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 f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned long rate,unsigned long type)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 f2fs_init_shrinker(void)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 f2fs_exit_shrinker(void)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_quota, 154 Opt_noquota, 155 Opt_usrquota, 156 Opt_grpquota, 157 Opt_prjquota, 158 Opt_usrjquota, 159 Opt_grpjquota, 160 Opt_prjjquota, 161 Opt_offusrjquota, 162 Opt_offgrpjquota, 163 Opt_offprjjquota, 164 Opt_jqfmt_vfsold, 165 Opt_jqfmt_vfsv0, 166 Opt_jqfmt_vfsv1, 167 Opt_alloc, 168 Opt_fsync, 169 Opt_test_dummy_encryption, 170 Opt_inlinecrypt, 171 Opt_checkpoint_disable, 172 Opt_checkpoint_disable_cap, 173 Opt_checkpoint_disable_cap_perc, 174 Opt_checkpoint_enable, 175 Opt_checkpoint_merge, 176 Opt_nocheckpoint_merge, 177 Opt_compress_algorithm, 178 Opt_compress_log_size, 179 Opt_compress_extension, 180 Opt_nocompress_extension, 181 Opt_compress_chksum, 182 Opt_compress_mode, 183 Opt_compress_cache, 184 Opt_atgc, 185 Opt_gc_merge, 186 Opt_nogc_merge, 187 Opt_discard_unit, 188 Opt_memory_mode, 189 Opt_age_extent_cache, 190 Opt_errors, 191 Opt_err, 192 }; 193 194 static match_table_t f2fs_tokens = { 195 {Opt_gc_background, "background_gc=%s"}, 196 {Opt_disable_roll_forward, "disable_roll_forward"}, 197 {Opt_norecovery, "norecovery"}, 198 {Opt_discard, "discard"}, 199 {Opt_nodiscard, "nodiscard"}, 200 {Opt_noheap, "no_heap"}, 201 {Opt_heap, "heap"}, 202 {Opt_user_xattr, "user_xattr"}, 203 {Opt_nouser_xattr, "nouser_xattr"}, 204 {Opt_acl, "acl"}, 205 {Opt_noacl, "noacl"}, 206 {Opt_active_logs, "active_logs=%u"}, 207 {Opt_disable_ext_identify, "disable_ext_identify"}, 208 {Opt_inline_xattr, "inline_xattr"}, 209 {Opt_noinline_xattr, "noinline_xattr"}, 210 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 211 {Opt_inline_data, "inline_data"}, 212 {Opt_inline_dentry, "inline_dentry"}, 213 {Opt_noinline_dentry, "noinline_dentry"}, 214 {Opt_flush_merge, "flush_merge"}, 215 {Opt_noflush_merge, "noflush_merge"}, 216 {Opt_barrier, "barrier"}, 217 {Opt_nobarrier, "nobarrier"}, 218 {Opt_fastboot, "fastboot"}, 219 {Opt_extent_cache, "extent_cache"}, 220 {Opt_noextent_cache, "noextent_cache"}, 221 {Opt_noinline_data, "noinline_data"}, 222 {Opt_data_flush, "data_flush"}, 223 {Opt_reserve_root, "reserve_root=%u"}, 224 {Opt_resgid, "resgid=%u"}, 225 {Opt_resuid, "resuid=%u"}, 226 {Opt_mode, "mode=%s"}, 227 {Opt_fault_injection, "fault_injection=%u"}, 228 {Opt_fault_type, "fault_type=%u"}, 229 {Opt_quota, "quota"}, 230 {Opt_noquota, "noquota"}, 231 {Opt_usrquota, "usrquota"}, 232 {Opt_grpquota, "grpquota"}, 233 {Opt_prjquota, "prjquota"}, 234 {Opt_usrjquota, "usrjquota=%s"}, 235 {Opt_grpjquota, "grpjquota=%s"}, 236 {Opt_prjjquota, "prjjquota=%s"}, 237 {Opt_offusrjquota, "usrjquota="}, 238 {Opt_offgrpjquota, "grpjquota="}, 239 {Opt_offprjjquota, "prjjquota="}, 240 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 241 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 242 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 243 {Opt_alloc, "alloc_mode=%s"}, 244 {Opt_fsync, "fsync_mode=%s"}, 245 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 246 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 247 {Opt_inlinecrypt, "inlinecrypt"}, 248 {Opt_checkpoint_disable, "checkpoint=disable"}, 249 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 250 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 251 {Opt_checkpoint_enable, "checkpoint=enable"}, 252 {Opt_checkpoint_merge, "checkpoint_merge"}, 253 {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, 254 {Opt_compress_algorithm, "compress_algorithm=%s"}, 255 {Opt_compress_log_size, "compress_log_size=%u"}, 256 {Opt_compress_extension, "compress_extension=%s"}, 257 {Opt_nocompress_extension, "nocompress_extension=%s"}, 258 {Opt_compress_chksum, "compress_chksum"}, 259 {Opt_compress_mode, "compress_mode=%s"}, 260 {Opt_compress_cache, "compress_cache"}, 261 {Opt_atgc, "atgc"}, 262 {Opt_gc_merge, "gc_merge"}, 263 {Opt_nogc_merge, "nogc_merge"}, 264 {Opt_discard_unit, "discard_unit=%s"}, 265 {Opt_memory_mode, "memory=%s"}, 266 {Opt_age_extent_cache, "age_extent_cache"}, 267 {Opt_errors, "errors=%s"}, 268 {Opt_err, NULL}, 269 }; 270 f2fs_printk(struct f2fs_sb_info * sbi,bool limit_rate,const char * fmt,...)271 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate, 272 const char *fmt, ...) 273 { 274 struct va_format vaf; 275 va_list args; 276 int level; 277 278 va_start(args, fmt); 279 280 level = printk_get_level(fmt); 281 vaf.fmt = printk_skip_level(fmt); 282 vaf.va = &args; 283 if (limit_rate) 284 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n", 285 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 286 else 287 printk("%c%cF2FS-fs (%s): %pV\n", 288 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 289 290 va_end(args); 291 } 292 293 #if IS_ENABLED(CONFIG_UNICODE) 294 static const struct f2fs_sb_encodings { 295 __u16 magic; 296 char *name; 297 unsigned int version; 298 } f2fs_sb_encoding_map[] = { 299 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 300 }; 301 302 static const struct f2fs_sb_encodings * f2fs_sb_read_encoding(const struct f2fs_super_block * sb)303 f2fs_sb_read_encoding(const struct f2fs_super_block *sb) 304 { 305 __u16 magic = le16_to_cpu(sb->s_encoding); 306 int i; 307 308 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 309 if (magic == f2fs_sb_encoding_map[i].magic) 310 return &f2fs_sb_encoding_map[i]; 311 312 return NULL; 313 } 314 315 struct kmem_cache *f2fs_cf_name_slab; f2fs_create_casefold_cache(void)316 static int __init f2fs_create_casefold_cache(void) 317 { 318 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 319 F2FS_NAME_LEN); 320 return f2fs_cf_name_slab ? 0 : -ENOMEM; 321 } 322 f2fs_destroy_casefold_cache(void)323 static void f2fs_destroy_casefold_cache(void) 324 { 325 kmem_cache_destroy(f2fs_cf_name_slab); 326 } 327 #else f2fs_create_casefold_cache(void)328 static int __init f2fs_create_casefold_cache(void) { return 0; } f2fs_destroy_casefold_cache(void)329 static void f2fs_destroy_casefold_cache(void) { } 330 #endif 331 limit_reserve_root(struct f2fs_sb_info * sbi)332 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 333 { 334 block_t limit = min((sbi->user_block_count >> 3), 335 sbi->user_block_count - sbi->reserved_blocks); 336 337 /* limit is 12.5% */ 338 if (test_opt(sbi, RESERVE_ROOT) && 339 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 340 F2FS_OPTION(sbi).root_reserved_blocks = limit; 341 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 342 F2FS_OPTION(sbi).root_reserved_blocks); 343 } 344 if (!test_opt(sbi, RESERVE_ROOT) && 345 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 346 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 347 !gid_eq(F2FS_OPTION(sbi).s_resgid, 348 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 349 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 350 from_kuid_munged(&init_user_ns, 351 F2FS_OPTION(sbi).s_resuid), 352 from_kgid_munged(&init_user_ns, 353 F2FS_OPTION(sbi).s_resgid)); 354 } 355 adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)356 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 357 { 358 if (!F2FS_OPTION(sbi).unusable_cap_perc) 359 return; 360 361 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 362 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 363 else 364 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 365 F2FS_OPTION(sbi).unusable_cap_perc; 366 367 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 368 F2FS_OPTION(sbi).unusable_cap, 369 F2FS_OPTION(sbi).unusable_cap_perc); 370 } 371 init_once(void * foo)372 static void init_once(void *foo) 373 { 374 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 375 376 inode_init_once(&fi->vfs_inode); 377 } 378 379 #ifdef CONFIG_QUOTA 380 static const char * const quotatypes[] = INITQFNAMES; 381 #define QTYPE2NAME(t) (quotatypes[t]) f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)382 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 383 substring_t *args) 384 { 385 struct f2fs_sb_info *sbi = F2FS_SB(sb); 386 char *qname; 387 int ret = -EINVAL; 388 389 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 390 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 391 return -EINVAL; 392 } 393 if (f2fs_sb_has_quota_ino(sbi)) { 394 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 395 return 0; 396 } 397 398 qname = match_strdup(args); 399 if (!qname) { 400 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 401 return -ENOMEM; 402 } 403 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 404 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 405 ret = 0; 406 else 407 f2fs_err(sbi, "%s quota file already specified", 408 QTYPE2NAME(qtype)); 409 goto errout; 410 } 411 if (strchr(qname, '/')) { 412 f2fs_err(sbi, "quotafile must be on filesystem root"); 413 goto errout; 414 } 415 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 416 set_opt(sbi, QUOTA); 417 return 0; 418 errout: 419 kfree(qname); 420 return ret; 421 } 422 f2fs_clear_qf_name(struct super_block * sb,int qtype)423 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 424 { 425 struct f2fs_sb_info *sbi = F2FS_SB(sb); 426 427 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 428 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 429 return -EINVAL; 430 } 431 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 432 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 433 return 0; 434 } 435 f2fs_check_quota_options(struct f2fs_sb_info * sbi)436 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 437 { 438 /* 439 * We do the test below only for project quotas. 'usrquota' and 440 * 'grpquota' mount options are allowed even without quota feature 441 * to support legacy quotas in quota files. 442 */ 443 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 444 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 445 return -1; 446 } 447 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 448 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 449 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 450 if (test_opt(sbi, USRQUOTA) && 451 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 452 clear_opt(sbi, USRQUOTA); 453 454 if (test_opt(sbi, GRPQUOTA) && 455 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 456 clear_opt(sbi, GRPQUOTA); 457 458 if (test_opt(sbi, PRJQUOTA) && 459 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 460 clear_opt(sbi, PRJQUOTA); 461 462 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 463 test_opt(sbi, PRJQUOTA)) { 464 f2fs_err(sbi, "old and new quota format mixing"); 465 return -1; 466 } 467 468 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 469 f2fs_err(sbi, "journaled quota format not specified"); 470 return -1; 471 } 472 } 473 474 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 475 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 476 F2FS_OPTION(sbi).s_jquota_fmt = 0; 477 } 478 return 0; 479 } 480 #endif 481 f2fs_set_test_dummy_encryption(struct super_block * sb,const char * opt,const substring_t * arg,bool is_remount)482 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 483 const char *opt, 484 const substring_t *arg, 485 bool is_remount) 486 { 487 struct f2fs_sb_info *sbi = F2FS_SB(sb); 488 struct fs_parameter param = { 489 .type = fs_value_is_string, 490 .string = arg->from ? arg->from : "", 491 }; 492 struct fscrypt_dummy_policy *policy = 493 &F2FS_OPTION(sbi).dummy_enc_policy; 494 int err; 495 496 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 497 f2fs_warn(sbi, "test_dummy_encryption option not supported"); 498 return -EINVAL; 499 } 500 501 if (!f2fs_sb_has_encrypt(sbi)) { 502 f2fs_err(sbi, "Encrypt feature is off"); 503 return -EINVAL; 504 } 505 506 /* 507 * This mount option is just for testing, and it's not worthwhile to 508 * implement the extra complexity (e.g. RCU protection) that would be 509 * needed to allow it to be set or changed during remount. We do allow 510 * it to be specified during remount, but only if there is no change. 511 */ 512 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) { 513 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 514 return -EINVAL; 515 } 516 517 err = fscrypt_parse_test_dummy_encryption(¶m, policy); 518 if (err) { 519 if (err == -EEXIST) 520 f2fs_warn(sbi, 521 "Can't change test_dummy_encryption on remount"); 522 else if (err == -EINVAL) 523 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 524 opt); 525 else 526 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 527 opt, err); 528 return -EINVAL; 529 } 530 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 531 return 0; 532 } 533 534 #ifdef CONFIG_F2FS_FS_COMPRESSION is_compress_extension_exist(struct f2fs_sb_info * sbi,const char * new_ext,bool is_ext)535 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi, 536 const char *new_ext, bool is_ext) 537 { 538 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 539 int ext_cnt; 540 int i; 541 542 if (is_ext) { 543 ext = F2FS_OPTION(sbi).extensions; 544 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 545 } else { 546 ext = F2FS_OPTION(sbi).noextensions; 547 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 548 } 549 550 for (i = 0; i < ext_cnt; i++) { 551 if (!strcasecmp(new_ext, ext[i])) 552 return true; 553 } 554 555 return false; 556 } 557 558 /* 559 * 1. The same extension name cannot not appear in both compress and non-compress extension 560 * at the same time. 561 * 2. If the compress extension specifies all files, the types specified by the non-compress 562 * extension will be treated as special cases and will not be compressed. 563 * 3. Don't allow the non-compress extension specifies all files. 564 */ f2fs_test_compress_extension(struct f2fs_sb_info * sbi)565 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi) 566 { 567 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 568 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 569 int ext_cnt, noext_cnt, index = 0, no_index = 0; 570 571 ext = F2FS_OPTION(sbi).extensions; 572 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 573 noext = F2FS_OPTION(sbi).noextensions; 574 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 575 576 if (!noext_cnt) 577 return 0; 578 579 for (no_index = 0; no_index < noext_cnt; no_index++) { 580 if (!strcasecmp("*", noext[no_index])) { 581 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files"); 582 return -EINVAL; 583 } 584 for (index = 0; index < ext_cnt; index++) { 585 if (!strcasecmp(ext[index], noext[no_index])) { 586 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension", 587 ext[index]); 588 return -EINVAL; 589 } 590 } 591 } 592 return 0; 593 } 594 595 #ifdef CONFIG_F2FS_FS_LZ4 f2fs_set_lz4hc_level(struct f2fs_sb_info * sbi,const char * str)596 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) 597 { 598 #ifdef CONFIG_F2FS_FS_LZ4HC 599 unsigned int level; 600 601 if (strlen(str) == 3) { 602 F2FS_OPTION(sbi).compress_level = 0; 603 return 0; 604 } 605 606 str += 3; 607 608 if (str[0] != ':') { 609 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 610 return -EINVAL; 611 } 612 if (kstrtouint(str + 1, 10, &level)) 613 return -EINVAL; 614 615 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) { 616 f2fs_info(sbi, "invalid lz4hc compress level: %d", level); 617 return -EINVAL; 618 } 619 620 F2FS_OPTION(sbi).compress_level = level; 621 return 0; 622 #else 623 if (strlen(str) == 3) { 624 F2FS_OPTION(sbi).compress_level = 0; 625 return 0; 626 } 627 f2fs_info(sbi, "kernel doesn't support lz4hc compression"); 628 return -EINVAL; 629 #endif 630 } 631 #endif 632 633 #ifdef CONFIG_F2FS_FS_ZSTD f2fs_set_zstd_level(struct f2fs_sb_info * sbi,const char * str)634 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) 635 { 636 int level; 637 int len = 4; 638 639 if (strlen(str) == len) { 640 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 641 return 0; 642 } 643 644 str += len; 645 646 if (str[0] != ':') { 647 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 648 return -EINVAL; 649 } 650 if (kstrtoint(str + 1, 10, &level)) 651 return -EINVAL; 652 653 /* f2fs does not support negative compress level now */ 654 if (level < 0) { 655 f2fs_info(sbi, "do not support negative compress level: %d", level); 656 return -ERANGE; 657 } 658 659 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) { 660 f2fs_info(sbi, "invalid zstd compress level: %d", level); 661 return -EINVAL; 662 } 663 664 F2FS_OPTION(sbi).compress_level = level; 665 return 0; 666 } 667 #endif 668 #endif 669 parse_options(struct super_block * sb,char * options,bool is_remount)670 static int parse_options(struct super_block *sb, char *options, bool is_remount) 671 { 672 struct f2fs_sb_info *sbi = F2FS_SB(sb); 673 substring_t args[MAX_OPT_ARGS]; 674 #ifdef CONFIG_F2FS_FS_COMPRESSION 675 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 676 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 677 int ext_cnt, noext_cnt; 678 #endif 679 char *p, *name; 680 int arg = 0; 681 kuid_t uid; 682 kgid_t gid; 683 int ret; 684 685 if (!options) 686 goto default_check; 687 688 while ((p = strsep(&options, ",")) != NULL) { 689 int token; 690 691 if (!*p) 692 continue; 693 /* 694 * Initialize args struct so we know whether arg was 695 * found; some options take optional arguments. 