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/statfs.h> 12 #include <linux/buffer_head.h> 13 #include <linux/backing-dev.h> 14 #include <linux/kthread.h> 15 #include <linux/parser.h> 16 #include <linux/mount.h> 17 #include <linux/seq_file.h> 18 #include <linux/proc_fs.h> 19 #include <linux/random.h> 20 #include <linux/exportfs.h> 21 #include <linux/blkdev.h> 22 #include <linux/quotaops.h> 23 #include <linux/f2fs_fs.h> 24 #include <linux/sysfs.h> 25 #include <linux/quota.h> 26 #include <linux/unicode.h> 27 #include <linux/part_stat.h> 28 #include <linux/zstd.h> 29 #include <linux/lz4.h> 30 31 #include "f2fs.h" 32 #include "node.h" 33 #include "segment.h" 34 #include "xattr.h" 35 #include "gc.h" 36 37 #define CREATE_TRACE_POINTS 38 #include <trace/events/f2fs.h> 39 40 static struct kmem_cache *f2fs_inode_cachep; 41 42 #ifdef CONFIG_F2FS_FAULT_INJECTION 43 44 const char *f2fs_fault_name[FAULT_MAX] = { 45 [FAULT_KMALLOC] = "kmalloc", 46 [FAULT_KVMALLOC] = "kvmalloc", 47 [FAULT_PAGE_ALLOC] = "page alloc", 48 [FAULT_PAGE_GET] = "page get", 49 [FAULT_ALLOC_NID] = "alloc nid", 50 [FAULT_ORPHAN] = "orphan", 51 [FAULT_BLOCK] = "no more block", 52 [FAULT_DIR_DEPTH] = "too big dir depth", 53 [FAULT_EVICT_INODE] = "evict_inode fail", 54 [FAULT_TRUNCATE] = "truncate fail", 55 [FAULT_READ_IO] = "read IO error", 56 [FAULT_CHECKPOINT] = "checkpoint error", 57 [FAULT_DISCARD] = "discard error", 58 [FAULT_WRITE_IO] = "write IO error", 59 }; 60 61 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, 62 unsigned int type) 63 { 64 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; 65 66 if (rate) { 67 atomic_set(&ffi->inject_ops, 0); 68 ffi->inject_rate = rate; 69 } 70 71 if (type) 72 ffi->inject_type = type; 73 74 if (!rate && !type) 75 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 76 } 77 #endif 78 79 /* f2fs-wide shrinker description */ 80 static struct shrinker f2fs_shrinker_info = { 81 .scan_objects = f2fs_shrink_scan, 82 .count_objects = f2fs_shrink_count, 83 .seeks = DEFAULT_SEEKS, 84 }; 85 86 enum { 87 Opt_gc_background, 88 Opt_disable_roll_forward, 89 Opt_norecovery, 90 Opt_discard, 91 Opt_nodiscard, 92 Opt_noheap, 93 Opt_heap, 94 Opt_user_xattr, 95 Opt_nouser_xattr, 96 Opt_acl, 97 Opt_noacl, 98 Opt_active_logs, 99 Opt_disable_ext_identify, 100 Opt_inline_xattr, 101 Opt_noinline_xattr, 102 Opt_inline_xattr_size, 103 Opt_inline_data, 104 Opt_inline_dentry, 105 Opt_noinline_dentry, 106 Opt_flush_merge, 107 Opt_noflush_merge, 108 Opt_nobarrier, 109 Opt_fastboot, 110 Opt_extent_cache, 111 Opt_noextent_cache, 112 Opt_noinline_data, 113 Opt_data_flush, 114 Opt_reserve_root, 115 Opt_resgid, 116 Opt_resuid, 117 Opt_mode, 118 Opt_io_size_bits, 119 Opt_fault_injection, 120 Opt_fault_type, 121 Opt_lazytime, 122 Opt_nolazytime, 123 Opt_quota, 124 Opt_noquota, 125 Opt_usrquota, 126 Opt_grpquota, 127 Opt_prjquota, 128 Opt_usrjquota, 129 Opt_grpjquota, 130 Opt_prjjquota, 131 Opt_offusrjquota, 132 Opt_offgrpjquota, 133 Opt_offprjjquota, 134 Opt_jqfmt_vfsold, 135 Opt_jqfmt_vfsv0, 136 Opt_jqfmt_vfsv1, 137 Opt_whint, 138 Opt_alloc, 139 Opt_fsync, 140 Opt_test_dummy_encryption, 141 Opt_inlinecrypt, 142 Opt_checkpoint_disable, 143 Opt_checkpoint_disable_cap, 144 Opt_checkpoint_disable_cap_perc, 145 Opt_checkpoint_enable, 146 Opt_checkpoint_merge, 147 Opt_nocheckpoint_merge, 148 Opt_compress_algorithm, 149 Opt_compress_log_size, 150 Opt_compress_extension, 151 Opt_nocompress_extension, 152 Opt_compress_chksum, 153 Opt_compress_mode, 154 Opt_compress_cache, 155 Opt_atgc, 156 Opt_gc_merge, 157 Opt_nogc_merge, 158 Opt_err, 159 }; 160 161 static match_table_t f2fs_tokens = { 162 {Opt_gc_background, "background_gc=%s"}, 163 {Opt_disable_roll_forward, "disable_roll_forward"}, 164 {Opt_norecovery, "norecovery"}, 165 {Opt_discard, "discard"}, 166 {Opt_nodiscard, "nodiscard"}, 167 {Opt_noheap, "no_heap"}, 168 {Opt_heap, "heap"}, 169 {Opt_user_xattr, "user_xattr"}, 170 {Opt_nouser_xattr, "nouser_xattr"}, 171 {Opt_acl, "acl"}, 172 {Opt_noacl, "noacl"}, 173 {Opt_active_logs, "active_logs=%u"}, 174 {Opt_disable_ext_identify, "disable_ext_identify"}, 175 {Opt_inline_xattr, "inline_xattr"}, 176 {Opt_noinline_xattr, "noinline_xattr"}, 177 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 178 {Opt_inline_data, "inline_data"}, 179 {Opt_inline_dentry, "inline_dentry"}, 180 {Opt_noinline_dentry, "noinline_dentry"}, 181 {Opt_flush_merge, "flush_merge"}, 182 {Opt_noflush_merge, "noflush_merge"}, 183 {Opt_nobarrier, "nobarrier"}, 184 {Opt_fastboot, "fastboot"}, 185 {Opt_extent_cache, "extent_cache"}, 186 {Opt_noextent_cache, "noextent_cache"}, 187 {Opt_noinline_data, "noinline_data"}, 188 {Opt_data_flush, "data_flush"}, 189 {Opt_reserve_root, "reserve_root=%u"}, 190 {Opt_resgid, "resgid=%u"}, 191 {Opt_resuid, "resuid=%u"}, 192 {Opt_mode, "mode=%s"}, 193 {Opt_io_size_bits, "io_bits=%u"}, 194 {Opt_fault_injection, "fault_injection=%u"}, 195 {Opt_fault_type, "fault_type=%u"}, 196 {Opt_lazytime, "lazytime"}, 197 {Opt_nolazytime, "nolazytime"}, 198 {Opt_quota, "quota"}, 199 {Opt_noquota, "noquota"}, 200 {Opt_usrquota, "usrquota"}, 201 {Opt_grpquota, "grpquota"}, 202 {Opt_prjquota, "prjquota"}, 203 {Opt_usrjquota, "usrjquota=%s"}, 204 {Opt_grpjquota, "grpjquota=%s"}, 205 {Opt_prjjquota, "prjjquota=%s"}, 206 {Opt_offusrjquota, "usrjquota="}, 207 {Opt_offgrpjquota, "grpjquota="}, 208 {Opt_offprjjquota, "prjjquota="}, 209 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 210 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 211 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 212 {Opt_whint, "whint_mode=%s"}, 213 {Opt_alloc, "alloc_mode=%s"}, 214 {Opt_fsync, "fsync_mode=%s"}, 215 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 216 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 217 {Opt_inlinecrypt, "inlinecrypt"}, 218 {Opt_checkpoint_disable, "checkpoint=disable"}, 219 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 220 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 221 {Opt_checkpoint_enable, "checkpoint=enable"}, 222 {Opt_checkpoint_merge, "checkpoint_merge"}, 223 {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, 224 {Opt_compress_algorithm, "compress_algorithm=%s"}, 225 {Opt_compress_log_size, "compress_log_size=%u"}, 226 {Opt_compress_extension, "compress_extension=%s"}, 227 {Opt_nocompress_extension, "nocompress_extension=%s"}, 228 {Opt_compress_chksum, "compress_chksum"}, 229 {Opt_compress_mode, "compress_mode=%s"}, 230 {Opt_compress_cache, "compress_cache"}, 231 {Opt_atgc, "atgc"}, 232 {Opt_gc_merge, "gc_merge"}, 233 {Opt_nogc_merge, "nogc_merge"}, 234 {Opt_err, NULL}, 235 }; 236 237 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 238 { 239 struct va_format vaf; 240 va_list args; 241 int level; 242 243 va_start(args, fmt); 244 245 level = printk_get_level(fmt); 246 vaf.fmt = printk_skip_level(fmt); 247 vaf.va = &args; 248 printk("%c%cF2FS-fs (%s): %pV\n", 249 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 250 251 va_end(args); 252 } 253 254 #ifdef CONFIG_UNICODE 255 static const struct f2fs_sb_encodings { 256 __u16 magic; 257 char *name; 258 char *version; 259 } f2fs_sb_encoding_map[] = { 260 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"}, 261 }; 262 263 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb, 264 const struct f2fs_sb_encodings **encoding, 265 __u16 *flags) 266 { 267 __u16 magic = le16_to_cpu(sb->s_encoding); 268 int i; 269 270 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 271 if (magic == f2fs_sb_encoding_map[i].magic) 272 break; 273 274 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map)) 275 return -EINVAL; 276 277 *encoding = &f2fs_sb_encoding_map[i]; 278 *flags = le16_to_cpu(sb->s_encoding_flags); 279 280 return 0; 281 } 282 283 struct kmem_cache *f2fs_cf_name_slab; 284 static int __init f2fs_create_casefold_cache(void) 285 { 286 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 287 F2FS_NAME_LEN); 288 if (!f2fs_cf_name_slab) 289 return -ENOMEM; 290 return 0; 291 } 292 293 static void f2fs_destroy_casefold_cache(void) 294 { 295 kmem_cache_destroy(f2fs_cf_name_slab); 296 } 297 #else 298 static int __init f2fs_create_casefold_cache(void) { return 0; } 299 static void f2fs_destroy_casefold_cache(void) { } 300 #endif 301 302 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 303 { 304 block_t limit = min((sbi->user_block_count << 1) / 1000, 305 sbi->user_block_count - sbi->reserved_blocks); 306 307 /* limit is 0.2% */ 308 if (test_opt(sbi, RESERVE_ROOT) && 309 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 310 F2FS_OPTION(sbi).root_reserved_blocks = limit; 311 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 312 F2FS_OPTION(sbi).root_reserved_blocks); 313 } 314 if (!test_opt(sbi, RESERVE_ROOT) && 315 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 316 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 317 !gid_eq(F2FS_OPTION(sbi).s_resgid, 318 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 319 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 320 from_kuid_munged(&init_user_ns, 321 F2FS_OPTION(sbi).s_resuid), 322 from_kgid_munged(&init_user_ns, 323 F2FS_OPTION(sbi).s_resgid)); 324 } 325 326 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 327 { 328 if (!F2FS_OPTION(sbi).unusable_cap_perc) 329 return; 330 331 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 332 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 333 else 334 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 335 F2FS_OPTION(sbi).unusable_cap_perc; 336 337 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 338 F2FS_OPTION(sbi).unusable_cap, 339 F2FS_OPTION(sbi).unusable_cap_perc); 340 } 341 342 static void init_once(void *foo) 343 { 344 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 345 346 inode_init_once(&fi->vfs_inode); 347 } 348 349 #ifdef CONFIG_QUOTA 350 static const char * const quotatypes[] = INITQFNAMES; 351 #define QTYPE2NAME(t) (quotatypes[t]) 352 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 353 substring_t *args) 354 { 355 struct f2fs_sb_info *sbi = F2FS_SB(sb); 356 char *qname; 357 int ret = -EINVAL; 358 359 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 360 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 361 return -EINVAL; 362 } 363 if (f2fs_sb_has_quota_ino(sbi)) { 364 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 365 return 0; 366 } 367 368 qname = match_strdup(args); 369 if (!qname) { 370 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 371 return -ENOMEM; 372 } 373 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 374 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 375 ret = 0; 376 else 377 f2fs_err(sbi, "%s quota file already specified", 378 QTYPE2NAME(qtype)); 379 goto errout; 380 } 381 if (strchr(qname, '/')) { 382 f2fs_err(sbi, "quotafile must be on filesystem root"); 383 goto errout; 384 } 385 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 386 set_opt(sbi, QUOTA); 387 return 0; 388 errout: 389 kfree(qname); 390 return ret; 391 } 392 393 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 394 { 395 struct f2fs_sb_info *sbi = F2FS_SB(sb); 396 397 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 398 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 399 return -EINVAL; 400 } 401 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 402 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 403 return 0; 404 } 405 406 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 407 { 408 /* 409 * We do the test below only for project quotas. 'usrquota' and 410 * 'grpquota' mount options are allowed even without quota feature 411 * to support legacy quotas in quota files. 412 */ 413 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 414 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 415 return -1; 416 } 417 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 418 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 419 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 420 if (test_opt(sbi, USRQUOTA) && 421 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 422 clear_opt(sbi, USRQUOTA); 423 424 if (test_opt(sbi, GRPQUOTA) && 425 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 426 clear_opt(sbi, GRPQUOTA); 427 428 if (test_opt(sbi, PRJQUOTA) && 429 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 430 clear_opt(sbi, PRJQUOTA); 431 432 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 433 test_opt(sbi, PRJQUOTA)) { 434 f2fs_err(sbi, "old and new quota format mixing"); 435 return -1; 436 } 437 438 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 439 f2fs_err(sbi, "journaled quota format not specified"); 440 return -1; 441 } 442 } 443 444 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 445 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 446 F2FS_OPTION(sbi).s_jquota_fmt = 0; 447 } 448 return 0; 449 } 450 #endif 451 452 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 453 const char *opt, 454 const substring_t *arg, 455 bool is_remount) 456 { 457 struct f2fs_sb_info *sbi = F2FS_SB(sb); 458 #ifdef CONFIG_FS_ENCRYPTION 459 int err; 460 461 if (!f2fs_sb_has_encrypt(sbi)) { 462 f2fs_err(sbi, "Encrypt feature is off"); 463 return -EINVAL; 464 } 465 466 /* 467 * This mount option is just for testing, and it's not worthwhile to 468 * implement the extra complexity (e.g. RCU protection) that would be 469 * needed to allow it to be set or changed during remount. We do allow 470 * it to be specified during remount, but only if there is no change. 471 */ 472 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) { 473 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 474 return -EINVAL; 475 } 476 err = fscrypt_set_test_dummy_encryption( 477 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy); 478 if (err) { 479 if (err == -EEXIST) 480 f2fs_warn(sbi, 481 "Can't change test_dummy_encryption on remount"); 482 else if (err == -EINVAL) 483 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 484 opt); 485 else 486 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 487 opt, err); 488 return -EINVAL; 489 } 490 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 491 #else 492 f2fs_warn(sbi, "Test dummy encryption mount option ignored"); 493 #endif 494 return 0; 495 } 496 497 #ifdef CONFIG_F2FS_FS_COMPRESSION 498 /* 499 * 1. The same extension name cannot not appear in both compress and non-compress extension 500 * at the same time. 501 * 2. If the compress extension specifies all files, the types specified by the non-compress 502 * extension will be treated as special cases and will not be compressed. 503 * 3. Don't allow the non-compress extension specifies all files. 504 */ 505 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi) 506 { 507 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 508 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 509 int ext_cnt, noext_cnt, index = 0, no_index = 0; 510 511 ext = F2FS_OPTION(sbi).extensions; 512 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 513 noext = F2FS_OPTION(sbi).noextensions; 514 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 515 516 if (!noext_cnt) 517 return 0; 518 519 for (no_index = 0; no_index < noext_cnt; no_index++) { 520 if (!strcasecmp("*", noext[no_index])) { 521 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files"); 522 return -EINVAL; 523 } 524 for (index = 0; index < ext_cnt; index++) { 525 if (!strcasecmp(ext[index], noext[no_index])) { 526 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension", 527 ext[index]); 528 return -EINVAL; 529 } 530 } 531 } 532 return 0; 533 } 534 535 #ifdef CONFIG_F2FS_FS_LZ4 536 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) 537 { 538 #ifdef CONFIG_F2FS_FS_LZ4HC 539 unsigned int level; 540 #endif 541 542 if (strlen(str) == 3) { 543 F2FS_OPTION(sbi).compress_level = 0; 544 return 0; 545 } 546 547 #ifdef CONFIG_F2FS_FS_LZ4HC 548 str += 3; 549 550 if (str[0] != ':') { 551 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 552 return -EINVAL; 553 } 554 if (kstrtouint(str + 1, 10, &level)) 555 return -EINVAL; 556 557 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) { 558 f2fs_info(sbi, "invalid lz4hc compress level: %d", level); 559 return -EINVAL; 560 } 561 562 F2FS_OPTION(sbi).compress_level = level; 563 return 0; 564 #else 565 f2fs_info(sbi, "kernel doesn't support lz4hc compression"); 566 return -EINVAL; 567 #endif 568 } 569 #endif 570 571 #ifdef CONFIG_F2FS_FS_ZSTD 572 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) 573 { 574 unsigned int level; 575 int len = 4; 576 577 if (strlen(str) == len) { 578 F2FS_OPTION(sbi).compress_level = 0; 579 return 0; 580 } 581 582 str += len; 583 584 if (str[0] != ':') { 585 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 586 return -EINVAL; 587 } 588 if (kstrtouint(str + 1, 10, &level)) 589 return -EINVAL; 590 591 if (!level || level > ZSTD_maxCLevel()) { 592 f2fs_info(sbi, "invalid zstd compress level: %d", level); 593 return -EINVAL; 594 } 595 596 F2FS_OPTION(sbi).compress_level = level; 597 return 0; 598 } 599 #endif 600 #endif 601 602 static int parse_options(struct super_block *sb, char *options, bool is_remount) 603 { 604 struct f2fs_sb_info *sbi = F2FS_SB(sb); 605 substring_t args[MAX_OPT_ARGS]; 606 #ifdef CONFIG_F2FS_FS_COMPRESSION 607 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 608 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 609 int ext_cnt, noext_cnt; 610 #endif 611 char *p, *name; 612 int arg = 0; 613 kuid_t uid; 614 kgid_t gid; 615 int ret; 616 617 if (!options) 618 goto default_check; 619 620 while ((p = strsep(&options, ",")) != NULL) { 621 int token; 622 623 if (!