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC)
2129                 seq_printf(seq, ",errors=%s", "panic");
2130
2131         return 0;
2132 }
2133
2134 static void default_options(struct f2fs_sb_info *sbi, bool remount)
2135 {
2136         /* init some FS parameters */
2137         if (!remount) {
2138                 set_opt(sbi, READ_EXTENT_CACHE);
2139                 clear_opt(sbi, DISABLE_CHECKPOINT);
2140
2141                 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi))
2142                         set_opt(sbi, DISCARD);
2143
2144                 if (f2fs_sb_has_blkzoned(sbi))
2145                         F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION;
2146                 else
2147                         F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK;
2148         }
2149
2150         if (f2fs_sb_has_readonly(sbi))
2151                 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE;
2152         else
2153                 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
2154
2155         F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
2156         if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <=
2157                                                         SMALL_VOLUME_SEGMENTS)
2158                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
2159         else
2160                 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
2161         F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
2162         F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
2163         F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
2164         if (f2fs_sb_has_compression(sbi)) {
2165                 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
2166                 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
2167                 F2FS_OPTION(sbi).compress_ext_cnt = 0;
2168                 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS;
2169         }
2170         F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
2171         F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL;
2172         F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE;
2173
2174         set_opt(sbi, INLINE_XATTR);
2175         set_opt(sbi, INLINE_DATA);
2176         set_opt(sbi, INLINE_DENTRY);
2177         set_opt(sbi, MERGE_CHECKPOINT);
2178         F2FS_OPTION(sbi).unusable_cap = 0;
2179         sbi->sb->s_flags |= SB_LAZYTIME;
2180         if (!f2fs_is_readonly(sbi))
2181                 set_opt(sbi, FLUSH_MERGE);
2182         if (f2fs_sb_has_blkzoned(sbi))
2183                 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
2184         else
2185                 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
2186
2187 #ifdef CONFIG_F2FS_FS_XATTR
2188         set_opt(sbi, XATTR_USER);
2189 #endif
2190 #ifdef CONFIG_F2FS_FS_POSIX_ACL
2191         set_opt(sbi, POSIX_ACL);
2192 #endif
2193
2194         f2fs_build_fault_attr(sbi, 0, 0);
2195 }
2196
2197 #ifdef CONFIG_QUOTA
2198 static int f2fs_enable_quotas(struct super_block *sb);
2199 #endif
2200
2201 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
2202 {
2203         unsigned int s_flags = sbi->sb->s_flags;
2204         struct cp_control cpc;
2205         unsigned int gc_mode = sbi->gc_mode;
2206         int err = 0;
2207         int ret;
2208         block_t unusable;
2209
2210         if (s_flags & SB_RDONLY) {
2211                 f2fs_err(sbi, "checkpoint=disable on readonly fs");
2212                 return -EINVAL;
2213         }
2214         sbi->sb->s_flags |= SB_ACTIVE;
2215
2216         /* check if we need more GC first */
2217         unusable = f2fs_get_unusable_blocks(sbi);
2218         if (!f2fs_disable_cp_again(sbi, unusable))
2219                 goto skip_gc;
2220
2221         f2fs_update_time(sbi, DISABLE_TIME);
2222
2223         sbi->gc_mode = GC_URGENT_HIGH;
2224
2225         while (!f2fs_time_over(sbi, DISABLE_TIME)) {
2226                 struct f2fs_gc_control gc_control = {
2227                         .victim_segno = NULL_SEGNO,
2228                         .init_gc_type = FG_GC,
2229                         .should_migrate_blocks = false,
2230                         .err_gc_skipped = true,
2231                         .no_bg_gc = true,
2232                         .nr_free_secs = 1 };
2233
2234                 f2fs_down_write(&sbi->gc_lock);
2235                 stat_inc_gc_call_count(sbi, FOREGROUND);
2236                 err = f2fs_gc(sbi, &gc_control);
2237                 if (err == -ENODATA) {
2238                         err = 0;
2239                         break;
2240                 }
2241                 if (err && err != -EAGAIN)
2242                         break;
2243         }
2244
2245         ret = sync_filesystem(sbi->sb);
2246         if (ret || err) {
2247                 err = ret ? ret : err;
2248                 goto restore_flag;
2249         }
2250
2251         unusable = f2fs_get_unusable_blocks(sbi);
2252         if (f2fs_disable_cp_again(sbi, unusable)) {
2253                 err = -EAGAIN;
2254                 goto restore_flag;
2255         }
2256
2257 skip_gc:
2258         f2fs_down_write(&sbi->gc_lock);
2259         cpc.reason = CP_PAUSE;
2260         set_sbi_flag(sbi, SBI_CP_DISABLED);
2261         stat_inc_cp_call_count(sbi, TOTAL_CALL);
2262         err = f2fs_write_checkpoint(sbi, &cpc);
2263         if (err)
2264                 goto out_unlock;
2265
2266         spin_lock(&sbi->stat_lock);
2267         sbi->unusable_block_count = unusable;
2268         spin_unlock(&sbi->stat_lock);
2269
2270 out_unlock:
2271         f2fs_up_write(&sbi->gc_lock);
2272 restore_flag:
2273         sbi->gc_mode = gc_mode;
2274         sbi->sb->s_flags = s_flags;     /* Restore SB_RDONLY status */
2275         return err;
2276 }
2277
2278 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
2279 {
2280         int retry = DEFAULT_RETRY_IO_COUNT;
2281
2282         /* we should flush all the data to keep data consistency */
2283         do {
2284                 sync_inodes_sb(sbi->sb);
2285                 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2286         } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
2287
2288         if (unlikely(retry < 0))
2289                 f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
2290
2291         f2fs_down_write(&sbi->gc_lock);
2292         f2fs_dirty_to_prefree(sbi);
2293
2294         clear_sbi_flag(sbi, SBI_CP_DISABLED);
2295         set_sbi_flag(sbi, SBI_IS_DIRTY);
2296         f2fs_up_write(&sbi->gc_lock);
2297
2298         f2fs_sync_fs(sbi->sb, 1);
2299
2300         /* Let's ensure there's no pending checkpoint anymore */
2301         f2fs_flush_ckpt_thread(sbi);
2302 }
2303
2304 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
2305 {
2306         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2307         struct f2fs_mount_info org_mount_opt;
2308         unsigned long old_sb_flags;
2309         int err;
2310         bool need_restart_gc = false, need_stop_gc = false;
2311         bool need_restart_flush = false, need_stop_flush = false;
2312         bool need_restart_discard = false, need_stop_discard = false;
2313         bool need_enable_checkpoint = false, need_disable_checkpoint = false;
2314         bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE);
2315         bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE);
2316         bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT);
2317         bool no_atgc = !test_opt(sbi, ATGC);
2318         bool no_discard = !test_opt(sbi, DISCARD);
2319         bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE);
2320         bool block_unit_discard = f2fs_block_unit_discard(sbi);
2321 #ifdef CONFIG_QUOTA
2322         int i, j;
2323 #endif
2324
2325         /*
2326          * Save the old mount options in case we
2327          * need to restore them.
2328          */
2329         org_mount_opt = sbi->mount_opt;
2330         old_sb_flags = sb->s_flags;
2331
2332 #ifdef CONFIG_QUOTA
2333         org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
2334         for (i = 0; i < MAXQUOTAS; i++) {
2335                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2336                         org_mount_opt.s_qf_names[i] =
2337                                 kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
2338                                 GFP_KERNEL);
2339                         if (!org_mount_opt.s_qf_names[i]) {
2340                                 for (j = 0; j < i; j++)
2341                                         kfree(org_mount_opt.s_qf_names[j]);
2342                                 return -ENOMEM;
2343                         }
2344                 } else {
2345                         org_mount_opt.s_qf_names[i] = NULL;
2346                 }
2347         }
2348 #endif
2349
2350         /* recover superblocks we couldn't write due to previous RO mount */
2351         if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
2352                 err = f2fs_commit_super(sbi, false);
2353                 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
2354                           err);
2355                 if (!err)
2356                         clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2357         }
2358
2359         default_options(sbi, true);
2360
2361         /* parse mount options */
2362         err = parse_options(sb, data, true);
2363         if (err)
2364                 goto restore_opts;
2365
2366 #ifdef CONFIG_BLK_DEV_ZONED
2367         if (f2fs_sb_has_blkzoned(sbi) &&
2368                 sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
2369                 f2fs_err(sbi,
2370                         "zoned: max open zones %u is too small, need at least %u open zones",
2371                                  sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
2372                 err = -EINVAL;
2373                 goto restore_opts;
2374         }
2375 #endif
2376
2377         /* flush outstanding errors before changing fs state */
2378         flush_work(&sbi->s_error_work);
2379
2380         /*
2381          * Previous and new state of filesystem is RO,
2382          * so skip checking GC and FLUSH_MERGE conditions.
2383          */
2384         if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
2385                 goto skip;
2386
2387         if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) {
2388                 err = -EROFS;
2389                 goto restore_opts;
2390         }
2391
2392 #ifdef CONFIG_QUOTA
2393         if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
2394                 err = dquot_suspend(sb, -1);
2395                 if (err < 0)
2396                         goto restore_opts;
2397         } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
2398                 /* dquot_resume needs RW */
2399                 sb->s_flags &= ~SB_RDONLY;
2400                 if (sb_any_quota_suspended(sb)) {
2401                         dquot_resume(sb, -1);
2402                 } else if (f2fs_sb_has_quota_ino(sbi)) {
2403                         err = f2fs_enable_quotas(sb);
2404                         if (err)
2405                                 goto restore_opts;
2406                 }
2407         }
2408 #endif
2409         if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) {
2410                 err = -EINVAL;
2411                 f2fs_warn(sbi, "LFS is not compatible with IPU");
2412                 goto restore_opts;
2413         }
2414
2415         /* disallow enable atgc dynamically */
2416         if (no_atgc == !!test_opt(sbi, ATGC)) {
2417                 err = -EINVAL;
2418                 f2fs_warn(sbi, "switch atgc option is not allowed");
2419                 goto restore_opts;
2420         }
2421
2422         /* disallow enable/disable extent_cache dynamically */
2423         if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) {
2424                 err = -EINVAL;
2425                 f2fs_warn(sbi, "switch extent_cache option is not allowed");
2426                 goto restore_opts;
2427         }
2428         /* disallow enable/disable age extent_cache dynamically */
2429         if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) {
2430                 err = -EINVAL;
2431                 f2fs_warn(sbi, "switch age_extent_cache option is not allowed");
2432                 goto restore_opts;
2433         }
2434
2435         if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) {
2436                 err = -EINVAL;
2437                 f2fs_warn(sbi, "switch compress_cache option is not allowed");
2438                 goto restore_opts;
2439         }
2440
2441         if (block_unit_discard != f2fs_block_unit_discard(sbi)) {
2442                 err = -EINVAL;
2443                 f2fs_warn(sbi, "switch discard_unit option is not allowed");
2444                 goto restore_opts;
2445         }
2446
2447         if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
2448                 err = -EINVAL;
2449                 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
2450                 goto restore_opts;
2451         }
2452
2453         /*
2454          * We stop the GC thread if FS is mounted as RO
2455          * or if background_gc = off is passed in mount
2456          * option. Also sync the filesystem.
2457          */
2458         if ((*flags & SB_RDONLY) ||
2459                         (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
2460                         !test_opt(sbi, GC_MERGE))) {
2461                 if (sbi->gc_thread) {
2462                         f2fs_stop_gc_thread(sbi);
2463                         need_restart_gc = true;
2464                 }
2465         } else if (!sbi->gc_thread) {
2466                 err = f2fs_start_gc_thread(sbi);
2467                 if (err)
2468                         goto restore_opts;
2469                 need_stop_gc = true;
2470         }
2471
2472         if (*flags & SB_RDONLY) {
2473                 sync_inodes_sb(sb);
2474
2475                 set_sbi_flag(sbi, SBI_IS_DIRTY);
2476                 set_sbi_flag(sbi, SBI_IS_CLOSE);
2477                 f2fs_sync_fs(sb, 1);
2478                 clear_sbi_flag(sbi, SBI_IS_CLOSE);
2479         }
2480
2481         /*
2482          * We stop issue flush thread if FS is mounted as RO
2483          * or if flush_merge is not passed in mount option.