696 */ 697 args[0].to = args[0].from = NULL; 698 token = match_token(p, f2fs_tokens, args); 699 700 switch (token) { 701 case Opt_gc_background: 702 name = match_strdup(&args[0]); 703 704 if (!name) 705 return -ENOMEM; 706 if (!strcmp(name, "on")) { 707 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 708 } else if (!strcmp(name, "off")) { 709 if (f2fs_sb_has_blkzoned(sbi)) { 710 f2fs_warn(sbi, "zoned devices need bggc"); 711 kfree(name); 712 return -EINVAL; 713 } 714 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 715 } else if (!strcmp(name, "sync")) { 716 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 717 } else { 718 kfree(name); 719 return -EINVAL; 720 } 721 kfree(name); 722 break; 723 case Opt_disable_roll_forward: 724 set_opt(sbi, DISABLE_ROLL_FORWARD); 725 break; 726 case Opt_norecovery: 727 /* this option mounts f2fs with ro */ 728 set_opt(sbi, NORECOVERY); 729 if (!f2fs_readonly(sb)) 730 return -EINVAL; 731 break; 732 case Opt_discard: 733 if (!f2fs_hw_support_discard(sbi)) { 734 f2fs_warn(sbi, "device does not support discard"); 735 break; 736 } 737 set_opt(sbi, DISCARD); 738 break; 739 case Opt_nodiscard: 740 if (f2fs_hw_should_discard(sbi)) { 741 f2fs_warn(sbi, "discard is required for zoned block devices"); 742 return -EINVAL; 743 } 744 clear_opt(sbi, DISCARD); 745 break; 746 case Opt_noheap: 747 case Opt_heap: 748 f2fs_warn(sbi, "heap/no_heap options were deprecated"); 749 break; 750 #ifdef CONFIG_F2FS_FS_XATTR 751 case Opt_user_xattr: 752 set_opt(sbi, XATTR_USER); 753 break; 754 case Opt_nouser_xattr: 755 clear_opt(sbi, XATTR_USER); 756 break; 757 case Opt_inline_xattr: 758 set_opt(sbi, INLINE_XATTR); 759 break; 760 case Opt_noinline_xattr: 761 clear_opt(sbi, INLINE_XATTR); 762 break; 763 case Opt_inline_xattr_size: 764 if (args->from && match_int(args, &arg)) 765 return -EINVAL; 766 set_opt(sbi, INLINE_XATTR_SIZE); 767 F2FS_OPTION(sbi).inline_xattr_size = arg; 768 break; 769 #else 770 case Opt_user_xattr: 771 f2fs_info(sbi, "user_xattr options not supported"); 772 break; 773 case Opt_nouser_xattr: 774 f2fs_info(sbi, "nouser_xattr options not supported"); 775 break; 776 case Opt_inline_xattr: 777 f2fs_info(sbi, "inline_xattr options not supported"); 778 break; 779 case Opt_noinline_xattr: 780 f2fs_info(sbi, "noinline_xattr options not supported"); 781 break; 782 #endif 783 #ifdef CONFIG_F2FS_FS_POSIX_ACL 784 case Opt_acl: 785 set_opt(sbi, POSIX_ACL); 786 break; 787 case Opt_noacl: 788 clear_opt(sbi, POSIX_ACL); 789 break; 790 #else 791 case Opt_acl: 792 f2fs_info(sbi, "acl options not supported"); 793 break; 794 case Opt_noacl: 795 f2fs_info(sbi, "noacl options not supported"); 796 break; 797 #endif 798 case Opt_active_logs: 799 if (args->from && match_int(args, &arg)) 800 return -EINVAL; 801 if (arg != 2 && arg != 4 && 802 arg != NR_CURSEG_PERSIST_TYPE) 803 return -EINVAL; 804 F2FS_OPTION(sbi).active_logs = arg; 805 break; 806 case Opt_disable_ext_identify: 807 set_opt(sbi, DISABLE_EXT_IDENTIFY); 808 break; 809 case Opt_inline_data: 810 set_opt(sbi, INLINE_DATA); 811 break; 812 case Opt_inline_dentry: 813 set_opt(sbi, INLINE_DENTRY); 814 break; 815 case Opt_noinline_dentry: 816 clear_opt(sbi, INLINE_DENTRY); 817 break; 818 case Opt_flush_merge: 819 set_opt(sbi, FLUSH_MERGE); 820 break; 821 case Opt_noflush_merge: 822 clear_opt(sbi, FLUSH_MERGE); 823 break; 824 case Opt_nobarrier: 825 set_opt(sbi, NOBARRIER); 826 break; 827 case Opt_barrier: 828 clear_opt(sbi, NOBARRIER); 829 break; 830 case Opt_fastboot: 831 set_opt(sbi, FASTBOOT); 832 break; 833 case Opt_extent_cache: 834 set_opt(sbi, READ_EXTENT_CACHE); 835 break; 836 case Opt_noextent_cache: 837 clear_opt(sbi, READ_EXTENT_CACHE); 838 break; 839 case Opt_noinline_data: 840 clear_opt(sbi, INLINE_DATA); 841 break; 842 case Opt_data_flush: 843 set_opt(sbi, DATA_FLUSH); 844 break; 845 case Opt_reserve_root: 846 if (args->from && match_int(args, &arg)) 847 return -EINVAL; 848 if (test_opt(sbi, RESERVE_ROOT)) { 849 f2fs_info(sbi, "Preserve previous reserve_root=%u", 850 F2FS_OPTION(sbi).root_reserved_blocks); 851 } else { 852 F2FS_OPTION(sbi).root_reserved_blocks = arg; 853 set_opt(sbi, RESERVE_ROOT); 854 } 855 break; 856 case Opt_resuid: 857 if (args->from && match_int(args, &arg)) 858 return -EINVAL; 859 uid = make_kuid(current_user_ns(), arg); 860 if (!uid_valid(uid)) { 861 f2fs_err(sbi, "Invalid uid value %d", arg); 862 return -EINVAL; 863 } 864 F2FS_OPTION(sbi).s_resuid = uid; 865 break; 866 case Opt_resgid: 867 if (args->from && match_int(args, &arg)) 868 return -EINVAL; 869 gid = make_kgid(current_user_ns(), arg); 870 if (!gid_valid(gid)) { 871 f2fs_err(sbi, "Invalid gid value %d", arg); 872 return -EINVAL; 873 } 874 F2FS_OPTION(sbi).s_resgid = gid; 875 break; 876 case Opt_mode: 877 name = match_strdup(&args[0]); 878 879 if (!name) 880 return -ENOMEM; 881 if (!strcmp(name, "adaptive")) { 882 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 883 } else if (!strcmp(name, "lfs")) { 884 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 885 } else if (!strcmp(name, "fragment:segment")) { 886 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG; 887 } else if (!strcmp(name, "fragment:block")) { 888 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK; 889 } else { 890 kfree(name); 891 return -EINVAL; 892 } 893 kfree(name); 894 break; 895 #ifdef CONFIG_F2FS_FAULT_INJECTION 896 case Opt_fault_injection: 897 if (args->from && match_int(args, &arg)) 898 return -EINVAL; 899 if (f2fs_build_fault_attr(sbi, arg, 900 F2FS_ALL_FAULT_TYPE)) 901 return -EINVAL; 902 set_opt(sbi, FAULT_INJECTION); 903 break; 904 905 case Opt_fault_type: 906 if (args->from && match_int(args, &arg)) 907 return -EINVAL; 908 if (f2fs_build_fault_attr(sbi, 0, arg)) 909 return -EINVAL; 910 set_opt(sbi, FAULT_INJECTION); 911 break; 912 #else 913 case Opt_fault_injection: 914 f2fs_info(sbi, "fault_injection options not supported"); 915 break; 916 917 case Opt_fault_type: 918 f2fs_info(sbi, "fault_type options not supported"); 919 break; 920 #endif 921 #ifdef CONFIG_QUOTA 922 case Opt_quota: 923 case Opt_usrquota: 924 set_opt(sbi, USRQUOTA); 925 break; 926 case Opt_grpquota: 927 set_opt(sbi, GRPQUOTA); 928 break; 929 case Opt_prjquota: 930 set_opt(sbi, PRJQUOTA); 931 break; 932 case Opt_usrjquota: 933 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 934 if (ret) 935 return ret; 936 break; 937 case Opt_grpjquota: 938 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 939 if (ret) 940 return ret; 941 break; 942 case Opt_prjjquota: 943 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 944 if (ret) 945 return ret; 946 break; 947 case Opt_offusrjquota: 948 ret = f2fs_clear_qf_name(sb, USRQUOTA); 949 if (ret) 950 return ret; 951 break; 952 case Opt_offgrpjquota: 953 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 954 if (ret) 955 return ret; 956 break; 957 case Opt_offprjjquota: 958 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 959 if (ret) 960 return ret; 961 break; 962 case Opt_jqfmt_vfsold: 963 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 964 break; 965 case Opt_jqfmt_vfsv0: 966 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 967 break; 968 case Opt_jqfmt_vfsv1: 969 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 970 break; 971 case Opt_noquota: 972 clear_opt(sbi, QUOTA); 973 clear_opt(sbi, USRQUOTA); 974 clear_opt(sbi, GRPQUOTA); 975 clear_opt(sbi, PRJQUOTA); 976 break; 977 #else 978 case Opt_quota: 979 case Opt_usrquota: 980 case Opt_grpquota: 981 case Opt_prjquota: 982 case Opt_usrjquota: 983 case Opt_grpjquota: 984 case Opt_prjjquota: 985 case Opt_offusrjquota: 986 case Opt_offgrpjquota: 987 case Opt_offprjjquota: 988 case Opt_jqfmt_vfsold: 989 case Opt_jqfmt_vfsv0: 990 case Opt_jqfmt_vfsv1: 991 case Opt_noquota: 992 f2fs_info(sbi, "quota operations not supported"); 993 break; 994 #endif 995 case Opt_alloc: 996 name = match_strdup(&args[0]); 997 if (!name) 998 return -ENOMEM; 999 1000 if (!strcmp(name, "default")) { 1001 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1002 } else if (!strcmp(name, "reuse")) { 1003 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 1004 } else { 1005 kfree(name); 1006 return -EINVAL; 1007 } 1008 kfree(name); 1009 break; 1010 case Opt_fsync: 1011 name = match_strdup(&args[0]); 1012 if (!name) 1013 return -ENOMEM; 1014 if (!strcmp(name, "posix")) { 1015 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1016 } else if (!strcmp(name, "strict")) { 1017 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 1018 } else if (!strcmp(name, "nobarrier")) { 1019 F2FS_OPTION(sbi).fsync_mode = 1020 FSYNC_MODE_NOBARRIER; 1021 } else { 1022 kfree(name); 1023 return -EINVAL; 1024 } 1025 kfree(name); 1026 break; 1027 case Opt_test_dummy_encryption: 1028 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 1029 is_remount); 1030 if (ret) 1031 return ret; 1032 break; 1033 case Opt_inlinecrypt: 1034 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 1035 sb->s_flags |= SB_INLINECRYPT; 1036 #else 1037 f2fs_info(sbi, "inline encryption not supported"); 1038 #endif 1039 break; 1040 case Opt_checkpoint_disable_cap_perc: 1041 if (args->from && match_int(args, &arg)) 1042 return -EINVAL; 1043 if (arg < 0 || arg > 100) 1044 return -EINVAL; 1045 F2FS_OPTION(sbi).unusable_cap_perc = arg; 1046 set_opt(sbi, DISABLE_CHECKPOINT); 1047 break; 1048 case Opt_checkpoint_disable_cap: 1049 if (args->from && match_int(args, &arg)) 1050 return -EINVAL; 1051 F2FS_OPTION(sbi).unusable_cap = arg; 1052 set_opt(sbi, DISABLE_CHECKPOINT); 1053 break; 1054 case Opt_checkpoint_disable: 1055 set_opt(sbi, DISABLE_CHECKPOINT); 1056 break; 1057 case Opt_checkpoint_enable: 1058 clear_opt(sbi, DISABLE_CHECKPOINT); 1059 break; 1060 case Opt_checkpoint_merge: 1061 set_opt(sbi, MERGE_CHECKPOINT); 1062 break; 1063 case Opt_nocheckpoint_merge: 1064 clear_opt(sbi, MERGE_CHECKPOINT); 1065 break; 1066 #ifdef CONFIG_F2FS_FS_COMPRESSION 1067 case Opt_compress_algorithm: 1068 if (!f2fs_sb_has_compression(sbi)) { 1069 f2fs_info(sbi, "Image doesn't support compression"); 1070 break; 1071 } 1072 name = match_strdup(&args[0]); 1073 if (!name) 1074 return -ENOMEM; 1075 if (!strcmp(name, "lzo")) { 1076 #ifdef CONFIG_F2FS_FS_LZO 1077 F2FS_OPTION(sbi).compress_level = 0; 1078 F2FS_OPTION(sbi).compress_algorithm = 1079 COMPRESS_LZO; 1080 #else 1081 f2fs_info(sbi, "kernel doesn't support lzo compression"); 1082 #endif 1083 } else if (!strncmp(name, "lz4", 3)) { 1084 #ifdef CONFIG_F2FS_FS_LZ4 1085 ret = f2fs_set_lz4hc_level(sbi, name); 1086 if (ret) { 1087 kfree(name); 1088 return -EINVAL; 1089 } 1090 F2FS_OPTION(sbi).compress_algorithm = 1091 COMPRESS_LZ4; 1092 #else 1093 f2fs_info(sbi, "kernel doesn't support lz4 compression"); 1094 #endif 1095 } else if (!strncmp(name, "zstd", 4)) { 1096 #ifdef CONFIG_F2FS_FS_ZSTD 1097 ret = f2fs_set_zstd_level(sbi, name); 1098 if (ret) { 1099 kfree(name); 1100 return -EINVAL; 1101 } 1102 F2FS_OPTION(sbi).compress_algorithm = 1103 COMPRESS_ZSTD; 1104 #else 1105 f2fs_info(sbi, "kernel doesn't support zstd compression"); 1106 #endif 1107 } else if (!strcmp(name, "lzo-rle")) { 1108 #ifdef CONFIG_F2FS_FS_LZORLE 1109 F2FS_OPTION(sbi).compress_level = 0; 1110 F2FS_OPTION(sbi).compress_algorithm = 1111 COMPRESS_LZORLE; 1112 #else 1113 f2fs_info(sbi, "kernel doesn't support lzorle compression"); 1114 #endif 1115 } else { 1116 kfree(name); 1117 return -EINVAL; 1118 } 1119 kfree(name); 1120 break; 1121 case Opt_compress_log_size: 1122 if (!f2fs_sb_has_compression(sbi)) { 1123 f2fs_info(sbi, "Image doesn't support compression"); 1124 break; 1125 } 1126 if (args->from && match_int(args, &arg)) 1127 return -EINVAL; 1128 if (arg < MIN_COMPRESS_LOG_SIZE || 1129 arg > MAX_COMPRESS_LOG_SIZE) { 1130 f2fs_err(sbi, 1131 "Compress cluster log size is out of range"); 1132 return -EINVAL; 1133 } 1134 F2FS_OPTION(sbi).compress_log_size = arg; 1135 break; 1136 case Opt_compress_extension: 1137 if (!f2fs_sb_has_compression(sbi)) { 1138 f2fs_info(sbi, "Image doesn't support compression"); 1139 break; 1140 } 1141 name = match_strdup(&args[0]); 1142 if (!name) 1143 return -ENOMEM; 1144 1145 ext = F2FS_OPTION(sbi).extensions; 1146 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1147 1148 if (strlen(name) >= F2FS_EXTENSION_LEN || 1149 ext_cnt >= COMPRESS_EXT_NUM) { 1150 f2fs_err(sbi, 1151 "invalid extension length/number"); 1152 kfree(name); 1153 return -EINVAL; 1154 } 1155 1156 if (is_compress_extension_exist(sbi, name, true)) { 1157 kfree(name); 1158 break; 1159 } 1160 1161 strcpy(ext[ext_cnt], name); 1162 F2FS_OPTION(sbi).compress_ext_cnt++; 1163 kfree(name); 1164 break; 1165 case Opt_nocompress_extension: 1166 if (!f2fs_sb_has_compression(sbi)) { 1167 f2fs_info(sbi, "Image doesn't support compression"); 1168 break; 1169 } 1170 name = match_strdup(&args[0]); 1171 if (!name) 1172 return -ENOMEM; 1173 1174 noext = F2FS_OPTION(sbi).noextensions; 1175 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1176 1177 if (strlen(name) >= F2FS_EXTENSION_LEN || 1178 noext_cnt >= COMPRESS_EXT_NUM) { 1179 f2fs_err(sbi, 1180 "invalid extension length/number"); 1181 kfree(name); 1182 return -EINVAL; 1183 } 1184 1185 if (is_compress_extension_exist(sbi, name, false)) { 1186 kfree(name); 1187 break; 1188 } 1189 1190 strcpy(noext[noext_cnt], name); 1191 F2FS_OPTION(sbi).nocompress_ext_cnt++; 1192 kfree(name); 1193 break; 1194 case Opt_compress_chksum: 1195 if (!f2fs_sb_has_compression(sbi)) { 1196 f2fs_info(sbi, "Image doesn't support compression"); 1197 break; 1198 } 1199 F2FS_OPTION(sbi).compress_chksum = true; 1200 break; 1201 case Opt_compress_mode: 1202 if (!f2fs_sb_has_compression(sbi)) { 1203 f2fs_info(sbi, "Image doesn't support compression"); 1204 break; 1205 } 1206 name = match_strdup(&args[0]); 1207 if (!name) 1208 return -ENOMEM; 1209 if (!strcmp(name, "fs")) { 1210 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1211 } else if (!strcmp(name, "user")) { 1212 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 1213 } else { 1214 kfree(name); 1215 return -EINVAL; 1216 } 1217 kfree(name); 1218 break; 1219 case Opt_compress_cache: 1220 if (!f2fs_sb_has_compression(sbi)) { 1221 f2fs_info(sbi, "Image doesn't support compression"); 1222 break; 1223 } 1224 set_opt(sbi, COMPRESS_CACHE); 1225 break; 1226 #else 1227 case Opt_compress_algorithm: 1228 case Opt_compress_log_size: 1229 case Opt_compress_extension: 1230 case Opt_nocompress_extension: 1231 case Opt_compress_chksum: 1232 case Opt_compress_mode: 1233 case Opt_compress_cache: 1234 f2fs_info(sbi, "compression options not supported"); 1235 break; 1236 #endif 1237 case Opt_atgc: 1238 set_opt(sbi, ATGC); 1239 break; 1240 case Opt_gc_merge: 1241 set_opt(sbi, GC_MERGE); 1242 break; 1243 case Opt_nogc_merge: 1244 clear_opt(sbi, GC_MERGE); 1245 break; 1246 case Opt_discard_unit: 1247 name = match_strdup(&args[0]); 1248 if (!name) 1249 return -ENOMEM; 1250 if (!strcmp(name, "block")) { 1251 F2FS_OPTION(sbi).discard_unit = 1252 DISCARD_UNIT_BLOCK; 1253 } else if (!strcmp(name, "segment")) { 1254 F2FS_OPTION(sbi).discard_unit = 1255 DISCARD_UNIT_SEGMENT; 1256 } else if (!strcmp(name, "section")) { 1257 F2FS_OPTION(sbi).discard_unit = 1258 DISCARD_UNIT_SECTION; 1259 } else { 1260 kfree(name); 1261 return -EINVAL; 1262 } 1263 kfree(name); 1264 break; 1265 case Opt_memory_mode: 1266 name = match_strdup(&args[0]); 1267 if (!name) 1268 return -ENOMEM; 1269 if (!strcmp(name, "normal")) { 1270 F2FS_OPTION(sbi).memory_mode = 1271 MEMORY_MODE_NORMAL; 1272 } else if (!strcmp(name, "low")) { 1273 F2FS_OPTION(sbi).memory_mode = 1274 MEMORY_MODE_LOW; 1275 } else { 1276 kfree(name); 1277 return -EINVAL; 1278 } 1279 kfree(name); 1280 break; 1281 case Opt_age_extent_cache: 1282 set_opt(sbi, AGE_EXTENT_CACHE); 1283 break; 1284 case Opt_errors: 1285 name = match_strdup(&args[0]); 1286 if (!name) 1287 return -ENOMEM; 1288 if (!strcmp(name, "remount-ro")) { 1289 F2FS_OPTION(sbi).errors = 1290 MOUNT_ERRORS_READONLY; 1291 } else if (!strcmp(name, "continue")) { 1292 F2FS_OPTION(sbi).errors = 1293 MOUNT_ERRORS_CONTINUE; 1294 } else if (!strcmp(name, "panic")) { 1295 F2FS_OPTION(sbi).errors = 1296 MOUNT_ERRORS_PANIC; 1297 } else { 1298 kfree(name); 1299 return -EINVAL; 1300 } 1301 kfree(name); 1302 break; 1303 default: 1304 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 1305 p); 1306 return -EINVAL; 1307 } 1308 } 1309 default_check: 1310 #ifdef CONFIG_QUOTA 1311 if (f2fs_check_quota_options(sbi)) 1312 return -EINVAL; 1313 #else 1314 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 1315 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1316 return -EINVAL; 1317 } 1318 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 1319 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1320 return -EINVAL; 1321 } 1322 #endif 1323 1324 if (!IS_ENABLED(CONFIG_UNICODE) && f2fs_sb_has_casefold(sbi)) { 1325 f2fs_err(sbi, 1326 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1327 return -EINVAL; 1328 } 1329 1330 /* 1331 * The BLKZONED feature indicates that the drive was formatted with 1332 * zone alignment optimization. This is optional for host-aware 1333 * devices, but mandatory for host-managed zoned block devices. 1334 */ 1335 if (f2fs_sb_has_blkzoned(sbi)) { 1336 #ifdef CONFIG_BLK_DEV_ZONED 1337 if (F2FS_OPTION(sbi).discard_unit != 1338 DISCARD_UNIT_SECTION) { 1339 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1340 F2FS_OPTION(sbi).discard_unit = 1341 DISCARD_UNIT_SECTION; 1342 } 1343 1344 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) { 1345 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature"); 1346 return -EINVAL; 1347 } 1348 #else 1349 f2fs_err(sbi, "Zoned block device support is not enabled"); 1350 return -EINVAL; 1351 #endif 1352 } 1353 1354 #ifdef CONFIG_F2FS_FS_COMPRESSION 1355 if (f2fs_test_compress_extension(sbi)) { 1356 f2fs_err(sbi, "invalid compress or nocompress extension"); 1357 return -EINVAL; 1358 } 1359 #endif 1360 1361 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1362 int min_size, max_size; 1363 1364 if (!f2fs_sb_has_extra_attr(sbi) || 1365 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1366 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1367 return -EINVAL; 1368 } 1369 if (!test_opt(sbi, INLINE_XATTR)) { 1370 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1371 return -EINVAL; 1372 } 1373 1374 min_size = MIN_INLINE_XATTR_SIZE; 1375 max_size = MAX_INLINE_XATTR_SIZE; 1376 1377 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1378 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1379 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1380 min_size, max_size); 1381 return -EINVAL; 1382 } 1383 } 1384 1385 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) { 1386 f2fs_err(sbi, "LFS is not compatible with ATGC"); 1387 return -EINVAL; 1388 } 1389 1390 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) { 1391 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode"); 1392 return -EINVAL; 1393 } 1394 1395 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1396 f2fs_err(sbi, "Allow to mount readonly mode only"); 1397 return -EROFS; 1398 } 1399 return 0; 1400 } 1401 f2fs_alloc_inode(struct super_block * sb)1402 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1403 { 1404 struct f2fs_inode_info *fi; 1405 1406 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) 1407 return NULL; 1408 1409 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO); 1410 if (!fi) 1411 return NULL; 1412 1413 init_once((void *) fi); 1414 1415 /* Initialize f2fs-specific inode info */ 1416 atomic_set(&fi->dirty_pages, 0); 1417 atomic_set(&fi->i_compr_blocks, 0); 1418 init_f2fs_rwsem(&fi->i_sem); 1419 spin_lock_init(&fi->i_size_lock); 1420 INIT_LIST_HEAD(&fi->dirty_list); 1421 INIT_LIST_HEAD(&fi->gdirty_list); 1422 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]); 1423 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]); 1424 init_f2fs_rwsem(&fi->i_xattr_sem); 1425 1426 /* Will be used by directory only */ 1427 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1428 1429 return &fi->vfs_inode; 1430 } 1431 f2fs_drop_inode(struct inode * inode)1432 static int f2fs_drop_inode(struct inode *inode) 1433 { 1434 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1435 int ret; 1436 1437 /* 1438 * during filesystem shutdown, if checkpoint is disabled, 1439 * drop useless meta/node dirty pages. 1440 */ 1441 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1442 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1443 inode->i_ino == F2FS_META_INO(sbi)) { 1444 trace_f2fs_drop_inode(inode, 1); 1445 return 1; 1446 } 1447 } 1448 1449 /* 1450 * This is to avoid a deadlock condition like below. 