*p) 624 continue; 625 /* 626 * Initialize args struct so we know whether arg was 627 * found; some options take optional arguments. 628 */ 629 args[0].to = args[0].from = NULL; 630 token = match_token(p, f2fs_tokens, args); 631 632 switch (token) { 633 case Opt_gc_background: 634 name = match_strdup(&args[0]); 635 636 if (!name) 637 return -ENOMEM; 638 if (!strcmp(name, "on")) { 639 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 640 } else if (!strcmp(name, "off")) { 641 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 642 } else if (!strcmp(name, "sync")) { 643 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 644 } else { 645 kfree(name); 646 return -EINVAL; 647 } 648 kfree(name); 649 break; 650 case Opt_disable_roll_forward: 651 set_opt(sbi, DISABLE_ROLL_FORWARD); 652 break; 653 case Opt_norecovery: 654 /* this option mounts f2fs with ro */ 655 set_opt(sbi, NORECOVERY); 656 if (!f2fs_readonly(sb)) 657 return -EINVAL; 658 break; 659 case Opt_discard: 660 set_opt(sbi, DISCARD); 661 break; 662 case Opt_nodiscard: 663 if (f2fs_sb_has_blkzoned(sbi)) { 664 f2fs_warn(sbi, "discard is required for zoned block devices"); 665 return -EINVAL; 666 } 667 clear_opt(sbi, DISCARD); 668 break; 669 case Opt_noheap: 670 set_opt(sbi, NOHEAP); 671 break; 672 case Opt_heap: 673 clear_opt(sbi, NOHEAP); 674 break; 675 #ifdef CONFIG_F2FS_FS_XATTR 676 case Opt_user_xattr: 677 set_opt(sbi, XATTR_USER); 678 break; 679 case Opt_nouser_xattr: 680 clear_opt(sbi, XATTR_USER); 681 break; 682 case Opt_inline_xattr: 683 set_opt(sbi, INLINE_XATTR); 684 break; 685 case Opt_noinline_xattr: 686 clear_opt(sbi, INLINE_XATTR); 687 break; 688 case Opt_inline_xattr_size: 689 if (args->from && match_int(args, &arg)) 690 return -EINVAL; 691 set_opt(sbi, INLINE_XATTR_SIZE); 692 F2FS_OPTION(sbi).inline_xattr_size = arg; 693 break; 694 #else 695 case Opt_user_xattr: 696 f2fs_info(sbi, "user_xattr options not supported"); 697 break; 698 case Opt_nouser_xattr: 699 f2fs_info(sbi, "nouser_xattr options not supported"); 700 break; 701 case Opt_inline_xattr: 702 f2fs_info(sbi, "inline_xattr options not supported"); 703 break; 704 case Opt_noinline_xattr: 705 f2fs_info(sbi, "noinline_xattr options not supported"); 706 break; 707 #endif 708 #ifdef CONFIG_F2FS_FS_POSIX_ACL 709 case Opt_acl: 710 set_opt(sbi, POSIX_ACL); 711 break; 712 case Opt_noacl: 713 clear_opt(sbi, POSIX_ACL); 714 break; 715 #else 716 case Opt_acl: 717 f2fs_info(sbi, "acl options not supported"); 718 break; 719 case Opt_noacl: 720 f2fs_info(sbi, "noacl options not supported"); 721 break; 722 #endif 723 case Opt_active_logs: 724 if (args->from && match_int(args, &arg)) 725 return -EINVAL; 726 if (arg != 2 && arg != 4 && 727 arg != NR_CURSEG_PERSIST_TYPE) 728 return -EINVAL; 729 F2FS_OPTION(sbi).active_logs = arg; 730 break; 731 case Opt_disable_ext_identify: 732 set_opt(sbi, DISABLE_EXT_IDENTIFY); 733 break; 734 case Opt_inline_data: 735 set_opt(sbi, INLINE_DATA); 736 break; 737 case Opt_inline_dentry: 738 set_opt(sbi, INLINE_DENTRY); 739 break; 740 case Opt_noinline_dentry: 741 clear_opt(sbi, INLINE_DENTRY); 742 break; 743 case Opt_flush_merge: 744 set_opt(sbi, FLUSH_MERGE); 745 break; 746 case Opt_noflush_merge: 747 clear_opt(sbi, FLUSH_MERGE); 748 break; 749 case Opt_nobarrier: 750 set_opt(sbi, NOBARRIER); 751 break; 752 case Opt_fastboot: 753 set_opt(sbi, FASTBOOT); 754 break; 755 case Opt_extent_cache: 756 set_opt(sbi, EXTENT_CACHE); 757 break; 758 case Opt_noextent_cache: 759 clear_opt(sbi, EXTENT_CACHE); 760 break; 761 case Opt_noinline_data: 762 clear_opt(sbi, INLINE_DATA); 763 break; 764 case Opt_data_flush: 765 set_opt(sbi, DATA_FLUSH); 766 break; 767 case Opt_reserve_root: 768 if (args->from && match_int(args, &arg)) 769 return -EINVAL; 770 if (test_opt(sbi, RESERVE_ROOT)) { 771 f2fs_info(sbi, "Preserve previous reserve_root=%u", 772 F2FS_OPTION(sbi).root_reserved_blocks); 773 } else { 774 F2FS_OPTION(sbi).root_reserved_blocks = arg; 775 set_opt(sbi, RESERVE_ROOT); 776 } 777 break; 778 case Opt_resuid: 779 if (args->from && match_int(args, &arg)) 780 return -EINVAL; 781 uid = make_kuid(current_user_ns(), arg); 782 if (!uid_valid(uid)) { 783 f2fs_err(sbi, "Invalid uid value %d", arg); 784 return -EINVAL; 785 } 786 F2FS_OPTION(sbi).s_resuid = uid; 787 break; 788 case Opt_resgid: 789 if (args->from && match_int(args, &arg)) 790 return -EINVAL; 791 gid = make_kgid(current_user_ns(), arg); 792 if (!gid_valid(gid)) { 793 f2fs_err(sbi, "Invalid gid value %d", arg); 794 return -EINVAL; 795 } 796 F2FS_OPTION(sbi).s_resgid = gid; 797 break; 798 case Opt_mode: 799 name = match_strdup(&args[0]); 800 801 if (!name) 802 return -ENOMEM; 803 if (!strcmp(name, "adaptive")) { 804 if (f2fs_sb_has_blkzoned(sbi)) { 805 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 806 kfree(name); 807 return -EINVAL; 808 } 809 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 810 } else if (!strcmp(name, "lfs")) { 811 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 812 } else { 813 kfree(name); 814 return -EINVAL; 815 } 816 kfree(name); 817 break; 818 case Opt_io_size_bits: 819 if (args->from && match_int(args, &arg)) 820 return -EINVAL; 821 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) { 822 f2fs_warn(sbi, "Not support %d, larger than %d", 823 1 << arg, BIO_MAX_VECS); 824 return -EINVAL; 825 } 826 F2FS_OPTION(sbi).write_io_size_bits = arg; 827 break; 828 #ifdef CONFIG_F2FS_FAULT_INJECTION 829 case Opt_fault_injection: 830 if (args->from && match_int(args, &arg)) 831 return -EINVAL; 832 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 833 set_opt(sbi, FAULT_INJECTION); 834 break; 835 836 case Opt_fault_type: 837 if (args->from && match_int(args, &arg)) 838 return -EINVAL; 839 f2fs_build_fault_attr(sbi, 0, arg); 840 set_opt(sbi, FAULT_INJECTION); 841 break; 842 #else 843 case Opt_fault_injection: 844 f2fs_info(sbi, "fault_injection options not supported"); 845 break; 846 847 case Opt_fault_type: 848 f2fs_info(sbi, "fault_type options not supported"); 849 break; 850 #endif 851 case Opt_lazytime: 852 sb->s_flags |= SB_LAZYTIME; 853 break; 854 case Opt_nolazytime: 855 sb->s_flags &= ~SB_LAZYTIME; 856 break; 857 #ifdef CONFIG_QUOTA 858 case Opt_quota: 859 case Opt_usrquota: 860 set_opt(sbi, USRQUOTA); 861 break; 862 case Opt_grpquota: 863 set_opt(sbi, GRPQUOTA); 864 break; 865 case Opt_prjquota: 866 set_opt(sbi, PRJQUOTA); 867 break; 868 case Opt_usrjquota: 869 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 870 if (ret) 871 return ret; 872 break; 873 case Opt_grpjquota: 874 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 875 if (ret) 876 return ret; 877 break; 878 case Opt_prjjquota: 879 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 880 if (ret) 881 return ret; 882 break; 883 case Opt_offusrjquota: 884 ret = f2fs_clear_qf_name(sb, USRQUOTA); 885 if (ret) 886 return ret; 887 break; 888 case Opt_offgrpjquota: 889 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 890 if (ret) 891 return ret; 892 break; 893 case Opt_offprjjquota: 894 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 895 if (ret) 896 return ret; 897 break; 898 case Opt_jqfmt_vfsold: 899 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 900 break; 901 case Opt_jqfmt_vfsv0: 902 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 903 break; 904 case Opt_jqfmt_vfsv1: 905 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 906 break; 907 case Opt_noquota: 908 clear_opt(sbi, QUOTA); 909 clear_opt(sbi, USRQUOTA); 910 clear_opt(sbi, GRPQUOTA); 911 clear_opt(sbi, PRJQUOTA); 912 break; 913 #else 914 case Opt_quota: 915 case Opt_usrquota: 916 case Opt_grpquota: 917 case Opt_prjquota: 918 case Opt_usrjquota: 919 case Opt_grpjquota: 920 case Opt_prjjquota: 921 case Opt_offusrjquota: 922 case Opt_offgrpjquota: 923 case Opt_offprjjquota: 924 case Opt_jqfmt_vfsold: 925 case Opt_jqfmt_vfsv0: 926 case Opt_jqfmt_vfsv1: 927 case Opt_noquota: 928 f2fs_info(sbi, "quota operations not supported"); 929 break; 930 #endif 931 case Opt_whint: 932 name = match_strdup(&args[0]); 933 if (!name) 934 return -ENOMEM; 935 if (!strcmp(name, "user-based")) { 936 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER; 937 } else if (!strcmp(name, "off")) { 938 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 939 } else if (!strcmp(name, "fs-based")) { 940 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS; 941 } else { 942 kfree(name); 943 return -EINVAL; 944 } 945 kfree(name); 946 break; 947 case Opt_alloc: 948 name = match_strdup(&args[0]); 949 if (!name) 950 return -ENOMEM; 951 952 if (!strcmp(name, "default")) { 953 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 954 } else if (!strcmp(name, "reuse")) { 955 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 956 } else { 957 kfree(name); 958 return -EINVAL; 959 } 960 kfree(name); 961 break; 962 case Opt_fsync: 963 name = match_strdup(&args[0]); 964 if (!name) 965 return -ENOMEM; 966 if (!strcmp(name, "posix")) { 967 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 968 } else if (!strcmp(name, "strict")) { 969 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 970 } else if (!strcmp(name, "nobarrier")) { 971 F2FS_OPTION(sbi).fsync_mode = 972 FSYNC_MODE_NOBARRIER; 973 } else { 974 kfree(name); 975 return -EINVAL; 976 } 977 kfree(name); 978 break; 979 case Opt_test_dummy_encryption: 980 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 981 is_remount); 982 if (ret) 983 return ret; 984 break; 985 case Opt_inlinecrypt: 986 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 987 sb->s_flags |= SB_INLINECRYPT; 988 #else 989 f2fs_info(sbi, "inline encryption not supported"); 990 #endif 991 break; 992 case Opt_checkpoint_disable_cap_perc: 993 if (args->from && match_int(args, &arg)) 994 return -EINVAL; 995 if (arg < 0 || arg > 100) 996 return -EINVAL; 997 F2FS_OPTION(sbi).unusable_cap_perc = arg; 998 set_opt(sbi, DISABLE_CHECKPOINT); 999 break; 1000 case Opt_checkpoint_disable_cap: 1001 if (args->from && match_int(args, &arg)) 1002 return -EINVAL; 1003 F2FS_OPTION(sbi).unusable_cap = arg; 1004 set_opt(sbi, DISABLE_CHECKPOINT); 1005 break; 1006 case Opt_checkpoint_disable: 1007 set_opt(sbi, DISABLE_CHECKPOINT); 1008 break; 1009 case Opt_checkpoint_enable: 1010 clear_opt(sbi, DISABLE_CHECKPOINT); 1011 break; 1012 case Opt_checkpoint_merge: 1013 set_opt(sbi, MERGE_CHECKPOINT); 1014 break; 1015 case Opt_nocheckpoint_merge: 1016 clear_opt(sbi, MERGE_CHECKPOINT); 1017 break; 1018 #ifdef CONFIG_F2FS_FS_COMPRESSION 1019 case Opt_compress_algorithm: 1020 if (!f2fs_sb_has_compression(sbi)) { 1021 f2fs_info(sbi, "Image doesn't support compression"); 1022 break; 1023 } 1024 name = match_strdup(&args[0]); 1025 if (!name) 1026 return -ENOMEM; 1027 if (!strcmp(name, "lzo")) { 1028 #ifdef CONFIG_F2FS_FS_LZO 1029 F2FS_OPTION(sbi).compress_level = 0; 1030 F2FS_OPTION(sbi).compress_algorithm = 1031 COMPRESS_LZO; 1032 #else 1033 f2fs_info(sbi, "kernel doesn't support lzo compression"); 1034 #endif 1035 } else if (!strncmp(name, "lz4", 3)) { 1036 #ifdef CONFIG_F2FS_FS_LZ4 1037 ret = f2fs_set_lz4hc_level(sbi, name); 1038 if (ret) { 1039 kfree(name); 1040 return -EINVAL; 1041 } 1042 F2FS_OPTION(sbi).compress_algorithm = 1043 COMPRESS_LZ4; 1044 #else 1045 f2fs_info(sbi, "kernel doesn't support lz4 compression"); 1046 #endif 1047 } else if (!strncmp(name, "zstd", 4)) { 1048 #ifdef CONFIG_F2FS_FS_ZSTD 1049 ret = f2fs_set_zstd_level(sbi, name); 1050 if (ret) { 1051 kfree(name); 1052 return -EINVAL; 1053 } 1054 F2FS_OPTION(sbi).compress_algorithm = 1055 COMPRESS_ZSTD; 1056 #else 1057 f2fs_info(sbi, "kernel doesn't support zstd compression"); 1058 #endif 1059 } else if (!strcmp(name, "lzo-rle")) { 1060 #ifdef CONFIG_F2FS_FS_LZORLE 1061 F2FS_OPTION(sbi).compress_level = 0; 1062 F2FS_OPTION(sbi).compress_algorithm = 1063 COMPRESS_LZORLE; 1064 #else 1065 f2fs_info(sbi, "kernel doesn't support lzorle compression"); 1066 #endif 1067 } else { 1068 kfree(name); 1069 return -EINVAL; 1070 } 1071 kfree(name); 1072 break; 1073 case Opt_compress_log_size: 1074 if (!f2fs_sb_has_compression(sbi)) { 1075 f2fs_info(sbi, "Image doesn't support compression"); 1076 break; 1077 } 1078 if (args->from && match_int(args, &arg)) 1079 return -EINVAL; 1080 if (arg < MIN_COMPRESS_LOG_SIZE || 1081 arg > MAX_COMPRESS_LOG_SIZE) { 1082 f2fs_err(sbi, 1083 "Compress cluster log size is out of range"); 1084 return -EINVAL; 1085 } 1086 F2FS_OPTION(sbi).compress_log_size = arg; 1087 break; 1088 case Opt_compress_extension: 1089 if (!f2fs_sb_has_compression(sbi)) { 1090 f2fs_info(sbi, "Image doesn't support compression"); 1091 break; 1092 } 1093 name = match_strdup(&args[0]); 1094 if (!name) 1095 return -ENOMEM; 1096 1097 ext = F2FS_OPTION(sbi).extensions; 1098 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1099 1100 if (strlen(name) >= F2FS_EXTENSION_LEN || 1101 ext_cnt >= COMPRESS_EXT_NUM) { 1102 f2fs_err(sbi, 1103 "invalid extension length/number"); 1104 kfree(name); 1105 return -EINVAL; 1106 } 1107 1108 strcpy(ext[ext_cnt], name); 1109 F2FS_OPTION(sbi).compress_ext_cnt++; 1110 kfree(name); 1111 break; 1112 case Opt_nocompress_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 noext = F2FS_OPTION(sbi).noextensions; 1122 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1123 1124 if (strlen(name) >= F2FS_EXTENSION_LEN || 1125 noext_cnt >= COMPRESS_EXT_NUM) { 1126 f2fs_err(sbi, 1127 "invalid extension length/number"); 1128 kfree(name); 1129 return -EINVAL; 1130 } 1131 1132 strcpy(noext[noext_cnt], name); 1133 F2FS_OPTION(sbi).nocompress_ext_cnt++; 1134 kfree(name); 1135 break; 1136 case Opt_compress_chksum: 1137 F2FS_OPTION(sbi).compress_chksum = true; 1138 break; 1139 case Opt_compress_mode: 1140 name = match_strdup(&args[0]); 1141 if (!name) 1142 return -ENOMEM; 1143 if (!strcmp(name, "fs")) { 1144 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1145 } else if (!strcmp(name, "user")) { 1146 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 1147 } else { 1148 kfree(name); 1149 return -EINVAL; 1150 } 1151 kfree(name); 1152 break; 1153 case Opt_compress_cache: 1154 set_opt(sbi, COMPRESS_CACHE); 1155 break; 1156 #else 1157 case Opt_compress_algorithm: 1158 case Opt_compress_log_size: 1159 case Opt_compress_extension: 1160 case Opt_nocompress_extension: 1161 case Opt_compress_chksum: 1162 case Opt_compress_mode: 1163 case Opt_compress_cache: 1164 f2fs_info(sbi, "compression options not supported"); 1165 break; 1166 #endif 1167 case Opt_atgc: 1168 set_opt(sbi, ATGC); 1169 break; 1170 case Opt_gc_merge: 1171 set_opt(sbi, GC_MERGE); 1172 break; 1173 case Opt_nogc_merge: 1174 clear_opt(sbi, GC_MERGE); 1175 break; 1176 default: 1177 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 1178 p); 1179 return -EINVAL; 1180 } 1181 } 1182 default_check: 1183 #ifdef CONFIG_QUOTA 1184 if (f2fs_check_quota_options(sbi)) 1185 return -EINVAL; 1186 #else 1187 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 1188 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1189 return -EINVAL; 1190 } 1191 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 1192 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1193 return -EINVAL; 1194 } 1195 #endif 1196 #ifndef CONFIG_UNICODE 1197 if (f2fs_sb_has_casefold(sbi)) { 1198 f2fs_err(sbi, 1199 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1200 return -EINVAL; 1201 } 1202 #endif 1203 /* 1204 * The BLKZONED feature indicates that the drive was formatted with 1205 * zone alignment optimization. This is optional for host-aware 1206 * devices, but mandatory for host-managed zoned block devices. 