2484          */
2485         if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2486                 clear_opt(sbi, FLUSH_MERGE);
2487                 f2fs_destroy_flush_cmd_control(sbi, false);
2488                 need_restart_flush = true;
2489         } else {
2490                 err = f2fs_create_flush_cmd_control(sbi);
2491                 if (err)
2492                         goto restore_gc;
2493                 need_stop_flush = true;
2494         }
2495
2496         if (no_discard == !!test_opt(sbi, DISCARD)) {
2497                 if (test_opt(sbi, DISCARD)) {
2498                         err = f2fs_start_discard_thread(sbi);
2499                         if (err)
2500                                 goto restore_flush;
2501                         need_stop_discard = true;
2502                 } else {
2503                         f2fs_stop_discard_thread(sbi);
2504                         f2fs_issue_discard_timeout(sbi);
2505                         need_restart_discard = true;
2506                 }
2507         }
2508
2509         adjust_unusable_cap_perc(sbi);
2510         if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) {
2511                 if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2512                         err = f2fs_disable_checkpoint(sbi);
2513                         if (err)
2514                                 goto restore_discard;
2515                         need_enable_checkpoint = true;
2516                 } else {
2517                         f2fs_enable_checkpoint(sbi);
2518                         need_disable_checkpoint = true;
2519                 }
2520         }
2521
2522         /*
2523          * Place this routine at the end, since a new checkpoint would be
2524          * triggered while remount and we need to take care of it before
2525          * returning from remount.
2526          */
2527         if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) ||
2528                         !test_opt(sbi, MERGE_CHECKPOINT)) {
2529                 f2fs_stop_ckpt_thread(sbi);
2530         } else {
2531                 /* Flush if the prevous checkpoint, if exists. */
2532                 f2fs_flush_ckpt_thread(sbi);
2533
2534                 err = f2fs_start_ckpt_thread(sbi);
2535                 if (err) {
2536                         f2fs_err(sbi,
2537                             "Failed to start F2FS issue_checkpoint_thread (%d)",
2538                             err);
2539                         goto restore_checkpoint;
2540                 }
2541         }
2542
2543 skip:
2544 #ifdef CONFIG_QUOTA
2545         /* Release old quota file names */
2546         for (i = 0; i < MAXQUOTAS; i++)
2547                 kfree(org_mount_opt.s_qf_names[i]);
2548 #endif
2549         /* Update the POSIXACL Flag */
2550         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2551                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2552
2553         limit_reserve_root(sbi);
2554         *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2555         return 0;
2556 restore_checkpoint:
2557         if (need_enable_checkpoint) {
2558                 f2fs_enable_checkpoint(sbi);
2559         } else if (need_disable_checkpoint) {
2560                 if (f2fs_disable_checkpoint(sbi))
2561                         f2fs_warn(sbi, "checkpoint has not been disabled");
2562         }
2563 restore_discard:
2564         if (need_restart_discard) {
2565                 if (f2fs_start_discard_thread(sbi))
2566                         f2fs_warn(sbi, "discard has been stopped");
2567         } else if (need_stop_discard) {
2568                 f2fs_stop_discard_thread(sbi);
2569         }
2570 restore_flush:
2571         if (need_restart_flush) {
2572                 if (f2fs_create_flush_cmd_control(sbi))
2573                         f2fs_warn(sbi, "background flush thread has stopped");
2574         } else if (need_stop_flush) {
2575                 clear_opt(sbi, FLUSH_MERGE);
2576                 f2fs_destroy_flush_cmd_control(sbi, false);
2577         }
2578 restore_gc:
2579         if (need_restart_gc) {
2580                 if (f2fs_start_gc_thread(sbi))
2581                         f2fs_warn(sbi, "background gc thread has stopped");
2582         } else if (need_stop_gc) {
2583                 f2fs_stop_gc_thread(sbi);
2584         }
2585 restore_opts:
2586 #ifdef CONFIG_QUOTA
2587         F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2588         for (i = 0; i < MAXQUOTAS; i++) {
2589                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2590                 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2591         }
2592 #endif
2593         sbi->mount_opt = org_mount_opt;
2594         sb->s_flags = old_sb_flags;
2595         return err;
2596 }
2597
2598 static void f2fs_shutdown(struct super_block *sb)
2599 {
2600         f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false, false);
2601 }
2602
2603 #ifdef CONFIG_QUOTA
2604 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi)
2605 {
2606         /* need to recovery orphan */
2607         if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
2608                 return true;
2609         /* need to recovery data */
2610         if (test_opt(sbi, DISABLE_ROLL_FORWARD))
2611                 return false;
2612         if (test_opt(sbi, NORECOVERY))
2613                 return false;
2614         return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG);
2615 }
2616
2617 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi)
2618 {
2619         bool readonly = f2fs_readonly(sbi->sb);
2620
2621         if (!f2fs_need_recovery(sbi))
2622                 return false;
2623
2624         /* it doesn't need to check f2fs_sb_has_readonly() */
2625         if (f2fs_hw_is_readonly(sbi))
2626                 return false;
2627
2628         if (readonly) {
2629                 sbi->sb->s_flags &= ~SB_RDONLY;
2630                 set_sbi_flag(sbi, SBI_IS_WRITABLE);
2631         }
2632
2633         /*
2634          * Turn on quotas which were not enabled for read-only mounts if
2635          * filesystem has quota feature, so that they are updated correctly.
2636          */
2637         return f2fs_enable_quota_files(sbi, readonly);
2638 }
2639
2640 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi,
2641                                                 bool quota_enabled)
2642 {
2643         if (quota_enabled)
2644                 f2fs_quota_off_umount(sbi->sb);
2645
2646         if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) {
2647                 clear_sbi_flag(sbi, SBI_IS_WRITABLE);
2648                 sbi->sb->s_flags |= SB_RDONLY;
2649         }
2650 }
2651
2652 /* Read data from quotafile */
2653 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2654                                size_t len, loff_t off)
2655 {
2656         struct inode *inode = sb_dqopt(sb)->files[type];
2657         struct address_space *mapping = inode->i_mapping;
2658         block_t blkidx = F2FS_BYTES_TO_BLK(off);
2659         int offset = off & (sb->s_blocksize - 1);
2660         int tocopy;
2661         size_t toread;
2662         loff_t i_size = i_size_read(inode);
2663         struct page *page;
2664
2665         if (off > i_size)
2666                 return 0;
2667
2668         if (off + len > i_size)
2669                 len = i_size - off;
2670         toread = len;
2671         while (toread > 0) {
2672                 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2673 repeat:
2674                 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2675                 if (IS_ERR(page)) {
2676                         if (PTR_ERR(page) == -ENOMEM) {
2677                                 memalloc_retry_wait(GFP_NOFS);
2678                                 goto repeat;
2679                         }
2680                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2681                         return PTR_ERR(page);
2682                 }
2683
2684                 lock_page(page);
2685
2686                 if (unlikely(page->mapping != mapping)) {
2687                         f2fs_put_page(page, 1);
2688                         goto repeat;
2689                 }
2690                 if (unlikely(!PageUptodate(page))) {
2691                         f2fs_put_page(page, 1);
2692                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2693                         return -EIO;
2694                 }
2695
2696                 memcpy_from_page(data, page, offset, tocopy);
2697                 f2fs_put_page(page, 1);
2698
2699                 offset = 0;
2700                 toread -= tocopy;
2701                 data += tocopy;
2702                 blkidx++;
2703         }
2704         return len;
2705 }
2706
2707 /* Write to quotafile */
2708 static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2709                                 const char *data, size_t len, loff_t off)
2710 {
2711         struct inode *inode = sb_dqopt(sb)->files[type];
2712         struct address_space *mapping = inode->i_mapping;
2713         const struct address_space_operations *a_ops = mapping->a_ops;
2714         int offset = off & (sb->s_blocksize - 1);
2715         size_t towrite = len;
2716         struct folio *folio;
2717         void *fsdata = NULL;
2718         int err = 0;
2719         int tocopy;
2720
2721         while (towrite > 0) {
2722                 tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2723                                                                 towrite);
2724 retry:
2725                 err = a_ops->write_begin(NULL, mapping, off, tocopy,
2726                                                         &folio, &fsdata);
2727                 if (unlikely(err)) {
2728                         if (err == -ENOMEM) {
2729                                 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT);
2730                                 goto retry;
2731                         }
2732                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2733                         break;
2734                 }
2735
2736                 memcpy_to_folio(folio, offset_in_folio(folio, off), data, tocopy);
2737
2738                 a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2739                                                 folio, fsdata);
2740                 offset = 0;
2741                 towrite -= tocopy;
2742                 off += tocopy;
2743                 data += tocopy;
2744                 cond_resched();
2745         }
2746
2747         if (len == towrite)
2748                 return err;
2749         inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
2750         f2fs_mark_inode_dirty_sync(inode, false);
2751         return len - towrite;
2752 }
2753
2754 int f2fs_dquot_initialize(struct inode *inode)
2755 {
2756         if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT))
2757                 return -ESRCH;
2758
2759         return dquot_initialize(inode);
2760 }
2761
2762 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode)
2763 {
2764         return F2FS_I(inode)->i_dquot;
2765 }
2766
2767 static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2768 {
2769         return &F2FS_I(inode)->i_reserved_quota;
2770 }
2771
2772 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2773 {
2774         if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2775                 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2776                 return 0;
2777         }
2778
2779         return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2780                                         F2FS_OPTION(sbi).s_jquota_fmt, type);
2781 }
2782
2783 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2784 {
2785         int enabled = 0;
2786         int i, err;
2787
2788         if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2789                 err = f2fs_enable_quotas(sbi->sb);
2790                 if (err) {
2791                         f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2792                         return 0;
2793                 }
2794                 return 1;
2795         }
2796
2797         for (i = 0; i < MAXQUOTAS; i++) {
2798                 if (F2FS_OPTION(sbi).s_qf_names[i]) {
2799                         err = f2fs_quota_on_mount(sbi, i);
2800                         if (!err) {
2801                                 enabled = 1;
2802                                 continue;
2803                         }
2804                         f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2805                                  err, i);
2806                 }
2807         }
2808         return enabled;
2809 }
2810
2811 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2812                              unsigned int flags)
2813 {
2814         struct inode *qf_inode;
2815         unsigned long qf_inum;
2816         unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL;
2817         int err;
2818
2819         BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2820
2821         qf_inum = f2fs_qf_ino(sb, type);
2822         if (!qf_inum)
2823                 return -EPERM;
2824
2825         qf_inode = f2fs_iget(sb, qf_inum);
2826         if (IS_ERR(qf_inode)) {
2827                 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2828                 return PTR_ERR(qf_inode);
2829         }
2830
2831         /* Don't account quota for quota files to avoid recursion */
2832         inode_lock(qf_inode);
2833         qf_inode->i_flags |= S_NOQUOTA;
2834
2835         if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) {
2836                 F2FS_I(qf_inode)->i_flags |= qf_flag;
2837                 f2fs_set_inode_flags(qf_inode);
2838         }
2839         inode_unlock(qf_inode);
2840
2841         err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2842         iput(qf_inode);
2843         return err;
2844 }
2845
2846 static int f2fs_enable_quotas(struct super_block *sb)
2847 {
2848         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2849         int type, err = 0;
2850         unsigned long qf_inum;
2851         bool quota_mopt[MAXQUOTAS] = {
2852                 test_opt(sbi, USRQUOTA),
2853                 test_opt(sbi, GRPQUOTA),
2854                 test_opt(sbi, PRJQUOTA),
2855         };
2856
2857         if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2858                 f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2859                 return 0;
2860         }
2861
2862         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2863
2864         for (type = 0; type < MAXQUOTAS; type++) {
2865                 qf_inum = f2fs_qf_ino(sb, type);
2866                 if (qf_inum) {
2867                         err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2868                                 DQUOT_USAGE_ENABLED |
2869                                 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2870                         if (err) {
2871                                 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2872                                          type, err);
2873                                 for (type--; type >= 0; type--)
2874                                         dquot_quota_off(sb, type);
2875                                 set_sbi_flag(F2FS_SB(sb),
2876                                                 SBI_QUOTA_NEED_REPAIR);
2877                                 return err;
2878                         }
2879                 }
2880         }
2881         return 0;
2882 }
2883
2884 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2885 {
2886         struct quota_info *dqopt = sb_dqopt(sbi->sb);
2887         struct address_space *mapping = dqopt->files[type]->i_mapping;
2888         int ret = 0;
2889
2890         ret = dquot_writeback_dquots(sbi->sb, type);
2891         if (ret)
2892                 goto out;
2893
2894         ret = filemap_fdatawrite(mapping);
2895         if (ret)
2896                 goto out;
2897
2898         /* if we are using journalled quota */
2899         if (is_journalled_quota(sbi))
2900                 goto out;
2901
2902         ret = filemap_fdatawait(mapping);
2903
2904         truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2905 out:
2906         if (ret)
2907                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2908         return ret;
2909 }
2910
2911 int f2fs_quota_sync(struct super_block *sb, int type)
2912 {
2913         struct f2fs_sb_info *sbi = F2FS_SB(sb);
2914         struct quota_info *dqopt = sb_dqopt(sb);
2915         int cnt;
2916         int ret = 0;
2917
2918         /*
2919          * Now when everything is written we can discard the pagecache so
2920          * that userspace sees the changes.