1451 * writeback_single_inode(inode) 1452 * - f2fs_write_data_page 1453 * - f2fs_gc -> iput -> evict 1454 * - inode_wait_for_writeback(inode) 1455 */ 1456 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1457 if (!inode->i_nlink && !is_bad_inode(inode)) { 1458 /* to avoid evict_inode call simultaneously */ 1459 atomic_inc(&inode->i_count); 1460 spin_unlock(&inode->i_lock); 1461 1462 /* should remain fi->extent_tree for writepage */ 1463 f2fs_destroy_extent_node(inode); 1464 1465 sb_start_intwrite(inode->i_sb); 1466 f2fs_i_size_write(inode, 0); 1467 1468 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1469 inode, NULL, 0, DATA); 1470 truncate_inode_pages_final(inode->i_mapping); 1471 1472 if (F2FS_HAS_BLOCKS(inode)) 1473 f2fs_truncate(inode); 1474 1475 sb_end_intwrite(inode->i_sb); 1476 1477 spin_lock(&inode->i_lock); 1478 atomic_dec(&inode->i_count); 1479 } 1480 trace_f2fs_drop_inode(inode, 0); 1481 return 0; 1482 } 1483 ret = generic_drop_inode(inode); 1484 if (!ret) 1485 ret = fscrypt_drop_inode(inode); 1486 trace_f2fs_drop_inode(inode, ret); 1487 return ret; 1488 } 1489 f2fs_inode_dirtied(struct inode * inode,bool sync)1490 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1491 { 1492 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1493 int ret = 0; 1494 1495 spin_lock(&sbi->inode_lock[DIRTY_META]); 1496 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1497 ret = 1; 1498 } else { 1499 set_inode_flag(inode, FI_DIRTY_INODE); 1500 stat_inc_dirty_inode(sbi, DIRTY_META); 1501 } 1502 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1503 list_add_tail(&F2FS_I(inode)->gdirty_list, 1504 &sbi->inode_list[DIRTY_META]); 1505 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1506 } 1507 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1508 return ret; 1509 } 1510 f2fs_inode_synced(struct inode * inode)1511 void f2fs_inode_synced(struct inode *inode) 1512 { 1513 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1514 1515 spin_lock(&sbi->inode_lock[DIRTY_META]); 1516 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1517 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1518 return; 1519 } 1520 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1521 list_del_init(&F2FS_I(inode)->gdirty_list); 1522 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1523 } 1524 clear_inode_flag(inode, FI_DIRTY_INODE); 1525 clear_inode_flag(inode, FI_AUTO_RECOVER); 1526 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1527 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1528 } 1529 1530 /* 1531 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1532 * 1533 * We should call set_dirty_inode to write the dirty inode through write_inode. 1534 */ f2fs_dirty_inode(struct inode * inode,int flags)1535 static void f2fs_dirty_inode(struct inode *inode, int flags) 1536 { 1537 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1538 1539 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1540 inode->i_ino == F2FS_META_INO(sbi)) 1541 return; 1542 1543 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1544 clear_inode_flag(inode, FI_AUTO_RECOVER); 1545 1546 f2fs_inode_dirtied(inode, false); 1547 } 1548 f2fs_free_inode(struct inode * inode)1549 static void f2fs_free_inode(struct inode *inode) 1550 { 1551 fscrypt_free_inode(inode); 1552 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1553 } 1554 destroy_percpu_info(struct f2fs_sb_info * sbi)1555 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1556 { 1557 percpu_counter_destroy(&sbi->total_valid_inode_count); 1558 percpu_counter_destroy(&sbi->rf_node_block_count); 1559 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1560 } 1561 destroy_device_list(struct f2fs_sb_info * sbi)1562 static void destroy_device_list(struct f2fs_sb_info *sbi) 1563 { 1564 int i; 1565 1566 for (i = 0; i < sbi->s_ndevs; i++) { 1567 if (i > 0) 1568 bdev_fput(FDEV(i).bdev_file); 1569 #ifdef CONFIG_BLK_DEV_ZONED 1570 kvfree(FDEV(i).blkz_seq); 1571 #endif 1572 } 1573 kvfree(sbi->devs); 1574 } 1575 f2fs_put_super(struct super_block * sb)1576 static void f2fs_put_super(struct super_block *sb) 1577 { 1578 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1579 int i; 1580 int err = 0; 1581 bool done; 1582 1583 /* unregister procfs/sysfs entries in advance to avoid race case */ 1584 f2fs_unregister_sysfs(sbi); 1585 1586 f2fs_quota_off_umount(sb); 1587 1588 /* prevent remaining shrinker jobs */ 1589 mutex_lock(&sbi->umount_mutex); 1590 1591 /* 1592 * flush all issued checkpoints and stop checkpoint issue thread. 1593 * after then, all checkpoints should be done by each process context. 1594 */ 1595 f2fs_stop_ckpt_thread(sbi); 1596 1597 /* 1598 * We don't need to do checkpoint when superblock is clean. 1599 * But, the previous checkpoint was not done by umount, it needs to do 1600 * clean checkpoint again. 1601 */ 1602 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1603 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1604 struct cp_control cpc = { 1605 .reason = CP_UMOUNT, 1606 }; 1607 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1608 err = f2fs_write_checkpoint(sbi, &cpc); 1609 } 1610 1611 /* be sure to wait for any on-going discard commands */ 1612 done = f2fs_issue_discard_timeout(sbi); 1613 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) { 1614 struct cp_control cpc = { 1615 .reason = CP_UMOUNT | CP_TRIMMED, 1616 }; 1617 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1618 err = f2fs_write_checkpoint(sbi, &cpc); 1619 } 1620 1621 /* 1622 * normally superblock is clean, so we need to release this. 1623 * In addition, EIO will skip do checkpoint, we need this as well. 1624 */ 1625 f2fs_release_ino_entry(sbi, true); 1626 1627 f2fs_leave_shrinker(sbi); 1628 mutex_unlock(&sbi->umount_mutex); 1629 1630 /* our cp_error case, we can wait for any writeback page */ 1631 f2fs_flush_merged_writes(sbi); 1632 1633 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1634 1635 if (err || f2fs_cp_error(sbi)) { 1636 truncate_inode_pages_final(NODE_MAPPING(sbi)); 1637 truncate_inode_pages_final(META_MAPPING(sbi)); 1638 } 1639 1640 for (i = 0; i < NR_COUNT_TYPE; i++) { 1641 if (!get_pages(sbi, i)) 1642 continue; 1643 f2fs_err(sbi, "detect filesystem reference count leak during " 1644 "umount, type: %d, count: %lld", i, get_pages(sbi, i)); 1645 f2fs_bug_on(sbi, 1); 1646 } 1647 1648 f2fs_bug_on(sbi, sbi->fsync_node_num); 1649 1650 f2fs_destroy_compress_inode(sbi); 1651 1652 iput(sbi->node_inode); 1653 sbi->node_inode = NULL; 1654 1655 iput(sbi->meta_inode); 1656 sbi->meta_inode = NULL; 1657 1658 /* 1659 * iput() can update stat information, if f2fs_write_checkpoint() 1660 * above failed with error. 1661 */ 1662 f2fs_destroy_stats(sbi); 1663 1664 /* destroy f2fs internal modules */ 1665 f2fs_destroy_node_manager(sbi); 1666 f2fs_destroy_segment_manager(sbi); 1667 1668 /* flush s_error_work before sbi destroy */ 1669 flush_work(&sbi->s_error_work); 1670 1671 f2fs_destroy_post_read_wq(sbi); 1672 1673 kvfree(sbi->ckpt); 1674 1675 if (sbi->s_chksum_driver) 1676 crypto_free_shash(sbi->s_chksum_driver); 1677 kfree(sbi->raw_super); 1678 1679 f2fs_destroy_page_array_cache(sbi); 1680 f2fs_destroy_xattr_caches(sbi); 1681 #ifdef CONFIG_QUOTA 1682 for (i = 0; i < MAXQUOTAS; i++) 1683 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1684 #endif 1685 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1686 destroy_percpu_info(sbi); 1687 f2fs_destroy_iostat(sbi); 1688 for (i = 0; i < NR_PAGE_TYPE; i++) 1689 kvfree(sbi->write_io[i]); 1690 #if IS_ENABLED(CONFIG_UNICODE) 1691 utf8_unload(sb->s_encoding); 1692 #endif 1693 } 1694 f2fs_sync_fs(struct super_block * sb,int sync)1695 int f2fs_sync_fs(struct super_block *sb, int sync) 1696 { 1697 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1698 int err = 0; 1699 1700 if (unlikely(f2fs_cp_error(sbi))) 1701 return 0; 1702 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1703 return 0; 1704 1705 trace_f2fs_sync_fs(sb, sync); 1706 1707 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1708 return -EAGAIN; 1709 1710 if (sync) { 1711 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1712 err = f2fs_issue_checkpoint(sbi); 1713 } 1714 1715 return err; 1716 } 1717 f2fs_freeze(struct super_block * sb)1718 static int f2fs_freeze(struct super_block *sb) 1719 { 1720 if (f2fs_readonly(sb)) 1721 return 0; 1722 1723 /* IO error happened before */ 1724 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1725 return -EIO; 1726 1727 /* must be clean, since sync_filesystem() was already called */ 1728 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1729 return -EINVAL; 1730 1731 /* Let's flush checkpoints and stop the thread. */ 1732 f2fs_flush_ckpt_thread(F2FS_SB(sb)); 1733 1734 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */ 1735 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1736 return 0; 1737 } 1738 f2fs_unfreeze(struct super_block * sb)1739 static int f2fs_unfreeze(struct super_block *sb) 1740 { 1741 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1742 return 0; 1743 } 1744 1745 #ifdef CONFIG_QUOTA f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1746 static int f2fs_statfs_project(struct super_block *sb, 1747 kprojid_t projid, struct kstatfs *buf) 1748 { 1749 struct kqid qid; 1750 struct dquot *dquot; 1751 u64 limit; 1752 u64 curblock; 1753 1754 qid = make_kqid_projid(projid); 1755 dquot = dqget(sb, qid); 1756 if (IS_ERR(dquot)) 1757 return PTR_ERR(dquot); 1758 spin_lock(&dquot->dq_dqb_lock); 1759 1760 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1761 dquot->dq_dqb.dqb_bhardlimit); 1762 if (limit) 1763 limit >>= sb->s_blocksize_bits; 1764 1765 if (limit && buf->f_blocks > limit) { 1766 curblock = (dquot->dq_dqb.dqb_curspace + 1767 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1768 buf->f_blocks = limit; 1769 buf->f_bfree = buf->f_bavail = 1770 (buf->f_blocks > curblock) ? 1771 (buf->f_blocks - curblock) : 0; 1772 } 1773 1774 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1775 dquot->dq_dqb.dqb_ihardlimit); 1776 1777 if (limit && buf->f_files > limit) { 1778 buf->f_files = limit; 1779 buf->f_ffree = 1780 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1781 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1782 } 1783 1784 spin_unlock(&dquot->dq_dqb_lock); 1785 dqput(dquot); 1786 return 0; 1787 } 1788 #endif 1789 f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1790 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1791 { 1792 struct super_block *sb = dentry->d_sb; 1793 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1794 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1795 block_t total_count, user_block_count, start_count; 1796 u64 avail_node_count; 1797 unsigned int total_valid_node_count; 1798 1799 total_count = le64_to_cpu(sbi->raw_super->block_count); 1800 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1801 buf->f_type = F2FS_SUPER_MAGIC; 1802 buf->f_bsize = sbi->blocksize; 1803 1804 buf->f_blocks = total_count - start_count; 1805 1806 spin_lock(&sbi->stat_lock); 1807 1808 user_block_count = sbi->user_block_count; 1809 total_valid_node_count = valid_node_count(sbi); 1810 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1811 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1812 sbi->current_reserved_blocks; 1813 1814 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1815 buf->f_bfree = 0; 1816 else 1817 buf->f_bfree -= sbi->unusable_block_count; 1818 spin_unlock(&sbi->stat_lock); 1819 1820 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1821 buf->f_bavail = buf->f_bfree - 1822 F2FS_OPTION(sbi).root_reserved_blocks; 1823 else 1824 buf->f_bavail = 0; 1825 1826 if (avail_node_count > user_block_count) { 1827 buf->f_files = user_block_count; 1828 buf->f_ffree = buf->f_bavail; 1829 } else { 1830 buf->f_files = avail_node_count; 1831 buf->f_ffree = min(avail_node_count - total_valid_node_count, 1832 buf->f_bavail); 1833 } 1834 1835 buf->f_namelen = F2FS_NAME_LEN; 1836 buf->f_fsid = u64_to_fsid(id); 1837 1838 #ifdef CONFIG_QUOTA 1839 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1840 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1841 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1842 } 1843 #endif 1844 return 0; 1845 } 1846 f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1847 static inline void f2fs_show_quota_options(struct seq_file *seq, 1848 struct super_block *sb) 1849 { 1850 #ifdef CONFIG_QUOTA 1851 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1852 1853 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1854 char *fmtname = ""; 1855 1856 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1857 case QFMT_VFS_OLD: 1858 fmtname = "vfsold"; 1859 break; 1860 case QFMT_VFS_V0: 1861 fmtname = "vfsv0"; 1862 break; 1863 case QFMT_VFS_V1: 1864 fmtname = "vfsv1"; 1865 break; 1866 } 1867 seq_printf(seq, ",jqfmt=%s", fmtname); 1868 } 1869 1870 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1871 seq_show_option(seq, "usrjquota", 1872 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1873 1874 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1875 seq_show_option(seq, "grpjquota", 1876 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1877 1878 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1879 seq_show_option(seq, "prjjquota", 1880 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1881 #endif 1882 } 1883 1884 #ifdef CONFIG_F2FS_FS_COMPRESSION f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1885 static inline void f2fs_show_compress_options(struct seq_file *seq, 1886 struct super_block *sb) 1887 { 1888 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1889 char *algtype = ""; 1890 int i; 1891 1892 if (!f2fs_sb_has_compression(sbi)) 1893 return; 1894 1895 switch (F2FS_OPTION(sbi).compress_algorithm) { 1896 case COMPRESS_LZO: 1897 algtype = "lzo"; 1898 break; 1899 case COMPRESS_LZ4: 1900 algtype = "lz4"; 1901 break; 1902 case COMPRESS_ZSTD: 1903 algtype = "zstd"; 1904 break; 1905 case COMPRESS_LZORLE: 1906 algtype = "lzo-rle"; 1907 break; 1908 } 1909 seq_printf(seq, ",compress_algorithm=%s", algtype); 1910 1911 if (F2FS_OPTION(sbi).compress_level) 1912 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1913 1914 seq_printf(seq, ",compress_log_size=%u", 1915 F2FS_OPTION(sbi).compress_log_size); 1916 1917 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1918 seq_printf(seq, ",compress_extension=%s", 1919 F2FS_OPTION(sbi).extensions[i]); 1920 } 1921 1922 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1923 seq_printf(seq, ",nocompress_extension=%s", 1924 F2FS_OPTION(sbi).noextensions[i]); 1925 } 1926 1927 if (F2FS_OPTION(sbi).compress_chksum) 1928 seq_puts(seq, ",compress_chksum"); 1929 1930 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1931 seq_printf(seq, ",compress_mode=%s", "fs"); 1932 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1933 seq_printf(seq, ",compress_mode=%s", "user"); 1934 1935 if (test_opt(sbi, COMPRESS_CACHE)) 1936 seq_puts(seq, ",compress_cache"); 1937 } 1938 #endif 1939 f2fs_show_options(struct seq_file * seq,struct dentry * root)1940 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1941 { 1942 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1943 1944 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1945 seq_printf(seq, ",background_gc=%s", "sync"); 1946 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1947 seq_printf(seq, ",background_gc=%s", "on"); 1948 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1949 seq_printf(seq, ",background_gc=%s", "off"); 1950 1951 if (test_opt(sbi, GC_MERGE)) 1952 seq_puts(seq, ",gc_merge"); 1953 else 1954 seq_puts(seq, ",nogc_merge"); 1955 1956 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1957 seq_puts(seq, ",disable_roll_forward"); 1958 if (test_opt(sbi, NORECOVERY)) 1959 seq_puts(seq, ",norecovery"); 1960 if (test_opt(sbi, DISCARD)) { 1961 seq_puts(seq, ",discard"); 1962 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 1963 seq_printf(seq, ",discard_unit=%s", "block"); 1964 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 1965 seq_printf(seq, ",discard_unit=%s", "segment"); 1966 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 1967 seq_printf(seq, ",discard_unit=%s", "section"); 1968 } else { 1969 seq_puts(seq, ",nodiscard"); 1970 } 1971 #ifdef CONFIG_F2FS_FS_XATTR 1972 if (test_opt(sbi, XATTR_USER)) 1973 seq_puts(seq, ",user_xattr"); 1974 else 1975 seq_puts(seq, ",nouser_xattr"); 1976 if (test_opt(sbi, INLINE_XATTR)) 1977 seq_puts(seq, ",inline_xattr"); 1978 else 1979 seq_puts(seq, ",noinline_xattr"); 1980 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1981 seq_printf(seq, ",inline_xattr_size=%u", 1982 F2FS_OPTION(sbi).inline_xattr_size); 1983 #endif 1984 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1985 if (test_opt(sbi, POSIX_ACL)) 1986 seq_puts(seq, ",acl"); 1987 else 1988 seq_puts(seq, ",noacl"); 1989 #endif 1990 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1991 seq_puts(seq, ",disable_ext_identify"); 1992 if (test_opt(sbi, INLINE_DATA)) 1993 seq_puts(seq, ",inline_data"); 1994 else 1995 seq_puts(seq, ",noinline_data"); 1996 if (test_opt(sbi, INLINE_DENTRY)) 1997 seq_puts(seq, ",inline_dentry"); 1998 else 1999 seq_puts(seq, ",noinline_dentry"); 2000 if (test_opt(sbi, FLUSH_MERGE)) 2001 seq_puts(seq, ",flush_merge"); 2002 else 2003 seq_puts(seq, ",noflush_merge"); 2004 if (test_opt(sbi, NOBARRIER)) 2005 seq_puts(seq, ",nobarrier"); 2006 else 2007 seq_puts(seq, ",barrier"); 2008 if (test_opt(sbi, FASTBOOT)) 2009 seq_puts(seq, ",fastboot"); 2010 if (test_opt(sbi, READ_EXTENT_CACHE)) 2011 seq_puts(seq, ",extent_cache"); 2012 else 2013 seq_puts(seq, ",noextent_cache"); 2014 if (test_opt(sbi, AGE_EXTENT_CACHE)) 2015 seq_puts(seq, ",age_extent_cache"); 2016 if (test_opt(sbi, DATA_FLUSH)) 2017 seq_puts(seq, ",data_flush"); 2018 2019 seq_puts(seq, ",mode="); 2020 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 2021 seq_puts(seq, "adaptive"); 2022 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 2023 seq_puts(seq, "lfs"); 2024 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 2025 seq_puts(seq, "fragment:segment"); 2026 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 2027 seq_puts(seq, "fragment:block"); 2028 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 2029 if (test_opt(sbi, RESERVE_ROOT)) 2030 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 2031 F2FS_OPTION(sbi).root_reserved_blocks, 2032 from_kuid_munged(&init_user_ns, 2033 F2FS_OPTION(sbi).s_resuid), 2034 from_kgid_munged(&init_user_ns, 2035 F2FS_OPTION(sbi).s_resgid)); 2036 #ifdef CONFIG_F2FS_FAULT_INJECTION 2037 if (test_opt(sbi, FAULT_INJECTION)) { 2038 seq_printf(seq, ",fault_injection=%u", 2039 F2FS_OPTION(sbi).fault_info.inject_rate); 2040 seq_printf(seq, ",fault_type=%u", 2041 F2FS_OPTION(sbi).fault_info.inject_type); 2042 } 2043 #endif 2044 #ifdef CONFIG_QUOTA 2045 if (test_opt(sbi, QUOTA)) 2046 seq_puts(seq, ",quota"); 2047 if (test_opt(sbi, USRQUOTA)) 2048 seq_puts(seq, ",usrquota"); 2049 