1207 */ 1208 #ifndef CONFIG_BLK_DEV_ZONED 1209 if (f2fs_sb_has_blkzoned(sbi)) { 1210 f2fs_err(sbi, "Zoned block device support is not enabled"); 1211 return -EINVAL; 1212 } 1213 #endif 1214 1215 #ifdef CONFIG_F2FS_FS_COMPRESSION 1216 if (f2fs_test_compress_extension(sbi)) { 1217 f2fs_err(sbi, "invalid compress or nocompress extension"); 1218 return -EINVAL; 1219 } 1220 #endif 1221 1222 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 1223 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO", 1224 F2FS_IO_SIZE_KB(sbi)); 1225 return -EINVAL; 1226 } 1227 1228 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1229 int min_size, max_size; 1230 1231 if (!f2fs_sb_has_extra_attr(sbi) || 1232 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1233 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1234 return -EINVAL; 1235 } 1236 if (!test_opt(sbi, INLINE_XATTR)) { 1237 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1238 return -EINVAL; 1239 } 1240 1241 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 1242 max_size = MAX_INLINE_XATTR_SIZE; 1243 1244 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1245 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1246 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1247 min_size, max_size); 1248 return -EINVAL; 1249 } 1250 } 1251 1252 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1253 f2fs_err(sbi, "LFS not compatible with checkpoint=disable"); 1254 return -EINVAL; 1255 } 1256 1257 /* Not pass down write hints if the number of active logs is lesser 1258 * than NR_CURSEG_PERSIST_TYPE. 1259 */ 1260 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE) 1261 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1262 1263 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1264 f2fs_err(sbi, "Allow to mount readonly mode only"); 1265 return -EROFS; 1266 } 1267 return 0; 1268 } 1269 1270 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1271 { 1272 struct f2fs_inode_info *fi; 1273 1274 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); 1275 if (!fi) 1276 return NULL; 1277 1278 init_once((void *) fi); 1279 1280 /* Initialize f2fs-specific inode info */ 1281 atomic_set(&fi->dirty_pages, 0); 1282 atomic_set(&fi->i_compr_blocks, 0); 1283 init_rwsem(&fi->i_sem); 1284 spin_lock_init(&fi->i_size_lock); 1285 INIT_LIST_HEAD(&fi->dirty_list); 1286 INIT_LIST_HEAD(&fi->gdirty_list); 1287 INIT_LIST_HEAD(&fi->inmem_ilist); 1288 INIT_LIST_HEAD(&fi->inmem_pages); 1289 mutex_init(&fi->inmem_lock); 1290 init_rwsem(&fi->i_gc_rwsem[READ]); 1291 init_rwsem(&fi->i_gc_rwsem[WRITE]); 1292 init_rwsem(&fi->i_mmap_sem); 1293 init_rwsem(&fi->i_xattr_sem); 1294 1295 /* Will be used by directory only */ 1296 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1297 1298 return &fi->vfs_inode; 1299 } 1300 1301 static int f2fs_drop_inode(struct inode *inode) 1302 { 1303 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1304 int ret; 1305 1306 /* 1307 * during filesystem shutdown, if checkpoint is disabled, 1308 * drop useless meta/node dirty pages. 1309 */ 1310 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1311 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1312 inode->i_ino == F2FS_META_INO(sbi)) { 1313 trace_f2fs_drop_inode(inode, 1); 1314 return 1; 1315 } 1316 } 1317 1318 /* 1319 * This is to avoid a deadlock condition like below. 1320 * writeback_single_inode(inode) 1321 * - f2fs_write_data_page 1322 * - f2fs_gc -> iput -> evict 1323 * - inode_wait_for_writeback(inode) 1324 */ 1325 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1326 if (!inode->i_nlink && !is_bad_inode(inode)) { 1327 /* to avoid evict_inode call simultaneously */ 1328 atomic_inc(&inode->i_count); 1329 spin_unlock(&inode->i_lock); 1330 1331 /* some remained atomic pages should discarded */ 1332 if (f2fs_is_atomic_file(inode)) 1333 f2fs_drop_inmem_pages(inode); 1334 1335 /* should remain fi->extent_tree for writepage */ 1336 f2fs_destroy_extent_node(inode); 1337 1338 sb_start_intwrite(inode->i_sb); 1339 f2fs_i_size_write(inode, 0); 1340 1341 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1342 inode, NULL, 0, DATA); 1343 truncate_inode_pages_final(inode->i_mapping); 1344 1345 if (F2FS_HAS_BLOCKS(inode)) 1346 f2fs_truncate(inode); 1347 1348 sb_end_intwrite(inode->i_sb); 1349 1350 spin_lock(&inode->i_lock); 1351 atomic_dec(&inode->i_count); 1352 } 1353 trace_f2fs_drop_inode(inode, 0); 1354 return 0; 1355 } 1356 ret = generic_drop_inode(inode); 1357 if (!ret) 1358 ret = fscrypt_drop_inode(inode); 1359 trace_f2fs_drop_inode(inode, ret); 1360 return ret; 1361 } 1362 1363 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1364 { 1365 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1366 int ret = 0; 1367 1368 spin_lock(&sbi->inode_lock[DIRTY_META]); 1369 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1370 ret = 1; 1371 } else { 1372 set_inode_flag(inode, FI_DIRTY_INODE); 1373 stat_inc_dirty_inode(sbi, DIRTY_META); 1374 } 1375 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1376 list_add_tail(&F2FS_I(inode)->gdirty_list, 1377 &sbi->inode_list[DIRTY_META]); 1378 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1379 } 1380 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1381 return ret; 1382 } 1383 1384 void f2fs_inode_synced(struct inode *inode) 1385 { 1386 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1387 1388 spin_lock(&sbi->inode_lock[DIRTY_META]); 1389 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1390 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1391 return; 1392 } 1393 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1394 list_del_init(&F2FS_I(inode)->gdirty_list); 1395 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1396 } 1397 clear_inode_flag(inode, FI_DIRTY_INODE); 1398 clear_inode_flag(inode, FI_AUTO_RECOVER); 1399 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1400 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1401 } 1402 1403 /* 1404 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1405 * 1406 * We should call set_dirty_inode to write the dirty inode through write_inode. 1407 */ 1408 static void f2fs_dirty_inode(struct inode *inode, int flags) 1409 { 1410 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1411 1412 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1413 inode->i_ino == F2FS_META_INO(sbi)) 1414 return; 1415 1416 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1417 clear_inode_flag(inode, FI_AUTO_RECOVER); 1418 1419 f2fs_inode_dirtied(inode, false); 1420 } 1421 1422 static void f2fs_free_inode(struct inode *inode) 1423 { 1424 fscrypt_free_inode(inode); 1425 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1426 } 1427 1428 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1429 { 1430 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1431 percpu_counter_destroy(&sbi->total_valid_inode_count); 1432 } 1433 1434 static void destroy_device_list(struct f2fs_sb_info *sbi) 1435 { 1436 int i; 1437 1438 for (i = 0; i < sbi->s_ndevs; i++) { 1439 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1440 #ifdef CONFIG_BLK_DEV_ZONED 1441 kvfree(FDEV(i).blkz_seq); 1442 kfree(FDEV(i).zone_capacity_blocks); 1443 #endif 1444 } 1445 kvfree(sbi->devs); 1446 } 1447 1448 static void f2fs_put_super(struct super_block *sb) 1449 { 1450 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1451 int i; 1452 bool dropped; 1453 1454 /* unregister procfs/sysfs entries in advance to avoid race case */ 1455 f2fs_unregister_sysfs(sbi); 1456 1457 f2fs_quota_off_umount(sb); 1458 1459 /* prevent remaining shrinker jobs */ 1460 mutex_lock(&sbi->umount_mutex); 1461 1462 /* 1463 * flush all issued checkpoints and stop checkpoint issue thread. 1464 * after then, all checkpoints should be done by each process context. 1465 */ 1466 f2fs_stop_ckpt_thread(sbi); 1467 1468 /* 1469 * We don't need to do checkpoint when superblock is clean. 1470 * But, the previous checkpoint was not done by umount, it needs to do 1471 * clean checkpoint again. 1472 */ 1473 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1474 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1475 struct cp_control cpc = { 1476 .reason = CP_UMOUNT, 1477 }; 1478 f2fs_write_checkpoint(sbi, &cpc); 1479 } 1480 1481 /* be sure to wait for any on-going discard commands */ 1482 dropped = f2fs_issue_discard_timeout(sbi); 1483 1484 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && 1485 !sbi->discard_blks && !dropped) { 1486 struct cp_control cpc = { 1487 .reason = CP_UMOUNT | CP_TRIMMED, 1488 }; 1489 f2fs_write_checkpoint(sbi, &cpc); 1490 } 1491 1492 /* 1493 * normally superblock is clean, so we need to release this. 1494 * In addition, EIO will skip do checkpoint, we need this as well. 1495 */ 1496 f2fs_release_ino_entry(sbi, true); 1497 1498 f2fs_leave_shrinker(sbi); 1499 mutex_unlock(&sbi->umount_mutex); 1500 1501 /* our cp_error case, we can wait for any writeback page */ 1502 f2fs_flush_merged_writes(sbi); 1503 1504 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1505 1506 f2fs_bug_on(sbi, sbi->fsync_node_num); 1507 1508 f2fs_destroy_compress_inode(sbi); 1509 1510 iput(sbi->node_inode); 1511 sbi->node_inode = NULL; 1512 1513 iput(sbi->meta_inode); 1514 sbi->meta_inode = NULL; 1515 1516 /* 1517 * iput() can update stat information, if f2fs_write_checkpoint() 1518 * above failed with error. 1519 */ 1520 f2fs_destroy_stats(sbi); 1521 1522 /* destroy f2fs internal modules */ 1523 f2fs_destroy_node_manager(sbi); 1524 f2fs_destroy_segment_manager(sbi); 1525 1526 f2fs_destroy_post_read_wq(sbi); 1527 1528 kvfree(sbi->ckpt); 1529 1530 sb->s_fs_info = NULL; 1531 if (sbi->s_chksum_driver) 1532 crypto_free_shash(sbi->s_chksum_driver); 1533 kfree(sbi->raw_super); 1534 1535 destroy_device_list(sbi); 1536 f2fs_destroy_page_array_cache(sbi); 1537 f2fs_destroy_xattr_caches(sbi); 1538 mempool_destroy(sbi->write_io_dummy); 1539 #ifdef CONFIG_QUOTA 1540 for (i = 0; i < MAXQUOTAS; i++) 1541 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1542 #endif 1543 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1544 destroy_percpu_info(sbi); 1545 for (i = 0; i < NR_PAGE_TYPE; i++) 1546 kvfree(sbi->write_io[i]); 1547 #ifdef CONFIG_UNICODE 1548 utf8_unload(sb->s_encoding); 1549 #endif 1550 kfree(sbi); 1551 } 1552 1553 int f2fs_sync_fs(struct super_block *sb, int sync) 1554 { 1555 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1556 int err = 0; 1557 1558 if (unlikely(f2fs_cp_error(sbi))) 1559 return 0; 1560 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1561 return 0; 1562 1563 trace_f2fs_sync_fs(sb, sync); 1564 1565 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1566 return -EAGAIN; 1567 1568 if (sync) 1569 err = f2fs_issue_checkpoint(sbi); 1570 1571 return err; 1572 } 1573 1574 static int f2fs_freeze(struct super_block *sb) 1575 { 1576 if (f2fs_readonly(sb)) 1577 return 0; 1578 1579 /* IO error happened before */ 1580 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1581 return -EIO; 1582 1583 /* must be clean, since sync_filesystem() was already called */ 1584 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1585 return -EINVAL; 1586 1587 /* ensure no checkpoint required */ 1588 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list)) 1589 return -EINVAL; 1590 return 0; 1591 } 1592 1593 static int f2fs_unfreeze(struct super_block *sb) 1594 { 1595 return 0; 1596 } 1597 1598 #ifdef CONFIG_QUOTA 1599 static int f2fs_statfs_project(struct super_block *sb, 1600 kprojid_t projid, struct kstatfs *buf) 1601 { 1602 struct kqid qid; 1603 struct dquot *dquot; 1604 u64 limit; 1605 u64 curblock; 1606 1607 qid = make_kqid_projid(projid); 1608 dquot = dqget(sb, qid); 1609 if (IS_ERR(dquot)) 1610 return PTR_ERR(dquot); 1611 spin_lock(&dquot->dq_dqb_lock); 1612 1613 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1614 dquot->dq_dqb.dqb_bhardlimit); 1615 if (limit) 1616 limit >>= sb->s_blocksize_bits; 1617 1618 if (limit && buf->f_blocks > limit) { 1619 curblock = (dquot->dq_dqb.dqb_curspace + 1620 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1621 buf->f_blocks = limit; 1622 buf->f_bfree = buf->f_bavail = 1623 (buf->f_blocks > curblock) ? 1624 (buf->f_blocks - curblock) : 0; 1625 } 1626 1627 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1628 dquot->dq_dqb.dqb_ihardlimit); 1629 1630 if (limit && buf->f_files > limit) { 1631 buf->f_files = limit; 1632 buf->f_ffree = 1633 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1634 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1635 } 1636 1637 spin_unlock(&dquot->dq_dqb_lock); 1638 dqput(dquot); 1639 return 0; 1640 } 1641 #endif 1642 1643 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1644 { 1645 struct super_block *sb = dentry->d_sb; 1646 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1647 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1648 block_t total_count, user_block_count, start_count; 1649 u64 avail_node_count; 1650 1651 total_count = le64_to_cpu(sbi->raw_super->block_count); 1652 user_block_count = sbi->user_block_count; 1653 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1654 buf->f_type = F2FS_SUPER_MAGIC; 1655 buf->f_bsize = sbi->blocksize; 1656 1657 buf->f_blocks = total_count - start_count; 1658 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1659 sbi->current_reserved_blocks; 1660 1661 spin_lock(&sbi->stat_lock); 1662 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1663 buf->f_bfree = 0; 1664 else 1665 buf->f_bfree -= sbi->unusable_block_count; 1666 spin_unlock(&sbi->stat_lock); 1667 1668 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1669 buf->f_bavail = buf->f_bfree - 1670 F2FS_OPTION(sbi).root_reserved_blocks; 1671 else 1672 buf->f_bavail = 0; 1673 1674 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1675 1676 if (avail_node_count > user_block_count) { 1677 buf->f_files = user_block_count; 1678 buf->f_ffree = buf->f_bavail; 1679 } else { 1680 buf->f_files = avail_node_count; 1681 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1682 buf->f_bavail); 1683 } 1684 1685 buf->f_namelen = F2FS_NAME_LEN; 1686 buf->f_fsid = u64_to_fsid(id); 1687 1688 #ifdef CONFIG_QUOTA 1689 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1690 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1691 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1692 } 1693 #endif 1694 return 0; 1695 } 1696 1697 static inline void f2fs_show_quota_options(struct seq_file *seq, 1698 struct super_block *sb) 1699 { 1700 #ifdef CONFIG_QUOTA 1701 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1702 1703 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1704 char *fmtname = ""; 1705 1706 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1707 case QFMT_VFS_OLD: 1708 fmtname = "vfsold"; 1709 break; 1710 case QFMT_VFS_V0: 1711 fmtname = "vfsv0"; 1712 break; 1713 case QFMT_VFS_V1: 1714 fmtname = "vfsv1"; 1715 break; 1716 } 1717 seq_printf(seq, ",jqfmt=%s", fmtname); 1718 } 1719 1720 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1721 seq_show_option(seq, "usrjquota", 1722 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1723 1724 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1725 seq_show_option(seq, "grpjquota", 1726 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1727 1728 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1729 seq_show_option(seq, "prjjquota", 1730 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1731 #endif 1732 } 1733 1734 #ifdef CONFIG_F2FS_FS_COMPRESSION 1735 static inline void f2fs_show_compress_options(struct seq_file *seq, 1736 struct super_block *sb) 1737 { 1738 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1739 char *algtype = ""; 1740 int i; 1741 1742 if (!f2fs_sb_has_compression(sbi)) 1743 return; 1744 1745 switch (F2FS_OPTION(sbi).compress_algorithm) { 1746 case COMPRESS_LZO: 1747 algtype = "lzo"; 1748 break; 1749 case COMPRESS_LZ4: 1750 algtype = "lz4"; 1751 break; 1752 case COMPRESS_ZSTD: 1753 algtype = "zstd"; 1754 break; 1755 case COMPRESS_LZORLE: 