2921          */
2922         for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2923
2924                 if (type != -1 && cnt != type)
2925                         continue;
2926
2927                 if (!sb_has_quota_active(sb, cnt))
2928                         continue;
2929
2930                 if (!f2fs_sb_has_quota_ino(sbi))
2931                         inode_lock(dqopt->files[cnt]);
2932
2933                 /*
2934                  * do_quotactl
2935                  *  f2fs_quota_sync
2936                  *  f2fs_down_read(quota_sem)
2937                  *  dquot_writeback_dquots()
2938                  *  f2fs_dquot_commit
2939                  *                            block_operation
2940                  *                            f2fs_down_read(quota_sem)
2941                  */
2942                 f2fs_lock_op(sbi);
2943                 f2fs_down_read(&sbi->quota_sem);
2944
2945                 ret = f2fs_quota_sync_file(sbi, cnt);
2946
2947                 f2fs_up_read(&sbi->quota_sem);
2948                 f2fs_unlock_op(sbi);
2949
2950                 if (!f2fs_sb_has_quota_ino(sbi))
2951                         inode_unlock(dqopt->files[cnt]);
2952
2953                 if (ret)
2954                         break;
2955         }
2956         return ret;
2957 }
2958
2959 static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2960                                                         const struct path *path)
2961 {
2962         struct inode *inode;
2963         int err;
2964
2965         /* if quota sysfile exists, deny enabling quota with specific file */
2966         if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2967                 f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2968                 return -EBUSY;
2969         }
2970
2971         if (path->dentry->d_sb != sb)
2972                 return -EXDEV;
2973
2974         err = f2fs_quota_sync(sb, type);
2975         if (err)
2976                 return err;
2977
2978         inode = d_inode(path->dentry);
2979
2980         err = filemap_fdatawrite(inode->i_mapping);
2981         if (err)
2982                 return err;
2983
2984         err = filemap_fdatawait(inode->i_mapping);
2985         if (err)
2986                 return err;
2987
2988         err = dquot_quota_on(sb, type, format_id, path);
2989         if (err)
2990                 return err;
2991
2992         inode_lock(inode);
2993         F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL;
2994         f2fs_set_inode_flags(inode);
2995         inode_unlock(inode);
2996         f2fs_mark_inode_dirty_sync(inode, false);
2997
2998         return 0;
2999 }
3000
3001 static int __f2fs_quota_off(struct super_block *sb, int type)
3002 {
3003         struct inode *inode = sb_dqopt(sb)->files[type];
3004         int err;
3005
3006         if (!inode || !igrab(inode))
3007                 return dquot_quota_off(sb, type);
3008
3009         err = f2fs_quota_sync(sb, type);
3010         if (err)
3011                 goto out_put;
3012
3013         err = dquot_quota_off(sb, type);
3014         if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
3015                 goto out_put;
3016
3017         inode_lock(inode);
3018         F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL;
3019         f2fs_set_inode_flags(inode);
3020         inode_unlock(inode);
3021         f2fs_mark_inode_dirty_sync(inode, false);
3022 out_put:
3023         iput(inode);
3024         return err;
3025 }
3026
3027 static int f2fs_quota_off(struct super_block *sb, int type)
3028 {
3029         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3030         int err;
3031
3032         err = __f2fs_quota_off(sb, type);
3033
3034         /*
3035          * quotactl can shutdown journalled quota, result in inconsistence
3036          * between quota record and fs data by following updates, tag the
3037          * flag to let fsck be aware of it.
3038          */
3039         if (is_journalled_quota(sbi))
3040                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3041         return err;
3042 }
3043
3044 void f2fs_quota_off_umount(struct super_block *sb)
3045 {
3046         int type;
3047         int err;
3048
3049         for (type = 0; type < MAXQUOTAS; type++) {
3050                 err = __f2fs_quota_off(sb, type);
3051                 if (err) {
3052                         int ret = dquot_quota_off(sb, type);
3053
3054                         f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
3055                                  type, err, ret);
3056                         set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
3057                 }
3058         }
3059         /*
3060          * In case of checkpoint=disable, we must flush quota blocks.
3061          * This can cause NULL exception for node_inode in end_io, since
3062          * put_super already dropped it.
3063          */
3064         sync_filesystem(sb);
3065 }
3066
3067 static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
3068 {
3069         struct quota_info *dqopt = sb_dqopt(sb);
3070         int type;
3071
3072         for (type = 0; type < MAXQUOTAS; type++) {
3073                 if (!dqopt->files[type])
3074                         continue;
3075                 f2fs_inode_synced(dqopt->files[type]);
3076         }
3077 }
3078
3079 static int f2fs_dquot_commit(struct dquot *dquot)
3080 {
3081         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3082         int ret;
3083
3084         f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
3085         ret = dquot_commit(dquot);
3086         if (ret < 0)
3087                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3088         f2fs_up_read(&sbi->quota_sem);
3089         return ret;
3090 }
3091
3092 static int f2fs_dquot_acquire(struct dquot *dquot)
3093 {
3094         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3095         int ret;
3096
3097         f2fs_down_read(&sbi->quota_sem);
3098         ret = dquot_acquire(dquot);
3099         if (ret < 0)
3100                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3101         f2fs_up_read(&sbi->quota_sem);
3102         return ret;
3103 }
3104
3105 static int f2fs_dquot_release(struct dquot *dquot)
3106 {
3107         struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
3108         int ret = dquot_release(dquot);
3109
3110         if (ret < 0)
3111                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3112         return ret;
3113 }
3114
3115 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
3116 {
3117         struct super_block *sb = dquot->dq_sb;
3118         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3119         int ret = dquot_mark_dquot_dirty(dquot);
3120
3121         /* if we are using journalled quota */
3122         if (is_journalled_quota(sbi))
3123                 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
3124
3125         return ret;
3126 }
3127
3128 static int f2fs_dquot_commit_info(struct super_block *sb, int type)
3129 {
3130         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3131         int ret = dquot_commit_info(sb, type);
3132
3133         if (ret < 0)
3134                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3135         return ret;
3136 }
3137
3138 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
3139 {
3140         *projid = F2FS_I(inode)->i_projid;
3141         return 0;
3142 }
3143
3144 static const struct dquot_operations f2fs_quota_operations = {
3145         .get_reserved_space = f2fs_get_reserved_space,
3146         .write_dquot    = f2fs_dquot_commit,
3147         .acquire_dquot  = f2fs_dquot_acquire,
3148         .release_dquot  = f2fs_dquot_release,
3149         .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
3150         .write_info     = f2fs_dquot_commit_info,
3151         .alloc_dquot    = dquot_alloc,
3152         .destroy_dquot  = dquot_destroy,
3153         .get_projid     = f2fs_get_projid,
3154         .get_next_id    = dquot_get_next_id,
3155 };
3156
3157 static const struct quotactl_ops f2fs_quotactl_ops = {
3158         .quota_on       = f2fs_quota_on,
3159         .quota_off      = f2fs_quota_off,
3160         .quota_sync     = f2fs_quota_sync,
3161         .get_state      = dquot_get_state,
3162         .set_info       = dquot_set_dqinfo,
3163         .get_dqblk      = dquot_get_dqblk,
3164         .set_dqblk      = dquot_set_dqblk,
3165         .get_nextdqblk  = dquot_get_next_dqblk,
3166 };
3167 #else
3168 int f2fs_dquot_initialize(struct inode *inode)
3169 {
3170         return 0;
3171 }
3172
3173 int f2fs_quota_sync(struct super_block *sb, int type)
3174 {
3175         return 0;
3176 }
3177
3178 void f2fs_quota_off_umount(struct super_block *sb)
3179 {
3180 }
3181 #endif
3182
3183 static const struct super_operations f2fs_sops = {
3184         .alloc_inode    = f2fs_alloc_inode,
3185         .free_inode     = f2fs_free_inode,
3186         .drop_inode     = f2fs_drop_inode,
3187         .write_inode    = f2fs_write_inode,
3188         .dirty_inode    = f2fs_dirty_inode,
3189         .show_options   = f2fs_show_options,
3190 #ifdef CONFIG_QUOTA
3191         .quota_read     = f2fs_quota_read,
3192         .quota_write    = f2fs_quota_write,
3193         .get_dquots     = f2fs_get_dquots,
3194 #endif
3195         .evict_inode    = f2fs_evict_inode,
3196         .put_super      = f2fs_put_super,
3197         .sync_fs        = f2fs_sync_fs,
3198         .freeze_fs      = f2fs_freeze,
3199         .unfreeze_fs    = f2fs_unfreeze,
3200         .statfs         = f2fs_statfs,
3201         .remount_fs     = f2fs_remount,
3202         .shutdown       = f2fs_shutdown,
3203 };
3204
3205 #ifdef CONFIG_FS_ENCRYPTION
3206 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
3207 {
3208         return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3209                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3210                                 ctx, len, NULL);
3211 }
3212
3213 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
3214                                                         void *fs_data)
3215 {
3216         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3217
3218         /*
3219          * Encrypting the root directory is not allowed because fsck
3220          * expects lost+found directory to exist and remain unencrypted
3221          * if LOST_FOUND feature is enabled.
3222          *
3223          */
3224         if (f2fs_sb_has_lost_found(sbi) &&
3225                         inode->i_ino == F2FS_ROOT_INO(sbi))
3226                 return -EPERM;
3227
3228         return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
3229                                 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
3230                                 ctx, len, fs_data, XATTR_CREATE);
3231 }
3232
3233 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
3234 {
3235         return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
3236 }
3237
3238 static bool f2fs_has_stable_inodes(struct super_block *sb)
3239 {
3240         return true;
3241 }
3242
3243 static struct block_device **f2fs_get_devices(struct super_block *sb,
3244                                               unsigned int *num_devs)
3245 {
3246         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3247         struct block_device **devs;
3248         int i;
3249
3250         if (!f2fs_is_multi_device(sbi))
3251                 return NULL;
3252
3253         devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL);
3254         if (!devs)
3255                 return ERR_PTR(-ENOMEM);
3256
3257         for (i = 0; i < sbi->s_ndevs; i++)
3258                 devs[i] = FDEV(i).bdev;
3259         *num_devs = sbi->s_ndevs;
3260         return devs;
3261 }
3262
3263 static const struct fscrypt_operations f2fs_cryptops = {
3264         .needs_bounce_pages     = 1,
3265         .has_32bit_inodes       = 1,
3266         .supports_subblock_data_units = 1,
3267         .legacy_key_prefix      = "f2fs:",
3268         .get_context            = f2fs_get_context,
3269         .set_context            = f2fs_set_context,
3270         .get_dummy_policy       = f2fs_get_dummy_policy,
3271         .empty_dir              = f2fs_empty_dir,
3272         .has_stable_inodes      = f2fs_has_stable_inodes,
3273         .get_devices            = f2fs_get_devices,
3274 };
3275 #endif
3276
3277 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
3278                 u64 ino, u32 generation)
3279 {
3280         struct f2fs_sb_info *sbi = F2FS_SB(sb);
3281         struct inode *inode;
3282
3283         if (f2fs_check_nid_range(sbi, ino))
3284                 return ERR_PTR(-ESTALE);
3285
3286         /*
3287          * f2fs_iget isn't quite right if the inode is currently unallocated!
3288          * However f2fs_iget currently does appropriate checks to handle stale
3289          * inodes so everything is OK.