if (test_opt(sbi, GRPQUOTA)) 2050 seq_puts(seq, ",grpquota"); 2051 if (test_opt(sbi, PRJQUOTA)) 2052 seq_puts(seq, ",prjquota"); 2053 #endif 2054 f2fs_show_quota_options(seq, sbi->sb); 2055 2056 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 2057 2058 if (sbi->sb->s_flags & SB_INLINECRYPT) 2059 seq_puts(seq, ",inlinecrypt"); 2060 2061 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 2062 seq_printf(seq, ",alloc_mode=%s", "default"); 2063 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 2064 seq_printf(seq, ",alloc_mode=%s", "reuse"); 2065 2066 if (test_opt(sbi, DISABLE_CHECKPOINT)) 2067 seq_printf(seq, ",checkpoint=disable:%u", 2068 F2FS_OPTION(sbi).unusable_cap); 2069 if (test_opt(sbi, MERGE_CHECKPOINT)) 2070 seq_puts(seq, ",checkpoint_merge"); 2071 else 2072 seq_puts(seq, ",nocheckpoint_merge"); 2073 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 2074 seq_printf(seq, ",fsync_mode=%s", "posix"); 2075 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 2076 seq_printf(seq, ",fsync_mode=%s", "strict"); 2077 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 2078 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 2079 2080 #ifdef CONFIG_F2FS_FS_COMPRESSION 2081 f2fs_show_compress_options(seq, sbi->sb); 2082 #endif 2083 2084 if (test_opt(sbi, ATGC)) 2085 seq_puts(seq, ",atgc"); 2086 2087 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL) 2088 seq_printf(seq, ",memory=%s", "normal"); 2089 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW) 2090 seq_printf(seq, ",memory=%s", "low"); 2091 2092 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY) 2093 seq_printf(seq, ",errors=%s", "remount-ro"); 2094 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE) 2095 seq_printf(seq, ",errors=%s", "continue"); 2096 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC) 2097 seq_printf(seq, ",errors=%s", "panic"); 2098 2099 return 0; 2100 } 2101 default_options(struct f2fs_sb_info * sbi,bool remount)2102 static void default_options(struct f2fs_sb_info *sbi, bool remount) 2103 { 2104 /* init some FS parameters */ 2105 if (!remount) { 2106 set_opt(sbi, READ_EXTENT_CACHE); 2107 clear_opt(sbi, DISABLE_CHECKPOINT); 2108 2109 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2110 set_opt(sbi, DISCARD); 2111 2112 if (f2fs_sb_has_blkzoned(sbi)) 2113 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2114 else 2115 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2116 } 2117 2118 if (f2fs_sb_has_readonly(sbi)) 2119 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2120 else 2121 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2122 2123 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2124 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <= 2125 SMALL_VOLUME_SEGMENTS) 2126 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 2127 else 2128 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2129 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2130 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2131 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2132 if (f2fs_sb_has_compression(sbi)) { 2133 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2134 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2135 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2136 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2137 } 2138 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2139 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL; 2140 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE; 2141 2142 set_opt(sbi, INLINE_XATTR); 2143 set_opt(sbi, INLINE_DATA); 2144 set_opt(sbi, INLINE_DENTRY); 2145 set_opt(sbi, MERGE_CHECKPOINT); 2146 F2FS_OPTION(sbi).unusable_cap = 0; 2147 sbi->sb->s_flags |= SB_LAZYTIME; 2148 if (!f2fs_is_readonly(sbi)) 2149 set_opt(sbi, FLUSH_MERGE); 2150 if (f2fs_sb_has_blkzoned(sbi)) 2151 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2152 else 2153 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2154 2155 #ifdef CONFIG_F2FS_FS_XATTR 2156 set_opt(sbi, XATTR_USER); 2157 #endif 2158 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2159 set_opt(sbi, POSIX_ACL); 2160 #endif 2161 2162 f2fs_build_fault_attr(sbi, 0, 0); 2163 } 2164 2165 #ifdef CONFIG_QUOTA 2166 static int f2fs_enable_quotas(struct super_block *sb); 2167 #endif 2168 f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)2169 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2170 { 2171 unsigned int s_flags = sbi->sb->s_flags; 2172 struct cp_control cpc; 2173 unsigned int gc_mode = sbi->gc_mode; 2174 int err = 0; 2175 int ret; 2176 block_t unusable; 2177 2178 if (s_flags & SB_RDONLY) { 2179 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2180 return -EINVAL; 2181 } 2182 sbi->sb->s_flags |= SB_ACTIVE; 2183 2184 /* check if we need more GC first */ 2185 unusable = f2fs_get_unusable_blocks(sbi); 2186 if (!f2fs_disable_cp_again(sbi, unusable)) 2187 goto skip_gc; 2188 2189 f2fs_update_time(sbi, DISABLE_TIME); 2190 2191 sbi->gc_mode = GC_URGENT_HIGH; 2192 2193 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2194 struct f2fs_gc_control gc_control = { 2195 .victim_segno = NULL_SEGNO, 2196 .init_gc_type = FG_GC, 2197 .should_migrate_blocks = false, 2198 .err_gc_skipped = true, 2199 .no_bg_gc = true, 2200 .nr_free_secs = 1 }; 2201 2202 f2fs_down_write(&sbi->gc_lock); 2203 stat_inc_gc_call_count(sbi, FOREGROUND); 2204 err = f2fs_gc(sbi, &gc_control); 2205 if (err == -ENODATA) { 2206 err = 0; 2207 break; 2208 } 2209 if (err && err != -EAGAIN) 2210 break; 2211 } 2212 2213 ret = sync_filesystem(sbi->sb); 2214 if (ret || err) { 2215 err = ret ? ret : err; 2216 goto restore_flag; 2217 } 2218 2219 unusable = f2fs_get_unusable_blocks(sbi); 2220 if (f2fs_disable_cp_again(sbi, unusable)) { 2221 err = -EAGAIN; 2222 goto restore_flag; 2223 } 2224 2225 skip_gc: 2226 f2fs_down_write(&sbi->gc_lock); 2227 cpc.reason = CP_PAUSE; 2228 set_sbi_flag(sbi, SBI_CP_DISABLED); 2229 stat_inc_cp_call_count(sbi, TOTAL_CALL); 2230 err = f2fs_write_checkpoint(sbi, &cpc); 2231 if (err) 2232 goto out_unlock; 2233 2234 spin_lock(&sbi->stat_lock); 2235 sbi->unusable_block_count = unusable; 2236 spin_unlock(&sbi->stat_lock); 2237 2238 out_unlock: 2239 f2fs_up_write(&sbi->gc_lock); 2240 restore_flag: 2241 sbi->gc_mode = gc_mode; 2242 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2243 return err; 2244 } 2245 f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2246 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2247 { 2248 int retry = DEFAULT_RETRY_IO_COUNT; 2249 2250 /* we should flush all the data to keep data consistency */ 2251 do { 2252 sync_inodes_sb(sbi->sb); 2253 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2254 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2255 2256 if (unlikely(retry < 0)) 2257 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2258 2259 f2fs_down_write(&sbi->gc_lock); 2260 f2fs_dirty_to_prefree(sbi); 2261 2262 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2263 set_sbi_flag(sbi, SBI_IS_DIRTY); 2264 f2fs_up_write(&sbi->gc_lock); 2265 2266 f2fs_sync_fs(sbi->sb, 1); 2267 2268 /* Let's ensure there's no pending checkpoint anymore */ 2269 f2fs_flush_ckpt_thread(sbi); 2270 } 2271 f2fs_remount(struct super_block * sb,int * flags,char * data)2272 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2273 { 2274 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2275 struct f2fs_mount_info org_mount_opt; 2276 unsigned long old_sb_flags; 2277 int err; 2278 bool need_restart_gc = false, need_stop_gc = false; 2279 bool need_restart_flush = false, need_stop_flush = false; 2280 bool need_restart_discard = false, need_stop_discard = false; 2281 bool need_enable_checkpoint = false, need_disable_checkpoint = false; 2282 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE); 2283 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE); 2284 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2285 bool no_atgc = !test_opt(sbi, ATGC); 2286 bool no_discard = !test_opt(sbi, DISCARD); 2287 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2288 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2289 #ifdef CONFIG_QUOTA 2290 int i, j; 2291 #endif 2292 2293 /* 2294 * Save the old mount options in case we 2295 * need to restore them. 2296 */ 2297 org_mount_opt = sbi->mount_opt; 2298 old_sb_flags = sb->s_flags; 2299 2300 #ifdef CONFIG_QUOTA 2301 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2302 for (i = 0; i < MAXQUOTAS; i++) { 2303 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2304 org_mount_opt.s_qf_names[i] = 2305 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2306 GFP_KERNEL); 2307 if (!org_mount_opt.s_qf_names[i]) { 2308 for (j = 0; j < i; j++) 2309 kfree(org_mount_opt.s_qf_names[j]); 2310 return -ENOMEM; 2311 } 2312 } else { 2313 org_mount_opt.s_qf_names[i] = NULL; 2314 } 2315 } 2316 #endif 2317 2318 /* recover superblocks we couldn't write due to previous RO mount */ 2319 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2320 err = f2fs_commit_super(sbi, false); 2321 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2322 err); 2323 if (!err) 2324 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2325 } 2326 2327 default_options(sbi, true); 2328 2329 /* parse mount options */ 2330 err = parse_options(sb, data, true); 2331 if (err) 2332 goto restore_opts; 2333 2334 #ifdef CONFIG_BLK_DEV_ZONED 2335 if (f2fs_sb_has_blkzoned(sbi) && 2336 sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) { 2337 f2fs_err(sbi, 2338 "zoned: max open zones %u is too small, need at least %u open zones", 2339 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs); 2340 err = -EINVAL; 2341 goto restore_opts; 2342 } 2343 #endif 2344 2345 /* flush outstanding errors before changing fs state */ 2346 flush_work(&sbi->s_error_work); 2347 2348 /* 2349 * Previous and new state of filesystem is RO, 2350 * so skip checking GC and FLUSH_MERGE conditions. 2351 */ 2352 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2353 goto skip; 2354 2355 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) { 2356 err = -EROFS; 2357 goto restore_opts; 2358 } 2359 2360 #ifdef CONFIG_QUOTA 2361 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2362 err = dquot_suspend(sb, -1); 2363 if (err < 0) 2364 goto restore_opts; 2365 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2366 /* dquot_resume needs RW */ 2367 sb->s_flags &= ~SB_RDONLY; 2368 if (sb_any_quota_suspended(sb)) { 2369 dquot_resume(sb, -1); 2370 } else if (f2fs_sb_has_quota_ino(sbi)) { 2371 err = f2fs_enable_quotas(sb); 2372 if (err) 2373 goto restore_opts; 2374 } 2375 } 2376 #endif 2377 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) { 2378 err = -EINVAL; 2379 f2fs_warn(sbi, "LFS is not compatible with IPU"); 2380 goto restore_opts; 2381 } 2382 2383 /* disallow enable atgc dynamically */ 2384 if (no_atgc == !!test_opt(sbi, ATGC)) { 2385 err = -EINVAL; 2386 f2fs_warn(sbi, "switch atgc option is not allowed"); 2387 goto restore_opts; 2388 } 2389 2390 /* disallow enable/disable extent_cache dynamically */ 2391 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) { 2392 err = -EINVAL; 2393 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2394 goto restore_opts; 2395 } 2396 /* disallow enable/disable age extent_cache dynamically */ 2397 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) { 2398 err = -EINVAL; 2399 f2fs_warn(sbi, "switch age_extent_cache option is not allowed"); 2400 goto restore_opts; 2401 } 2402 2403 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2404 err = -EINVAL; 2405 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2406 goto restore_opts; 2407 } 2408 2409 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2410 err = -EINVAL; 2411 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2412 goto restore_opts; 2413 } 2414 2415 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2416 err = -EINVAL; 2417 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2418 goto restore_opts; 2419 } 2420 2421 /* 2422 * We stop the GC thread if FS is mounted as RO 2423 * or if background_gc = off is passed in mount 2424 * option. Also sync the filesystem. 2425 */ 2426 if ((*flags & SB_RDONLY) || 2427 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2428 !test_opt(sbi, GC_MERGE))) { 2429 if (sbi->gc_thread) { 2430 f2fs_stop_gc_thread(sbi); 2431 need_restart_gc = true; 2432 } 2433 } else if (!sbi->gc_thread) { 2434 err = f2fs_start_gc_thread(sbi); 2435 if (err) 2436 goto restore_opts; 2437 need_stop_gc = true; 2438 } 2439 2440 if (*flags & SB_RDONLY) { 2441 sync_inodes_sb(sb); 2442 2443 set_sbi_flag(sbi, SBI_IS_DIRTY); 2444 set_sbi_flag(sbi, SBI_IS_CLOSE); 2445 f2fs_sync_fs(sb, 1); 2446 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2447 } 2448 2449 /* 2450 * We stop issue flush thread if FS is mounted as RO 2451 * or if flush_merge is not passed in mount option. 2452 */ 2453 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2454 clear_opt(sbi, FLUSH_MERGE); 2455 f2fs_destroy_flush_cmd_control(sbi, false); 2456 need_restart_flush = true; 2457 } else { 2458 err = f2fs_create_flush_cmd_control(sbi); 2459 if (err) 2460 goto restore_gc; 2461 need_stop_flush = true; 2462 } 2463 2464 if (no_discard == !!test_opt(sbi, DISCARD)) { 2465 if (test_opt(sbi, DISCARD)) { 2466 err = f2fs_start_discard_thread(sbi); 2467 if (err) 2468 goto restore_flush; 2469 need_stop_discard = true; 2470 } else { 2471 f2fs_stop_discard_thread(sbi); 2472 f2fs_issue_discard_timeout(sbi); 2473 need_restart_discard = true; 2474 } 2475 } 2476 2477 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2478 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2479 err = f2fs_disable_checkpoint(sbi); 2480 if (err) 2481 goto restore_discard; 2482 need_enable_checkpoint = true; 2483 } else { 2484 f2fs_enable_checkpoint(sbi); 2485 need_disable_checkpoint = true; 2486 } 2487 } 2488 2489 /* 2490 * Place this routine at the end, since a new checkpoint would be 2491 * triggered while remount and we need to take care of it before 2492 * returning from remount. 2493 */ 2494 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2495 !test_opt(sbi, MERGE_CHECKPOINT)) { 2496 f2fs_stop_ckpt_thread(sbi); 2497 } else { 2498 /* Flush if the prevous checkpoint, if exists. */ 2499 f2fs_flush_ckpt_thread(sbi); 2500 2501 err = f2fs_start_ckpt_thread(sbi); 2502 if (err) { 2503 f2fs_err(sbi, 2504 "Failed to start F2FS issue_checkpoint_thread (%d)", 2505 err); 2506 goto restore_checkpoint; 2507 } 2508 } 2509 2510 skip: 2511 #ifdef CONFIG_QUOTA 2512 /* Release old quota file names */ 2513 for (i = 0; i < MAXQUOTAS; i++) 2514 kfree(org_mount_opt.s_qf_names[i]); 2515 #endif 2516 /* Update the POSIXACL Flag */ 2517 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2518 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2519 2520 limit_reserve_root(sbi); 2521 adjust_unusable_cap_perc(sbi); 2522 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2523 return 0; 2524 restore_checkpoint: 2525 if (need_enable_checkpoint) { 2526 f2fs_enable_checkpoint(sbi); 2527 } else if (need_disable_checkpoint) { 2528 if (f2fs_disable_checkpoint(sbi)) 2529 f2fs_warn(sbi, "checkpoint has not been disabled"); 2530 } 2531 restore_discard: 2532 if (need_restart_discard) { 2533 if (f2fs_start_discard_thread(sbi)) 2534 f2fs_warn(sbi, "discard has been stopped"); 2535 } else if (need_stop_discard) { 2536 f2fs_stop_discard_thread(sbi); 2537 } 2538 restore_flush: 2539 if (need_restart_flush) { 2540 if (f2fs_create_flush_cmd_control(sbi)) 2541 f2fs_warn(sbi, "background flush thread has stopped"); 2542 } else if (need_stop_flush) { 2543 clear_opt(sbi, FLUSH_MERGE); 2544 f2fs_destroy_flush_cmd_control(sbi, false); 2545 } 2546 restore_gc: 2547 if (need_restart_gc) { 2548 if (f2fs_start_gc_thread(sbi)) 2549 f2fs_warn(sbi, "background gc thread has stopped"); 2550 } else if (need_stop_gc) { 2551 f2fs_stop_gc_thread(sbi); 2552 } 2553 restore_opts: 2554 #ifdef CONFIG_QUOTA 2555 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2556 for (i = 0; i < MAXQUOTAS; i++) { 2557 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2558 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2559 } 2560 #endif 2561 sbi->mount_opt = org_mount_opt; 2562 sb->s_flags = old_sb_flags; 2563 return err; 2564 } 2565 f2fs_shutdown(struct super_block * sb)2566 static void f2fs_shutdown(struct super_block *sb) 2567 { 2568 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false, false); 2569 } 2570 2571 #ifdef CONFIG_QUOTA f2fs_need_recovery(struct f2fs_sb_info * sbi)2572 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi) 2573 { 2574 /* need to recovery orphan */ 2575 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) 2576 return true; 2577 /* need to recovery data */ 2578 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 2579 return false; 2580 if (test_opt(sbi, NORECOVERY)) 2581 return false; 2582 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG); 2583 } 2584 f2fs_recover_quota_begin(struct f2fs_sb_info * sbi)2585 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi) 2586 { 2587 bool readonly = f2fs_readonly(sbi->sb); 2588 2589 if (!f2fs_need_recovery(sbi)) 2590 return false; 2591 2592 /* it doesn't need to check f2fs_sb_has_readonly() */ 2593 if (f2fs_hw_is_readonly(sbi)) 2594 return false; 2595 2596 if (readonly) { 2597 sbi->sb->s_flags &= ~SB_RDONLY; 2598 set_sbi_flag(sbi, SBI_IS_WRITABLE); 2599 } 2600 2601 /* 2602 * Turn on quotas which were not enabled for read-only mounts if 2603 * filesystem has quota feature, so that they are updated correctly. 