1756 algtype = "lzo-rle"; 1757 break; 1758 } 1759 seq_printf(seq, ",compress_algorithm=%s", algtype); 1760 1761 if (F2FS_OPTION(sbi).compress_level) 1762 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1763 1764 seq_printf(seq, ",compress_log_size=%u", 1765 F2FS_OPTION(sbi).compress_log_size); 1766 1767 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1768 seq_printf(seq, ",compress_extension=%s", 1769 F2FS_OPTION(sbi).extensions[i]); 1770 } 1771 1772 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1773 seq_printf(seq, ",nocompress_extension=%s", 1774 F2FS_OPTION(sbi).noextensions[i]); 1775 } 1776 1777 if (F2FS_OPTION(sbi).compress_chksum) 1778 seq_puts(seq, ",compress_chksum"); 1779 1780 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1781 seq_printf(seq, ",compress_mode=%s", "fs"); 1782 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1783 seq_printf(seq, ",compress_mode=%s", "user"); 1784 1785 if (test_opt(sbi, COMPRESS_CACHE)) 1786 seq_puts(seq, ",compress_cache"); 1787 } 1788 #endif 1789 1790 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1791 { 1792 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1793 1794 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1795 seq_printf(seq, ",background_gc=%s", "sync"); 1796 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1797 seq_printf(seq, ",background_gc=%s", "on"); 1798 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1799 seq_printf(seq, ",background_gc=%s", "off"); 1800 1801 if (test_opt(sbi, GC_MERGE)) 1802 seq_puts(seq, ",gc_merge"); 1803 1804 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1805 seq_puts(seq, ",disable_roll_forward"); 1806 if (test_opt(sbi, NORECOVERY)) 1807 seq_puts(seq, ",norecovery"); 1808 if (test_opt(sbi, DISCARD)) 1809 seq_puts(seq, ",discard"); 1810 else 1811 seq_puts(seq, ",nodiscard"); 1812 if (test_opt(sbi, NOHEAP)) 1813 seq_puts(seq, ",no_heap"); 1814 else 1815 seq_puts(seq, ",heap"); 1816 #ifdef CONFIG_F2FS_FS_XATTR 1817 if (test_opt(sbi, XATTR_USER)) 1818 seq_puts(seq, ",user_xattr"); 1819 else 1820 seq_puts(seq, ",nouser_xattr"); 1821 if (test_opt(sbi, INLINE_XATTR)) 1822 seq_puts(seq, ",inline_xattr"); 1823 else 1824 seq_puts(seq, ",noinline_xattr"); 1825 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1826 seq_printf(seq, ",inline_xattr_size=%u", 1827 F2FS_OPTION(sbi).inline_xattr_size); 1828 #endif 1829 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1830 if (test_opt(sbi, POSIX_ACL)) 1831 seq_puts(seq, ",acl"); 1832 else 1833 seq_puts(seq, ",noacl"); 1834 #endif 1835 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1836 seq_puts(seq, ",disable_ext_identify"); 1837 if (test_opt(sbi, INLINE_DATA)) 1838 seq_puts(seq, ",inline_data"); 1839 else 1840 seq_puts(seq, ",noinline_data"); 1841 if (test_opt(sbi, INLINE_DENTRY)) 1842 seq_puts(seq, ",inline_dentry"); 1843 else 1844 seq_puts(seq, ",noinline_dentry"); 1845 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1846 seq_puts(seq, ",flush_merge"); 1847 if (test_opt(sbi, NOBARRIER)) 1848 seq_puts(seq, ",nobarrier"); 1849 if (test_opt(sbi, FASTBOOT)) 1850 seq_puts(seq, ",fastboot"); 1851 if (test_opt(sbi, EXTENT_CACHE)) 1852 seq_puts(seq, ",extent_cache"); 1853 else 1854 seq_puts(seq, ",noextent_cache"); 1855 if (test_opt(sbi, DATA_FLUSH)) 1856 seq_puts(seq, ",data_flush"); 1857 1858 seq_puts(seq, ",mode="); 1859 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1860 seq_puts(seq, "adaptive"); 1861 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1862 seq_puts(seq, "lfs"); 1863 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1864 if (test_opt(sbi, RESERVE_ROOT)) 1865 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1866 F2FS_OPTION(sbi).root_reserved_blocks, 1867 from_kuid_munged(&init_user_ns, 1868 F2FS_OPTION(sbi).s_resuid), 1869 from_kgid_munged(&init_user_ns, 1870 F2FS_OPTION(sbi).s_resgid)); 1871 if (F2FS_IO_SIZE_BITS(sbi)) 1872 seq_printf(seq, ",io_bits=%u", 1873 F2FS_OPTION(sbi).write_io_size_bits); 1874 #ifdef CONFIG_F2FS_FAULT_INJECTION 1875 if (test_opt(sbi, FAULT_INJECTION)) { 1876 seq_printf(seq, ",fault_injection=%u", 1877 F2FS_OPTION(sbi).fault_info.inject_rate); 1878 seq_printf(seq, ",fault_type=%u", 1879 F2FS_OPTION(sbi).fault_info.inject_type); 1880 } 1881 #endif 1882 #ifdef CONFIG_QUOTA 1883 if (test_opt(sbi, QUOTA)) 1884 seq_puts(seq, ",quota"); 1885 if (test_opt(sbi, USRQUOTA)) 1886 seq_puts(seq, ",usrquota"); 1887 if (test_opt(sbi, GRPQUOTA)) 1888 seq_puts(seq, ",grpquota"); 1889 if (test_opt(sbi, PRJQUOTA)) 1890 seq_puts(seq, ",prjquota"); 1891 #endif 1892 f2fs_show_quota_options(seq, sbi->sb); 1893 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1894 seq_printf(seq, ",whint_mode=%s", "user-based"); 1895 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1896 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1897 1898 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1899 1900 if (sbi->sb->s_flags & SB_INLINECRYPT) 1901 seq_puts(seq, ",inlinecrypt"); 1902 1903 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1904 seq_printf(seq, ",alloc_mode=%s", "default"); 1905 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1906 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1907 1908 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1909 seq_printf(seq, ",checkpoint=disable:%u", 1910 F2FS_OPTION(sbi).unusable_cap); 1911 if (test_opt(sbi, MERGE_CHECKPOINT)) 1912 seq_puts(seq, ",checkpoint_merge"); 1913 else 1914 seq_puts(seq, ",nocheckpoint_merge"); 1915 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1916 seq_printf(seq, ",fsync_mode=%s", "posix"); 1917 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1918 seq_printf(seq, ",fsync_mode=%s", "strict"); 1919 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1920 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 1921 1922 #ifdef CONFIG_F2FS_FS_COMPRESSION 1923 f2fs_show_compress_options(seq, sbi->sb); 1924 #endif 1925 1926 if (test_opt(sbi, ATGC)) 1927 seq_puts(seq, ",atgc"); 1928 return 0; 1929 } 1930 1931 static void default_options(struct f2fs_sb_info *sbi) 1932 { 1933 /* init some FS parameters */ 1934 if (f2fs_sb_has_readonly(sbi)) 1935 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 1936 else 1937 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 1938 1939 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 1940 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1941 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1942 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1943 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 1944 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 1945 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 1946 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 1947 F2FS_OPTION(sbi).compress_ext_cnt = 0; 1948 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1949 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 1950 1951 sbi->sb->s_flags &= ~SB_INLINECRYPT; 1952 1953 set_opt(sbi, INLINE_XATTR); 1954 set_opt(sbi, INLINE_DATA); 1955 set_opt(sbi, INLINE_DENTRY); 1956 set_opt(sbi, EXTENT_CACHE); 1957 set_opt(sbi, NOHEAP); 1958 clear_opt(sbi, DISABLE_CHECKPOINT); 1959 set_opt(sbi, MERGE_CHECKPOINT); 1960 F2FS_OPTION(sbi).unusable_cap = 0; 1961 sbi->sb->s_flags |= SB_LAZYTIME; 1962 set_opt(sbi, FLUSH_MERGE); 1963 set_opt(sbi, DISCARD); 1964 if (f2fs_sb_has_blkzoned(sbi)) 1965 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 1966 else 1967 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 1968 1969 #ifdef CONFIG_F2FS_FS_XATTR 1970 set_opt(sbi, XATTR_USER); 1971 #endif 1972 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1973 set_opt(sbi, POSIX_ACL); 1974 #endif 1975 1976 f2fs_build_fault_attr(sbi, 0, 0); 1977 } 1978 1979 #ifdef CONFIG_QUOTA 1980 static int f2fs_enable_quotas(struct super_block *sb); 1981 #endif 1982 1983 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 1984 { 1985 unsigned int s_flags = sbi->sb->s_flags; 1986 struct cp_control cpc; 1987 int err = 0; 1988 int ret; 1989 block_t unusable; 1990 1991 if (s_flags & SB_RDONLY) { 1992 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 1993 return -EINVAL; 1994 } 1995 sbi->sb->s_flags |= SB_ACTIVE; 1996 1997 f2fs_update_time(sbi, DISABLE_TIME); 1998 1999 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2000 down_write(&sbi->gc_lock); 2001 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO); 2002 if (err == -ENODATA) { 2003 err = 0; 2004 break; 2005 } 2006 if (err && err != -EAGAIN) 2007 break; 2008 } 2009 2010 ret = sync_filesystem(sbi->sb); 2011 if (ret || err) { 2012 err = ret ? ret : err; 2013 goto restore_flag; 2014 } 2015 2016 unusable = f2fs_get_unusable_blocks(sbi); 2017 if (f2fs_disable_cp_again(sbi, unusable)) { 2018 err = -EAGAIN; 2019 goto restore_flag; 2020 } 2021 2022 down_write(&sbi->gc_lock); 2023 cpc.reason = CP_PAUSE; 2024 set_sbi_flag(sbi, SBI_CP_DISABLED); 2025 err = f2fs_write_checkpoint(sbi, &cpc); 2026 if (err) 2027 goto out_unlock; 2028 2029 spin_lock(&sbi->stat_lock); 2030 sbi->unusable_block_count = unusable; 2031 spin_unlock(&sbi->stat_lock); 2032 2033 out_unlock: 2034 up_write(&sbi->gc_lock); 2035 restore_flag: 2036 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2037 return err; 2038 } 2039 2040 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2041 { 2042 /* we should flush all the data to keep data consistency */ 2043 sync_inodes_sb(sbi->sb); 2044 2045 down_write(&sbi->gc_lock); 2046 f2fs_dirty_to_prefree(sbi); 2047 2048 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2049 set_sbi_flag(sbi, SBI_IS_DIRTY); 2050 up_write(&sbi->gc_lock); 2051 2052 f2fs_sync_fs(sbi->sb, 1); 2053 } 2054 2055 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2056 { 2057 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2058 struct f2fs_mount_info org_mount_opt; 2059 unsigned long old_sb_flags; 2060 int err; 2061 bool need_restart_gc = false, need_stop_gc = false; 2062 bool need_restart_ckpt = false, need_stop_ckpt = false; 2063 bool need_restart_flush = false, need_stop_flush = false; 2064 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 2065 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT); 2066 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 2067 bool no_atgc = !test_opt(sbi, ATGC); 2068 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2069 bool checkpoint_changed; 2070 #ifdef CONFIG_QUOTA 2071 int i, j; 2072 #endif 2073 2074 /* 2075 * Save the old mount options in case we 2076 * need to restore them. 2077 */ 2078 org_mount_opt = sbi->mount_opt; 2079 old_sb_flags = sb->s_flags; 2080 2081 #ifdef CONFIG_QUOTA 2082 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2083 for (i = 0; i < MAXQUOTAS; i++) { 2084 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2085 org_mount_opt.s_qf_names[i] = 2086 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2087 GFP_KERNEL); 2088 if (!org_mount_opt.s_qf_names[i]) { 2089 for (j = 0; j < i; j++) 2090 kfree(org_mount_opt.s_qf_names[j]); 2091 return -ENOMEM; 2092 } 2093 } else { 2094 org_mount_opt.s_qf_names[i] = NULL; 2095 } 2096 } 2097 #endif 2098 2099 /* recover superblocks we couldn't write due to previous RO mount */ 2100 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2101 err = f2fs_commit_super(sbi, false); 2102 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2103 err); 2104 if (!err) 2105 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2106 } 2107 2108 default_options(sbi); 2109 2110 /* parse mount options */ 2111 err = parse_options(sb, data, true); 2112 if (err) 2113 goto restore_opts; 2114 checkpoint_changed = 2115 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT); 2116 2117 /* 2118 * Previous and new state of filesystem is RO, 2119 * so skip checking GC and FLUSH_MERGE conditions. 2120 */ 2121 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2122 goto skip; 2123 2124 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) { 2125 err = -EROFS; 2126 goto restore_opts; 2127 } 2128 2129 #ifdef CONFIG_QUOTA 2130 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2131 err = dquot_suspend(sb, -1); 2132 if (err < 0) 2133 goto restore_opts; 2134 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2135 /* dquot_resume needs RW */ 2136 sb->s_flags &= ~SB_RDONLY; 2137 if (sb_any_quota_suspended(sb)) { 2138 dquot_resume(sb, -1); 2139 } else if (f2fs_sb_has_quota_ino(sbi)) { 2140 err = f2fs_enable_quotas(sb); 2141 if (err) 2142 goto restore_opts; 2143 } 2144 } 2145 #endif 2146 /* disallow enable atgc dynamically */ 2147 if (no_atgc == !!test_opt(sbi, ATGC)) { 2148 err = -EINVAL; 2149 f2fs_warn(sbi, "switch atgc option is not allowed"); 2150 goto restore_opts; 2151 } 2152 2153 /* disallow enable/disable extent_cache dynamically */ 2154 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 2155 err = -EINVAL; 2156 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2157 goto restore_opts; 2158 } 2159 2160 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 2161 err = -EINVAL; 2162 f2fs_warn(sbi, "switch io_bits option is not allowed"); 2163 goto restore_opts; 2164 } 2165 2166 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2167 err = -EINVAL; 2168 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2169 goto restore_opts; 2170 } 2171 2172 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2173 err = -EINVAL; 2174 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2175 goto restore_opts; 2176 } 2177 2178 /* 2179 * We stop the GC thread if FS is mounted as RO 2180 * or if background_gc = off is passed in mount 2181 * option. Also sync the filesystem. 2182 */ 2183 if ((*flags & SB_RDONLY) || 2184 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2185 !test_opt(sbi, GC_MERGE))) { 2186 if (sbi->gc_thread) { 2187 f2fs_stop_gc_thread(sbi); 2188 need_restart_gc = true; 2189 } 2190 } else if (!sbi->gc_thread) { 2191 err = f2fs_start_gc_thread(sbi); 2192 if (err) 2193 goto restore_opts; 2194 need_stop_gc = true; 2195 } 2196 2197 if (*flags & SB_RDONLY || 2198 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 2199 sync_inodes_sb(sb); 2200 2201 set_sbi_flag(sbi, SBI_IS_DIRTY); 2202 set_sbi_flag(sbi, SBI_IS_CLOSE); 2203 f2fs_sync_fs(sb, 1); 2204 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2205 } 2206 2207 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2208 !test_opt(sbi, MERGE_CHECKPOINT)) { 2209 f2fs_stop_ckpt_thread(sbi); 2210 need_restart_ckpt = true; 2211 } else { 2212 err = f2fs_start_ckpt_thread(sbi); 2213 if (err) { 2214 f2fs_err(sbi, 2215 "Failed to start F2FS issue_checkpoint_thread (%d)", 2216 err); 2217 goto restore_gc; 2218 } 2219 need_stop_ckpt = true; 2220 } 2221 2222 /* 2223 * We stop issue flush thread if FS is mounted as RO 2224 * or if flush_merge is not passed in mount option. 