3290          */
3291         inode = f2fs_iget(sb, ino);
3292         if (IS_ERR(inode))
3293                 return ERR_CAST(inode);
3294         if (unlikely(generation && inode->i_generation != generation)) {
3295                 /* we didn't find the right inode.. */
3296                 iput(inode);
3297                 return ERR_PTR(-ESTALE);
3298         }
3299         return inode;
3300 }
3301
3302 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
3303                 int fh_len, int fh_type)
3304 {
3305         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
3306                                     f2fs_nfs_get_inode);
3307 }
3308
3309 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
3310                 int fh_len, int fh_type)
3311 {
3312         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
3313                                     f2fs_nfs_get_inode);
3314 }
3315
3316 static const struct export_operations f2fs_export_ops = {
3317         .encode_fh = generic_encode_ino32_fh,
3318         .fh_to_dentry = f2fs_fh_to_dentry,
3319         .fh_to_parent = f2fs_fh_to_parent,
3320         .get_parent = f2fs_get_parent,
3321 };
3322
3323 loff_t max_file_blocks(struct inode *inode)
3324 {
3325         loff_t result = 0;
3326         loff_t leaf_count;
3327
3328         /*
3329          * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
3330          * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
3331          * space in inode.i_addr, it will be more safe to reassign
3332          * result as zero.
3333          */
3334
3335         if (inode && f2fs_compressed_file(inode))
3336                 leaf_count = ADDRS_PER_BLOCK(inode);
3337         else
3338                 leaf_count = DEF_ADDRS_PER_BLOCK;
3339
3340         /* two direct node blocks */
3341         result += (leaf_count * 2);
3342
3343         /* two indirect node blocks */
3344         leaf_count *= NIDS_PER_BLOCK;
3345         result += (leaf_count * 2);
3346
3347         /* one double indirect node block */
3348         leaf_count *= NIDS_PER_BLOCK;
3349         result += leaf_count;
3350
3351         /*
3352          * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with
3353          * a 4K crypto data unit, we must restrict the max filesize to what can
3354          * fit within U32_MAX + 1 data units.
3355          */
3356
3357         result = umin(result, F2FS_BYTES_TO_BLK(((loff_t)U32_MAX + 1) * 4096));
3358
3359         return result;
3360 }
3361
3362 static int __f2fs_commit_super(struct f2fs_sb_info *sbi, struct folio *folio,
3363                                                 pgoff_t index, bool update)
3364 {
3365         struct bio *bio;
3366         /* it's rare case, we can do fua all the time */
3367         blk_opf_t opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH | REQ_FUA;
3368         int ret;
3369
3370         folio_lock(folio);
3371         folio_wait_writeback(folio);
3372         if (update)
3373                 memcpy(F2FS_SUPER_BLOCK(folio, index), F2FS_RAW_SUPER(sbi),
3374                                         sizeof(struct f2fs_super_block));
3375         folio_mark_dirty(folio);
3376         folio_clear_dirty_for_io(folio);
3377         folio_start_writeback(folio);
3378         folio_unlock(folio);
3379
3380         bio = bio_alloc(sbi->sb->s_bdev, 1, opf, GFP_NOFS);
3381
3382         /* it doesn't need to set crypto context for superblock update */
3383         bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(folio_index(folio));
3384
3385         if (!bio_add_folio(bio, folio, folio_size(folio), 0))
3386                 f2fs_bug_on(sbi, 1);
3387
3388         ret = submit_bio_wait(bio);
3389         folio_end_writeback(folio);
3390
3391         return ret;
3392 }
3393
3394 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
3395                                         struct folio *folio, pgoff_t index)
3396 {
3397         struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index);
3398         struct super_block *sb = sbi->sb;
3399         u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
3400         u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
3401         u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
3402         u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
3403         u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
3404         u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
3405         u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
3406         u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
3407         u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
3408         u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
3409         u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3410         u32 segment_count = le32_to_cpu(raw_super->segment_count);
3411         u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3412         u64 main_end_blkaddr = main_blkaddr +
3413                                 ((u64)segment_count_main << log_blocks_per_seg);
3414         u64 seg_end_blkaddr = segment0_blkaddr +
3415                                 ((u64)segment_count << log_blocks_per_seg);
3416
3417         if (segment0_blkaddr != cp_blkaddr) {
3418                 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
3419                           segment0_blkaddr, cp_blkaddr);
3420                 return true;
3421         }
3422
3423         if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
3424                                                         sit_blkaddr) {
3425                 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
3426                           cp_blkaddr, sit_blkaddr,
3427                           segment_count_ckpt << log_blocks_per_seg);
3428                 return true;
3429         }
3430
3431         if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
3432                                                         nat_blkaddr) {
3433                 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
3434                           sit_blkaddr, nat_blkaddr,
3435                           segment_count_sit << log_blocks_per_seg);
3436                 return true;
3437         }
3438
3439         if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
3440                                                         ssa_blkaddr) {
3441                 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
3442                           nat_blkaddr, ssa_blkaddr,
3443                           segment_count_nat << log_blocks_per_seg);
3444                 return true;
3445         }
3446
3447         if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
3448                                                         main_blkaddr) {
3449                 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
3450                           ssa_blkaddr, main_blkaddr,
3451                           segment_count_ssa << log_blocks_per_seg);
3452                 return true;
3453         }
3454
3455         if (main_end_blkaddr > seg_end_blkaddr) {
3456                 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
3457                           main_blkaddr, seg_end_blkaddr,
3458                           segment_count_main << log_blocks_per_seg);
3459                 return true;
3460         } else if (main_end_blkaddr < seg_end_blkaddr) {
3461                 int err = 0;
3462                 char *res;
3463
3464                 /* fix in-memory information all the time */
3465                 raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
3466                                 segment0_blkaddr) >> log_blocks_per_seg);
3467
3468                 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) {
3469                         set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3470                         res = "internally";
3471                 } else {
3472                         err = __f2fs_commit_super(sbi, folio, index, false);
3473                         res = err ? "failed" : "done";
3474                 }
3475                 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
3476                           res, main_blkaddr, seg_end_blkaddr,
3477                           segment_count_main << log_blocks_per_seg);
3478                 if (err)
3479                         return true;
3480         }
3481         return false;
3482 }
3483
3484 static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
3485                                         struct folio *folio, pgoff_t index)
3486 {
3487         block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
3488         block_t total_sections, blocks_per_seg;
3489         struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index);
3490         size_t crc_offset = 0;
3491         __u32 crc = 0;
3492
3493         if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
3494                 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
3495                           F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
3496                 return -EINVAL;
3497         }
3498
3499         /* Check checksum_offset and crc in superblock */
3500         if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
3501                 crc_offset = le32_to_cpu(raw_super->checksum_offset);
3502                 if (crc_offset !=
3503                         offsetof(struct f2fs_super_block, crc)) {
3504                         f2fs_info(sbi, "Invalid SB checksum offset: %zu",
3505                                   crc_offset);
3506                         return -EFSCORRUPTED;
3507                 }
3508                 crc = le32_to_cpu(raw_super->crc);
3509                 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
3510                         f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
3511                         return -EFSCORRUPTED;
3512                 }
3513         }
3514
3515         /* only support block_size equals to PAGE_SIZE */
3516         if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
3517                 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
3518                           le32_to_cpu(raw_super->log_blocksize),
3519                           F2FS_BLKSIZE_BITS);
3520                 return -EFSCORRUPTED;
3521         }
3522
3523         /* check log blocks per segment */
3524         if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
3525                 f2fs_info(sbi, "Invalid log blocks per segment (%u)",
3526                           le32_to_cpu(raw_super->log_blocks_per_seg));
3527                 return -EFSCORRUPTED;
3528         }
3529
3530         /* Currently, support 512/1024/2048/4096/16K bytes sector size */
3531         if (le32_to_cpu(raw_super->log_sectorsize) >
3532                                 F2FS_MAX_LOG_SECTOR_SIZE ||
3533                 le32_to_cpu(raw_super->log_sectorsize) <
3534                                 F2FS_MIN_LOG_SECTOR_SIZE) {
3535                 f2fs_info(sbi, "Invalid log sectorsize (%u)",
3536                           le32_to_cpu(raw_super->log_sectorsize));
3537                 return -EFSCORRUPTED;
3538         }
3539         if (le32_to_cpu(raw_super->log_sectors_per_block) +
3540                 le32_to_cpu(raw_super->log_sectorsize) !=
3541                         F2FS_MAX_LOG_SECTOR_SIZE) {
3542                 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
3543                           le32_to_cpu(raw_super->log_sectors_per_block),
3544                           le32_to_cpu(raw_super->log_sectorsize));
3545                 return -EFSCORRUPTED;
3546         }
3547
3548         segment_count = le32_to_cpu(raw_super->segment_count);
3549         segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3550         segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3551         secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3552         total_sections = le32_to_cpu(raw_super->section_count);
3553
3554         /* blocks_per_seg should be 512, given the above check */
3555         blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg));
3556
3557         if (segment_count > F2FS_MAX_SEGMENT ||
3558                                 segment_count < F2FS_MIN_SEGMENTS) {
3559                 f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
3560                 return -EFSCORRUPTED;
3561         }
3562
3563         if (total_sections > segment_count_main || total_sections < 1 ||
3564                         segs_per_sec > segment_count || !segs_per_sec) {
3565                 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
3566                           segment_count, total_sections, segs_per_sec);
3567                 return -EFSCORRUPTED;
3568         }
3569
3570         if (segment_count_main != total_sections * segs_per_sec) {
3571                 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
3572                           segment_count_main, total_sections, segs_per_sec);
3573                 return -EFSCORRUPTED;
3574         }
3575
3576         if ((segment_count / segs_per_sec) < total_sections) {
3577                 f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
3578                           segment_count, segs_per_sec, total_sections);
3579                 return -EFSCORRUPTED;
3580         }
3581
3582         if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
3583                 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
3584                           segment_count, le64_to_cpu(raw_super->block_count));
3585                 return -EFSCORRUPTED;
3586         }
3587
3588         if (RDEV(0).path[0]) {
3589                 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
3590                 int i = 1;
3591
3592                 while (i < MAX_DEVICES && RDEV(i).path[0]) {
3593                         dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
3594                         i++;
3595                 }
3596                 if (segment_count != dev_seg_count) {
3597                         f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
3598                                         segment_count, dev_seg_count);
3599                         return -EFSCORRUPTED;
3600                 }
3601         } else {
3602                 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
3603                                         !bdev_is_zoned(sbi->sb->s_bdev)) {
3604                         f2fs_info(sbi, "Zoned block device path is missing");
3605                         return -EFSCORRUPTED;
3606                 }
3607         }
3608
3609         if (secs_per_zone > total_sections || !secs_per_zone) {
3610                 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
3611                           secs_per_zone, total_sections);
3612                 return -EFSCORRUPTED;
3613         }
3614         if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
3615                         raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
3616                         (le32_to_cpu(raw_super->extension_count) +
3617                         raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
3618                 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
3619                           le32_to_cpu(raw_super->extension_count),
3620                           raw_super->hot_ext_count,
3621                           F2FS_MAX_EXTENSION);
3622                 return -EFSCORRUPTED;
3623         }
3624
3625         if (le32_to_cpu(raw_super->cp_payload) >=
3626                                 (blocks_per_seg - F2FS_CP_PACKS -
3627                                 NR_CURSEG_PERSIST_TYPE)) {
3628                 f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
3629                           le32_to_cpu(raw_super->cp_payload),
3630                           blocks_per_seg - F2FS_CP_PACKS -
3631                           NR_CURSEG_PERSIST_TYPE);
3632                 return -EFSCORRUPTED;
3633         }
3634
3635         /* check reserved ino info */