2604 */ 2605 return f2fs_enable_quota_files(sbi, readonly); 2606 } 2607 f2fs_recover_quota_end(struct f2fs_sb_info * sbi,bool quota_enabled)2608 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi, 2609 bool quota_enabled) 2610 { 2611 if (quota_enabled) 2612 f2fs_quota_off_umount(sbi->sb); 2613 2614 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) { 2615 clear_sbi_flag(sbi, SBI_IS_WRITABLE); 2616 sbi->sb->s_flags |= SB_RDONLY; 2617 } 2618 } 2619 2620 /* Read data from quotafile */ f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2621 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2622 size_t len, loff_t off) 2623 { 2624 struct inode *inode = sb_dqopt(sb)->files[type]; 2625 struct address_space *mapping = inode->i_mapping; 2626 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2627 int offset = off & (sb->s_blocksize - 1); 2628 int tocopy; 2629 size_t toread; 2630 loff_t i_size = i_size_read(inode); 2631 struct page *page; 2632 2633 if (off > i_size) 2634 return 0; 2635 2636 if (off + len > i_size) 2637 len = i_size - off; 2638 toread = len; 2639 while (toread > 0) { 2640 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2641 repeat: 2642 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2643 if (IS_ERR(page)) { 2644 if (PTR_ERR(page) == -ENOMEM) { 2645 memalloc_retry_wait(GFP_NOFS); 2646 goto repeat; 2647 } 2648 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2649 return PTR_ERR(page); 2650 } 2651 2652 lock_page(page); 2653 2654 if (unlikely(page->mapping != mapping)) { 2655 f2fs_put_page(page, 1); 2656 goto repeat; 2657 } 2658 if (unlikely(!PageUptodate(page))) { 2659 f2fs_put_page(page, 1); 2660 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2661 return -EIO; 2662 } 2663 2664 memcpy_from_page(data, page, offset, tocopy); 2665 f2fs_put_page(page, 1); 2666 2667 offset = 0; 2668 toread -= tocopy; 2669 data += tocopy; 2670 blkidx++; 2671 } 2672 return len; 2673 } 2674 2675 /* Write to quotafile */ f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2676 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2677 const char *data, size_t len, loff_t off) 2678 { 2679 struct inode *inode = sb_dqopt(sb)->files[type]; 2680 struct address_space *mapping = inode->i_mapping; 2681 const struct address_space_operations *a_ops = mapping->a_ops; 2682 int offset = off & (sb->s_blocksize - 1); 2683 size_t towrite = len; 2684 struct folio *folio; 2685 void *fsdata = NULL; 2686 int err = 0; 2687 int tocopy; 2688 2689 while (towrite > 0) { 2690 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2691 towrite); 2692 retry: 2693 err = a_ops->write_begin(NULL, mapping, off, tocopy, 2694 &folio, &fsdata); 2695 if (unlikely(err)) { 2696 if (err == -ENOMEM) { 2697 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2698 goto retry; 2699 } 2700 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2701 break; 2702 } 2703 2704 memcpy_to_folio(folio, offset_in_folio(folio, off), data, tocopy); 2705 2706 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2707 folio, fsdata); 2708 offset = 0; 2709 towrite -= tocopy; 2710 off += tocopy; 2711 data += tocopy; 2712 cond_resched(); 2713 } 2714 2715 if (len == towrite) 2716 return err; 2717 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 2718 f2fs_mark_inode_dirty_sync(inode, false); 2719 return len - towrite; 2720 } 2721 f2fs_dquot_initialize(struct inode * inode)2722 int f2fs_dquot_initialize(struct inode *inode) 2723 { 2724 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) 2725 return -ESRCH; 2726 2727 return dquot_initialize(inode); 2728 } 2729 f2fs_get_dquots(struct inode * inode)2730 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode) 2731 { 2732 return F2FS_I(inode)->i_dquot; 2733 } 2734 f2fs_get_reserved_space(struct inode * inode)2735 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2736 { 2737 return &F2FS_I(inode)->i_reserved_quota; 2738 } 2739 f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2740 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2741 { 2742 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2743 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2744 return 0; 2745 } 2746 2747 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2748 F2FS_OPTION(sbi).s_jquota_fmt, type); 2749 } 2750 f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2751 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2752 { 2753 int enabled = 0; 2754 int i, err; 2755 2756 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2757 err = f2fs_enable_quotas(sbi->sb); 2758 if (err) { 2759 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2760 return 0; 2761 } 2762 return 1; 2763 } 2764 2765 for (i = 0; i < MAXQUOTAS; i++) { 2766 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2767 err = f2fs_quota_on_mount(sbi, i); 2768 if (!err) { 2769 enabled = 1; 2770 continue; 2771 } 2772 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2773 err, i); 2774 } 2775 } 2776 return enabled; 2777 } 2778 f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2779 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2780 unsigned int flags) 2781 { 2782 struct inode *qf_inode; 2783 unsigned long qf_inum; 2784 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL; 2785 int err; 2786 2787 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2788 2789 qf_inum = f2fs_qf_ino(sb, type); 2790 if (!qf_inum) 2791 return -EPERM; 2792 2793 qf_inode = f2fs_iget(sb, qf_inum); 2794 if (IS_ERR(qf_inode)) { 2795 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2796 return PTR_ERR(qf_inode); 2797 } 2798 2799 /* Don't account quota for quota files to avoid recursion */ 2800 inode_lock(qf_inode); 2801 qf_inode->i_flags |= S_NOQUOTA; 2802 2803 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) { 2804 F2FS_I(qf_inode)->i_flags |= qf_flag; 2805 f2fs_set_inode_flags(qf_inode); 2806 } 2807 inode_unlock(qf_inode); 2808 2809 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2810 iput(qf_inode); 2811 return err; 2812 } 2813 f2fs_enable_quotas(struct super_block * sb)2814 static int f2fs_enable_quotas(struct super_block *sb) 2815 { 2816 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2817 int type, err = 0; 2818 unsigned long qf_inum; 2819 bool quota_mopt[MAXQUOTAS] = { 2820 test_opt(sbi, USRQUOTA), 2821 test_opt(sbi, GRPQUOTA), 2822 test_opt(sbi, PRJQUOTA), 2823 }; 2824 2825 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2826 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2827 return 0; 2828 } 2829 2830 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2831 2832 for (type = 0; type < MAXQUOTAS; type++) { 2833 qf_inum = f2fs_qf_ino(sb, type); 2834 if (qf_inum) { 2835 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2836 DQUOT_USAGE_ENABLED | 2837 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2838 if (err) { 2839 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2840 type, err); 2841 for (type--; type >= 0; type--) 2842 dquot_quota_off(sb, type); 2843 set_sbi_flag(F2FS_SB(sb), 2844 SBI_QUOTA_NEED_REPAIR); 2845 return err; 2846 } 2847 } 2848 } 2849 return 0; 2850 } 2851 f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2852 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 2853 { 2854 struct quota_info *dqopt = sb_dqopt(sbi->sb); 2855 struct address_space *mapping = dqopt->files[type]->i_mapping; 2856 int ret = 0; 2857 2858 ret = dquot_writeback_dquots(sbi->sb, type); 2859 if (ret) 2860 goto out; 2861 2862 ret = filemap_fdatawrite(mapping); 2863 if (ret) 2864 goto out; 2865 2866 /* if we are using journalled quota */ 2867 if (is_journalled_quota(sbi)) 2868 goto out; 2869 2870 ret = filemap_fdatawait(mapping); 2871 2872 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 2873 out: 2874 if (ret) 2875 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2876 return ret; 2877 } 2878 f2fs_quota_sync(struct super_block * sb,int type)2879 int f2fs_quota_sync(struct super_block *sb, int type) 2880 { 2881 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2882 struct quota_info *dqopt = sb_dqopt(sb); 2883 int cnt; 2884 int ret = 0; 2885 2886 /* 2887 * Now when everything is written we can discard the pagecache so 2888 * that userspace sees the changes. 2889 */ 2890 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2891 2892 if (type != -1 && cnt != type) 2893 continue; 2894 2895 if (!sb_has_quota_active(sb, cnt)) 2896 continue; 2897 2898 if (!f2fs_sb_has_quota_ino(sbi)) 2899 inode_lock(dqopt->files[cnt]); 2900 2901 /* 2902 * do_quotactl 2903 * f2fs_quota_sync 2904 * f2fs_down_read(quota_sem) 2905 * dquot_writeback_dquots() 2906 * f2fs_dquot_commit 2907 * block_operation 2908 * f2fs_down_read(quota_sem) 2909 */ 2910 f2fs_lock_op(sbi); 2911 f2fs_down_read(&sbi->quota_sem); 2912 2913 ret = f2fs_quota_sync_file(sbi, cnt); 2914 2915 f2fs_up_read(&sbi->quota_sem); 2916 f2fs_unlock_op(sbi); 2917 2918 if (!f2fs_sb_has_quota_ino(sbi)) 2919 inode_unlock(dqopt->files[cnt]); 2920 2921 if (ret) 2922 break; 2923 } 2924 return ret; 2925 } 2926 f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2927 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2928 const struct path *path) 2929 { 2930 struct inode *inode; 2931 int err; 2932 2933 /* if quota sysfile exists, deny enabling quota with specific file */ 2934 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2935 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2936 return -EBUSY; 2937 } 2938 2939 if (path->dentry->d_sb != sb) 2940 return -EXDEV; 2941 2942 err = f2fs_quota_sync(sb, type); 2943 if (err) 2944 return err; 2945 2946 inode = d_inode(path->dentry); 2947 2948 err = filemap_fdatawrite(inode->i_mapping); 2949 if (err) 2950 return err; 2951 2952 err = filemap_fdatawait(inode->i_mapping); 2953 if (err) 2954 return err; 2955 2956 err = dquot_quota_on(sb, type, format_id, path); 2957 if (err) 2958 return err; 2959 2960 inode_lock(inode); 2961 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL; 2962 f2fs_set_inode_flags(inode); 2963 inode_unlock(inode); 2964 f2fs_mark_inode_dirty_sync(inode, false); 2965 2966 return 0; 2967 } 2968 __f2fs_quota_off(struct super_block * sb,int type)2969 static int __f2fs_quota_off(struct super_block *sb, int type) 2970 { 2971 struct inode *inode = sb_dqopt(sb)->files[type]; 2972 int err; 2973 2974 if (!inode || !igrab(inode)) 2975 return dquot_quota_off(sb, type); 2976 2977 err = f2fs_quota_sync(sb, type); 2978 if (err) 2979 goto out_put; 2980 2981 err = dquot_quota_off(sb, type); 2982 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2983 goto out_put; 2984 2985 inode_lock(inode); 2986 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL; 2987 f2fs_set_inode_flags(inode); 2988 inode_unlock(inode); 2989 f2fs_mark_inode_dirty_sync(inode, false); 2990 out_put: 2991 iput(inode); 2992 return err; 2993 } 2994 f2fs_quota_off(struct super_block * sb,int type)2995 static int f2fs_quota_off(struct super_block *sb, int type) 2996 { 2997 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2998 int err; 2999 3000 err = __f2fs_quota_off(sb, type); 3001 3002 /* 3003 * quotactl can shutdown journalled quota, result in inconsistence 3004 * between quota record and fs data by following updates, tag the 3005 * flag to let fsck be aware of it. 3006 */ 3007 if (is_journalled_quota(sbi)) 3008 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3009 return err; 3010 } 3011 f2fs_quota_off_umount(struct super_block * sb)3012 void f2fs_quota_off_umount(struct super_block *sb) 3013 { 3014 int type; 3015 int err; 3016 3017 for (type = 0; type < MAXQUOTAS; type++) { 3018 err = __f2fs_quota_off(sb, type); 3019 if (err) { 3020 int ret = dquot_quota_off(sb, type); 3021 3022 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 3023 type, err, ret); 3024 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 3025 } 3026 } 3027 /* 3028 * In case of checkpoint=disable, we must flush quota blocks. 3029 * This can cause NULL exception for node_inode in end_io, since 3030 * put_super already dropped it. 3031 */ 3032 sync_filesystem(sb); 3033 } 3034 f2fs_truncate_quota_inode_pages(struct super_block * sb)3035 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 3036 { 3037 struct quota_info *dqopt = sb_dqopt(sb); 3038 int type; 3039 3040 for (type = 0; type < MAXQUOTAS; type++) { 3041 if (!dqopt->files[type]) 3042 continue; 3043 f2fs_inode_synced(dqopt->files[type]); 3044 } 3045 } 3046 f2fs_dquot_commit(struct dquot * dquot)3047 static int f2fs_dquot_commit(struct dquot *dquot) 3048 { 3049 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3050 int ret; 3051 3052 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 3053 ret = dquot_commit(dquot); 3054 if (ret < 0) 3055 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3056 f2fs_up_read(&sbi->quota_sem); 3057 return ret; 3058 } 3059 f2fs_dquot_acquire(struct dquot * dquot)3060 static int f2fs_dquot_acquire(struct dquot *dquot) 3061 { 3062 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3063 int ret; 3064 3065 f2fs_down_read(&sbi->quota_sem); 3066 ret = dquot_acquire(dquot); 3067 if (ret < 0) 3068 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3069 f2fs_up_read(&sbi->quota_sem); 3070 return ret; 3071 } 3072 f2fs_dquot_release(struct dquot * dquot)3073 static int f2fs_dquot_release(struct dquot *dquot) 3074 { 3075 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3076 int ret = dquot_release(dquot); 3077 3078 if (ret < 0) 3079 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3080 return ret; 3081 } 3082 f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)3083 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 3084 { 3085 struct super_block *sb = dquot->dq_sb; 3086 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3087 int ret = dquot_mark_dquot_dirty(dquot); 3088 3089 /* if we are using journalled quota */ 3090 if (is_journalled_quota(sbi)) 3091 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 3092 3093 return ret; 3094 } 3095 f2fs_dquot_commit_info(struct super_block * sb,int type)3096 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 3097 { 3098 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3099 int ret = dquot_commit_info(sb, type); 3100 3101 if (ret < 0) 3102 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3103 return ret; 3104 } 3105 f2fs_get_projid(struct inode * inode,kprojid_t * projid)3106 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 3107 { 3108 *projid = F2FS_I(inode)->i_projid; 3109 return 0; 3110 } 3111 3112 static const struct dquot_operations f2fs_quota_operations = { 3113 .get_reserved_space = f2fs_get_reserved_space, 3114 .write_dquot = f2fs_dquot_commit, 3115 .acquire_dquot = f2fs_dquot_acquire, 3116 .release_dquot = f2fs_dquot_release, 3117 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 3118 .write_info = f2fs_dquot_commit_info, 3119 .alloc_dquot = dquot_alloc, 3120 .destroy_dquot = dquot_destroy, 3121 .get_projid = f2fs_get_projid, 3122 .get_next_id = dquot_get_next_id, 3123 }; 3124 3125 static const struct quotactl_ops f2fs_quotactl_ops = { 3126 .quota_on = f2fs_quota_on, 3127 .quota_off = f2fs_quota_off, 3128 .quota_sync = f2fs_quota_sync, 3129 .get_state = dquot_get_state, 3130 .set_info = dquot_set_dqinfo, 3131 .get_dqblk = dquot_get_dqblk, 3132 .set_dqblk = dquot_set_dqblk, 3133 .get_nextdqblk = dquot_get_next_dqblk, 3134 }; 3135 #else f2fs_dquot_initialize(struct inode * inode)3136 int f2fs_dquot_initialize(struct inode *inode) 3137 { 3138 return 0; 3139 } 3140 f2fs_quota_sync(struct super_block * sb,int type)3141 int f2fs_quota_sync(struct super_block *sb, int type) 3142 { 3143 return 0; 3144 } 3145 f2fs_quota_off_umount(struct super_block * sb)3146 void f2fs_quota_off_umount(struct super_block *sb) 3147 { 3148 } 3149 #endif 3150 3151 static const struct super_operations f2fs_sops = { 3152 .alloc_inode = f2fs_alloc_inode, 3153 .free_inode = f2fs_free_inode, 3154 .drop_inode = f2fs_drop_inode, 3155 .write_inode = f2fs_write_inode, 3156 .dirty_inode = f2fs_dirty_inode, 3157 .show_options = f2fs_show_options, 3158 #ifdef CONFIG_QUOTA 3159 .quota_read = f2fs_quota_read, 3160 .quota_write = f2fs_quota_write, 3161 .get_dquots = f2fs_get_dquots, 3162 #endif 3163 .evict_inode = f2fs_evict_inode, 3164 .put_super = f2fs_put_super, 3165 .sync_fs = f2fs_sync_fs, 3166 .freeze_fs = f2fs_freeze, 3167 .unfreeze_fs = f2fs_unfreeze, 3168 .statfs = f2fs_statfs, 3169 .remount_fs = f2fs_remount, 3170 .shutdown = f2fs_shutdown, 3171 }; 3172 3173 #ifdef CONFIG_FS_ENCRYPTION f2fs_get_context(struct inode * inode,void * ctx,size_t len)3174 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 3175 { 3176 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3177 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3178 ctx, len, NULL); 3179 } 3180 f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)3181 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 3182 void *fs_data) 3183 { 3184 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3185 3186 /* 3187 * Encrypting the root directory is not allowed because fsck 3188 * expects lost+found directory to exist and remain unencrypted 3189 * if LOST_FOUND feature is enabled. 3190 * 3191 */ 3192 if (f2fs_sb_has_lost_found(sbi) && 3193 inode->i_ino == F2FS_ROOT_INO(sbi)) 3194 return -EPERM; 3195 3196 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3197 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3198 ctx, len, fs_data, XATTR_CREATE); 3199 } 3200 f2fs_get_dummy_policy(struct super_block * sb)3201 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 3202 { 3203 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 3204 } 3205 f2fs_has_stable_inodes(struct super_block * sb)3206 static bool f2fs_has_stable_inodes(struct super_block *sb) 3207 { 3208 return true; 3209 } 3210 f2fs_get_devices(struct super_block * sb,unsigned int * num_devs)3211 static struct block_device **f2fs_get_devices(struct super_block *sb, 3212 unsigned int *num_devs) 3213 { 3214 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3215 struct block_device **devs; 3216 int i; 3217 3218 if (!f2fs_is_multi_device(sbi)) 3219 return NULL; 3220 3221 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL); 3222 if (!devs) 3223 return ERR_PTR(-ENOMEM); 3224 3225 for (i = 0; i < sbi->s_ndevs; i++) 3226 devs[i] = FDEV(i).bdev; 3227 *num_devs = sbi->s_ndevs; 3228 return devs; 3229 } 3230 3231 static const struct fscrypt_operations f2fs_cryptops = { 3232 .needs_bounce_pages = 1, 3233 .has_32bit_inodes = 1, 3234 .supports_subblock_data_units = 1, 3235 .legacy_key_prefix = "f2fs:", 3236 .get_context = f2fs_get_context, 3237 .set_context = f2fs_set_context, 3238 .get_dummy_policy = f2fs_get_dummy_policy, 3239 .empty_dir = f2fs_empty_dir, 3240 .has_stable_inodes = f2fs_has_stable_inodes, 3241 .get_devices = f2fs_get_devices, 3242 }; 3243 #endif 3244 f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)3245 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3246 u64 ino, u32 generation) 3247 { 3248 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3249 struct inode *inode; 3250 3251 if (f2fs_check_nid_range(sbi, ino)) 3252 return ERR_PTR(-ESTALE); 3253 3254 /* 3255 * f2fs_iget isn't quite right if the inode is currently unallocated! 3256 * However f2fs_iget currently does appropriate checks to handle stale 3257 * inodes so everything is OK. 3258 */ 3259 inode = f2fs_iget(sb, ino); 3260 if (IS_ERR(inode)) 3261 return ERR_CAST(inode); 3262 if (unlikely(generation && inode->i_generation != generation)) { 3263 /* we didn't find the right inode.. */ 3264 iput(inode); 3265 return ERR_PTR(-ESTALE); 3266 } 3267 return inode; 3268 } 3269 f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3270 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3271 int fh_len, int fh_type) 3272 { 3273 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3274 f2fs_nfs_get_inode); 3275 } 3276 f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3277 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3278 int fh_len, int fh_type) 3279 { 3280 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3281 f2fs_nfs_get_inode); 3282 } 3283 3284 static const struct export_operations f2fs_export_ops = { 3285 .encode_fh = generic_encode_ino32_fh, 3286 .fh_to_dentry = f2fs_fh_to_dentry, 3287 .fh_to_parent = f2fs_fh_to_parent, 3288 .get_parent = f2fs_get_parent, 3289 }; 3290 max_file_blocks(struct inode * inode)3291 loff_t max_file_blocks(struct inode *inode) 3292 { 3293 loff_t result = 0; 3294 loff_t leaf_count; 3295 3296 /* 3297 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3298 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3299 * space in inode.i_addr, it will be more safe to reassign 3300 * result as zero. 3301 */ 3302 3303 if (inode && f2fs_compressed_file(inode)) 3304 leaf_count = ADDRS_PER_BLOCK(inode); 3305 else 3306 leaf_count = DEF_ADDRS_PER_BLOCK; 3307 3308 /* two direct node blocks */ 3309 result += (leaf_count * 2); 3310 3311 /* two indirect node blocks */ 3312 leaf_count *= NIDS_PER_BLOCK; 3313 result += (leaf_count * 2); 3314 3315 /* one double indirect node block */ 3316 leaf_count *= NIDS_PER_BLOCK; 3317 result += leaf_count; 3318 3319 /* 3320 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with 3321 * a 4K crypto data unit, we must restrict the max filesize to what can 3322 * fit within U32_MAX + 1 data units. 