2225 */ 2226 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2227 clear_opt(sbi, FLUSH_MERGE); 2228 f2fs_destroy_flush_cmd_control(sbi, false); 2229 need_restart_flush = true; 2230 } else { 2231 err = f2fs_create_flush_cmd_control(sbi); 2232 if (err) 2233 goto restore_ckpt; 2234 need_stop_flush = true; 2235 } 2236 2237 if (checkpoint_changed) { 2238 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2239 err = f2fs_disable_checkpoint(sbi); 2240 if (err) 2241 goto restore_flush; 2242 } else { 2243 f2fs_enable_checkpoint(sbi); 2244 } 2245 } 2246 2247 skip: 2248 #ifdef CONFIG_QUOTA 2249 /* Release old quota file names */ 2250 for (i = 0; i < MAXQUOTAS; i++) 2251 kfree(org_mount_opt.s_qf_names[i]); 2252 #endif 2253 /* Update the POSIXACL Flag */ 2254 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2255 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2256 2257 limit_reserve_root(sbi); 2258 adjust_unusable_cap_perc(sbi); 2259 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2260 return 0; 2261 restore_flush: 2262 if (need_restart_flush) { 2263 if (f2fs_create_flush_cmd_control(sbi)) 2264 f2fs_warn(sbi, "background flush thread has stopped"); 2265 } else if (need_stop_flush) { 2266 clear_opt(sbi, FLUSH_MERGE); 2267 f2fs_destroy_flush_cmd_control(sbi, false); 2268 } 2269 restore_ckpt: 2270 if (need_restart_ckpt) { 2271 if (f2fs_start_ckpt_thread(sbi)) 2272 f2fs_warn(sbi, "background ckpt thread has stopped"); 2273 } else if (need_stop_ckpt) { 2274 f2fs_stop_ckpt_thread(sbi); 2275 } 2276 restore_gc: 2277 if (need_restart_gc) { 2278 if (f2fs_start_gc_thread(sbi)) 2279 f2fs_warn(sbi, "background gc thread has stopped"); 2280 } else if (need_stop_gc) { 2281 f2fs_stop_gc_thread(sbi); 2282 } 2283 restore_opts: 2284 #ifdef CONFIG_QUOTA 2285 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2286 for (i = 0; i < MAXQUOTAS; i++) { 2287 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2288 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2289 } 2290 #endif 2291 sbi->mount_opt = org_mount_opt; 2292 sb->s_flags = old_sb_flags; 2293 return err; 2294 } 2295 2296 #ifdef CONFIG_QUOTA 2297 /* Read data from quotafile */ 2298 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2299 size_t len, loff_t off) 2300 { 2301 struct inode *inode = sb_dqopt(sb)->files[type]; 2302 struct address_space *mapping = inode->i_mapping; 2303 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2304 int offset = off & (sb->s_blocksize - 1); 2305 int tocopy; 2306 size_t toread; 2307 loff_t i_size = i_size_read(inode); 2308 struct page *page; 2309 char *kaddr; 2310 2311 if (off > i_size) 2312 return 0; 2313 2314 if (off + len > i_size) 2315 len = i_size - off; 2316 toread = len; 2317 while (toread > 0) { 2318 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2319 repeat: 2320 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2321 if (IS_ERR(page)) { 2322 if (PTR_ERR(page) == -ENOMEM) { 2323 congestion_wait(BLK_RW_ASYNC, 2324 DEFAULT_IO_TIMEOUT); 2325 goto repeat; 2326 } 2327 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2328 return PTR_ERR(page); 2329 } 2330 2331 lock_page(page); 2332 2333 if (unlikely(page->mapping != mapping)) { 2334 f2fs_put_page(page, 1); 2335 goto repeat; 2336 } 2337 if (unlikely(!PageUptodate(page))) { 2338 f2fs_put_page(page, 1); 2339 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2340 return -EIO; 2341 } 2342 2343 kaddr = kmap_atomic(page); 2344 memcpy(data, kaddr + offset, tocopy); 2345 kunmap_atomic(kaddr); 2346 f2fs_put_page(page, 1); 2347 2348 offset = 0; 2349 toread -= tocopy; 2350 data += tocopy; 2351 blkidx++; 2352 } 2353 return len; 2354 } 2355 2356 /* Write to quotafile */ 2357 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2358 const char *data, size_t len, loff_t off) 2359 { 2360 struct inode *inode = sb_dqopt(sb)->files[type]; 2361 struct address_space *mapping = inode->i_mapping; 2362 const struct address_space_operations *a_ops = mapping->a_ops; 2363 int offset = off & (sb->s_blocksize - 1); 2364 size_t towrite = len; 2365 struct page *page; 2366 void *fsdata = NULL; 2367 char *kaddr; 2368 int err = 0; 2369 int tocopy; 2370 2371 while (towrite > 0) { 2372 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2373 towrite); 2374 retry: 2375 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2376 &page, &fsdata); 2377 if (unlikely(err)) { 2378 if (err == -ENOMEM) { 2379 congestion_wait(BLK_RW_ASYNC, 2380 DEFAULT_IO_TIMEOUT); 2381 goto retry; 2382 } 2383 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2384 break; 2385 } 2386 2387 kaddr = kmap_atomic(page); 2388 memcpy(kaddr + offset, data, tocopy); 2389 kunmap_atomic(kaddr); 2390 flush_dcache_page(page); 2391 2392 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2393 page, fsdata); 2394 offset = 0; 2395 towrite -= tocopy; 2396 off += tocopy; 2397 data += tocopy; 2398 cond_resched(); 2399 } 2400 2401 if (len == towrite) 2402 return err; 2403 inode->i_mtime = inode->i_ctime = current_time(inode); 2404 f2fs_mark_inode_dirty_sync(inode, false); 2405 return len - towrite; 2406 } 2407 2408 static struct dquot **f2fs_get_dquots(struct inode *inode) 2409 { 2410 return F2FS_I(inode)->i_dquot; 2411 } 2412 2413 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2414 { 2415 return &F2FS_I(inode)->i_reserved_quota; 2416 } 2417 2418 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2419 { 2420 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2421 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2422 return 0; 2423 } 2424 2425 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2426 F2FS_OPTION(sbi).s_jquota_fmt, type); 2427 } 2428 2429 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2430 { 2431 int enabled = 0; 2432 int i, err; 2433 2434 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2435 err = f2fs_enable_quotas(sbi->sb); 2436 if (err) { 2437 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2438 return 0; 2439 } 2440 return 1; 2441 } 2442 2443 for (i = 0; i < MAXQUOTAS; i++) { 2444 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2445 err = f2fs_quota_on_mount(sbi, i); 2446 if (!err) { 2447 enabled = 1; 2448 continue; 2449 } 2450 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2451 err, i); 2452 } 2453 } 2454 return enabled; 2455 } 2456 2457 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2458 unsigned int flags) 2459 { 2460 struct inode *qf_inode; 2461 unsigned long qf_inum; 2462 int err; 2463 2464 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2465 2466 qf_inum = f2fs_qf_ino(sb, type); 2467 if (!qf_inum) 2468 return -EPERM; 2469 2470 qf_inode = f2fs_iget(sb, qf_inum); 2471 if (IS_ERR(qf_inode)) { 2472 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2473 return PTR_ERR(qf_inode); 2474 } 2475 2476 /* Don't account quota for quota files to avoid recursion */ 2477 qf_inode->i_flags |= S_NOQUOTA; 2478 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2479 iput(qf_inode); 2480 return err; 2481 } 2482 2483 static int f2fs_enable_quotas(struct super_block *sb) 2484 { 2485 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2486 int type, err = 0; 2487 unsigned long qf_inum; 2488 bool quota_mopt[MAXQUOTAS] = { 2489 test_opt(sbi, USRQUOTA), 2490 test_opt(sbi, GRPQUOTA), 2491 test_opt(sbi, PRJQUOTA), 2492 }; 2493 2494 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2495 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2496 return 0; 2497 } 2498 2499 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2500 2501 for (type = 0; type < MAXQUOTAS; type++) { 2502 qf_inum = f2fs_qf_ino(sb, type); 2503 if (qf_inum) { 2504 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2505 DQUOT_USAGE_ENABLED | 2506 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2507 if (err) { 2508 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2509 type, err); 2510 for (type--; type >= 0; type--) 2511 dquot_quota_off(sb, type); 2512 set_sbi_flag(F2FS_SB(sb), 2513 SBI_QUOTA_NEED_REPAIR); 2514 return err; 2515 } 2516 } 2517 } 2518 return 0; 2519 } 2520 2521 int f2fs_quota_sync(struct super_block *sb, int type) 2522 { 2523 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2524 struct quota_info *dqopt = sb_dqopt(sb); 2525 int cnt; 2526 int ret; 2527 2528 /* 2529 * do_quotactl 2530 * f2fs_quota_sync 2531 * down_read(quota_sem) 2532 * dquot_writeback_dquots() 2533 * f2fs_dquot_commit 2534 * block_operation 2535 * down_read(quota_sem) 2536 */ 2537 f2fs_lock_op(sbi); 2538 2539 down_read(&sbi->quota_sem); 2540 ret = dquot_writeback_dquots(sb, type); 2541 if (ret) 2542 goto out; 2543 2544 /* 2545 * Now when everything is written we can discard the pagecache so 2546 * that userspace sees the changes. 2547 */ 2548 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2549 struct address_space *mapping; 2550 2551 if (type != -1 && cnt != type) 2552 continue; 2553 if (!sb_has_quota_active(sb, cnt)) 2554 continue; 2555 2556 mapping = dqopt->files[cnt]->i_mapping; 2557 2558 ret = filemap_fdatawrite(mapping); 2559 if (ret) 2560 goto out; 2561 2562 /* if we are using journalled quota */ 2563 if (is_journalled_quota(sbi)) 2564 continue; 2565 2566 ret = filemap_fdatawait(mapping); 2567 if (ret) 2568 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2569 2570 inode_lock(dqopt->files[cnt]); 2571 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0); 2572 inode_unlock(dqopt->files[cnt]); 2573 } 2574 out: 2575 if (ret) 2576 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2577 up_read(&sbi->quota_sem); 2578 f2fs_unlock_op(sbi); 2579 return ret; 2580 } 2581 2582 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2583 const struct path *path) 2584 { 2585 struct inode *inode; 2586 int err; 2587 2588 /* if quota sysfile exists, deny enabling quota with specific file */ 2589 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2590 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2591 return -EBUSY; 2592 } 2593 2594 err = f2fs_quota_sync(sb, type); 2595 if (err) 2596 return err; 2597 2598 err = dquot_quota_on(sb, type, format_id, path); 2599 if (err) 2600 return err; 2601 2602 inode = d_inode(path->dentry); 2603 2604 inode_lock(inode); 2605 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2606 f2fs_set_inode_flags(inode); 2607 inode_unlock(inode); 2608 f2fs_mark_inode_dirty_sync(inode, false); 2609 2610 return 0; 2611 } 2612 2613 static int __f2fs_quota_off(struct super_block *sb, int type) 2614 { 2615 struct inode *inode = sb_dqopt(sb)->files[type]; 2616 int err; 2617 2618 if (!inode || !igrab(inode)) 2619 return dquot_quota_off(sb, type); 2620 2621 err = f2fs_quota_sync(sb, type); 2622 if (err) 2623 goto out_put; 2624 2625 err = dquot_quota_off(sb, type); 2626 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2627 goto out_put; 2628 2629 inode_lock(inode); 2630 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2631 f2fs_set_inode_flags(inode); 2632 inode_unlock(inode); 2633 f2fs_mark_inode_dirty_sync(inode, false); 2634 out_put: 2635 iput(inode); 2636 return err; 2637 } 2638 2639 static int f2fs_quota_off(struct super_block *sb, int type) 2640 { 2641 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2642 int err; 2643 2644 err = __f2fs_quota_off(sb, type); 2645 2646 /* 2647 * quotactl can shutdown journalled quota, result in inconsistence 2648 * between quota record and fs data by following updates, tag the 2649 * flag to let fsck be aware of it. 2650 */ 2651 if (is_journalled_quota(sbi)) 2652 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2653 return err; 2654 } 2655 2656 void f2fs_quota_off_umount(struct super_block *sb) 2657 { 2658 int type; 2659 int err; 2660 2661 for (type = 0; type < MAXQUOTAS; type++) { 2662 err = __f2fs_quota_off(sb, type); 2663 if (err) { 2664 int ret = dquot_quota_off(sb, type); 2665 2666 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2667 type, err, ret); 2668 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2669 } 2670 } 2671 /* 2672 * In case of checkpoint=disable, we must flush quota blocks. 2673 * This can cause NULL exception for node_inode in end_io, since 2674 * put_super already dropped it. 2675 */ 2676 sync_filesystem(sb); 2677 } 2678 2679 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2680 { 2681 struct quota_info *dqopt = sb_dqopt(sb); 2682 int type; 2683 2684 for (type = 0; type < MAXQUOTAS; type++) { 2685 if (!dqopt->files[type]) 2686 continue; 2687 f2fs_inode_synced(dqopt->files[type]); 2688 } 2689 } 2690 2691 static int f2fs_dquot_commit(struct dquot *dquot) 2692 { 2693 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2694 int ret; 2695 2696 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2697 ret = dquot_commit(dquot); 2698 if (ret < 0) 2699 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2700 up_read(&sbi->quota_sem); 2701 return ret; 2702 } 2703 2704 static int f2fs_dquot_acquire(struct dquot *dquot) 2705 { 2706 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2707 int ret; 2708 2709 down_read(&sbi->quota_sem); 2710 ret = dquot_acquire(dquot); 2711 if (ret < 0) 2712 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2713 up_read(&sbi->quota_sem); 2714 return ret; 2715 } 2716 2717 static int f2fs_dquot_release(struct dquot *dquot) 2718 { 2719 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2720 int ret = dquot_release(dquot); 2721 2722 if (ret < 0) 2723 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2724 return ret; 2725 } 2726 2727 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2728 { 2729 struct super_block *sb = dquot->dq_sb; 2730 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2731 int ret = dquot_mark_dquot_dirty(dquot); 2732 2733 /* if we are using journalled quota */ 2734 if (is_journalled_quota(sbi)) 2735 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2736 2737 return ret; 2738 } 2739 2740 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2741 { 2742 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2743 int ret = dquot_commit_info(sb, type); 2744 2745 if (ret < 0) 2746 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2747 return ret; 2748 } 2749 2750 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2751 { 2752 *projid = F2FS_I(inode)->i_projid; 2753 return 0; 2754 } 2755 2756 static const struct dquot_operations f2fs_quota_operations = { 2757 .get_reserved_space = f2fs_get_reserved_space, 2758 .write_dquot = f2fs_dquot_commit, 2759 .acquire_dquot = f2fs_dquot_acquire, 2760 .release_dquot = f2fs_dquot_release, 2761 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2762 .write_info = f2fs_dquot_commit_info, 2763 .alloc_dquot = dquot_alloc, 2764 .destroy_dquot = dquot_destroy, 2765 .get_projid = f2fs_get_projid, 2766 .get_next_id = dquot_get_next_id, 2767 }; 2768 2769 static const struct quotactl_ops f2fs_quotactl_ops = { 2770 .quota_on = f2fs_quota_on, 2771 .quota_off = f2fs_quota_off, 2772 .quota_sync = f2fs_quota_sync, 2773 .get_state = dquot_get_state, 2774 .set_info = dquot_set_dqinfo, 2775 .get_dqblk = dquot_get_dqblk, 2776 .set_dqblk = dquot_set_dqblk, 2777 .get_nextdqblk = dquot_get_next_dqblk, 2778 }; 2779 #else 2780 int f2fs_quota_sync(struct super_block *sb, int type) 2781 { 2782 return 0; 2783 } 2784 2785 void f2fs_quota_off_umount(struct super_block *sb) 2786 { 2787 } 2788 #endif 2789 2790 static const struct super_operations f2fs_sops = { 2791 .alloc_inode = f2fs_alloc_inode, 2792 .free_inode = f2fs_free_inode, 2793 .drop_inode = f2fs_drop_inode, 2794 .write_inode = f2fs_write_inode, 2795 .dirty_inode = f2fs_dirty_inode, 2796 .show_options = f2fs_show_options, 2797 #ifdef CONFIG_QUOTA 2798 .quota_read = f2fs_quota_read, 2799 .quota_write = f2fs_quota_write, 2800 .get_dquots = f2fs_get_dquots, 2801 #endif 2802 .evict_inode = f2fs_evict_inode, 2803 .put_super = f2fs_put_super, 2804 .sync_fs = f2fs_sync_fs, 2805 .freeze_fs = f2fs_freeze, 2806 .unfreeze_fs = f2fs_unfreeze, 2807 .statfs = f2fs_statfs, 2808 .remount_fs = f2fs_remount, 2809 }; 2810 2811 #ifdef CONFIG_FS_ENCRYPTION 2812 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2813 { 2814 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2815 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2816 ctx, len, NULL); 2817 } 2818 2819 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2820 void *fs_data) 2821 { 2822 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2823 2824 /* 2825 * Encrypting the root directory is not allowed because fsck 2826 * expects lost+found directory to exist and remain unencrypted 2827 * if LOST_FOUND feature is enabled. 2828 * 2829 */ 2830 if (f2fs_sb_has_lost_found(sbi) && 2831 inode->i_ino == F2FS_ROOT_INO(sbi)) 2832 return -EPERM; 2833 2834 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2835 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2836 ctx, len, fs_data, XATTR_CREATE); 2837 } 2838 2839 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 2840 { 2841 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 2842 } 2843 2844 static bool f2fs_has_stable_inodes(struct super_block *sb) 2845 { 2846 return true; 2847 } 2848 2849 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 2850 int *ino_bits_ret, int *lblk_bits_ret) 2851 { 2852 *ino_bits_ret = 8 * sizeof(nid_t); 2853 *lblk_bits_ret = 8 * sizeof(block_t); 2854 } 2855 2856 static int f2fs_get_num_devices(struct super_block *sb) 2857 { 2858 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2859 2860 if (f2fs_is_multi_device(sbi)) 2861 return sbi->s_ndevs; 2862 return 1; 2863 } 2864 2865 static void f2fs_get_devices(struct super_block *sb, 2866 struct request_queue **devs) 2867 { 2868 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2869 int i; 2870 2871 for (i = 0; i < sbi->s_ndevs; i++) 2872 devs[i] = bdev_get_queue(FDEV(i).bdev); 2873 } 2874 2875 static const struct fscrypt_operations f2fs_cryptops = { 2876 .key_prefix = "f2fs:", 2877 .get_context = f2fs_get_context, 2878 .set_context = f2fs_set_context, 2879 .get_dummy_policy = f2fs_get_dummy_policy, 2880 .empty_dir = f2fs_empty_dir, 2881 .max_namelen = F2FS_NAME_LEN, 2882 .has_stable_inodes = f2fs_has_stable_inodes, 2883 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 2884 .get_num_devices = f2fs_get_num_devices, 2885 .get_devices = f2fs_get_devices, 2886 }; 2887 #endif 2888 2889 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 2890 u64 ino, u32 generation) 2891 { 2892 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2893 struct inode *inode; 2894 2895 if (f2fs_check_nid_range(sbi, ino)) 2896 return ERR_PTR(-ESTALE); 2897 2898 /* 2899 * f2fs_iget isn't quite right if the inode is currently unallocated! 