3636         if (le32_to_cpu(raw_super->node_ino) != 1 ||
3637                 le32_to_cpu(raw_super->meta_ino) != 2 ||
3638                 le32_to_cpu(raw_super->root_ino) != 3) {
3639                 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3640                           le32_to_cpu(raw_super->node_ino),
3641                           le32_to_cpu(raw_super->meta_ino),
3642                           le32_to_cpu(raw_super->root_ino));
3643                 return -EFSCORRUPTED;
3644         }
3645
3646         /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3647         if (sanity_check_area_boundary(sbi, folio, index))
3648                 return -EFSCORRUPTED;
3649
3650         return 0;
3651 }
3652
3653 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3654 {
3655         unsigned int total, fsmeta;
3656         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3657         struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3658         unsigned int ovp_segments, reserved_segments;
3659         unsigned int main_segs, blocks_per_seg;
3660         unsigned int sit_segs, nat_segs;
3661         unsigned int sit_bitmap_size, nat_bitmap_size;
3662         unsigned int log_blocks_per_seg;
3663         unsigned int segment_count_main;
3664         unsigned int cp_pack_start_sum, cp_payload;
3665         block_t user_block_count, valid_user_blocks;
3666         block_t avail_node_count, valid_node_count;
3667         unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3668         int i, j;
3669
3670         total = le32_to_cpu(raw_super->segment_count);
3671         fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3672         sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3673         fsmeta += sit_segs;
3674         nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3675         fsmeta += nat_segs;
3676         fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3677         fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3678
3679         if (unlikely(fsmeta >= total))
3680                 return 1;
3681
3682         ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3683         reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3684
3685         if (!f2fs_sb_has_readonly(sbi) &&
3686                         unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3687                         ovp_segments == 0 || reserved_segments == 0)) {
3688                 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3689                 return 1;
3690         }
3691         user_block_count = le64_to_cpu(ckpt->user_block_count);
3692         segment_count_main = le32_to_cpu(raw_super->segment_count_main) +
3693                         (f2fs_sb_has_readonly(sbi) ? 1 : 0);
3694         log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3695         if (!user_block_count || user_block_count >=
3696                         segment_count_main << log_blocks_per_seg) {
3697                 f2fs_err(sbi, "Wrong user_block_count: %u",
3698                          user_block_count);
3699                 return 1;
3700         }
3701
3702         valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3703         if (valid_user_blocks > user_block_count) {
3704                 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3705                          valid_user_blocks, user_block_count);
3706                 return 1;
3707         }
3708
3709         valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3710         avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3711         if (valid_node_count > avail_node_count) {
3712                 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3713                          valid_node_count, avail_node_count);
3714                 return 1;
3715         }
3716
3717         main_segs = le32_to_cpu(raw_super->segment_count_main);
3718         blocks_per_seg = BLKS_PER_SEG(sbi);
3719
3720         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3721                 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3722                         le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3723                         return 1;
3724
3725                 if (f2fs_sb_has_readonly(sbi))
3726                         goto check_data;
3727
3728                 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3729                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3730                                 le32_to_cpu(ckpt->cur_node_segno[j])) {
3731                                 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3732                                          i, j,
3733                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3734                                 return 1;
3735                         }
3736                 }
3737         }
3738 check_data:
3739         for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3740                 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3741                         le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3742                         return 1;
3743
3744                 if (f2fs_sb_has_readonly(sbi))
3745                         goto skip_cross;
3746
3747                 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3748                         if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3749                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3750                                 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3751                                          i, j,
3752                                          le32_to_cpu(ckpt->cur_data_segno[i]));
3753                                 return 1;
3754                         }
3755                 }
3756         }
3757         for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3758                 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3759                         if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3760                                 le32_to_cpu(ckpt->cur_data_segno[j])) {
3761                                 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3762                                          i, j,
3763                                          le32_to_cpu(ckpt->cur_node_segno[i]));
3764                                 return 1;
3765                         }
3766                 }
3767         }
3768 skip_cross:
3769         sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3770         nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3771
3772         if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3773                 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3774                 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3775                          sit_bitmap_size, nat_bitmap_size);
3776                 return 1;
3777         }
3778
3779         cp_pack_start_sum = __start_sum_addr(sbi);
3780         cp_payload = __cp_payload(sbi);
3781         if (cp_pack_start_sum < cp_payload + 1 ||
3782                 cp_pack_start_sum > blocks_per_seg - 1 -
3783                         NR_CURSEG_PERSIST_TYPE) {
3784                 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3785                          cp_pack_start_sum);
3786                 return 1;
3787         }
3788
3789         if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3790                 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3791                 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3792                           "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3793                           "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3794                           le32_to_cpu(ckpt->checksum_offset));
3795                 return 1;
3796         }
3797
3798         nat_blocks = nat_segs << log_blocks_per_seg;
3799         nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3800         nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3801         if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3802                 (cp_payload + F2FS_CP_PACKS +
3803                 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3804                 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3805                           cp_payload, nat_bits_blocks);
3806                 return 1;
3807         }
3808
3809         if (unlikely(f2fs_cp_error(sbi))) {
3810                 f2fs_err(sbi, "A bug case: need to run fsck");
3811                 return 1;
3812         }
3813         return 0;
3814 }
3815
3816 static void init_sb_info(struct f2fs_sb_info *sbi)
3817 {
3818         struct f2fs_super_block *raw_super = sbi->raw_super;
3819         int i;
3820
3821         sbi->log_sectors_per_block =
3822                 le32_to_cpu(raw_super->log_sectors_per_block);
3823         sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3824         sbi->blocksize = BIT(sbi->log_blocksize);
3825         sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3826         sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg);
3827         sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3828         sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3829         sbi->total_sections = le32_to_cpu(raw_super->section_count);
3830         sbi->total_node_count = SEGS_TO_BLKS(sbi,
3831                         ((le32_to_cpu(raw_super->segment_count_nat) / 2) *
3832                         NAT_ENTRY_PER_BLOCK));
3833         F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino);
3834         F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino);
3835         F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino);
3836         sbi->cur_victim_sec = NULL_SECNO;
3837         sbi->gc_mode = GC_NORMAL;
3838         sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3839         sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3840         sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3841         sbi->migration_granularity = SEGS_PER_SEC(sbi);
3842         sbi->migration_window_granularity = f2fs_sb_has_blkzoned(sbi) ?
3843                 DEF_MIGRATION_WINDOW_GRANULARITY_ZONED : SEGS_PER_SEC(sbi);
3844         sbi->seq_file_ra_mul = MIN_RA_MUL;
3845         sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
3846         sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
3847         spin_lock_init(&sbi->gc_remaining_trials_lock);
3848         atomic64_set(&sbi->current_atomic_write, 0);
3849
3850         sbi->dir_level = DEF_DIR_LEVEL;
3851         sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3852         sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3853         sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3854         sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3855         sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3856         sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3857                                 DEF_UMOUNT_DISCARD_TIMEOUT;
3858         clear_sbi_flag(sbi, SBI_NEED_FSCK);
3859
3860         for (i = 0; i < NR_COUNT_TYPE; i++)
3861                 atomic_set(&sbi->nr_pages[i], 0);
3862
3863         for (i = 0; i < META; i++)
3864                 atomic_set(&sbi->wb_sync_req[i], 0);
3865
3866         INIT_LIST_HEAD(&sbi->s_list);
3867         mutex_init(&sbi->umount_mutex);
3868         init_f2fs_rwsem(&sbi->io_order_lock);
3869         spin_lock_init(&sbi->cp_lock);
3870
3871         sbi->dirty_device = 0;
3872         spin_lock_init(&sbi->dev_lock);
3873
3874         init_f2fs_rwsem(&sbi->sb_lock);
3875         init_f2fs_rwsem(&sbi->pin_sem);
3876 }
3877
3878 static int init_percpu_info(struct f2fs_sb_info *sbi)
3879 {
3880         int err;
3881
3882         err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3883         if (err)
3884                 return err;
3885
3886         err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL);
3887         if (err)
3888                 goto err_valid_block;
3889
3890         err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3891                                                                 GFP_KERNEL);
3892         if (err)
3893                 goto err_node_block;
3894         return 0;
3895
3896 err_node_block:
3897         percpu_counter_destroy(&sbi->rf_node_block_count);
3898 err_valid_block:
3899         percpu_counter_destroy(&sbi->alloc_valid_block_count);
3900         return err;
3901 }
3902
3903 #ifdef CONFIG_BLK_DEV_ZONED
3904
3905 struct f2fs_report_zones_args {
3906         struct f2fs_sb_info *sbi;
3907         struct f2fs_dev_info *dev;
3908 };
3909
3910 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3911                               void *data)
3912 {
3913         struct f2fs_report_zones_args *rz_args = data;
3914         block_t unusable_blocks = (zone->len - zone->capacity) >>
3915                                         F2FS_LOG_SECTORS_PER_BLOCK;
3916
3917         if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3918                 return 0;
3919
3920         set_bit(idx, rz_args->dev->blkz_seq);
3921         if (!rz_args->sbi->unusable_blocks_per_sec) {
3922                 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3923                 return 0;
3924         }
3925         if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3926                 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3927                 return -EINVAL;
3928         }
3929         return 0;
3930 }
3931
3932 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3933 {
3934         struct block_device *bdev = FDEV(devi).bdev;
3935         sector_t nr_sectors = bdev_nr_sectors(bdev);
3936         struct f2fs_report_zones_args rep_zone_arg;
3937         u64 zone_sectors;
3938         unsigned int max_open_zones;
3939         int ret;
3940
3941         if (!f2fs_sb_has_blkzoned(sbi))
3942                 return 0;
3943
3944         if (bdev_is_zoned(FDEV(devi).bdev)) {
3945                 max_open_zones = bdev_max_open_zones(bdev);
3946                 if (max_open_zones && (max_open_zones < sbi->max_open_zones))
3947                         sbi->max_open_zones = max_open_zones;
3948                 if (sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) {
3949                         f2fs_err(sbi,
3950                                 "zoned: max open zones %u is too small, need at least %u open zones",
3951                                 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs);
3952                         return -EINVAL;
3953                 }
3954         }
3955
3956         zone_sectors = bdev_zone_sectors(bdev);
3957         if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3958                                 SECTOR_TO_BLOCK(zone_sectors))
3959                 return -EINVAL;
3960         sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors);
3961         FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors),
3962                                         sbi->blocks_per_blkz);
3963         if (nr_sectors & (zone_sectors - 1))
3964                 FDEV(devi).nr_blkz++;
3965
3966         FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3967                                         BITS_TO_LONGS(FDEV(devi).nr_blkz)
3968                                         * sizeof(unsigned long),
3969                                         GFP_KERNEL);
3970         if (!FDEV(devi).blkz_seq)
3971                 return -ENOMEM;
3972
3973         rep_zone_arg.sbi = sbi;
3974         rep_zone_arg.dev = &FDEV(devi);
3975
3976         ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3977                                   &rep_zone_arg);
3978         if (ret < 0)
3979                 return ret;
3980         return 0;
3981 }
3982 #endif
3983
3984 /*
3985  * Read f2fs raw super block.
3986  * Because we have two copies of super block, so read both of them
3987  * to get the first valid one. If any one of them is broken, we pass
3988  * them recovery flag back to the caller.