3323 */ 3324 3325 result = min(result, F2FS_BYTES_TO_BLK(((loff_t)U32_MAX + 1) * 4096)); 3326 3327 return result; 3328 } 3329 __f2fs_commit_super(struct f2fs_sb_info * sbi,struct folio * folio,pgoff_t index,bool update)3330 static int __f2fs_commit_super(struct f2fs_sb_info *sbi, struct folio *folio, 3331 pgoff_t index, bool update) 3332 { 3333 struct bio *bio; 3334 /* it's rare case, we can do fua all the time */ 3335 blk_opf_t opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH | REQ_FUA; 3336 int ret; 3337 3338 folio_lock(folio); 3339 folio_wait_writeback(folio); 3340 if (update) 3341 memcpy(F2FS_SUPER_BLOCK(folio, index), F2FS_RAW_SUPER(sbi), 3342 sizeof(struct f2fs_super_block)); 3343 folio_mark_dirty(folio); 3344 folio_clear_dirty_for_io(folio); 3345 folio_start_writeback(folio); 3346 folio_unlock(folio); 3347 3348 bio = bio_alloc(sbi->sb->s_bdev, 1, opf, GFP_NOFS); 3349 3350 /* it doesn't need to set crypto context for superblock update */ 3351 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(folio_index(folio)); 3352 3353 if (!bio_add_folio(bio, folio, folio_size(folio), 0)) 3354 f2fs_bug_on(sbi, 1); 3355 3356 ret = submit_bio_wait(bio); 3357 folio_end_writeback(folio); 3358 3359 return ret; 3360 } 3361 sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct folio * folio,pgoff_t index)3362 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3363 struct folio *folio, pgoff_t index) 3364 { 3365 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index); 3366 struct super_block *sb = sbi->sb; 3367 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3368 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3369 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3370 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3371 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3372 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3373 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3374 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3375 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3376 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3377 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3378 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3379 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3380 u64 main_end_blkaddr = main_blkaddr + 3381 ((u64)segment_count_main << log_blocks_per_seg); 3382 u64 seg_end_blkaddr = segment0_blkaddr + 3383 ((u64)segment_count << log_blocks_per_seg); 3384 3385 if (segment0_blkaddr != cp_blkaddr) { 3386 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3387 segment0_blkaddr, cp_blkaddr); 3388 return true; 3389 } 3390 3391 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3392 sit_blkaddr) { 3393 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3394 cp_blkaddr, sit_blkaddr, 3395 segment_count_ckpt << log_blocks_per_seg); 3396 return true; 3397 } 3398 3399 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3400 nat_blkaddr) { 3401 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3402 sit_blkaddr, nat_blkaddr, 3403 segment_count_sit << log_blocks_per_seg); 3404 return true; 3405 } 3406 3407 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3408 ssa_blkaddr) { 3409 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3410 nat_blkaddr, ssa_blkaddr, 3411 segment_count_nat << log_blocks_per_seg); 3412 return true; 3413 } 3414 3415 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3416 main_blkaddr) { 3417 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3418 ssa_blkaddr, main_blkaddr, 3419 segment_count_ssa << log_blocks_per_seg); 3420 return true; 3421 } 3422 3423 if (main_end_blkaddr > seg_end_blkaddr) { 3424 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3425 main_blkaddr, seg_end_blkaddr, 3426 segment_count_main << log_blocks_per_seg); 3427 return true; 3428 } else if (main_end_blkaddr < seg_end_blkaddr) { 3429 int err = 0; 3430 char *res; 3431 3432 /* fix in-memory information all the time */ 3433 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3434 segment0_blkaddr) >> log_blocks_per_seg); 3435 3436 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) { 3437 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3438 res = "internally"; 3439 } else { 3440 err = __f2fs_commit_super(sbi, folio, index, false); 3441 res = err ? "failed" : "done"; 3442 } 3443 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3444 res, main_blkaddr, seg_end_blkaddr, 3445 segment_count_main << log_blocks_per_seg); 3446 if (err) 3447 return true; 3448 } 3449 return false; 3450 } 3451 sanity_check_raw_super(struct f2fs_sb_info * sbi,struct folio * folio,pgoff_t index)3452 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3453 struct folio *folio, pgoff_t index) 3454 { 3455 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3456 block_t total_sections, blocks_per_seg; 3457 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index); 3458 size_t crc_offset = 0; 3459 __u32 crc = 0; 3460 3461 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3462 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3463 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3464 return -EINVAL; 3465 } 3466 3467 /* Check checksum_offset and crc in superblock */ 3468 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3469 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3470 if (crc_offset != 3471 offsetof(struct f2fs_super_block, crc)) { 3472 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3473 crc_offset); 3474 return -EFSCORRUPTED; 3475 } 3476 crc = le32_to_cpu(raw_super->crc); 3477 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3478 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3479 return -EFSCORRUPTED; 3480 } 3481 } 3482 3483 /* only support block_size equals to PAGE_SIZE */ 3484 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3485 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3486 le32_to_cpu(raw_super->log_blocksize), 3487 F2FS_BLKSIZE_BITS); 3488 return -EFSCORRUPTED; 3489 } 3490 3491 /* check log blocks per segment */ 3492 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3493 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3494 le32_to_cpu(raw_super->log_blocks_per_seg)); 3495 return -EFSCORRUPTED; 3496 } 3497 3498 /* Currently, support 512/1024/2048/4096/16K bytes sector size */ 3499 if (le32_to_cpu(raw_super->log_sectorsize) > 3500 F2FS_MAX_LOG_SECTOR_SIZE || 3501 le32_to_cpu(raw_super->log_sectorsize) < 3502 F2FS_MIN_LOG_SECTOR_SIZE) { 3503 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3504 le32_to_cpu(raw_super->log_sectorsize)); 3505 return -EFSCORRUPTED; 3506 } 3507 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3508 le32_to_cpu(raw_super->log_sectorsize) != 3509 F2FS_MAX_LOG_SECTOR_SIZE) { 3510 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3511 le32_to_cpu(raw_super->log_sectors_per_block), 3512 le32_to_cpu(raw_super->log_sectorsize)); 3513 return -EFSCORRUPTED; 3514 } 3515 3516 segment_count = le32_to_cpu(raw_super->segment_count); 3517 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3518 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3519 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3520 total_sections = le32_to_cpu(raw_super->section_count); 3521 3522 /* blocks_per_seg should be 512, given the above check */ 3523 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg)); 3524 3525 if (segment_count > F2FS_MAX_SEGMENT || 3526 segment_count < F2FS_MIN_SEGMENTS) { 3527 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3528 return -EFSCORRUPTED; 3529 } 3530 3531 if (total_sections > segment_count_main || total_sections < 1 || 3532 segs_per_sec > segment_count || !segs_per_sec) { 3533 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3534 segment_count, total_sections, segs_per_sec); 3535 return -EFSCORRUPTED; 3536 } 3537 3538 if (segment_count_main != total_sections * segs_per_sec) { 3539 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3540 segment_count_main, total_sections, segs_per_sec); 3541 return -EFSCORRUPTED; 3542 } 3543 3544 if ((segment_count / segs_per_sec) < total_sections) { 3545 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3546 segment_count, segs_per_sec, total_sections); 3547 return -EFSCORRUPTED; 3548 } 3549 3550 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3551 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3552 segment_count, le64_to_cpu(raw_super->block_count)); 3553 return -EFSCORRUPTED; 3554 } 3555 3556 if (RDEV(0).path[0]) { 3557 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3558 int i = 1; 3559 3560 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3561 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3562 i++; 3563 } 3564 if (segment_count != dev_seg_count) { 3565 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3566 segment_count, dev_seg_count); 3567 return -EFSCORRUPTED; 3568 } 3569 } else { 3570 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3571 !bdev_is_zoned(sbi->sb->s_bdev)) { 3572 f2fs_info(sbi, "Zoned block device path is missing"); 3573 return -EFSCORRUPTED; 3574 } 3575 } 3576 3577 if (secs_per_zone > total_sections || !secs_per_zone) { 3578 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3579 secs_per_zone, total_sections); 3580 return -EFSCORRUPTED; 3581 } 3582 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3583 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3584 (le32_to_cpu(raw_super->extension_count) + 3585 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3586 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3587 le32_to_cpu(raw_super->extension_count), 3588 raw_super->hot_ext_count, 3589 F2FS_MAX_EXTENSION); 3590 return -EFSCORRUPTED; 3591 } 3592 3593 if (le32_to_cpu(raw_super->cp_payload) >= 3594 (blocks_per_seg - F2FS_CP_PACKS - 3595 NR_CURSEG_PERSIST_TYPE)) { 3596 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3597 le32_to_cpu(raw_super->cp_payload), 3598 blocks_per_seg - F2FS_CP_PACKS - 3599 NR_CURSEG_PERSIST_TYPE); 3600 return -EFSCORRUPTED; 3601 } 3602 3603 /* check reserved ino info */ 3604 if (le32_to_cpu(raw_super->node_ino) != 1 || 3605 le32_to_cpu(raw_super->meta_ino) != 2 || 3606 le32_to_cpu(raw_super->root_ino) != 3) { 3607 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3608 le32_to_cpu(raw_super->node_ino), 3609 le32_to_cpu(raw_super->meta_ino), 3610 le32_to_cpu(raw_super->root_ino)); 3611 return -EFSCORRUPTED; 3612 } 3613 3614 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3615 if (sanity_check_area_boundary(sbi, folio, index)) 3616 return -EFSCORRUPTED; 3617 3618 return 0; 3619 } 3620 f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3621 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3622 { 3623 unsigned int total, fsmeta; 3624 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3625 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3626 unsigned int ovp_segments, reserved_segments; 3627 unsigned int main_segs, blocks_per_seg; 3628 unsigned int sit_segs, nat_segs; 3629 unsigned int sit_bitmap_size, nat_bitmap_size; 3630 unsigned int log_blocks_per_seg; 3631 unsigned int segment_count_main; 3632 unsigned int cp_pack_start_sum, cp_payload; 3633 block_t user_block_count, valid_user_blocks; 3634 block_t avail_node_count, valid_node_count; 3635 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3636 int i, j; 3637 3638 total = le32_to_cpu(raw_super->segment_count); 3639 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3640 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3641 fsmeta += sit_segs; 3642 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3643 fsmeta += nat_segs; 3644 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3645 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3646 3647 if (unlikely(fsmeta >= total)) 3648 return 1; 3649 3650 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3651 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3652 3653 if (!f2fs_sb_has_readonly(sbi) && 3654 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3655 ovp_segments == 0 || reserved_segments == 0)) { 3656 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3657 return 1; 3658 } 3659 user_block_count = le64_to_cpu(ckpt->user_block_count); 3660 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3661 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3662 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3663 if (!user_block_count || user_block_count >= 3664 segment_count_main << log_blocks_per_seg) { 3665 f2fs_err(sbi, "Wrong user_block_count: %u", 3666 user_block_count); 3667 return 1; 3668 } 3669 3670 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3671 if (valid_user_blocks > user_block_count) { 3672 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3673 valid_user_blocks, user_block_count); 3674 return 1; 3675 } 3676 3677 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3678 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3679 if (valid_node_count > avail_node_count) { 3680 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3681 valid_node_count, avail_node_count); 3682 return 1; 3683 } 3684 3685 main_segs = le32_to_cpu(raw_super->segment_count_main); 3686 blocks_per_seg = BLKS_PER_SEG(sbi); 3687 3688 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3689 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3690 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3691 return 1; 3692 3693 if (f2fs_sb_has_readonly(sbi)) 3694 goto check_data; 3695 3696 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3697 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3698 le32_to_cpu(ckpt->cur_node_segno[j])) { 3699 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3700 i, j, 3701 le32_to_cpu(ckpt->cur_node_segno[i])); 3702 return 1; 3703 } 3704 } 3705 } 3706 check_data: 3707 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3708 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3709 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3710 return 1; 3711 3712 if (f2fs_sb_has_readonly(sbi)) 3713 goto skip_cross; 3714 3715 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3716 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3717 le32_to_cpu(ckpt->cur_data_segno[j])) { 3718 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3719 i, j, 3720 le32_to_cpu(ckpt->cur_data_segno[i])); 3721 return 1; 3722 } 3723 } 3724 } 3725 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3726 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3727 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3728 le32_to_cpu(ckpt->cur_data_segno[j])) { 3729 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3730 i, j, 3731 le32_to_cpu(ckpt->cur_node_segno[i])); 3732 return 1; 3733 } 3734 } 3735 } 3736 skip_cross: 3737 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3738 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3739 3740 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3741 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3742 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3743 sit_bitmap_size, nat_bitmap_size); 3744 return 1; 3745 } 3746 3747 cp_pack_start_sum = __start_sum_addr(sbi); 3748 cp_payload = __cp_payload(sbi); 3749 if (cp_pack_start_sum < cp_payload + 1 || 3750 cp_pack_start_sum > blocks_per_seg - 1 - 3751 NR_CURSEG_PERSIST_TYPE) { 3752 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3753 cp_pack_start_sum); 3754 return 1; 3755 } 3756 3757 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3758 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3759 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3760 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3761 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3762 le32_to_cpu(ckpt->checksum_offset)); 3763 return 1; 3764 } 3765 3766 nat_blocks = nat_segs << log_blocks_per_seg; 3767 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 3768 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 3769 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 3770 (cp_payload + F2FS_CP_PACKS + 3771 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 3772 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 3773 cp_payload, nat_bits_blocks); 3774 return 1; 3775 } 3776 3777 if (unlikely(f2fs_cp_error(sbi))) { 3778 f2fs_err(sbi, "A bug case: need to run fsck"); 3779 return 1; 3780 } 3781 return 0; 3782 } 3783 init_sb_info(struct f2fs_sb_info * sbi)3784 static void init_sb_info(struct f2fs_sb_info *sbi) 3785 { 3786 struct f2fs_super_block *raw_super = sbi->raw_super; 3787 int i; 3788 3789 sbi->log_sectors_per_block = 3790 le32_to_cpu(raw_super->log_sectors_per_block); 3791 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3792 sbi->blocksize = BIT(sbi->log_blocksize); 3793 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3794 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg); 3795 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3796 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3797 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3798 sbi->total_node_count = SEGS_TO_BLKS(sbi, 3799 ((le32_to_cpu(raw_super->segment_count_nat) / 2) * 3800 NAT_ENTRY_PER_BLOCK)); 3801 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3802 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3803 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3804 sbi->cur_victim_sec = NULL_SECNO; 3805 sbi->gc_mode = GC_NORMAL; 3806 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3807 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3808 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3809 sbi->migration_granularity = SEGS_PER_SEC(sbi); 3810 sbi->migration_window_granularity = f2fs_sb_has_blkzoned(sbi) ? 