2900 * However f2fs_iget currently does appropriate checks to handle stale 2901 * inodes so everything is OK. 2902 */ 2903 inode = f2fs_iget(sb, ino); 2904 if (IS_ERR(inode)) 2905 return ERR_CAST(inode); 2906 if (unlikely(generation && inode->i_generation != generation)) { 2907 /* we didn't find the right inode.. */ 2908 iput(inode); 2909 return ERR_PTR(-ESTALE); 2910 } 2911 return inode; 2912 } 2913 2914 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 2915 int fh_len, int fh_type) 2916 { 2917 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 2918 f2fs_nfs_get_inode); 2919 } 2920 2921 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 2922 int fh_len, int fh_type) 2923 { 2924 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 2925 f2fs_nfs_get_inode); 2926 } 2927 2928 static const struct export_operations f2fs_export_ops = { 2929 .fh_to_dentry = f2fs_fh_to_dentry, 2930 .fh_to_parent = f2fs_fh_to_parent, 2931 .get_parent = f2fs_get_parent, 2932 }; 2933 2934 loff_t max_file_blocks(struct inode *inode) 2935 { 2936 loff_t result = 0; 2937 loff_t leaf_count; 2938 2939 /* 2940 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 2941 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 2942 * space in inode.i_addr, it will be more safe to reassign 2943 * result as zero. 2944 */ 2945 2946 if (inode && f2fs_compressed_file(inode)) 2947 leaf_count = ADDRS_PER_BLOCK(inode); 2948 else 2949 leaf_count = DEF_ADDRS_PER_BLOCK; 2950 2951 /* two direct node blocks */ 2952 result += (leaf_count * 2); 2953 2954 /* two indirect node blocks */ 2955 leaf_count *= NIDS_PER_BLOCK; 2956 result += (leaf_count * 2); 2957 2958 /* one double indirect node block */ 2959 leaf_count *= NIDS_PER_BLOCK; 2960 result += leaf_count; 2961 2962 return result; 2963 } 2964 2965 static int __f2fs_commit_super(struct buffer_head *bh, 2966 struct f2fs_super_block *super) 2967 { 2968 lock_buffer(bh); 2969 if (super) 2970 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 2971 set_buffer_dirty(bh); 2972 unlock_buffer(bh); 2973 2974 /* it's rare case, we can do fua all the time */ 2975 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 2976 } 2977 2978 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 2979 struct buffer_head *bh) 2980 { 2981 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2982 (bh->b_data + F2FS_SUPER_OFFSET); 2983 struct super_block *sb = sbi->sb; 2984 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 2985 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 2986 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 2987 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 2988 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 2989 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 2990 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 2991 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 2992 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 2993 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 2994 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2995 u32 segment_count = le32_to_cpu(raw_super->segment_count); 2996 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2997 u64 main_end_blkaddr = main_blkaddr + 2998 (segment_count_main << log_blocks_per_seg); 2999 u64 seg_end_blkaddr = segment0_blkaddr + 3000 (segment_count << log_blocks_per_seg); 3001 3002 if (segment0_blkaddr != cp_blkaddr) { 3003 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3004 segment0_blkaddr, cp_blkaddr); 3005 return true; 3006 } 3007 3008 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3009 sit_blkaddr) { 3010 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3011 cp_blkaddr, sit_blkaddr, 3012 segment_count_ckpt << log_blocks_per_seg); 3013 return true; 3014 } 3015 3016 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3017 nat_blkaddr) { 3018 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3019 sit_blkaddr, nat_blkaddr, 3020 segment_count_sit << log_blocks_per_seg); 3021 return true; 3022 } 3023 3024 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3025 ssa_blkaddr) { 3026 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3027 nat_blkaddr, ssa_blkaddr, 3028 segment_count_nat << log_blocks_per_seg); 3029 return true; 3030 } 3031 3032 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3033 main_blkaddr) { 3034 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3035 ssa_blkaddr, main_blkaddr, 3036 segment_count_ssa << log_blocks_per_seg); 3037 return true; 3038 } 3039 3040 if (main_end_blkaddr > seg_end_blkaddr) { 3041 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3042 main_blkaddr, seg_end_blkaddr, 3043 segment_count_main << log_blocks_per_seg); 3044 return true; 3045 } else if (main_end_blkaddr < seg_end_blkaddr) { 3046 int err = 0; 3047 char *res; 3048 3049 /* fix in-memory information all the time */ 3050 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3051 segment0_blkaddr) >> log_blocks_per_seg); 3052 3053 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 3054 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3055 res = "internally"; 3056 } else { 3057 err = __f2fs_commit_super(bh, NULL); 3058 res = err ? "failed" : "done"; 3059 } 3060 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3061 res, main_blkaddr, seg_end_blkaddr, 3062 segment_count_main << log_blocks_per_seg); 3063 if (err) 3064 return true; 3065 } 3066 return false; 3067 } 3068 3069 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3070 struct buffer_head *bh) 3071 { 3072 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3073 block_t total_sections, blocks_per_seg; 3074 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3075 (bh->b_data + F2FS_SUPER_OFFSET); 3076 size_t crc_offset = 0; 3077 __u32 crc = 0; 3078 3079 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3080 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3081 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3082 return -EINVAL; 3083 } 3084 3085 /* Check checksum_offset and crc in superblock */ 3086 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3087 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3088 if (crc_offset != 3089 offsetof(struct f2fs_super_block, crc)) { 3090 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3091 crc_offset); 3092 return -EFSCORRUPTED; 3093 } 3094 crc = le32_to_cpu(raw_super->crc); 3095 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3096 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3097 return -EFSCORRUPTED; 3098 } 3099 } 3100 3101 /* Currently, support only 4KB block size */ 3102 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3103 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3104 le32_to_cpu(raw_super->log_blocksize), 3105 F2FS_BLKSIZE_BITS); 3106 return -EFSCORRUPTED; 3107 } 3108 3109 /* check log blocks per segment */ 3110 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3111 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3112 le32_to_cpu(raw_super->log_blocks_per_seg)); 3113 return -EFSCORRUPTED; 3114 } 3115 3116 /* Currently, support 512/1024/2048/4096 bytes sector size */ 3117 if (le32_to_cpu(raw_super->log_sectorsize) > 3118 F2FS_MAX_LOG_SECTOR_SIZE || 3119 le32_to_cpu(raw_super->log_sectorsize) < 3120 F2FS_MIN_LOG_SECTOR_SIZE) { 3121 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3122 le32_to_cpu(raw_super->log_sectorsize)); 3123 return -EFSCORRUPTED; 3124 } 3125 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3126 le32_to_cpu(raw_super->log_sectorsize) != 3127 F2FS_MAX_LOG_SECTOR_SIZE) { 3128 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3129 le32_to_cpu(raw_super->log_sectors_per_block), 3130 le32_to_cpu(raw_super->log_sectorsize)); 3131 return -EFSCORRUPTED; 3132 } 3133 3134 segment_count = le32_to_cpu(raw_super->segment_count); 3135 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3136 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3137 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3138 total_sections = le32_to_cpu(raw_super->section_count); 3139 3140 /* blocks_per_seg should be 512, given the above check */ 3141 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 3142 3143 if (segment_count > F2FS_MAX_SEGMENT || 3144 segment_count < F2FS_MIN_SEGMENTS) { 3145 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3146 return -EFSCORRUPTED; 3147 } 3148 3149 if (total_sections > segment_count_main || total_sections < 1 || 3150 segs_per_sec > segment_count || !segs_per_sec) { 3151 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3152 segment_count, total_sections, segs_per_sec); 3153 return -EFSCORRUPTED; 3154 } 3155 3156 if (segment_count_main != total_sections * segs_per_sec) { 3157 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3158 segment_count_main, total_sections, segs_per_sec); 3159 return -EFSCORRUPTED; 3160 } 3161 3162 if ((segment_count / segs_per_sec) < total_sections) { 3163 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3164 segment_count, segs_per_sec, total_sections); 3165 return -EFSCORRUPTED; 3166 } 3167 3168 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3169 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3170 segment_count, le64_to_cpu(raw_super->block_count)); 3171 return -EFSCORRUPTED; 3172 } 3173 3174 if (RDEV(0).path[0]) { 3175 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3176 int i = 1; 3177 3178 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3179 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3180 i++; 3181 } 3182 if (segment_count != dev_seg_count) { 3183 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3184 segment_count, dev_seg_count); 3185 return -EFSCORRUPTED; 3186 } 3187 } else { 3188 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3189 !bdev_is_zoned(sbi->sb->s_bdev)) { 3190 f2fs_info(sbi, "Zoned block device path is missing"); 3191 return -EFSCORRUPTED; 3192 } 3193 } 3194 3195 if (secs_per_zone > total_sections || !secs_per_zone) { 3196 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3197 secs_per_zone, total_sections); 3198 return -EFSCORRUPTED; 3199 } 3200 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3201 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3202 (le32_to_cpu(raw_super->extension_count) + 3203 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3204 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3205 le32_to_cpu(raw_super->extension_count), 3206 raw_super->hot_ext_count, 3207 F2FS_MAX_EXTENSION); 3208 return -EFSCORRUPTED; 3209 } 3210 3211 if (le32_to_cpu(raw_super->cp_payload) > 3212 (blocks_per_seg - F2FS_CP_PACKS)) { 3213 f2fs_info(sbi, "Insane cp_payload (%u > %u)", 3214 le32_to_cpu(raw_super->cp_payload), 3215 blocks_per_seg - F2FS_CP_PACKS); 3216 return -EFSCORRUPTED; 3217 } 3218 3219 /* check reserved ino info */ 3220 if (le32_to_cpu(raw_super->node_ino) != 1 || 3221 le32_to_cpu(raw_super->meta_ino) != 2 || 3222 le32_to_cpu(raw_super->root_ino) != 3) { 3223 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3224 le32_to_cpu(raw_super->node_ino), 3225 le32_to_cpu(raw_super->meta_ino), 3226 le32_to_cpu(raw_super->root_ino)); 3227 return -EFSCORRUPTED; 3228 } 3229 3230 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3231 if (sanity_check_area_boundary(sbi, bh)) 3232 return -EFSCORRUPTED; 3233 3234 return 0; 3235 } 3236 3237 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3238 { 3239 unsigned int total, fsmeta; 3240 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3241 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3242 unsigned int ovp_segments, reserved_segments; 3243 unsigned int main_segs, blocks_per_seg; 3244 unsigned int sit_segs, nat_segs; 3245 unsigned int sit_bitmap_size, nat_bitmap_size; 3246 unsigned int log_blocks_per_seg; 3247 unsigned int segment_count_main; 3248 unsigned int cp_pack_start_sum, cp_payload; 3249 block_t user_block_count, valid_user_blocks; 3250 block_t avail_node_count, valid_node_count; 3251 int i, j; 3252 3253 total = le32_to_cpu(raw_super->segment_count); 3254 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3255 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3256 fsmeta += sit_segs; 3257 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3258 fsmeta += nat_segs; 3259 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3260 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3261 3262 if (unlikely(fsmeta >= total)) 3263 return 1; 3264 3265 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3266 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3267 3268 if (!f2fs_sb_has_readonly(sbi) && 3269 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3270 ovp_segments == 0 || reserved_segments == 0)) { 3271 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3272 return 1; 3273 } 3274 user_block_count = le64_to_cpu(ckpt->user_block_count); 3275 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3276 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3277 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3278 if (!user_block_count || user_block_count >= 3279 segment_count_main << log_blocks_per_seg) { 3280 f2fs_err(sbi, "Wrong user_block_count: %u", 3281 user_block_count); 3282 return 1; 3283 } 3284 3285 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3286 if (valid_user_blocks > user_block_count) { 3287 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3288 valid_user_blocks, user_block_count); 3289 return 1; 3290 } 3291 3292 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3293 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3294 if (valid_node_count > avail_node_count) { 3295 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3296 valid_node_count, avail_node_count); 3297 return 1; 3298 } 3299 3300 main_segs = le32_to_cpu(raw_super->segment_count_main); 3301 blocks_per_seg = sbi->blocks_per_seg; 3302 3303 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3304 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3305 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3306 return 1; 3307 3308 if (f2fs_sb_has_readonly(sbi)) 3309 goto check_data; 3310 3311 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3312 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3313 le32_to_cpu(ckpt->cur_node_segno[j])) { 3314 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3315 i, j, 3316 le32_to_cpu(ckpt->cur_node_segno[i])); 3317 return 1; 3318 } 3319 } 3320 } 3321 check_data: 3322 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3323 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3324 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3325 return 1; 3326 3327 if (f2fs_sb_has_readonly(sbi)) 3328 goto skip_cross; 3329 3330 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3331 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3332 le32_to_cpu(ckpt->cur_data_segno[j])) { 3333 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3334 i, j, 3335 le32_to_cpu(ckpt->cur_data_segno[i])); 3336 return 1; 3337 } 3338 } 3339 } 3340 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3341 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3342 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3343 le32_to_cpu(ckpt->cur_data_segno[j])) { 3344 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3345 i, j, 3346 le32_to_cpu(ckpt->cur_node_segno[i])); 3347 return 1; 3348 } 3349 } 3350 } 3351 skip_cross: 3352 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3353 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3354 3355 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3356 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3357 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3358 sit_bitmap_size, nat_bitmap_size); 3359 return 1; 3360 } 3361 3362 cp_pack_start_sum = __start_sum_addr(sbi); 3363 cp_payload = __cp_payload(sbi); 3364 if (cp_pack_start_sum < cp_payload + 1 || 3365 cp_pack_start_sum > blocks_per_seg - 1 - 3366 NR_CURSEG_PERSIST_TYPE) { 3367 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3368 cp_pack_start_sum); 3369 return 1; 3370 } 3371 3372 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3373 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3374 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3375 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3376 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3377 le32_to_cpu(ckpt->checksum_offset)); 3378 return 1; 3379 } 3380 3381 if (unlikely(f2fs_cp_error(sbi))) { 3382 f2fs_err(sbi, "A bug case: need to run fsck"); 3383 return 1; 3384 } 3385 return 0; 3386 } 3387 3388 static void init_sb_info(struct f2fs_sb_info *sbi) 3389 { 3390 struct f2fs_super_block *raw_super = sbi->raw_super; 3391 int i; 3392 3393 sbi->log_sectors_per_block = 3394 le32_to_cpu(raw_super->log_sectors_per_block); 3395 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3396 sbi->blocksize = 1 << sbi->log_blocksize; 3397 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3398 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3399 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3400 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3401 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3402 sbi->total_node_count = 3403 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3404 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3405 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3406 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3407 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3408 sbi->cur_victim_sec = NULL_SECNO; 3409 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3410 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3411 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3412 sbi->migration_granularity = sbi->segs_per_sec; 3413 3414 sbi->dir_level = DEF_DIR_LEVEL; 3415 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3416 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3417 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3418 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3419 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3420 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3421 DEF_UMOUNT_DISCARD_TIMEOUT; 3422 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3423 3424 for (i = 0; i < NR_COUNT_TYPE; i++) 3425 atomic_set(&sbi->nr_pages[i], 0); 3426 3427 for (i = 0; i < META; i++) 3428 atomic_set(&sbi->wb_sync_req[i], 0); 3429 3430 INIT_LIST_HEAD(&sbi->s_list); 3431 mutex_init(&sbi->umount_mutex); 3432 init_rwsem(&sbi->io_order_lock); 3433 spin_lock_init(&sbi->cp_lock); 3434 3435 sbi->dirty_device = 0; 3436 spin_lock_init(&sbi->dev_lock); 3437 3438 init_rwsem(&sbi->sb_lock); 3439 init_rwsem(&sbi->pin_sem); 3440 } 3441 3442 static int init_percpu_info(struct f2fs_sb_info *sbi) 3443 { 3444 int err; 3445 3446 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3447 if (err) 3448 return err; 3449 3450 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3451 GFP_KERNEL); 3452 if (err) 3453 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3454 3455 return err; 3456 } 3457 3458 #ifdef CONFIG_BLK_DEV_ZONED 3459 3460 struct f2fs_report_zones_args { 3461 struct f2fs_dev_info *dev; 3462 bool zone_cap_mismatch; 3463 }; 3464 3465 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3466 void *data) 3467 { 3468 struct f2fs_report_zones_args *rz_args = data; 3469 3470 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3471 return 0; 3472 3473 set_bit(idx, rz_args->dev->blkz_seq); 3474 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >> 3475 F2FS_LOG_SECTORS_PER_BLOCK; 3476 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch) 3477 rz_args->zone_cap_mismatch = true; 3478 3479 return 0; 3480 } 3481 3482 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3483 { 3484 struct block_device *bdev = FDEV(devi).bdev; 3485 sector_t nr_sectors = bdev_nr_sectors(bdev); 3486 struct f2fs_report_zones_args rep_zone_arg; 3487 int ret; 3488 3489 if (!f2fs_sb_has_blkzoned(sbi)) 3490 return 0; 3491 3492 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3493 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 3494 return -EINVAL; 3495 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 3496 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3497 __ilog2_u32(sbi->blocks_per_blkz)) 3498 return -EINVAL; 3499 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3500 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3501 sbi->log_blocks_per_blkz; 3502 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 3503 FDEV(devi).nr_blkz++; 3504 3505 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3506 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3507 * sizeof(unsigned long), 3508 GFP_KERNEL); 3509 if (!FDEV(devi).blkz_seq) 3510 return -ENOMEM; 3511 3512 /* Get block zones type and zone-capacity */ 3513 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi, 3514 FDEV(devi).nr_blkz * sizeof(block_t), 3515 GFP_KERNEL); 3516 if (!FDEV(devi).zone_capacity_blocks) 3517 return -ENOMEM; 3518 3519 rep_zone_arg.dev = &FDEV(devi); 3520 rep_zone_arg.zone_cap_mismatch = false; 3521 3522 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3523 &rep_zone_arg); 3524 if (ret < 0) 3525 return ret; 3526 3527 if (!rep_zone_arg.zone_cap_mismatch) { 3528 kfree(FDEV(devi).zone_capacity_blocks); 3529 FDEV(devi).zone_capacity_blocks = NULL; 3530 } 3531 3532 return 0; 3533 } 3534 #endif 3535 3536 /* 3537 * Read f2fs raw super block. 3538 * Because we have two copies of super block, so read both of them 3539 * to get the first valid one. If any one of them is broken, we pass 3540 * them recovery flag back to the caller. 3541 */ 3542 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3543 struct f2fs_super_block **raw_super, 3544 int *valid_super_block, int *recovery) 3545 { 3546 struct super_block *sb = sbi->sb; 3547 int block; 3548 struct buffer_head *bh; 3549 struct f2fs_super_block *super; 3550 int err = 0; 3551 3552 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3553 if (!super) 3554 return -ENOMEM; 3555 3556 for (block = 0; block < 2; block++) { 3557 bh = sb_bread(sb, block); 3558 if (!bh) { 3559 f2fs_err(sbi, "Unable to read %dth superblock", 3560 block + 1); 3561 err = -EIO; 3562 *recovery = 1; 3563 continue; 3564 } 3565 3566 /* sanity checking of raw super */ 3567 err = sanity_check_raw_super(sbi, bh); 3568 if (err) { 3569 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3570 block + 1); 3571 brelse(bh); 3572 *recovery = 1; 3573 continue; 3574 } 3575 3576 if (!*raw_super) { 3577 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3578 sizeof(*super)); 3579 *valid_super_block = block; 3580 *raw_super = super; 3581 } 3582 brelse(bh); 3583 } 3584 3585 /* No valid superblock */ 3586 if (!*raw_super) 3587 kfree(super); 3588 else 3589 err = 0; 3590 3591 return err; 3592 } 3593 3594 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3595 { 3596 struct buffer_head *bh; 3597 __u32 crc = 0; 3598 int err; 3599 3600 if ((recover && f2fs_readonly(sbi->sb)) || 3601 bdev_read_only(sbi->sb->s_bdev)) { 3602 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3603 return -EROFS; 3604 } 3605 3606 /* we should update superblock crc here */ 3607 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3608 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3609 offsetof(struct f2fs_super_block, crc)); 3610 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3611 } 3612 3613 /* write back-up superblock first */ 3614 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3615 if (!bh) 3616 return -EIO; 3617 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3618 brelse(bh); 3619 3620 /* if we are in recovery path, skip writing valid superblock */ 3621 if (recover || err) 3622 return err; 3623 3624 /* write current valid superblock */ 3625 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3626 if (!bh) 3627 return -EIO; 3628 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3629 brelse(bh); 3630 return err; 3631 } 3632 3633 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3634 { 3635 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3636 unsigned int max_devices = MAX_DEVICES; 3637 int i; 3638 3639 /* Initialize single device information */ 3640 if (!RDEV(0).path[0]) { 3641 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3642 return 0; 3643 max_devices = 1; 3644 } 3645 3646 /* 3647 * Initialize multiple devices information, or single 3648 * zoned block device information. 3649 */ 3650 sbi->devs = f2fs_kzalloc(sbi, 3651 array_size(max_devices, 3652 sizeof(struct f2fs_dev_info)), 3653 GFP_KERNEL); 3654 if (!sbi->devs) 3655 return -ENOMEM; 3656 3657 for (i = 0; i < max_devices; i++) { 3658 3659 if (i > 0 && !RDEV(i).path[0]) 3660 break; 3661 3662 if (max_devices == 1) { 3663 /* Single zoned block device mount */ 3664 FDEV(0).bdev = 3665 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3666 sbi->sb->s_mode, sbi->sb->s_type); 3667 } else { 3668 /* Multi-device mount */ 3669 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3670 FDEV(i).total_segments = 3671 le32_to_cpu(RDEV(i).total_segments); 3672 if (i == 0) { 3673 FDEV(i).start_blk = 0; 3674 FDEV(i).end_blk = FDEV(i).start_blk + 3675 (FDEV(i).total_segments << 3676 sbi->log_blocks_per_seg) - 1 + 3677 le32_to_cpu(raw_super->segment0_blkaddr); 3678 } else { 3679 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3680 FDEV(i).end_blk = FDEV(i).start_blk + 3681 (FDEV(i).total_segments << 3682 sbi->log_blocks_per_seg) - 1; 3683 } 3684 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3685 sbi->sb->s_mode, sbi->sb->s_type); 3686 } 3687 if (IS_ERR(FDEV(i).bdev)) 3688 return PTR_ERR(FDEV(i).bdev); 3689 3690 /* to release errored devices */ 3691 sbi->s_ndevs = i + 1; 3692 3693 #ifdef CONFIG_BLK_DEV_ZONED 3694 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3695 !f2fs_sb_has_blkzoned(sbi)) { 3696 f2fs_err(sbi, "Zoned block device feature not enabled"); 3697 return -EINVAL; 3698 } 3699 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3700 if (init_blkz_info(sbi, i)) { 3701 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3702 return -EINVAL; 3703 } 3704 if (max_devices == 1) 3705 break; 3706 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3707 i, FDEV(i).path, 3708 FDEV(i).total_segments, 3709 FDEV(i).start_blk, FDEV(i).end_blk, 3710 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3711 "Host-aware" : "Host-managed"); 3712 continue; 3713 } 3714 #endif 3715 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3716 i, FDEV(i).path, 3717 FDEV(i).total_segments, 3718 FDEV(i).start_blk, FDEV(i).end_blk); 3719 } 3720 f2fs_info(sbi, 3721 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3722 return 0; 3723 } 3724 3725 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3726 { 3727 #ifdef CONFIG_UNICODE 3728 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 3729 const struct f2fs_sb_encodings *encoding_info; 3730 struct unicode_map *encoding; 3731 __u16 encoding_flags; 3732 3733 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info, 3734 &encoding_flags)) { 3735 f2fs_err(sbi, 3736 "Encoding requested by superblock is unknown"); 3737 return -EINVAL; 3738 } 3739 3740 encoding = utf8_load(encoding_info->version); 3741 if (IS_ERR(encoding)) { 3742 f2fs_err(sbi, 3743 "can't mount with superblock charset: %s-%s " 3744 "not supported by the kernel. flags: 0x%x.", 3745 encoding_info->name, encoding_info->version, 3746 encoding_flags); 3747 return PTR_ERR(encoding); 3748 } 3749 f2fs_info(sbi, "Using encoding defined by superblock: " 3750 "%s-%s with flags 0x%hx", encoding_info->name, 3751 encoding_info->version?:"\b", encoding_flags); 3752 3753 sbi->sb->s_encoding = encoding; 3754 sbi->sb->s_encoding_flags = encoding_flags; 3755 } 3756 #else 3757 if (f2fs_sb_has_casefold(sbi)) { 3758 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3759 return -EINVAL; 3760 } 3761 #endif 3762 return 0; 3763 } 3764 3765 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3766 { 3767 struct f2fs_sm_info *sm_i = SM_I(sbi); 3768 3769 /* adjust parameters according to the volume size */ 3770 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3771 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3772 sm_i->dcc_info->discard_granularity = 1; 3773 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 3774 } 3775 3776 sbi->readdir_ra = 1; 3777 } 3778 3779 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3780 { 3781 struct f2fs_sb_info *sbi; 3782 struct f2fs_super_block *raw_super; 3783 struct inode *root; 3784 int err; 3785 bool skip_recovery = false, need_fsck = false; 3786 char *options = NULL; 3787 int recovery, i, valid_super_block; 3788 struct curseg_info *seg_i; 3789 int retry_cnt = 1; 3790 3791 try_onemore: 3792 err = -EINVAL; 3793 raw_super = NULL; 3794 valid_super_block = -1; 3795 recovery = 0; 3796 3797 /* allocate memory for f2fs-specific super block info */ 3798 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 3799 if (!sbi) 3800 return -ENOMEM; 3801 3802 sbi->sb = sb; 3803 3804 /* Load the checksum driver */ 3805 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 3806 if (IS_ERR(sbi->s_chksum_driver)) { 3807 f2fs_err(sbi, "Cannot load crc32 driver."); 3808 err = PTR_ERR(sbi->s_chksum_driver); 3809 sbi->s_chksum_driver = NULL; 3810 goto free_sbi; 3811 } 3812 3813 /* set a block size */ 3814 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 3815 f2fs_err(sbi, "unable to set blocksize"); 3816 goto free_sbi; 3817 } 3818 3819 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 3820 &recovery); 3821 if (err) 3822 goto free_sbi; 3823 3824 sb->s_fs_info = sbi; 3825 sbi->raw_super = raw_super; 3826 3827 /* precompute checksum seed for metadata */ 3828 if (f2fs_sb_has_inode_chksum(sbi)) 3829 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 3830 sizeof(raw_super->uuid)); 3831 3832 default_options(sbi); 3833 /* parse mount options */ 3834 options = kstrdup((const char *)data, GFP_KERNEL); 3835 if (data && !options) { 3836 err = -ENOMEM; 3837 goto free_sb_buf; 3838 } 3839 3840 err = parse_options(sb, options, false); 3841 if (err) 3842 goto free_options; 3843 3844 sb->s_maxbytes = max_file_blocks(NULL) << 3845 le32_to_cpu(raw_super->log_blocksize); 3846 sb->s_max_links = F2FS_LINK_MAX; 3847 3848 err = f2fs_setup_casefold(sbi); 3849 if (err) 3850 goto free_options; 3851 3852 #ifdef CONFIG_QUOTA 3853 sb->dq_op = &f2fs_quota_operations; 3854 sb->s_qcop = &f2fs_quotactl_ops; 3855 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 3856 3857 if (f2fs_sb_has_quota_ino(sbi)) { 3858 for (i = 0; i < MAXQUOTAS; i++) { 3859 if (f2fs_qf_ino(sbi->sb, i)) 3860 sbi->nquota_files++; 3861 } 3862 } 3863 #endif 3864 3865 sb->s_op = &f2fs_sops; 3866 #ifdef CONFIG_FS_ENCRYPTION 3867 sb->s_cop = &f2fs_cryptops; 3868 #endif 3869 #ifdef CONFIG_FS_VERITY 3870 sb->s_vop = &f2fs_verityops; 3871 #endif 3872 sb->s_xattr = f2fs_xattr_handlers; 3873 sb->s_export_op = &f2fs_export_ops; 3874 sb->s_magic = F2FS_SUPER_MAGIC; 3875 sb->s_time_gran = 1; 3876 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3877 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 3878 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 3879 sb->s_iflags |= SB_I_CGROUPWB; 3880 3881 /* init f2fs-specific super block info */ 3882 sbi->valid_super_block = valid_super_block; 3883 init_rwsem(&sbi->gc_lock); 3884 mutex_init(&sbi->writepages); 3885 init_rwsem(&sbi->cp_global_sem); 3886 init_rwsem(&sbi->node_write); 3887 init_rwsem(&sbi->node_change); 3888 3889 /* disallow all the data/node/meta page writes */ 3890 set_sbi_flag(sbi, SBI_POR_DOING); 3891 spin_lock_init(&sbi->stat_lock); 3892 3893 /* init iostat info */ 3894 spin_lock_init(&sbi->iostat_lock); 3895 sbi->iostat_enable = false; 3896 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS; 3897 3898 for (i = 0; i < NR_PAGE_TYPE; i++) { 3899 int n = (i == META) ? 