3989  */
3990 static int read_raw_super_block(struct f2fs_sb_info *sbi,
3991                         struct f2fs_super_block **raw_super,
3992                         int *valid_super_block, int *recovery)
3993 {
3994         struct super_block *sb = sbi->sb;
3995         int block;
3996         struct folio *folio;
3997         struct f2fs_super_block *super;
3998         int err = 0;
3999
4000         super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
4001         if (!super)
4002                 return -ENOMEM;
4003
4004         for (block = 0; block < 2; block++) {
4005                 folio = read_mapping_folio(sb->s_bdev->bd_mapping, block, NULL);
4006                 if (IS_ERR(folio)) {
4007                         f2fs_err(sbi, "Unable to read %dth superblock",
4008                                  block + 1);
4009                         err = PTR_ERR(folio);
4010                         *recovery = 1;
4011                         continue;
4012                 }
4013
4014                 /* sanity checking of raw super */
4015                 err = sanity_check_raw_super(sbi, folio, block);
4016                 if (err) {
4017                         f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
4018                                  block + 1);
4019                         folio_put(folio);
4020                         *recovery = 1;
4021                         continue;
4022                 }
4023
4024                 if (!*raw_super) {
4025                         memcpy(super, F2FS_SUPER_BLOCK(folio, block),
4026                                                         sizeof(*super));
4027                         *valid_super_block = block;
4028                         *raw_super = super;
4029                 }
4030                 folio_put(folio);
4031         }
4032
4033         /* No valid superblock */
4034         if (!*raw_super)
4035                 kfree(super);
4036         else
4037                 err = 0;
4038
4039         return err;
4040 }
4041
4042 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
4043 {
4044         struct folio *folio;
4045         pgoff_t index;
4046         __u32 crc = 0;
4047         int err;
4048
4049         if ((recover && f2fs_readonly(sbi->sb)) ||
4050                                 f2fs_hw_is_readonly(sbi)) {
4051                 set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
4052                 return -EROFS;
4053         }
4054
4055         /* we should update superblock crc here */
4056         if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
4057                 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
4058                                 offsetof(struct f2fs_super_block, crc));
4059                 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
4060         }
4061
4062         /* write back-up superblock first */
4063         index = sbi->valid_super_block ? 0 : 1;
4064         folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL);
4065         if (IS_ERR(folio))
4066                 return PTR_ERR(folio);
4067         err = __f2fs_commit_super(sbi, folio, index, true);
4068         folio_put(folio);
4069
4070         /* if we are in recovery path, skip writing valid superblock */
4071         if (recover || err)
4072                 return err;
4073
4074         /* write current valid superblock */
4075         index = sbi->valid_super_block;
4076         folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL);
4077         if (IS_ERR(folio))
4078                 return PTR_ERR(folio);
4079         err = __f2fs_commit_super(sbi, folio, index, true);
4080         folio_put(folio);
4081         return err;
4082 }
4083
4084 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason)
4085 {
4086         unsigned long flags;
4087
4088         spin_lock_irqsave(&sbi->error_lock, flags);
4089         if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0))
4090                 sbi->stop_reason[reason]++;
4091         spin_unlock_irqrestore(&sbi->error_lock, flags);
4092 }
4093
4094 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi)
4095 {
4096         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4097         unsigned long flags;
4098         int err;
4099
4100         f2fs_down_write(&sbi->sb_lock);
4101
4102         spin_lock_irqsave(&sbi->error_lock, flags);
4103         if (sbi->error_dirty) {
4104                 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4105                                                         MAX_F2FS_ERRORS);
4106                 sbi->error_dirty = false;
4107         }
4108         memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON);
4109         spin_unlock_irqrestore(&sbi->error_lock, flags);
4110
4111         err = f2fs_commit_super(sbi, false);
4112
4113         f2fs_up_write(&sbi->sb_lock);
4114         if (err)
4115                 f2fs_err_ratelimited(sbi,
4116                         "f2fs_commit_super fails to record stop_reason, err:%d",
4117                         err);
4118 }
4119
4120 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag)
4121 {
4122         unsigned long flags;
4123
4124         spin_lock_irqsave(&sbi->error_lock, flags);
4125         if (!test_bit(flag, (unsigned long *)sbi->errors)) {
4126                 set_bit(flag, (unsigned long *)sbi->errors);
4127                 sbi->error_dirty = true;
4128         }
4129         spin_unlock_irqrestore(&sbi->error_lock, flags);
4130 }
4131
4132 static bool f2fs_update_errors(struct f2fs_sb_info *sbi)
4133 {
4134         unsigned long flags;
4135         bool need_update = false;
4136
4137         spin_lock_irqsave(&sbi->error_lock, flags);
4138         if (sbi->error_dirty) {
4139                 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors,
4140                                                         MAX_F2FS_ERRORS);
4141                 sbi->error_dirty = false;
4142                 need_update = true;
4143         }
4144         spin_unlock_irqrestore(&sbi->error_lock, flags);
4145
4146         return need_update;
4147 }
4148
4149 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error)
4150 {
4151         int err;
4152
4153         f2fs_down_write(&sbi->sb_lock);
4154
4155         if (!f2fs_update_errors(sbi))
4156                 goto out_unlock;
4157
4158         err = f2fs_commit_super(sbi, false);
4159         if (err)
4160                 f2fs_err_ratelimited(sbi,
4161                         "f2fs_commit_super fails to record errors:%u, err:%d",
4162                         error, err);
4163 out_unlock:
4164         f2fs_up_write(&sbi->sb_lock);
4165 }
4166
4167 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error)
4168 {
4169         f2fs_save_errors(sbi, error);
4170         f2fs_record_errors(sbi, error);
4171 }
4172
4173 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error)
4174 {
4175         f2fs_save_errors(sbi, error);
4176
4177         if (!sbi->error_dirty)
4178                 return;
4179         if (!test_bit(error, (unsigned long *)sbi->errors))
4180                 return;
4181         schedule_work(&sbi->s_error_work);
4182 }
4183
4184 static bool system_going_down(void)
4185 {
4186         return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
4187                 || system_state == SYSTEM_RESTART;
4188 }
4189
4190 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason)
4191 {
4192         struct super_block *sb = sbi->sb;
4193         bool shutdown = reason == STOP_CP_REASON_SHUTDOWN;
4194         bool continue_fs = !shutdown &&
4195                         F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE;
4196
4197         set_ckpt_flags(sbi, CP_ERROR_FLAG);
4198
4199         if (!f2fs_hw_is_readonly(sbi)) {
4200                 save_stop_reason(sbi, reason);
4201
4202                 /*
4203                  * always create an asynchronous task to record stop_reason
4204                  * in order to avoid potential deadlock when running into
4205                  * f2fs_record_stop_reason() synchronously.
4206                  */
4207                 schedule_work(&sbi->s_error_work);
4208         }
4209
4210         /*
4211          * We force ERRORS_RO behavior when system is rebooting. Otherwise we
4212          * could panic during 'reboot -f' as the underlying device got already
4213          * disabled.
4214          */
4215         if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC &&
4216                                 !shutdown && !system_going_down() &&
4217                                 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN))
4218                 panic("F2FS-fs (device %s): panic forced after error\n",
4219                                                         sb->s_id);
4220
4221         if (shutdown)
4222                 set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
4223
4224         /*
4225          * Continue filesystem operators if errors=continue. Should not set
4226          * RO by shutdown, since RO bypasses thaw_super which can hang the
4227          * system.
4228          */
4229         if (continue_fs || f2fs_readonly(sb) || shutdown) {
4230                 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason);
4231                 return;
4232         }
4233
4234         f2fs_warn(sbi, "Remounting filesystem read-only");
4235
4236         /*
4237          * We have already set CP_ERROR_FLAG flag to stop all updates
4238          * to filesystem, so it doesn't need to set SB_RDONLY flag here
4239          * because the flag should be set covered w/ sb->s_umount semaphore
4240          * via remount procedure, otherwise, it will confuse code like
4241          * freeze_super() which will lead to deadlocks and other problems.
4242          */
4243 }
4244
4245 static void f2fs_record_error_work(struct work_struct *work)
4246 {
4247         struct f2fs_sb_info *sbi = container_of(work,
4248                                         struct f2fs_sb_info, s_error_work);
4249
4250         f2fs_record_stop_reason(sbi);
4251 }
4252
4253 static inline unsigned int get_first_zoned_segno(struct f2fs_sb_info *sbi)
4254 {
4255         int devi;
4256
4257         for (devi = 0; devi < sbi->s_ndevs; devi++)
4258                 if (bdev_is_zoned(FDEV(devi).bdev))
4259                         return GET_SEGNO(sbi, FDEV(devi).start_blk);
4260         return 0;
4261 }
4262
4263 static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
4264 {
4265         struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
4266         unsigned int max_devices = MAX_DEVICES;
4267         unsigned int logical_blksize;
4268         blk_mode_t mode = sb_open_mode(sbi->sb->s_flags);
4269         int i;
4270
4271         /* Initialize single device information */
4272         if (!RDEV(0).path[0]) {
4273                 if (!bdev_is_zoned(sbi->sb->s_bdev))
4274                         return 0;
4275                 max_devices = 1;
4276         }
4277
4278         /*
4279          * Initialize multiple devices information, or single
4280          * zoned block device information.
4281          */
4282         sbi->devs = f2fs_kzalloc(sbi,
4283                                  array_size(max_devices,
4284                                             sizeof(struct f2fs_dev_info)),
4285                                  GFP_KERNEL);
4286         if (!sbi->devs)
4287                 return -ENOMEM;
4288
4289         logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev);
4290         sbi->aligned_blksize = true;
4291 #ifdef CONFIG_BLK_DEV_ZONED
4292         sbi->max_open_zones = UINT_MAX;
4293         sbi->blkzone_alloc_policy = BLKZONE_ALLOC_PRIOR_SEQ;
4294 #endif
4295
4296         for (i = 0; i < max_devices; i++) {
4297                 if (i == 0)
4298                         FDEV(0).bdev_file = sbi->sb->s_bdev_file;
4299                 else if (!RDEV(i).path[0])
4300                         break;
4301
4302                 if (max_devices > 1) {
4303                         /* Multi-device mount */
4304                         memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
4305                         FDEV(i).total_segments =
4306                                 le32_to_cpu(RDEV(i).total_segments);
4307                         if (i == 0) {
4308                                 FDEV(i).start_blk = 0;
4309                                 FDEV(i).end_blk = FDEV(i).start_blk +
4310                                         SEGS_TO_BLKS(sbi,
4311                                         FDEV(i).total_segments) - 1 +
4312                                         le32_to_cpu(raw_super->segment0_blkaddr);
4313                         } else {
4314                                 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
4315                                 FDEV(i).end_blk = FDEV(i).start_blk +
4316                                                 SEGS_TO_BLKS(sbi,
4317                                                 FDEV(i).total_segments) - 1;
4318                                 FDEV(i).bdev_file = bdev_file_open_by_path(
4319                                         FDEV(i).path, mode, sbi->sb, NULL);
4320                         }
4321                 }
4322                 if (IS_ERR(FDEV(i).bdev_file))
4323                         return PTR_ERR(FDEV(i).bdev_file);
4324
4325                 FDEV(i).bdev = file_bdev(FDEV(i).bdev_file);
4326                 /* to release errored devices */
4327                 sbi->s_ndevs = i + 1;
4328
4329                 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev))
4330                         sbi->aligned_blksize = false;
4331
4332 #ifdef CONFIG_BLK_DEV_ZONED
4333                 if (bdev_is_zoned(FDEV(i).bdev)) {
4334                         if (!f2fs_sb_has_blkzoned(sbi)) {
4335                                 f2fs_err(sbi, "Zoned block device feature not enabled");
4336                                 return -EINVAL;
4337                         }
4338                         if (init_blkz_info(sbi, i)) {
4339                                 f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
4340                                 return -EINVAL;
4341                         }
4342                         if (max_devices == 1)
4343                                 break;
4344                         f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)",
4345                                   i, FDEV(i).path,
4346                                   FDEV(i).total_segments,
4347                                   FDEV(i).start_blk, FDEV(i).end_blk);
4348                         continue;
4349                 }
4350 #endif
4351                 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
4352                           i, FDEV(i).path,
4353                           FDEV(i).total_segments,
4354                           FDEV(i).start_blk, FDEV(i).end_blk);
4355         }
4356         return 0;
4357 }
4358
4359 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
4360 {
4361 #if IS_ENABLED(CONFIG_UNICODE)
4362         if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
4363                 const struct f2fs_sb_encodings *encoding_info;
4364                 struct unicode_map *encoding;
4365                 __u16 encoding_flags;
4366
4367                 encoding_info = f2fs_sb_read_encoding(sbi->raw_super);
4368                 if (!encoding_info) {
4369                         f2fs_err(sbi,
4370                                  "Encoding requested by superblock is unknown");
4371                         return -EINVAL;
4372                 }
4373
4374                 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags);
4375                 encoding = utf8_load(encoding_info->version);
4376                 if (IS_ERR(encoding)) {
4377                         f2fs_err(sbi,
4378                                  "can't mount with superblock charset: %s-%u.%u.%u "
4379                                  "not supported by the kernel. flags: 0x%x.",
4380                                  encoding_info->name,
4381                                  unicode_major(encoding_info->version),