3811 DEF_MIGRATION_WINDOW_GRANULARITY_ZONED : SEGS_PER_SEC(sbi); 3812 sbi->seq_file_ra_mul = MIN_RA_MUL; 3813 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 3814 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 3815 spin_lock_init(&sbi->gc_remaining_trials_lock); 3816 atomic64_set(&sbi->current_atomic_write, 0); 3817 3818 sbi->dir_level = DEF_DIR_LEVEL; 3819 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3820 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3821 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3822 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3823 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3824 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3825 DEF_UMOUNT_DISCARD_TIMEOUT; 3826 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3827 3828 for (i = 0; i < NR_COUNT_TYPE; i++) 3829 atomic_set(&sbi->nr_pages[i], 0); 3830 3831 for (i = 0; i < META; i++) 3832 atomic_set(&sbi->wb_sync_req[i], 0); 3833 3834 INIT_LIST_HEAD(&sbi->s_list); 3835 mutex_init(&sbi->umount_mutex); 3836 init_f2fs_rwsem(&sbi->io_order_lock); 3837 spin_lock_init(&sbi->cp_lock); 3838 3839 sbi->dirty_device = 0; 3840 spin_lock_init(&sbi->dev_lock); 3841 3842 init_f2fs_rwsem(&sbi->sb_lock); 3843 init_f2fs_rwsem(&sbi->pin_sem); 3844 } 3845 init_percpu_info(struct f2fs_sb_info * sbi)3846 static int init_percpu_info(struct f2fs_sb_info *sbi) 3847 { 3848 int err; 3849 3850 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3851 if (err) 3852 return err; 3853 3854 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL); 3855 if (err) 3856 goto err_valid_block; 3857 3858 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3859 GFP_KERNEL); 3860 if (err) 3861 goto err_node_block; 3862 return 0; 3863 3864 err_node_block: 3865 percpu_counter_destroy(&sbi->rf_node_block_count); 3866 err_valid_block: 3867 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3868 return err; 3869 } 3870 3871 #ifdef CONFIG_BLK_DEV_ZONED 3872 3873 struct f2fs_report_zones_args { 3874 struct f2fs_sb_info *sbi; 3875 struct f2fs_dev_info *dev; 3876 }; 3877 f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3878 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3879 void *data) 3880 { 3881 struct f2fs_report_zones_args *rz_args = data; 3882 block_t unusable_blocks = (zone->len - zone->capacity) >> 3883 F2FS_LOG_SECTORS_PER_BLOCK; 3884 3885 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3886 return 0; 3887 3888 set_bit(idx, rz_args->dev->blkz_seq); 3889 if (!rz_args->sbi->unusable_blocks_per_sec) { 3890 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks; 3891 return 0; 3892 } 3893 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) { 3894 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n"); 3895 return -EINVAL; 3896 } 3897 return 0; 3898 } 3899 init_blkz_info(struct f2fs_sb_info * sbi,int devi)3900 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3901 { 3902 struct block_device *bdev = FDEV(devi).bdev; 3903 sector_t nr_sectors = bdev_nr_sectors(bdev); 3904 struct f2fs_report_zones_args rep_zone_arg; 3905 u64 zone_sectors; 3906 unsigned int max_open_zones; 3907 int ret; 3908 3909 if (!f2fs_sb_has_blkzoned(sbi)) 3910 return 0; 3911 3912 if (bdev_is_zoned(FDEV(devi).bdev)) { 3913 max_open_zones = bdev_max_open_zones(bdev); 3914 if (max_open_zones && (max_open_zones < sbi->max_open_zones)) 3915 sbi->max_open_zones = max_open_zones; 3916 if (sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) { 3917 f2fs_err(sbi, 3918 "zoned: max open zones %u is too small, need at least %u open zones", 3919 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs); 3920 return -EINVAL; 3921 } 3922 } 3923 3924 zone_sectors = bdev_zone_sectors(bdev); 3925 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3926 SECTOR_TO_BLOCK(zone_sectors)) 3927 return -EINVAL; 3928 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors); 3929 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors), 3930 sbi->blocks_per_blkz); 3931 if (nr_sectors & (zone_sectors - 1)) 3932 FDEV(devi).nr_blkz++; 3933 3934 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3935 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3936 * sizeof(unsigned long), 3937 GFP_KERNEL); 3938 if (!FDEV(devi).blkz_seq) 3939 return -ENOMEM; 3940 3941 rep_zone_arg.sbi = sbi; 3942 rep_zone_arg.dev = &FDEV(devi); 3943 3944 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3945 &rep_zone_arg); 3946 if (ret < 0) 3947 return ret; 3948 return 0; 3949 } 3950 #endif 3951 3952 /* 3953 * Read f2fs raw super block. 3954 * Because we have two copies of super block, so read both of them 3955 * to get the first valid one. If any one of them is broken, we pass 3956 * them recovery flag back to the caller. 3957 */ read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3958 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3959 struct f2fs_super_block **raw_super, 3960 int *valid_super_block, int *recovery) 3961 { 3962 struct super_block *sb = sbi->sb; 3963 int block; 3964 struct folio *folio; 3965 struct f2fs_super_block *super; 3966 int err = 0; 3967 3968 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3969 if (!super) 3970 return -ENOMEM; 3971 3972 for (block = 0; block < 2; block++) { 3973 folio = read_mapping_folio(sb->s_bdev->bd_mapping, block, NULL); 3974 if (IS_ERR(folio)) { 3975 f2fs_err(sbi, "Unable to read %dth superblock", 3976 block + 1); 3977 err = PTR_ERR(folio); 3978 *recovery = 1; 3979 continue; 3980 } 3981 3982 /* sanity checking of raw super */ 3983 err = sanity_check_raw_super(sbi, folio, block); 3984 if (err) { 3985 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3986 block + 1); 3987 folio_put(folio); 3988 *recovery = 1; 3989 continue; 3990 } 3991 3992 if (!*raw_super) { 3993 memcpy(super, F2FS_SUPER_BLOCK(folio, block), 3994 sizeof(*super)); 3995 *valid_super_block = block; 3996 *raw_super = super; 3997 } 3998 folio_put(folio); 3999 } 4000 4001 /* No valid superblock */ 4002 if (!*raw_super) 4003 kfree(super); 4004 else 4005 err = 0; 4006 4007 return err; 4008 } 4009 f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)4010 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 4011 { 4012 struct folio *folio; 4013 pgoff_t index; 4014 __u32 crc = 0; 4015 int err; 4016 4017 if ((recover && f2fs_readonly(sbi->sb)) || 4018 f2fs_hw_is_readonly(sbi)) { 4019 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 4020 return -EROFS; 4021 } 4022 4023 /* we should update superblock crc here */ 4024 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 4025 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 4026 offsetof(struct f2fs_super_block, crc)); 4027 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 4028 } 4029 4030 /* write back-up superblock first */ 4031 index = sbi->valid_super_block ? 0 : 1; 4032 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL); 4033 if (IS_ERR(folio)) 4034 return PTR_ERR(folio); 4035 err = __f2fs_commit_super(sbi, folio, index, true); 4036 folio_put(folio); 4037 4038 /* if we are in recovery path, skip writing valid superblock */ 4039 if (recover || err) 4040 return err; 4041 4042 /* write current valid superblock */ 4043 index = sbi->valid_super_block; 4044 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL); 4045 if (IS_ERR(folio)) 4046 return PTR_ERR(folio); 4047 err = __f2fs_commit_super(sbi, folio, index, true); 4048 folio_put(folio); 4049 return err; 4050 } 4051 save_stop_reason(struct f2fs_sb_info * sbi,unsigned char reason)4052 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason) 4053 { 4054 unsigned long flags; 4055 4056 spin_lock_irqsave(&sbi->error_lock, flags); 4057 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0)) 4058 sbi->stop_reason[reason]++; 4059 spin_unlock_irqrestore(&sbi->error_lock, flags); 4060 } 4061 f2fs_record_stop_reason(struct f2fs_sb_info * sbi)4062 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi) 4063 { 4064 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4065 unsigned long flags; 4066 int err; 4067 4068 f2fs_down_write(&sbi->sb_lock); 4069 4070 spin_lock_irqsave(&sbi->error_lock, flags); 4071 if (sbi->error_dirty) { 4072 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4073 MAX_F2FS_ERRORS); 4074 sbi->error_dirty = false; 4075 } 4076 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON); 4077 spin_unlock_irqrestore(&sbi->error_lock, flags); 4078 4079 err = f2fs_commit_super(sbi, false); 4080 4081 f2fs_up_write(&sbi->sb_lock); 4082 if (err) 4083 f2fs_err_ratelimited(sbi, 4084 "f2fs_commit_super fails to record stop_reason, err:%d", 4085 err); 4086 } 4087 f2fs_save_errors(struct f2fs_sb_info * sbi,unsigned char flag)4088 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag) 4089 { 4090 unsigned long flags; 4091 4092 spin_lock_irqsave(&sbi->error_lock, flags); 4093 if (!test_bit(flag, (unsigned long *)sbi->errors)) { 4094 set_bit(flag, (unsigned long *)sbi->errors); 4095 sbi->error_dirty = true; 4096 } 4097 spin_unlock_irqrestore(&sbi->error_lock, flags); 4098 } 4099 f2fs_update_errors(struct f2fs_sb_info * sbi)4100 static bool f2fs_update_errors(struct f2fs_sb_info *sbi) 4101 { 4102 unsigned long flags; 4103 bool need_update = false; 4104 4105 spin_lock_irqsave(&sbi->error_lock, flags); 4106 if (sbi->error_dirty) { 4107 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4108 MAX_F2FS_ERRORS); 4109 sbi->error_dirty = false; 4110 need_update = true; 4111 } 4112 spin_unlock_irqrestore(&sbi->error_lock, flags); 4113 4114 return need_update; 4115 } 4116 f2fs_record_errors(struct f2fs_sb_info * sbi,unsigned char error)4117 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error) 4118 { 4119 int err; 4120 4121 f2fs_down_write(&sbi->sb_lock); 4122 4123 if (!f2fs_update_errors(sbi)) 4124 goto out_unlock; 4125 4126 err = f2fs_commit_super(sbi, false); 4127 if (err) 4128 f2fs_err_ratelimited(sbi, 4129 "f2fs_commit_super fails to record errors:%u, err:%d", 4130 error, err); 4131 out_unlock: 4132 f2fs_up_write(&sbi->sb_lock); 4133 } 4134 f2fs_handle_error(struct f2fs_sb_info * sbi,unsigned char error)4135 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error) 4136 { 4137 f2fs_save_errors(sbi, error); 4138 f2fs_record_errors(sbi, error); 4139 } 4140 f2fs_handle_error_async(struct f2fs_sb_info * sbi,unsigned char error)4141 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error) 4142 { 4143 f2fs_save_errors(sbi, error); 4144 4145 if (!sbi->error_dirty) 4146 return; 4147 if (!test_bit(error, (unsigned long *)sbi->errors)) 4148 return; 4149 schedule_work(&sbi->s_error_work); 4150 } 4151 system_going_down(void)4152 static bool system_going_down(void) 4153 { 4154 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 4155 || system_state == SYSTEM_RESTART; 4156 } 4157 f2fs_handle_critical_error(struct f2fs_sb_info * sbi,unsigned char reason,bool irq_context)4158 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason, 4159 bool irq_context) 4160 { 4161 struct super_block *sb = sbi->sb; 4162 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN; 4163 bool continue_fs = !shutdown && 4164 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE; 4165 4166 set_ckpt_flags(sbi, CP_ERROR_FLAG); 4167 4168 if (!f2fs_hw_is_readonly(sbi)) { 4169 save_stop_reason(sbi, reason); 4170 4171 if (irq_context && !shutdown) 4172 schedule_work(&sbi->s_error_work); 4173 else 4174 f2fs_record_stop_reason(sbi); 4175 } 4176 4177 /* 4178 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 4179 * could panic during 'reboot -f' as the underlying device got already 4180 * disabled. 4181 */ 4182 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC && 4183 !shutdown && !system_going_down() && 4184 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)) 4185 panic("F2FS-fs (device %s): panic forced after error\n", 4186 sb->s_id); 4187 4188 if (shutdown) 4189 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 4190 4191 /* 4192 * Continue filesystem operators if errors=continue. Should not set 4193 * RO by shutdown, since RO bypasses thaw_super which can hang the 4194 * system. 4195 */ 4196 if (continue_fs || f2fs_readonly(sb) || shutdown) { 4197 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason); 4198 return; 4199 } 4200 4201 f2fs_warn(sbi, "Remounting filesystem read-only"); 4202 4203 /* 4204 * We have already set CP_ERROR_FLAG flag to stop all updates 4205 * to filesystem, so it doesn't need to set SB_RDONLY flag here 4206 * because the flag should be set covered w/ sb->s_umount semaphore 4207 * via remount procedure, otherwise, it will confuse code like 4208 * freeze_super() which will lead to deadlocks and other problems. 4209 */ 4210 } 4211 f2fs_record_error_work(struct work_struct * work)4212 static void f2fs_record_error_work(struct work_struct *work) 4213 { 4214 struct f2fs_sb_info *sbi = container_of(work, 4215 struct f2fs_sb_info, s_error_work); 4216 4217 f2fs_record_stop_reason(sbi); 4218 } 4219 f2fs_scan_devices(struct f2fs_sb_info * sbi)4220 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 4221 { 4222 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4223 unsigned int max_devices = MAX_DEVICES; 4224 unsigned int logical_blksize; 4225 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags); 4226 int i; 4227 4228 /* Initialize single device information */ 4229 if (!RDEV(0).path[0]) { 4230 if (!bdev_is_zoned(sbi->sb->s_bdev)) 4231 return 0; 4232 max_devices = 1; 4233 } 4234 4235 /* 4236 * Initialize multiple devices information, or single 4237 * zoned block device information. 4238 */ 4239 sbi->devs = f2fs_kzalloc(sbi, 4240 array_size(max_devices, 4241 sizeof(struct f2fs_dev_info)), 4242 GFP_KERNEL); 4243 if (!sbi->devs) 4244 return -ENOMEM; 4245 4246 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 4247 sbi->aligned_blksize = true; 4248 #ifdef CONFIG_BLK_DEV_ZONED 4249 sbi->max_open_zones = UINT_MAX; 4250 sbi->blkzone_alloc_policy = BLKZONE_ALLOC_PRIOR_SEQ; 4251 #endif 4252 4253 for (i = 0; i < max_devices; i++) { 4254 if (i == 0) 4255 FDEV(0).bdev_file = sbi->sb->s_bdev_file; 4256 else if (!RDEV(i).path[0]) 4257 break; 4258 4259 if (max_devices > 1) { 4260 /* Multi-device mount */ 4261 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 4262 FDEV(i).total_segments = 4263 le32_to_cpu(RDEV(i).total_segments); 4264 if (i == 0) { 4265 FDEV(i).start_blk = 0; 4266 FDEV(i).end_blk = FDEV(i).start_blk + 4267 SEGS_TO_BLKS(sbi, 4268 FDEV(i).total_segments) - 1 + 4269 le32_to_cpu(raw_super->segment0_blkaddr); 4270 } else { 4271 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 4272 FDEV(i).end_blk = FDEV(i).start_blk + 4273 SEGS_TO_BLKS(sbi, 4274 FDEV(i).total_segments) - 1; 4275 FDEV(i).bdev_file = bdev_file_open_by_path( 4276 FDEV(i).path, mode, sbi->sb, NULL); 4277 } 4278 } 4279 if (IS_ERR(FDEV(i).bdev_file)) 4280 return PTR_ERR(FDEV(i).bdev_file); 4281 4282 FDEV(i).bdev = file_bdev(FDEV(i).bdev_file); 4283 /* to release errored devices */ 4284 sbi->s_ndevs = i + 1; 4285 4286 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 4287 sbi->aligned_blksize = false; 4288 4289 #ifdef CONFIG_BLK_DEV_ZONED 4290 if (bdev_is_zoned(FDEV(i).bdev)) { 4291 if (!f2fs_sb_has_blkzoned(sbi)) { 4292 f2fs_err(sbi, "Zoned block device feature not enabled"); 4293 return -EINVAL; 4294 } 4295 if (init_blkz_info(sbi, i)) { 4296 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 4297 return -EINVAL; 4298 } 4299 if (max_devices == 1) 4300 break; 4301 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)", 4302 i, FDEV(i).path, 4303 FDEV(i).total_segments, 4304 FDEV(i).start_blk, FDEV(i).end_blk); 4305 continue; 4306 } 4307 #endif 4308 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 4309 i, FDEV(i).path, 4310 FDEV(i).total_segments, 4311 FDEV(i).start_blk, FDEV(i).end_blk); 4312 } 4313 return 0; 4314 } 4315 f2fs_setup_casefold(struct f2fs_sb_info * sbi)4316 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 4317 { 4318 #if IS_ENABLED(CONFIG_UNICODE) 4319 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 4320 const struct f2fs_sb_encodings *encoding_info; 4321 struct unicode_map *encoding; 4322 __u16 encoding_flags; 4323 4324 encoding_info = f2fs_sb_read_encoding(sbi->raw_super); 4325 if (!encoding_info) { 4326 f2fs_err(sbi, 4327 "Encoding requested by superblock is unknown"); 4328 return -EINVAL; 4329 } 4330 4331 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags); 4332 encoding = utf8_load(encoding_info->version); 4333 if (IS_ERR(encoding)) { 4334 f2fs_err(sbi, 4335 "can't mount with superblock charset: %s-%u.%u.%u " 4336 "not supported by the kernel. flags: 0x%x.", 4337 encoding_info->name, 4338 unicode_major(encoding_info->version), 4339 unicode_minor(encoding_info->version), 4340 unicode_rev(encoding_info->version), 4341 encoding_flags); 4342 return PTR_ERR(encoding); 4343 } 4344 f2fs_info(sbi, "Using encoding defined by superblock: " 4345 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4346 unicode_major(encoding_info->version), 4347 unicode_minor(encoding_info->version), 4348 unicode_rev(encoding_info->version), 4349 encoding_flags); 4350 4351 sbi->sb->s_encoding = encoding; 4352 sbi->sb->s_encoding_flags = encoding_flags; 4353 } 4354 #else 4355 if (f2fs_sb_has_casefold(sbi)) { 4356 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 4357 return -EINVAL; 4358 } 4359 #endif 4360 return 0; 4361 } 4362 f2fs_tuning_parameters(struct f2fs_sb_info * sbi)4363 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 4364 { 4365 /* adjust parameters according to the volume size */ 4366 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) { 4367 if (f2fs_block_unit_discard(sbi)) 4368 SM_I(sbi)->dcc_info->discard_granularity = 4369 MIN_DISCARD_GRANULARITY; 4370 if (!f2fs_lfs_mode(sbi)) 4371 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) | 4372 BIT(F2FS_IPU_HONOR_OPU_WRITE); 4373 } 4374 4375 sbi->readdir_ra = true; 4376 } 4377 f2fs_fill_super(struct super_block * sb,void * data,int silent)4378 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 4379 { 4380 struct f2fs_sb_info *sbi; 4381 struct f2fs_super_block *raw_super; 4382 struct inode *root; 4383 int err; 4384 bool skip_recovery = false, need_fsck = false; 4385 char *options = NULL; 4386 int recovery, i, valid_super_block; 4387 struct curseg_info *seg_i; 4388 int retry_cnt = 1; 4389 #ifdef CONFIG_QUOTA 4390 bool quota_enabled = false; 4391 #endif 4392 4393 try_onemore: 4394 err = -EINVAL; 4395 raw_super = NULL; 4396 valid_super_block = -1; 4397 recovery = 0; 4398 4399 /* allocate memory for f2fs-specific super block info */ 4400 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 4401 if (!sbi) 4402 return -ENOMEM; 4403 4404 sbi->sb = sb; 4405 4406 /* initialize locks within allocated memory */ 4407 init_f2fs_rwsem(&sbi->gc_lock); 4408 mutex_init(&sbi->writepages); 4409 init_f2fs_rwsem(&sbi->cp_global_sem); 4410 init_f2fs_rwsem(&sbi->node_write); 4411 init_f2fs_rwsem(&sbi->node_change); 4412 spin_lock_init(&sbi->stat_lock); 4413 init_f2fs_rwsem(&sbi->cp_rwsem); 4414 init_f2fs_rwsem(&sbi->quota_sem); 4415 init_waitqueue_head(&sbi->cp_wait); 4416 spin_lock_init(&sbi->error_lock); 4417 4418 for (i = 0; i < NR_INODE_TYPE; i++) { 4419 INIT_LIST_HEAD(&sbi->inode_list[i]); 4420 spin_lock_init(&sbi->inode_lock[i]); 4421 } 4422 mutex_init(&sbi->flush_lock); 4423 4424 /* Load the checksum driver */ 4425 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 4426 if (IS_ERR(sbi->s_chksum_driver)) { 4427 f2fs_err(sbi, "Cannot load crc32 driver."); 4428 err = PTR_ERR(sbi->s_chksum_driver); 4429 sbi->s_chksum_driver = NULL; 4430 goto free_sbi; 4431 } 