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