1 : NR_TEMP_TYPE; 3900 int j; 3901 3902 sbi->write_io[i] = 3903 f2fs_kmalloc(sbi, 3904 array_size(n, 3905 sizeof(struct f2fs_bio_info)), 3906 GFP_KERNEL); 3907 if (!sbi->write_io[i]) { 3908 err = -ENOMEM; 3909 goto free_bio_info; 3910 } 3911 3912 for (j = HOT; j < n; j++) { 3913 init_rwsem(&sbi->write_io[i][j].io_rwsem); 3914 sbi->write_io[i][j].sbi = sbi; 3915 sbi->write_io[i][j].bio = NULL; 3916 spin_lock_init(&sbi->write_io[i][j].io_lock); 3917 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 3918 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); 3919 init_rwsem(&sbi->write_io[i][j].bio_list_lock); 3920 } 3921 } 3922 3923 init_rwsem(&sbi->cp_rwsem); 3924 init_rwsem(&sbi->quota_sem); 3925 init_waitqueue_head(&sbi->cp_wait); 3926 init_sb_info(sbi); 3927 3928 err = init_percpu_info(sbi); 3929 if (err) 3930 goto free_bio_info; 3931 3932 if (F2FS_IO_ALIGNED(sbi)) { 3933 sbi->write_io_dummy = 3934 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 3935 if (!sbi->write_io_dummy) { 3936 err = -ENOMEM; 3937 goto free_percpu; 3938 } 3939 } 3940 3941 /* init per sbi slab cache */ 3942 err = f2fs_init_xattr_caches(sbi); 3943 if (err) 3944 goto free_io_dummy; 3945 err = f2fs_init_page_array_cache(sbi); 3946 if (err) 3947 goto free_xattr_cache; 3948 3949 /* get an inode for meta space */ 3950 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 3951 if (IS_ERR(sbi->meta_inode)) { 3952 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 3953 err = PTR_ERR(sbi->meta_inode); 3954 goto free_page_array_cache; 3955 } 3956 3957 err = f2fs_get_valid_checkpoint(sbi); 3958 if (err) { 3959 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 3960 goto free_meta_inode; 3961 } 3962 3963 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 3964 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3965 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 3966 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3967 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 3968 } 3969 3970 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 3971 set_sbi_flag(sbi, SBI_NEED_FSCK); 3972 3973 /* Initialize device list */ 3974 err = f2fs_scan_devices(sbi); 3975 if (err) { 3976 f2fs_err(sbi, "Failed to find devices"); 3977 goto free_devices; 3978 } 3979 3980 err = f2fs_init_post_read_wq(sbi); 3981 if (err) { 3982 f2fs_err(sbi, "Failed to initialize post read workqueue"); 3983 goto free_devices; 3984 } 3985 3986 sbi->total_valid_node_count = 3987 le32_to_cpu(sbi->ckpt->valid_node_count); 3988 percpu_counter_set(&sbi->total_valid_inode_count, 3989 le32_to_cpu(sbi->ckpt->valid_inode_count)); 3990 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 3991 sbi->total_valid_block_count = 3992 le64_to_cpu(sbi->ckpt->valid_block_count); 3993 sbi->last_valid_block_count = sbi->total_valid_block_count; 3994 sbi->reserved_blocks = 0; 3995 sbi->current_reserved_blocks = 0; 3996 limit_reserve_root(sbi); 3997 adjust_unusable_cap_perc(sbi); 3998 3999 for (i = 0; i < NR_INODE_TYPE; i++) { 4000 INIT_LIST_HEAD(&sbi->inode_list[i]); 4001 spin_lock_init(&sbi->inode_lock[i]); 4002 } 4003 mutex_init(&sbi->flush_lock); 4004 4005 f2fs_init_extent_cache_info(sbi); 4006 4007 f2fs_init_ino_entry_info(sbi); 4008 4009 f2fs_init_fsync_node_info(sbi); 4010 4011 /* setup checkpoint request control and start checkpoint issue thread */ 4012 f2fs_init_ckpt_req_control(sbi); 4013 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4014 test_opt(sbi, MERGE_CHECKPOINT)) { 4015 err = f2fs_start_ckpt_thread(sbi); 4016 if (err) { 4017 f2fs_err(sbi, 4018 "Failed to start F2FS issue_checkpoint_thread (%d)", 4019 err); 4020 goto stop_ckpt_thread; 4021 } 4022 } 4023 4024 /* setup f2fs internal modules */ 4025 err = f2fs_build_segment_manager(sbi); 4026 if (err) { 4027 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4028 err); 4029 goto free_sm; 4030 } 4031 err = f2fs_build_node_manager(sbi); 4032 if (err) { 4033 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4034 err); 4035 goto free_nm; 4036 } 4037 4038 /* For write statistics */ 4039 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4040 4041 /* Read accumulated write IO statistics if exists */ 4042 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4043 if (__exist_node_summaries(sbi)) 4044 sbi->kbytes_written = 4045 le64_to_cpu(seg_i->journal->info.kbytes_written); 4046 4047 f2fs_build_gc_manager(sbi); 4048 4049 err = f2fs_build_stats(sbi); 4050 if (err) 4051 goto free_nm; 4052 4053 /* get an inode for node space */ 4054 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4055 if (IS_ERR(sbi->node_inode)) { 4056 f2fs_err(sbi, "Failed to read node inode"); 4057 err = PTR_ERR(sbi->node_inode); 4058 goto free_stats; 4059 } 4060 4061 /* read root inode and dentry */ 4062 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4063 if (IS_ERR(root)) { 4064 f2fs_err(sbi, "Failed to read root inode"); 4065 err = PTR_ERR(root); 4066 goto free_node_inode; 4067 } 4068 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4069 !root->i_size || !root->i_nlink) { 4070 iput(root); 4071 err = -EINVAL; 4072 goto free_node_inode; 4073 } 4074 4075 sb->s_root = d_make_root(root); /* allocate root dentry */ 4076 if (!sb->s_root) { 4077 err = -ENOMEM; 4078 goto free_node_inode; 4079 } 4080 4081 err = f2fs_init_compress_inode(sbi); 4082 if (err) 4083 goto free_root_inode; 4084 4085 err = f2fs_register_sysfs(sbi); 4086 if (err) 4087 goto free_compress_inode; 4088 4089 #ifdef CONFIG_QUOTA 4090 /* Enable quota usage during mount */ 4091 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4092 err = f2fs_enable_quotas(sb); 4093 if (err) 4094 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4095 } 4096 #endif 4097 /* if there are any orphan inodes, free them */ 4098 err = f2fs_recover_orphan_inodes(sbi); 4099 if (err) 4100 goto free_meta; 4101 4102 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4103 goto reset_checkpoint; 4104 4105 /* recover fsynced data */ 4106 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4107 !test_opt(sbi, NORECOVERY)) { 4108 /* 4109 * mount should be failed, when device has readonly mode, and 4110 * previous checkpoint was not done by clean system shutdown. 4111 */ 4112 if (f2fs_hw_is_readonly(sbi)) { 4113 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4114 err = f2fs_recover_fsync_data(sbi, true); 4115 if (err > 0) { 4116 err = -EROFS; 4117 f2fs_err(sbi, "Need to recover fsync data, but " 4118 "write access unavailable, please try " 4119 "mount w/ disable_roll_forward or norecovery"); 4120 } 4121 if (err < 0) 4122 goto free_meta; 4123 } 4124 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4125 goto reset_checkpoint; 4126 } 4127 4128 if (need_fsck) 4129 set_sbi_flag(sbi, SBI_NEED_FSCK); 4130 4131 if (skip_recovery) 4132 goto reset_checkpoint; 4133 4134 err = f2fs_recover_fsync_data(sbi, false); 4135 if (err < 0) { 4136 if (err != -ENOMEM) 4137 skip_recovery = true; 4138 need_fsck = true; 4139 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4140 err); 4141 goto free_meta; 4142 } 4143 } else { 4144 err = f2fs_recover_fsync_data(sbi, true); 4145 4146 if (!f2fs_readonly(sb) && err > 0) { 4147 err = -EINVAL; 4148 f2fs_err(sbi, "Need to recover fsync data"); 4149 goto free_meta; 4150 } 4151 } 4152 4153 /* 4154 * If the f2fs is not readonly and fsync data recovery succeeds, 4155 * check zoned block devices' write pointer consistency. 4156 */ 4157 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 4158 err = f2fs_check_write_pointer(sbi); 4159 if (err) 4160 goto free_meta; 4161 } 4162 4163 reset_checkpoint: 4164 f2fs_init_inmem_curseg(sbi); 4165 4166 /* f2fs_recover_fsync_data() cleared this already */ 4167 clear_sbi_flag(sbi, SBI_POR_DOING); 4168 4169 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4170 err = f2fs_disable_checkpoint(sbi); 4171 if (err) 4172 goto sync_free_meta; 4173 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4174 f2fs_enable_checkpoint(sbi); 4175 } 4176 4177 /* 4178 * If filesystem is not mounted as read-only then 4179 * do start the gc_thread. 4180 */ 4181 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4182 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4183 /* After POR, we can run background GC thread.*/ 4184 err = f2fs_start_gc_thread(sbi); 4185 if (err) 4186 goto sync_free_meta; 4187 } 4188 kvfree(options); 4189 4190 /* recover broken superblock */ 4191 if (recovery) { 4192 err = f2fs_commit_super(sbi, true); 4193 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4194 sbi->valid_super_block ? 1 : 2, err); 4195 } 4196 4197 f2fs_join_shrinker(sbi); 4198 4199 f2fs_tuning_parameters(sbi); 4200 4201 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4202 cur_cp_version(F2FS_CKPT(sbi))); 4203 f2fs_update_time(sbi, CP_TIME); 4204 f2fs_update_time(sbi, REQ_TIME); 4205 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4206 return 0; 4207 4208 sync_free_meta: 4209 /* safe to flush all the data */ 4210 sync_filesystem(sbi->sb); 4211 retry_cnt = 0; 4212 4213 free_meta: 4214 #ifdef CONFIG_QUOTA 4215 f2fs_truncate_quota_inode_pages(sb); 4216 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4217 f2fs_quota_off_umount(sbi->sb); 4218 #endif 4219 /* 4220 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4221 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4222 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4223 * falls into an infinite loop in f2fs_sync_meta_pages(). 4224 */ 4225 truncate_inode_pages_final(META_MAPPING(sbi)); 4226 /* evict some inodes being cached by GC */ 4227 evict_inodes(sb); 4228 f2fs_unregister_sysfs(sbi); 4229 free_compress_inode: 4230 f2fs_destroy_compress_inode(sbi); 4231 free_root_inode: 4232 dput(sb->s_root); 4233 sb->s_root = NULL; 4234 free_node_inode: 4235 f2fs_release_ino_entry(sbi, true); 4236 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4237 iput(sbi->node_inode); 4238 sbi->node_inode = NULL; 4239 free_stats: 4240 f2fs_destroy_stats(sbi); 4241 free_nm: 4242 f2fs_destroy_node_manager(sbi); 4243 free_sm: 4244 f2fs_destroy_segment_manager(sbi); 4245 f2fs_destroy_post_read_wq(sbi); 4246 stop_ckpt_thread: 4247 f2fs_stop_ckpt_thread(sbi); 4248 free_devices: 4249 destroy_device_list(sbi); 4250 kvfree(sbi->ckpt); 4251 free_meta_inode: 4252 make_bad_inode(sbi->meta_inode); 4253 iput(sbi->meta_inode); 4254 sbi->meta_inode = NULL; 4255 free_page_array_cache: 4256 f2fs_destroy_page_array_cache(sbi); 4257 free_xattr_cache: 4258 f2fs_destroy_xattr_caches(sbi); 4259 free_io_dummy: 4260 mempool_destroy(sbi->write_io_dummy); 4261 free_percpu: 4262 destroy_percpu_info(sbi); 4263 free_bio_info: 4264 for (i = 0; i < NR_PAGE_TYPE; i++) 4265 kvfree(sbi->write_io[i]); 4266 4267 #ifdef CONFIG_UNICODE 4268 utf8_unload(sb->s_encoding); 4269 sb->s_encoding = NULL; 4270 #endif 4271 free_options: 4272 #ifdef CONFIG_QUOTA 4273 for (i = 0; i < MAXQUOTAS; i++) 4274 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4275 #endif 4276 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4277 kvfree(options); 4278 free_sb_buf: 4279 kfree(raw_super); 4280 free_sbi: 4281 if (sbi->s_chksum_driver) 4282 crypto_free_shash(sbi->s_chksum_driver); 4283 kfree(sbi); 4284 4285 /* give only one another chance */ 4286 if (retry_cnt > 0 && skip_recovery) { 4287 retry_cnt--; 4288 shrink_dcache_sb(sb); 4289 goto try_onemore; 4290 } 4291 return err; 4292 } 4293 4294 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4295 const char *dev_name, void *data) 4296 { 4297 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4298 } 4299 4300 static void kill_f2fs_super(struct super_block *sb) 4301 { 4302 if (sb->s_root) { 4303 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4304 4305 set_sbi_flag(sbi, SBI_IS_CLOSE); 4306 f2fs_stop_gc_thread(sbi); 4307 f2fs_stop_discard_thread(sbi); 4308 4309 #ifdef CONFIG_F2FS_FS_COMPRESSION 4310 /* 4311 * latter evict_inode() can bypass checking and invalidating 4312 * compress inode cache. 4313 */ 4314 if (test_opt(sbi, COMPRESS_CACHE)) 4315 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4316 #endif 4317 4318 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4319 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4320 struct cp_control cpc = { 4321 .reason = CP_UMOUNT, 4322 }; 4323 f2fs_write_checkpoint(sbi, &cpc); 4324 } 4325 4326 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4327 sb->s_flags &= ~SB_RDONLY; 4328 } 4329 kill_block_super(sb); 4330 } 4331 4332 static struct file_system_type f2fs_fs_type = { 4333 .owner = THIS_MODULE, 4334 .name = "f2fs", 4335 .mount = f2fs_mount, 4336 .kill_sb = kill_f2fs_super, 4337 .fs_flags = FS_REQUIRES_DEV, 4338 }; 4339 MODULE_ALIAS_FS("f2fs"); 4340 4341 static int __init init_inodecache(void) 4342 { 4343 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4344 sizeof(struct f2fs_inode_info), 0, 4345 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4346 if (!f2fs_inode_cachep) 4347 return -ENOMEM; 4348 return 0; 4349 } 4350 4351 static void destroy_inodecache(void) 4352 { 4353 /* 4354 * Make sure all delayed rcu free inodes are flushed before we 4355 * destroy cache. 4356 */ 4357 rcu_barrier(); 4358 kmem_cache_destroy(f2fs_inode_cachep); 4359 } 4360 4361 static int __init init_f2fs_fs(void) 4362 { 4363 int err; 4364 4365 if (PAGE_SIZE != F2FS_BLKSIZE) { 4366 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4367 PAGE_SIZE, F2FS_BLKSIZE); 4368 return -EINVAL; 4369 } 4370 4371 err = init_inodecache(); 4372 if (err) 4373 goto fail; 4374 err = f2fs_create_node_manager_caches(); 4375 if (err) 4376 goto free_inodecache; 4377 err = f2fs_create_segment_manager_caches(); 4378 if (err) 4379 goto free_node_manager_caches; 4380 err = f2fs_create_checkpoint_caches(); 4381 if (err) 4382 goto free_segment_manager_caches; 4383 err = f2fs_create_recovery_cache(); 4384 if (err) 4385 goto free_checkpoint_caches; 4386 err = f2fs_create_extent_cache(); 4387 if (err) 4388 goto free_recovery_cache; 4389 err = f2fs_create_garbage_collection_cache(); 4390 if (err) 4391 goto free_extent_cache; 4392 err = f2fs_init_sysfs(); 4393 if (err) 4394 goto free_garbage_collection_cache; 4395 err = register_shrinker(&f2fs_shrinker_info); 4396 if (err) 4397 goto free_sysfs; 4398 err = register_filesystem(&f2fs_fs_type); 4399 if (err) 4400 goto free_shrinker; 4401 f2fs_create_root_stats(); 4402 err = f2fs_init_post_read_processing(); 4403 if (err) 4404 goto free_root_stats; 4405 err = f2fs_init_bio_entry_cache(); 4406 if (err) 4407 goto free_post_read; 4408 err = f2fs_init_bioset(); 4409 if (err) 4410 goto free_bio_enrty_cache; 4411 err = f2fs_init_compress_mempool(); 4412 if (err) 4413 goto free_bioset; 4414 err = f2fs_init_compress_cache(); 4415 if (err) 4416 goto free_compress_mempool; 4417 err = f2fs_create_casefold_cache(); 4418 if (err) 4419 goto free_compress_cache; 4420 return 0; 4421 free_compress_cache: 4422 f2fs_destroy_compress_cache(); 4423 free_compress_mempool: 4424 f2fs_destroy_compress_mempool(); 4425 free_bioset: 4426 f2fs_destroy_bioset(); 4427 free_bio_enrty_cache: 4428 f2fs_destroy_bio_entry_cache(); 4429 free_post_read: 4430 f2fs_destroy_post_read_processing(); 4431 free_root_stats: 4432 f2fs_destroy_root_stats(); 4433 unregister_filesystem(&f2fs_fs_type); 4434 free_shrinker: 4435 unregister_shrinker(&f2fs_shrinker_info); 4436 free_sysfs: 4437 f2fs_exit_sysfs(); 4438 free_garbage_collection_cache: 4439 f2fs_destroy_garbage_collection_cache(); 4440 free_extent_cache: 4441 f2fs_destroy_extent_cache(); 4442 free_recovery_cache: 4443 f2fs_destroy_recovery_cache(); 4444 free_checkpoint_caches: 4445 f2fs_destroy_checkpoint_caches(); 4446 free_segment_manager_caches: 4447 f2fs_destroy_segment_manager_caches(); 4448 free_node_manager_caches: 4449 f2fs_destroy_node_manager_caches(); 4450 free_inodecache: 4451 destroy_inodecache(); 4452 fail: 4453 return err; 4454 } 4455 4456 static void __exit exit_f2fs_fs(void) 4457 { 4458 f2fs_destroy_casefold_cache(); 4459 f2fs_destroy_compress_cache(); 4460 f2fs_destroy_compress_mempool(); 4461 f2fs_destroy_bioset(); 4462 f2fs_destroy_bio_entry_cache(); 4463 f2fs_destroy_post_read_processing(); 4464 f2fs_destroy_root_stats(); 4465 unregister_filesystem(&f2fs_fs_type); 4466 unregister_shrinker(&f2fs_shrinker_info); 4467 f2fs_exit_sysfs(); 4468 f2fs_destroy_garbage_collection_cache(); 4469 f2fs_destroy_extent_cache(); 4470 f2fs_destroy_recovery_cache(); 4471 f2fs_destroy_checkpoint_caches(); 4472 f2fs_destroy_segment_manager_caches(); 4473 f2fs_destroy_node_manager_caches(); 4474 destroy_inodecache(); 4475 } 4476 4477 module_init(init_f2fs_fs) 4478 module_exit(exit_f2fs_fs) 4479 4480 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4481 MODULE_DESCRIPTION("Flash Friendly File System"); 4482 MODULE_LICENSE("GPL"); 4483 MODULE_SOFTDEP("pre: crc32"); 4484 4485