4382                                  unicode_minor(encoding_info->version),
4383                                  unicode_rev(encoding_info->version),
4384                                  encoding_flags);
4385                         return PTR_ERR(encoding);
4386                 }
4387                 f2fs_info(sbi, "Using encoding defined by superblock: "
4388                          "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4389                          unicode_major(encoding_info->version),
4390                          unicode_minor(encoding_info->version),
4391                          unicode_rev(encoding_info->version),
4392                          encoding_flags);
4393
4394                 sbi->sb->s_encoding = encoding;
4395                 sbi->sb->s_encoding_flags = encoding_flags;
4396         }
4397 #else
4398         if (f2fs_sb_has_casefold(sbi)) {
4399                 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
4400                 return -EINVAL;
4401         }
4402 #endif
4403         return 0;
4404 }
4405
4406 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
4407 {
4408         /* adjust parameters according to the volume size */
4409         if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) {
4410                 if (f2fs_block_unit_discard(sbi))
4411                         SM_I(sbi)->dcc_info->discard_granularity =
4412                                                 MIN_DISCARD_GRANULARITY;
4413                 if (!f2fs_lfs_mode(sbi))
4414                         SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) |
4415                                                 BIT(F2FS_IPU_HONOR_OPU_WRITE);
4416         }
4417
4418         sbi->readdir_ra = true;
4419 }
4420
4421 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
4422 {
4423         struct f2fs_sb_info *sbi;
4424         struct f2fs_super_block *raw_super;
4425         struct inode *root;
4426         int err;
4427         bool skip_recovery = false, need_fsck = false;
4428         char *options = NULL;
4429         int recovery, i, valid_super_block;
4430         struct curseg_info *seg_i;
4431         int retry_cnt = 1;
4432 #ifdef CONFIG_QUOTA
4433         bool quota_enabled = false;
4434 #endif
4435
4436 try_onemore:
4437         err = -EINVAL;
4438         raw_super = NULL;
4439         valid_super_block = -1;
4440         recovery = 0;
4441
4442         /* allocate memory for f2fs-specific super block info */
4443         sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
4444         if (!sbi)
4445                 return -ENOMEM;
4446
4447         sbi->sb = sb;
4448
4449         /* initialize locks within allocated memory */
4450         init_f2fs_rwsem(&sbi->gc_lock);
4451         mutex_init(&sbi->writepages);
4452         init_f2fs_rwsem(&sbi->cp_global_sem);
4453         init_f2fs_rwsem(&sbi->node_write);
4454         init_f2fs_rwsem(&sbi->node_change);
4455         spin_lock_init(&sbi->stat_lock);
4456         init_f2fs_rwsem(&sbi->cp_rwsem);
4457         init_f2fs_rwsem(&sbi->quota_sem);
4458         init_waitqueue_head(&sbi->cp_wait);
4459         spin_lock_init(&sbi->error_lock);
4460
4461         for (i = 0; i < NR_INODE_TYPE; i++) {
4462                 INIT_LIST_HEAD(&sbi->inode_list[i]);
4463                 spin_lock_init(&sbi->inode_lock[i]);
4464         }
4465         mutex_init(&sbi->flush_lock);
4466
4467         /* set a block size */
4468         if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
4469                 f2fs_err(sbi, "unable to set blocksize");
4470                 goto free_sbi;
4471         }
4472
4473         err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
4474                                                                 &recovery);
4475         if (err)
4476                 goto free_sbi;
4477
4478         sb->s_fs_info = sbi;
4479         sbi->raw_super = raw_super;
4480
4481         INIT_WORK(&sbi->s_error_work, f2fs_record_error_work);
4482         memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS);
4483         memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON);
4484
4485         /* precompute checksum seed for metadata */
4486         if (f2fs_sb_has_inode_chksum(sbi))
4487                 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
4488                                                 sizeof(raw_super->uuid));
4489
4490         default_options(sbi, false);
4491         /* parse mount options */
4492         options = kstrdup((const char *)data, GFP_KERNEL);
4493         if (data && !options) {
4494                 err = -ENOMEM;
4495                 goto free_sb_buf;
4496         }
4497
4498         err = parse_options(sb, options, false);
4499         if (err)
4500                 goto free_options;
4501
4502         sb->s_maxbytes = max_file_blocks(NULL) <<
4503                                 le32_to_cpu(raw_super->log_blocksize);
4504         sb->s_max_links = F2FS_LINK_MAX;
4505
4506         err = f2fs_setup_casefold(sbi);
4507         if (err)
4508                 goto free_options;
4509
4510 #ifdef CONFIG_QUOTA
4511         sb->dq_op = &f2fs_quota_operations;
4512         sb->s_qcop = &f2fs_quotactl_ops;
4513         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4514
4515         if (f2fs_sb_has_quota_ino(sbi)) {
4516                 for (i = 0; i < MAXQUOTAS; i++) {
4517                         if (f2fs_qf_ino(sbi->sb, i))
4518                                 sbi->nquota_files++;
4519                 }
4520         }
4521 #endif
4522
4523         sb->s_op = &f2fs_sops;
4524 #ifdef CONFIG_FS_ENCRYPTION
4525         sb->s_cop = &f2fs_cryptops;
4526 #endif
4527 #ifdef CONFIG_FS_VERITY
4528         sb->s_vop = &f2fs_verityops;
4529 #endif
4530         sb->s_xattr = f2fs_xattr_handlers;
4531         sb->s_export_op = &f2fs_export_ops;
4532         sb->s_magic = F2FS_SUPER_MAGIC;
4533         sb->s_time_gran = 1;
4534         sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4535                 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
4536         super_set_uuid(sb, (void *) raw_super->uuid, sizeof(raw_super->uuid));
4537         super_set_sysfs_name_bdev(sb);
4538         sb->s_iflags |= SB_I_CGROUPWB;
4539
4540         /* init f2fs-specific super block info */
4541         sbi->valid_super_block = valid_super_block;
4542
4543         /* disallow all the data/node/meta page writes */
4544         set_sbi_flag(sbi, SBI_POR_DOING);
4545
4546         err = f2fs_init_write_merge_io(sbi);
4547         if (err)
4548                 goto free_bio_info;
4549
4550         init_sb_info(sbi);
4551
4552         err = f2fs_init_iostat(sbi);
4553         if (err)
4554                 goto free_bio_info;
4555
4556         err = init_percpu_info(sbi);
4557         if (err)
4558                 goto free_iostat;
4559
4560         /* init per sbi slab cache */
4561         err = f2fs_init_xattr_caches(sbi);
4562         if (err)
4563                 goto free_percpu;
4564         err = f2fs_init_page_array_cache(sbi);
4565         if (err)
4566                 goto free_xattr_cache;
4567
4568         /* get an inode for meta space */
4569         sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
4570         if (IS_ERR(sbi->meta_inode)) {
4571                 f2fs_err(sbi, "Failed to read F2FS meta data inode");
4572                 err = PTR_ERR(sbi->meta_inode);
4573                 goto free_page_array_cache;
4574         }
4575
4576         err = f2fs_get_valid_checkpoint(sbi);
4577         if (err) {
4578                 f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
4579                 goto free_meta_inode;
4580         }
4581
4582         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
4583                 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
4584         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
4585                 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4586                 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
4587         }
4588
4589         if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
4590                 set_sbi_flag(sbi, SBI_NEED_FSCK);
4591
4592         /* Initialize device list */
4593         err = f2fs_scan_devices(sbi);
4594         if (err) {
4595                 f2fs_err(sbi, "Failed to find devices");
4596                 goto free_devices;
4597         }
4598
4599         err = f2fs_init_post_read_wq(sbi);
4600         if (err) {
4601                 f2fs_err(sbi, "Failed to initialize post read workqueue");
4602                 goto free_devices;
4603         }
4604
4605         sbi->total_valid_node_count =
4606                                 le32_to_cpu(sbi->ckpt->valid_node_count);
4607         percpu_counter_set(&sbi->total_valid_inode_count,
4608                                 le32_to_cpu(sbi->ckpt->valid_inode_count));
4609         sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
4610         sbi->total_valid_block_count =
4611                                 le64_to_cpu(sbi->ckpt->valid_block_count);
4612         sbi->last_valid_block_count = sbi->total_valid_block_count;
4613         sbi->reserved_blocks = 0;
4614         sbi->current_reserved_blocks = 0;
4615         limit_reserve_root(sbi);
4616         adjust_unusable_cap_perc(sbi);
4617
4618         f2fs_init_extent_cache_info(sbi);
4619
4620         f2fs_init_ino_entry_info(sbi);
4621
4622         f2fs_init_fsync_node_info(sbi);
4623
4624         /* setup checkpoint request control and start checkpoint issue thread */
4625         f2fs_init_ckpt_req_control(sbi);
4626         if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) &&
4627                         test_opt(sbi, MERGE_CHECKPOINT)) {
4628                 err = f2fs_start_ckpt_thread(sbi);
4629                 if (err) {
4630                         f2fs_err(sbi,
4631                             "Failed to start F2FS issue_checkpoint_thread (%d)",
4632                             err);
4633                         goto stop_ckpt_thread;
4634                 }
4635         }
4636
4637         /* setup f2fs internal modules */
4638         err = f2fs_build_segment_manager(sbi);
4639         if (err) {
4640                 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
4641                          err);
4642                 goto free_sm;
4643         }
4644         err = f2fs_build_node_manager(sbi);
4645         if (err) {
4646                 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
4647                          err);
4648                 goto free_nm;
4649         }
4650
4651         /* For write statistics */
4652         sbi->sectors_written_start = f2fs_get_sectors_written(sbi);
4653
4654         /* get segno of first zoned block device */
4655         sbi->first_zoned_segno = get_first_zoned_segno(sbi);
4656
4657         /* Read accumulated write IO statistics if exists */
4658         seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
4659         if (__exist_node_summaries(sbi))
4660                 sbi->kbytes_written =
4661                         le64_to_cpu(seg_i->journal->info.kbytes_written);
4662
4663         f2fs_build_gc_manager(sbi);
4664
4665         err = f2fs_build_stats(sbi);
4666         if (err)
4667                 goto free_nm;
4668
4669         /* get an inode for node space */
4670         sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
4671         if (IS_ERR(sbi->node_inode)) {
4672                 f2fs_err(sbi, "Failed to read node inode");
4673                 err = PTR_ERR(sbi->node_inode);
4674                 goto free_stats;
4675         }
4676
4677         /* read root inode and dentry */
4678         root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
4679         if (IS_ERR(root)) {
4680                 f2fs_err(sbi, "Failed to read root inode");
4681                 err = PTR_ERR(root);
4682                 goto free_node_inode;
4683         }
4684         if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
4685                         !root->i_size || !root->i_nlink) {
4686                 iput(root);
4687                 err = -EINVAL;
4688                 goto free_node_inode;
4689         }
4690
4691         generic_set_sb_d_ops(sb);
4692         sb->s_root = d_make_root(root); /* allocate root dentry */
4693         if (!sb->s_root) {
4694                 err = -ENOMEM;
4695                 goto free_node_inode;
4696         }
4697
4698         err = f2fs_init_compress_inode(sbi);
4699         if (err)
4700                 goto free_root_inode;
4701
4702         err = f2fs_register_sysfs(sbi);
4703         if (err)
4704                 goto free_compress_inode;
4705
4706 #ifdef CONFIG_QUOTA
4707         /* Enable quota usage during mount */
4708         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
4709                 err = f2fs_enable_quotas(sb);
4710                 if (err)
4711                         f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
4712         }
4713
4714         quota_enabled = f2fs_recover_quota_begin(sbi);
4715 #endif
4716         /* if there are any orphan inodes, free them */
4717         err = f2fs_recover_orphan_inodes(sbi);
4718         if (err)
4719                 goto free_meta;
4720
4721         if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
4722                 goto reset_checkpoint;
4723
4724         /* recover fsynced data */
4725         if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
4726                         !test_opt(sbi, NORECOVERY)) {
4727                 /*
4728                  * mount should be failed, when device has readonly mode, and
4729                  * previous checkpoint was not done by clean system shutdown.
4730                  */
4731                 if (f2fs_hw_is_readonly(sbi)) {
4732                         if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4733                                 err = f2fs_recover_fsync_data(sbi, true);
4734                                 if (err > 0) {
4735                                         err = -EROFS;
4736                                         f2fs_err(sbi, "Need to recover fsync data, but "
4737                                                 "write access unavailable, please try "
4738                                                 "mount w/ disable_roll_forward or norecovery");
4739                                 }
4740                                 if (err < 0)
4741                                         goto free_meta;
4742                         }
4743                         f2fs_info(sbi, "write access unavailable, skipping recovery");
4744                         goto reset_checkpoint;
4745                 }
4746
4747                 if (need_fsck)
4748                         set_sbi_flag(sbi, SBI_NEED_FSCK);
4749
4750                 if (skip_recovery)
4751                         goto reset_checkpoint;
4752
4753                 err = f2fs_recover_fsync_data(sbi, false);
4754                 if (err < 0) {
4755                         if (err != -ENOMEM)
4756                                 skip_recovery = true;
4757                         need_fsck = true;
4758                         f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
4759                                  err);
4760                         goto free_meta;
4761                 }
4762         } else {
4763                 err = f2fs_recover_fsync_data(sbi, true);
4764
4765                 if (!f2fs_readonly(sb) && err > 0) {
4766                         err = -EINVAL;
4767                         f2fs_err(sbi, "Need to recover fsync data");
4768                         goto free_meta;
4769                 }
4770         }
4771
4772 #ifdef CONFIG_QUOTA
4773         f2fs_recover_quota_end(sbi, quota_enabled);
4774 #endif
4775 reset_checkpoint:
4776         /*
4777          * If the f2fs is not readonly and fsync data recovery succeeds,
4778          * write pointer consistency of cursegs and other zones are already
4779          * checked and fixed during recovery. However, if recovery fails,
4780          * write pointers are left untouched, and retry-mount should check
4781          * them here.