4432 4433 /* set a block size */ 4434 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4435 f2fs_err(sbi, "unable to set blocksize"); 4436 goto free_sbi; 4437 } 4438 4439 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4440 &recovery); 4441 if (err) 4442 goto free_sbi; 4443 4444 sb->s_fs_info = sbi; 4445 sbi->raw_super = raw_super; 4446 4447 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work); 4448 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS); 4449 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON); 4450 4451 /* precompute checksum seed for metadata */ 4452 if (f2fs_sb_has_inode_chksum(sbi)) 4453 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 4454 sizeof(raw_super->uuid)); 4455 4456 default_options(sbi, false); 4457 /* parse mount options */ 4458 options = kstrdup((const char *)data, GFP_KERNEL); 4459 if (data && !options) { 4460 err = -ENOMEM; 4461 goto free_sb_buf; 4462 } 4463 4464 err = parse_options(sb, options, false); 4465 if (err) 4466 goto free_options; 4467 4468 sb->s_maxbytes = max_file_blocks(NULL) << 4469 le32_to_cpu(raw_super->log_blocksize); 4470 sb->s_max_links = F2FS_LINK_MAX; 4471 4472 err = f2fs_setup_casefold(sbi); 4473 if (err) 4474 goto free_options; 4475 4476 #ifdef CONFIG_QUOTA 4477 sb->dq_op = &f2fs_quota_operations; 4478 sb->s_qcop = &f2fs_quotactl_ops; 4479 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4480 4481 if (f2fs_sb_has_quota_ino(sbi)) { 4482 for (i = 0; i < MAXQUOTAS; i++) { 4483 if (f2fs_qf_ino(sbi->sb, i)) 4484 sbi->nquota_files++; 4485 } 4486 } 4487 #endif 4488 4489 sb->s_op = &f2fs_sops; 4490 #ifdef CONFIG_FS_ENCRYPTION 4491 sb->s_cop = &f2fs_cryptops; 4492 #endif 4493 #ifdef CONFIG_FS_VERITY 4494 sb->s_vop = &f2fs_verityops; 4495 #endif 4496 sb->s_xattr = f2fs_xattr_handlers; 4497 sb->s_export_op = &f2fs_export_ops; 4498 sb->s_magic = F2FS_SUPER_MAGIC; 4499 sb->s_time_gran = 1; 4500 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4501 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4502 super_set_uuid(sb, (void *) raw_super->uuid, sizeof(raw_super->uuid)); 4503 super_set_sysfs_name_bdev(sb); 4504 sb->s_iflags |= SB_I_CGROUPWB; 4505 4506 /* init f2fs-specific super block info */ 4507 sbi->valid_super_block = valid_super_block; 4508 4509 /* disallow all the data/node/meta page writes */ 4510 set_sbi_flag(sbi, SBI_POR_DOING); 4511 4512 err = f2fs_init_write_merge_io(sbi); 4513 if (err) 4514 goto free_bio_info; 4515 4516 init_sb_info(sbi); 4517 4518 err = f2fs_init_iostat(sbi); 4519 if (err) 4520 goto free_bio_info; 4521 4522 err = init_percpu_info(sbi); 4523 if (err) 4524 goto free_iostat; 4525 4526 /* init per sbi slab cache */ 4527 err = f2fs_init_xattr_caches(sbi); 4528 if (err) 4529 goto free_percpu; 4530 err = f2fs_init_page_array_cache(sbi); 4531 if (err) 4532 goto free_xattr_cache; 4533 4534 /* get an inode for meta space */ 4535 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4536 if (IS_ERR(sbi->meta_inode)) { 4537 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4538 err = PTR_ERR(sbi->meta_inode); 4539 goto free_page_array_cache; 4540 } 4541 4542 err = f2fs_get_valid_checkpoint(sbi); 4543 if (err) { 4544 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4545 goto free_meta_inode; 4546 } 4547 4548 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4549 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4550 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4551 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4552 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4553 } 4554 4555 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4556 set_sbi_flag(sbi, SBI_NEED_FSCK); 4557 4558 /* Initialize device list */ 4559 err = f2fs_scan_devices(sbi); 4560 if (err) { 4561 f2fs_err(sbi, "Failed to find devices"); 4562 goto free_devices; 4563 } 4564 4565 err = f2fs_init_post_read_wq(sbi); 4566 if (err) { 4567 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4568 goto free_devices; 4569 } 4570 4571 sbi->total_valid_node_count = 4572 le32_to_cpu(sbi->ckpt->valid_node_count); 4573 percpu_counter_set(&sbi->total_valid_inode_count, 4574 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4575 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4576 sbi->total_valid_block_count = 4577 le64_to_cpu(sbi->ckpt->valid_block_count); 4578 sbi->last_valid_block_count = sbi->total_valid_block_count; 4579 sbi->reserved_blocks = 0; 4580 sbi->current_reserved_blocks = 0; 4581 limit_reserve_root(sbi); 4582 adjust_unusable_cap_perc(sbi); 4583 4584 f2fs_init_extent_cache_info(sbi); 4585 4586 f2fs_init_ino_entry_info(sbi); 4587 4588 f2fs_init_fsync_node_info(sbi); 4589 4590 /* setup checkpoint request control and start checkpoint issue thread */ 4591 f2fs_init_ckpt_req_control(sbi); 4592 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4593 test_opt(sbi, MERGE_CHECKPOINT)) { 4594 err = f2fs_start_ckpt_thread(sbi); 4595 if (err) { 4596 f2fs_err(sbi, 4597 "Failed to start F2FS issue_checkpoint_thread (%d)", 4598 err); 4599 goto stop_ckpt_thread; 4600 } 4601 } 4602 4603 /* setup f2fs internal modules */ 4604 err = f2fs_build_segment_manager(sbi); 4605 if (err) { 4606 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4607 err); 4608 goto free_sm; 4609 } 4610 err = f2fs_build_node_manager(sbi); 4611 if (err) { 4612 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4613 err); 4614 goto free_nm; 4615 } 4616 4617 /* For write statistics */ 4618 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4619 4620 /* Read accumulated write IO statistics if exists */ 4621 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4622 if (__exist_node_summaries(sbi)) 4623 sbi->kbytes_written = 4624 le64_to_cpu(seg_i->journal->info.kbytes_written); 4625 4626 f2fs_build_gc_manager(sbi); 4627 4628 err = f2fs_build_stats(sbi); 4629 if (err) 4630 goto free_nm; 4631 4632 /* get an inode for node space */ 4633 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4634 if (IS_ERR(sbi->node_inode)) { 4635 f2fs_err(sbi, "Failed to read node inode"); 4636 err = PTR_ERR(sbi->node_inode); 4637 goto free_stats; 4638 } 4639 4640 /* read root inode and dentry */ 4641 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4642 if (IS_ERR(root)) { 4643 f2fs_err(sbi, "Failed to read root inode"); 4644 err = PTR_ERR(root); 4645 goto free_node_inode; 4646 } 4647 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4648 !root->i_size || !root->i_nlink) { 4649 iput(root); 4650 err = -EINVAL; 4651 goto free_node_inode; 4652 } 4653 4654 generic_set_sb_d_ops(sb); 4655 sb->s_root = d_make_root(root); /* allocate root dentry */ 4656 if (!sb->s_root) { 4657 err = -ENOMEM; 4658 goto free_node_inode; 4659 } 4660 4661 err = f2fs_init_compress_inode(sbi); 4662 if (err) 4663 goto free_root_inode; 4664 4665 err = f2fs_register_sysfs(sbi); 4666 if (err) 4667 goto free_compress_inode; 4668 4669 #ifdef CONFIG_QUOTA 4670 /* Enable quota usage during mount */ 4671 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4672 err = f2fs_enable_quotas(sb); 4673 if (err) 4674 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4675 } 4676 4677 quota_enabled = f2fs_recover_quota_begin(sbi); 4678 #endif 4679 /* if there are any orphan inodes, free them */ 4680 err = f2fs_recover_orphan_inodes(sbi); 4681 if (err) 4682 goto free_meta; 4683 4684 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4685 goto reset_checkpoint; 4686 4687 /* recover fsynced data */ 4688 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4689 !test_opt(sbi, NORECOVERY)) { 4690 /* 4691 * mount should be failed, when device has readonly mode, and 4692 * previous checkpoint was not done by clean system shutdown. 4693 */ 4694 if (f2fs_hw_is_readonly(sbi)) { 4695 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4696 err = f2fs_recover_fsync_data(sbi, true); 4697 if (err > 0) { 4698 err = -EROFS; 4699 f2fs_err(sbi, "Need to recover fsync data, but " 4700 "write access unavailable, please try " 4701 "mount w/ disable_roll_forward or norecovery"); 4702 } 4703 if (err < 0) 4704 goto free_meta; 4705 } 4706 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4707 goto reset_checkpoint; 4708 } 4709 4710 if (need_fsck) 4711 set_sbi_flag(sbi, SBI_NEED_FSCK); 4712 4713 if (skip_recovery) 4714 goto reset_checkpoint; 4715 4716 err = f2fs_recover_fsync_data(sbi, false); 4717 if (err < 0) { 4718 if (err != -ENOMEM) 4719 skip_recovery = true; 4720 need_fsck = true; 4721 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4722 err); 4723 goto free_meta; 4724 } 4725 } else { 4726 err = f2fs_recover_fsync_data(sbi, true); 4727 4728 if (!f2fs_readonly(sb) && err > 0) { 4729 err = -EINVAL; 4730 f2fs_err(sbi, "Need to recover fsync data"); 4731 goto free_meta; 4732 } 4733 } 4734 4735 #ifdef CONFIG_QUOTA 4736 f2fs_recover_quota_end(sbi, quota_enabled); 4737 #endif 4738 reset_checkpoint: 4739 /* 4740 * If the f2fs is not readonly and fsync data recovery succeeds, 4741 * check zoned block devices' write pointer consistency. 4742 */ 4743 if (f2fs_sb_has_blkzoned(sbi) && !f2fs_readonly(sb)) { 4744 int err2; 4745 4746 f2fs_notice(sbi, "Checking entire write pointers"); 4747 err2 = f2fs_check_write_pointer(sbi); 4748 if (err2) 4749 err = err2; 4750 } 4751 if (err) 4752 goto free_meta; 4753 4754 err = f2fs_init_inmem_curseg(sbi); 4755 if (err) 4756 goto sync_free_meta; 4757 4758 /* f2fs_recover_fsync_data() cleared this already */ 4759 clear_sbi_flag(sbi, SBI_POR_DOING); 4760 4761 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4762 err = f2fs_disable_checkpoint(sbi); 4763 if (err) 4764 goto sync_free_meta; 4765 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4766 f2fs_enable_checkpoint(sbi); 4767 } 4768 4769 /* 4770 * If filesystem is not mounted as read-only then 4771 * do start the gc_thread. 4772 */ 4773 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4774 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4775 /* After POR, we can run background GC thread.*/ 4776 err = f2fs_start_gc_thread(sbi); 4777 if (err) 4778 goto sync_free_meta; 4779 } 4780 kvfree(options); 4781 4782 /* recover broken superblock */ 4783 if (recovery) { 4784 err = f2fs_commit_super(sbi, true); 4785 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4786 sbi->valid_super_block ? 1 : 2, err); 4787 } 4788 4789 f2fs_join_shrinker(sbi); 4790 4791 f2fs_tuning_parameters(sbi); 4792 4793 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4794 cur_cp_version(F2FS_CKPT(sbi))); 4795 f2fs_update_time(sbi, CP_TIME); 4796 f2fs_update_time(sbi, REQ_TIME); 4797 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4798 return 0; 4799 4800 sync_free_meta: 4801 /* safe to flush all the data */ 4802 sync_filesystem(sbi->sb); 4803 retry_cnt = 0; 4804 4805 free_meta: 4806 #ifdef CONFIG_QUOTA 4807 f2fs_truncate_quota_inode_pages(sb); 4808 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4809 f2fs_quota_off_umount(sbi->sb); 4810 #endif 4811 /* 4812 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4813 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4814 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4815 * falls into an infinite loop in f2fs_sync_meta_pages(). 4816 */ 4817 truncate_inode_pages_final(META_MAPPING(sbi)); 4818 /* evict some inodes being cached by GC */ 4819 evict_inodes(sb); 4820 f2fs_unregister_sysfs(sbi); 4821 free_compress_inode: 4822 f2fs_destroy_compress_inode(sbi); 4823 free_root_inode: 4824 dput(sb->s_root); 4825 sb->s_root = NULL; 4826 free_node_inode: 4827 f2fs_release_ino_entry(sbi, true); 4828 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4829 iput(sbi->node_inode); 4830 sbi->node_inode = NULL; 4831 free_stats: 4832 f2fs_destroy_stats(sbi); 4833 free_nm: 4834 /* stop discard thread before destroying node manager */ 4835 f2fs_stop_discard_thread(sbi); 4836 f2fs_destroy_node_manager(sbi); 4837 free_sm: 4838 f2fs_destroy_segment_manager(sbi); 4839 stop_ckpt_thread: 4840 f2fs_stop_ckpt_thread(sbi); 4841 /* flush s_error_work before sbi destroy */ 4842 flush_work(&sbi->s_error_work); 4843 f2fs_destroy_post_read_wq(sbi); 4844 free_devices: 4845 destroy_device_list(sbi); 4846 kvfree(sbi->ckpt); 4847 free_meta_inode: 4848 make_bad_inode(sbi->meta_inode); 4849 iput(sbi->meta_inode); 4850 sbi->meta_inode = NULL; 4851 free_page_array_cache: 4852 f2fs_destroy_page_array_cache(sbi); 4853 free_xattr_cache: 4854 f2fs_destroy_xattr_caches(sbi); 4855 free_percpu: 4856 destroy_percpu_info(sbi); 4857 free_iostat: 4858 f2fs_destroy_iostat(sbi); 4859 free_bio_info: 4860 for (i = 0; i < NR_PAGE_TYPE; i++) 4861 kvfree(sbi->write_io[i]); 4862 4863 #if IS_ENABLED(CONFIG_UNICODE) 4864 utf8_unload(sb->s_encoding); 4865 sb->s_encoding = NULL; 4866 #endif 4867 free_options: 4868 #ifdef CONFIG_QUOTA 4869 for (i = 0; i < MAXQUOTAS; i++) 4870 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4871 #endif 4872 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4873 kvfree(options); 4874 free_sb_buf: 4875 kfree(raw_super); 4876 free_sbi: 4877 if (sbi->s_chksum_driver) 4878 crypto_free_shash(sbi->s_chksum_driver); 4879 kfree(sbi); 4880 sb->s_fs_info = NULL; 4881 4882 /* give only one another chance */ 4883 if (retry_cnt > 0 && skip_recovery) { 4884 retry_cnt--; 4885 shrink_dcache_sb(sb); 4886 goto try_onemore; 4887 } 4888 return err; 4889 } 4890 f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4891 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4892 const char *dev_name, void *data) 4893 { 4894 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4895 } 4896 kill_f2fs_super(struct super_block * sb)4897 static void kill_f2fs_super(struct super_block *sb) 4898 { 4899 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4900 4901 if (sb->s_root) { 4902 set_sbi_flag(sbi, SBI_IS_CLOSE); 4903 f2fs_stop_gc_thread(sbi); 4904 f2fs_stop_discard_thread(sbi); 4905 4906 #ifdef CONFIG_F2FS_FS_COMPRESSION 4907 /* 4908 * latter evict_inode() can bypass checking and invalidating 4909 * compress inode cache. 4910 */ 4911 if (test_opt(sbi, COMPRESS_CACHE)) 4912 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4913 #endif 4914 4915 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4916 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4917 struct cp_control cpc = { 4918 .reason = CP_UMOUNT, 4919 }; 4920 stat_inc_cp_call_count(sbi, TOTAL_CALL); 4921 f2fs_write_checkpoint(sbi, &cpc); 4922 } 4923 4924 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4925 sb->s_flags &= ~SB_RDONLY; 4926 } 4927 kill_block_super(sb); 4928 /* Release block devices last, after fscrypt_destroy_keyring(). */ 4929 if (sbi) { 4930 destroy_device_list(sbi); 4931 kfree(sbi); 4932 sb->s_fs_info = NULL; 4933 } 4934 } 4935 4936 static struct file_system_type f2fs_fs_type = { 4937 .owner = THIS_MODULE, 4938 .name = "f2fs", 4939 .mount = f2fs_mount, 4940 .kill_sb = kill_f2fs_super, 4941 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 4942 }; 4943 MODULE_ALIAS_FS("f2fs"); 4944 init_inodecache(void)4945 static int __init init_inodecache(void) 4946 { 4947 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4948 sizeof(struct f2fs_inode_info), 0, 4949 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4950 return f2fs_inode_cachep ? 0 : -ENOMEM; 4951 } 4952 destroy_inodecache(void)4953 static void destroy_inodecache(void) 4954 { 4955 /* 4956 * Make sure all delayed rcu free inodes are flushed before we 4957 * destroy cache. 4958 */ 4959 rcu_barrier(); 4960 kmem_cache_destroy(f2fs_inode_cachep); 4961 } 4962 init_f2fs_fs(void)4963 static int __init init_f2fs_fs(void) 4964 { 4965 int err; 4966 4967 err = init_inodecache(); 4968 if (err) 4969 goto fail; 4970 err = f2fs_create_node_manager_caches(); 4971 if (err) 4972 goto free_inodecache; 4973 err = f2fs_create_segment_manager_caches(); 4974 if (err) 4975 goto free_node_manager_caches; 4976 err = f2fs_create_checkpoint_caches(); 4977 if (err) 4978 goto free_segment_manager_caches; 4979 err = f2fs_create_recovery_cache(); 4980 if (err) 4981 goto free_checkpoint_caches; 4982 err = f2fs_create_extent_cache(); 4983 if (err) 4984 goto free_recovery_cache; 4985 err = f2fs_create_garbage_collection_cache(); 4986 if (err) 4987 goto free_extent_cache; 4988 err = f2fs_init_sysfs(); 4989 if (err) 4990 goto free_garbage_collection_cache; 4991 err = f2fs_init_shrinker(); 4992 if (err) 4993 goto free_sysfs; 4994 err = register_filesystem(&f2fs_fs_type); 4995 if (err) 4996 goto free_shrinker; 4997 f2fs_create_root_stats(); 4998 err = f2fs_init_post_read_processing(); 4999 if (err) 5000 goto free_root_stats; 5001 err = f2fs_init_iostat_processing(); 5002 if (err) 5003 goto free_post_read; 5004 err = f2fs_init_bio_entry_cache(); 5005 if (err) 5006 goto free_iostat; 5007 err = f2fs_init_bioset(); 5008 if (err) 5009 goto free_bio_entry_cache; 5010 err = f2fs_init_compress_mempool(); 5011 if (err) 5012 goto free_bioset; 5013 err = f2fs_init_compress_cache(); 5014 if (err) 5015 goto free_compress_mempool; 5016 err = f2fs_create_casefold_cache(); 5017 if (err) 5018 goto free_compress_cache; 5019 return 0; 5020 free_compress_cache: 5021 f2fs_destroy_compress_cache(); 5022 free_compress_mempool: 5023 f2fs_destroy_compress_mempool(); 5024 free_bioset: 5025 f2fs_destroy_bioset(); 5026 free_bio_entry_cache: 5027 f2fs_destroy_bio_entry_cache(); 5028 free_iostat: 5029 f2fs_destroy_iostat_processing(); 5030 free_post_read: 5031 f2fs_destroy_post_read_processing(); 5032 free_root_stats: 5033 f2fs_destroy_root_stats(); 5034 unregister_filesystem(&f2fs_fs_type); 5035 free_shrinker: 5036 f2fs_exit_shrinker(); 5037 free_sysfs: 5038 f2fs_exit_sysfs(); 5039 free_garbage_collection_cache: 5040 f2fs_destroy_garbage_collection_cache(); 5041 free_extent_cache: 5042 f2fs_destroy_extent_cache(); 5043 free_recovery_cache: 5044 f2fs_destroy_recovery_cache(); 5045 free_checkpoint_caches: 5046 f2fs_destroy_checkpoint_caches(); 5047 free_segment_manager_caches: 5048 f2fs_destroy_segment_manager_caches(); 5049 free_node_manager_caches: 5050 f2fs_destroy_node_manager_caches(); 5051 free_inodecache: 5052 destroy_inodecache(); 5053 fail: 5054 return err; 5055 } 5056 exit_f2fs_fs(void)5057 static void __exit exit_f2fs_fs(void) 5058 { 5059 f2fs_destroy_casefold_cache(); 5060 f2fs_destroy_compress_cache(); 5061 f2fs_destroy_compress_mempool(); 5062 f2fs_destroy_bioset(); 5063 f2fs_destroy_bio_entry_cache(); 5064 f2fs_destroy_iostat_processing(); 5065 f2fs_destroy_post_read_processing(); 5066 f2fs_destroy_root_stats(); 5067 unregister_filesystem(&f2fs_fs_type); 5068 f2fs_exit_shrinker(); 5069 f2fs_exit_sysfs(); 5070 f2fs_destroy_garbage_collection_cache(); 5071 f2fs_destroy_extent_cache(); 5072 f2fs_destroy_recovery_cache(); 5073 f2fs_destroy_checkpoint_caches(); 5074 f2fs_destroy_segment_manager_caches(); 5075 f2fs_destroy_node_manager_caches(); 5076 destroy_inodecache(); 5077 } 5078 5079 module_init(init_f2fs_fs) 5080 module_exit(exit_f2fs_fs) 5081 5082 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 5083 MODULE_DESCRIPTION("Flash Friendly File System"); 5084 MODULE_LICENSE("GPL"); 5085 MODULE_SOFTDEP("pre: crc32"); 5086 5087