4782          */
4783         if (skip_recovery)
4784                 err = f2fs_check_and_fix_write_pointer(sbi);
4785         if (err)
4786                 goto free_meta;
4787
4788         /* f2fs_recover_fsync_data() cleared this already */
4789         clear_sbi_flag(sbi, SBI_POR_DOING);
4790
4791         err = f2fs_init_inmem_curseg(sbi);
4792         if (err)
4793                 goto sync_free_meta;
4794
4795         if (test_opt(sbi, DISABLE_CHECKPOINT)) {
4796                 err = f2fs_disable_checkpoint(sbi);
4797                 if (err)
4798                         goto sync_free_meta;
4799         } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
4800                 f2fs_enable_checkpoint(sbi);
4801         }
4802
4803         /*
4804          * If filesystem is not mounted as read-only then
4805          * do start the gc_thread.
4806          */
4807         if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
4808                 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
4809                 /* After POR, we can run background GC thread.*/
4810                 err = f2fs_start_gc_thread(sbi);
4811                 if (err)
4812                         goto sync_free_meta;
4813         }
4814         kvfree(options);
4815
4816         /* recover broken superblock */
4817         if (recovery) {
4818                 err = f2fs_commit_super(sbi, true);
4819                 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
4820                           sbi->valid_super_block ? 1 : 2, err);
4821         }
4822
4823         f2fs_join_shrinker(sbi);
4824
4825         f2fs_tuning_parameters(sbi);
4826
4827         f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
4828                     cur_cp_version(F2FS_CKPT(sbi)));
4829         f2fs_update_time(sbi, CP_TIME);
4830         f2fs_update_time(sbi, REQ_TIME);
4831         clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
4832         return 0;
4833
4834 sync_free_meta:
4835         /* safe to flush all the data */
4836         sync_filesystem(sbi->sb);
4837         retry_cnt = 0;
4838
4839 free_meta:
4840 #ifdef CONFIG_QUOTA
4841         f2fs_truncate_quota_inode_pages(sb);
4842         if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4843                 f2fs_quota_off_umount(sbi->sb);
4844 #endif
4845         /*
4846          * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4847          * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4848          * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4849          * falls into an infinite loop in f2fs_sync_meta_pages().
4850          */
4851         truncate_inode_pages_final(META_MAPPING(sbi));
4852         /* evict some inodes being cached by GC */
4853         evict_inodes(sb);
4854         f2fs_unregister_sysfs(sbi);
4855 free_compress_inode:
4856         f2fs_destroy_compress_inode(sbi);
4857 free_root_inode:
4858         dput(sb->s_root);
4859         sb->s_root = NULL;
4860 free_node_inode:
4861         f2fs_release_ino_entry(sbi, true);
4862         truncate_inode_pages_final(NODE_MAPPING(sbi));
4863         iput(sbi->node_inode);
4864         sbi->node_inode = NULL;
4865 free_stats:
4866         f2fs_destroy_stats(sbi);
4867 free_nm:
4868         /* stop discard thread before destroying node manager */
4869         f2fs_stop_discard_thread(sbi);
4870         f2fs_destroy_node_manager(sbi);
4871 free_sm:
4872         f2fs_destroy_segment_manager(sbi);
4873 stop_ckpt_thread:
4874         f2fs_stop_ckpt_thread(sbi);
4875         /* flush s_error_work before sbi destroy */
4876         flush_work(&sbi->s_error_work);
4877         f2fs_destroy_post_read_wq(sbi);
4878 free_devices:
4879         destroy_device_list(sbi);
4880         kvfree(sbi->ckpt);
4881 free_meta_inode:
4882         make_bad_inode(sbi->meta_inode);
4883         iput(sbi->meta_inode);
4884         sbi->meta_inode = NULL;
4885 free_page_array_cache:
4886         f2fs_destroy_page_array_cache(sbi);
4887 free_xattr_cache:
4888         f2fs_destroy_xattr_caches(sbi);
4889 free_percpu:
4890         destroy_percpu_info(sbi);
4891 free_iostat:
4892         f2fs_destroy_iostat(sbi);
4893 free_bio_info:
4894         for (i = 0; i < NR_PAGE_TYPE; i++)
4895                 kvfree(sbi->write_io[i]);
4896
4897 #if IS_ENABLED(CONFIG_UNICODE)
4898         utf8_unload(sb->s_encoding);
4899         sb->s_encoding = NULL;
4900 #endif
4901 free_options:
4902 #ifdef CONFIG_QUOTA
4903         for (i = 0; i < MAXQUOTAS; i++)
4904                 kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4905 #endif
4906         fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4907         kvfree(options);
4908 free_sb_buf:
4909         kfree(raw_super);
4910 free_sbi:
4911         kfree(sbi);
4912         sb->s_fs_info = NULL;
4913
4914         /* give only one another chance */
4915         if (retry_cnt > 0 && skip_recovery) {
4916                 retry_cnt--;
4917                 shrink_dcache_sb(sb);
4918                 goto try_onemore;
4919         }
4920         return err;
4921 }
4922
4923 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4924                         const char *dev_name, void *data)
4925 {
4926         return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4927 }
4928
4929 static void kill_f2fs_super(struct super_block *sb)
4930 {
4931         struct f2fs_sb_info *sbi = F2FS_SB(sb);
4932
4933         if (sb->s_root) {
4934                 set_sbi_flag(sbi, SBI_IS_CLOSE);
4935                 f2fs_stop_gc_thread(sbi);
4936                 f2fs_stop_discard_thread(sbi);
4937
4938 #ifdef CONFIG_F2FS_FS_COMPRESSION
4939                 /*
4940                  * latter evict_inode() can bypass checking and invalidating
4941                  * compress inode cache.
4942                  */
4943                 if (test_opt(sbi, COMPRESS_CACHE))
4944                         truncate_inode_pages_final(COMPRESS_MAPPING(sbi));
4945 #endif
4946
4947                 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4948                                 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4949                         struct cp_control cpc = {
4950                                 .reason = CP_UMOUNT,
4951                         };
4952                         stat_inc_cp_call_count(sbi, TOTAL_CALL);
4953                         f2fs_write_checkpoint(sbi, &cpc);
4954                 }
4955
4956                 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4957                         sb->s_flags &= ~SB_RDONLY;
4958         }
4959         kill_block_super(sb);
4960         /* Release block devices last, after fscrypt_destroy_keyring(). */
4961         if (sbi) {
4962                 destroy_device_list(sbi);
4963                 kfree(sbi);
4964                 sb->s_fs_info = NULL;
4965         }
4966 }
4967
4968 static struct file_system_type f2fs_fs_type = {
4969         .owner          = THIS_MODULE,
4970         .name           = "f2fs",
4971         .mount          = f2fs_mount,
4972         .kill_sb        = kill_f2fs_super,
4973         .fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
4974 };
4975 MODULE_ALIAS_FS("f2fs");
4976
4977 static int __init init_inodecache(void)
4978 {
4979         f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4980                         sizeof(struct f2fs_inode_info), 0,
4981                         SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4982         return f2fs_inode_cachep ? 0 : -ENOMEM;
4983 }
4984
4985 static void destroy_inodecache(void)
4986 {
4987         /*
4988          * Make sure all delayed rcu free inodes are flushed before we
4989          * destroy cache.
4990          */
4991         rcu_barrier();
4992         kmem_cache_destroy(f2fs_inode_cachep);
4993 }
4994
4995 static int __init init_f2fs_fs(void)
4996 {
4997         int err;
4998
4999         err = init_inodecache();
5000         if (err)
5001                 goto fail;
5002         err = f2fs_create_node_manager_caches();
5003         if (err)
5004                 goto free_inodecache;
5005         err = f2fs_create_segment_manager_caches();
5006         if (err)
5007                 goto free_node_manager_caches;
5008         err = f2fs_create_checkpoint_caches();
5009         if (err)
5010                 goto free_segment_manager_caches;
5011         err = f2fs_create_recovery_cache();
5012         if (err)
5013                 goto free_checkpoint_caches;
5014         err = f2fs_create_extent_cache();
5015         if (err)
5016                 goto free_recovery_cache;
5017         err = f2fs_create_garbage_collection_cache();
5018         if (err)
5019                 goto free_extent_cache;
5020         err = f2fs_init_sysfs();
5021         if (err)
5022                 goto free_garbage_collection_cache;
5023         err = f2fs_init_shrinker();
5024         if (err)
5025                 goto free_sysfs;
5026         f2fs_create_root_stats();
5027         err = f2fs_init_post_read_processing();
5028         if (err)
5029                 goto free_root_stats;
5030         err = f2fs_init_iostat_processing();
5031         if (err)
5032                 goto free_post_read;
5033         err = f2fs_init_bio_entry_cache();
5034         if (err)
5035                 goto free_iostat;
5036         err = f2fs_init_bioset();
5037         if (err)
5038                 goto free_bio_entry_cache;
5039         err = f2fs_init_compress_mempool();
5040         if (err)
5041                 goto free_bioset;
5042         err = f2fs_init_compress_cache();
5043         if (err)
5044                 goto free_compress_mempool;
5045         err = f2fs_create_casefold_cache();
5046         if (err)
5047                 goto free_compress_cache;
5048         err = register_filesystem(&f2fs_fs_type);
5049         if (err)
5050                 goto free_casefold_cache;
5051         return 0;
5052 free_casefold_cache:
5053         f2fs_destroy_casefold_cache();
5054 free_compress_cache:
5055         f2fs_destroy_compress_cache();
5056 free_compress_mempool:
5057         f2fs_destroy_compress_mempool();
5058 free_bioset:
5059         f2fs_destroy_bioset();
5060 free_bio_entry_cache:
5061         f2fs_destroy_bio_entry_cache();
5062 free_iostat:
5063         f2fs_destroy_iostat_processing();
5064 free_post_read:
5065         f2fs_destroy_post_read_processing();
5066 free_root_stats:
5067         f2fs_destroy_root_stats();
5068         f2fs_exit_shrinker();
5069 free_sysfs:
5070         f2fs_exit_sysfs();
5071 free_garbage_collection_cache:
5072         f2fs_destroy_garbage_collection_cache();
5073 free_extent_cache:
5074         f2fs_destroy_extent_cache();
5075 free_recovery_cache:
5076         f2fs_destroy_recovery_cache();
5077 free_checkpoint_caches:
5078         f2fs_destroy_checkpoint_caches();
5079 free_segment_manager_caches:
5080         f2fs_destroy_segment_manager_caches();
5081 free_node_manager_caches:
5082         f2fs_destroy_node_manager_caches();
5083 free_inodecache:
5084         destroy_inodecache();
5085 fail:
5086         return err;
5087 }
5088
5089 static void __exit exit_f2fs_fs(void)
5090 {
5091         unregister_filesystem(&f2fs_fs_type);
5092         f2fs_destroy_casefold_cache();
5093         f2fs_destroy_compress_cache();
5094         f2fs_destroy_compress_mempool();
5095         f2fs_destroy_bioset();
5096         f2fs_destroy_bio_entry_cache();
5097         f2fs_destroy_iostat_processing();
5098         f2fs_destroy_post_read_processing();
5099         f2fs_destroy_root_stats();
5100         f2fs_exit_shrinker();
5101         f2fs_exit_sysfs();
5102         f2fs_destroy_garbage_collection_cache();
5103         f2fs_destroy_extent_cache();
5104         f2fs_destroy_recovery_cache();
5105         f2fs_destroy_checkpoint_caches();
5106         f2fs_destroy_segment_manager_caches();
5107         f2fs_destroy_node_manager_caches();
5108         destroy_inodecache();
5109 }
5110
5111 module_init(init_f2fs_fs)
5112 module_exit(exit_f2fs_fs)
5113
5114 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
5115 MODULE_DESCRIPTION("Flash Friendly File System");
5116 MODULE_LICENSE("GPL");