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 29 #include "f2fs.h" 30 #include "node.h" 31 #include "segment.h" 32 #include "xattr.h" 33 #include "gc.h" 34 #include "trace.h" 35 36 #define CREATE_TRACE_POINTS 37 #include <trace/events/f2fs.h> 38 39 static struct kmem_cache *f2fs_inode_cachep; 40 41 #ifdef CONFIG_F2FS_FAULT_INJECTION 42 43 const char *f2fs_fault_name[FAULT_MAX] = { 44 [FAULT_KMALLOC] = "kmalloc", 45 [FAULT_KVMALLOC] = "kvmalloc", 46 [FAULT_PAGE_ALLOC] = "page alloc", 47 [FAULT_PAGE_GET] = "page get", 48 [FAULT_ALLOC_BIO] = "alloc bio", 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_compress_algorithm, 147 Opt_compress_log_size, 148 Opt_compress_extension, 149 Opt_compress_chksum, 150 Opt_compress_mode, 151 Opt_atgc, 152 Opt_err, 153 }; 154 155 static match_table_t f2fs_tokens = { 156 {Opt_gc_background, "background_gc=%s"}, 157 {Opt_disable_roll_forward, "disable_roll_forward"}, 158 {Opt_norecovery, "norecovery"}, 159 {Opt_discard, "discard"}, 160 {Opt_nodiscard, "nodiscard"}, 161 {Opt_noheap, "no_heap"}, 162 {Opt_heap, "heap"}, 163 {Opt_user_xattr, "user_xattr"}, 164 {Opt_nouser_xattr, "nouser_xattr"}, 165 {Opt_acl, "acl"}, 166 {Opt_noacl, "noacl"}, 167 {Opt_active_logs, "active_logs=%u"}, 168 {Opt_disable_ext_identify, "disable_ext_identify"}, 169 {Opt_inline_xattr, "inline_xattr"}, 170 {Opt_noinline_xattr, "noinline_xattr"}, 171 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 172 {Opt_inline_data, "inline_data"}, 173 {Opt_inline_dentry, "inline_dentry"}, 174 {Opt_noinline_dentry, "noinline_dentry"}, 175 {Opt_flush_merge, "flush_merge"}, 176 {Opt_noflush_merge, "noflush_merge"}, 177 {Opt_nobarrier, "nobarrier"}, 178 {Opt_fastboot, "fastboot"}, 179 {Opt_extent_cache, "extent_cache"}, 180 {Opt_noextent_cache, "noextent_cache"}, 181 {Opt_noinline_data, "noinline_data"}, 182 {Opt_data_flush, "data_flush"}, 183 {Opt_reserve_root, "reserve_root=%u"}, 184 {Opt_resgid, "resgid=%u"}, 185 {Opt_resuid, "resuid=%u"}, 186 {Opt_mode, "mode=%s"}, 187 {Opt_io_size_bits, "io_bits=%u"}, 188 {Opt_fault_injection, "fault_injection=%u"}, 189 {Opt_fault_type, "fault_type=%u"}, 190 {Opt_lazytime, "lazytime"}, 191 {Opt_nolazytime, "nolazytime"}, 192 {Opt_quota, "quota"}, 193 {Opt_noquota, "noquota"}, 194 {Opt_usrquota, "usrquota"}, 195 {Opt_grpquota, "grpquota"}, 196 {Opt_prjquota, "prjquota"}, 197 {Opt_usrjquota, "usrjquota=%s"}, 198 {Opt_grpjquota, "grpjquota=%s"}, 199 {Opt_prjjquota, "prjjquota=%s"}, 200 {Opt_offusrjquota, "usrjquota="}, 201 {Opt_offgrpjquota, "grpjquota="}, 202 {Opt_offprjjquota, "prjjquota="}, 203 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 204 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 205 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 206 {Opt_whint, "whint_mode=%s"}, 207 {Opt_alloc, "alloc_mode=%s"}, 208 {Opt_fsync, "fsync_mode=%s"}, 209 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 210 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 211 {Opt_inlinecrypt, "inlinecrypt"}, 212 {Opt_checkpoint_disable, "checkpoint=disable"}, 213 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 214 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 215 {Opt_checkpoint_enable, "checkpoint=enable"}, 216 {Opt_compress_algorithm, "compress_algorithm=%s"}, 217 {Opt_compress_log_size, "compress_log_size=%u"}, 218 {Opt_compress_extension, "compress_extension=%s"}, 219 {Opt_compress_chksum, "compress_chksum"}, 220 {Opt_compress_mode, "compress_mode=%s"}, 221 {Opt_atgc, "atgc"}, 222 {Opt_err, NULL}, 223 }; 224 225 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 226 { 227 struct va_format vaf; 228 va_list args; 229 int level; 230 231 va_start(args, fmt); 232 233 level = printk_get_level(fmt); 234 vaf.fmt = printk_skip_level(fmt); 235 vaf.va = &args; 236 printk("%c%cF2FS-fs (%s): %pV\n", 237 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 238 239 va_end(args); 240 } 241 242 #ifdef CONFIG_UNICODE 243 static const struct f2fs_sb_encodings { 244 __u16 magic; 245 char *name; 246 char *version; 247 } f2fs_sb_encoding_map[] = { 248 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"}, 249 }; 250 251 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb, 252 const struct f2fs_sb_encodings **encoding, 253 __u16 *flags) 254 { 255 __u16 magic = le16_to_cpu(sb->s_encoding); 256 int i; 257 258 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 259 if (magic == f2fs_sb_encoding_map[i].magic) 260 break; 261 262 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map)) 263 return -EINVAL; 264 265 *encoding = &f2fs_sb_encoding_map[i]; 266 *flags = le16_to_cpu(sb->s_encoding_flags); 267 268 return 0; 269 } 270 #endif 271 272 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 273 { 274 block_t limit = min((sbi->user_block_count << 1) / 1000, 275 sbi->user_block_count - sbi->reserved_blocks); 276 277 /* limit is 0.2% */ 278 if (test_opt(sbi, RESERVE_ROOT) && 279 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 280 F2FS_OPTION(sbi).root_reserved_blocks = limit; 281 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 282 F2FS_OPTION(sbi).root_reserved_blocks); 283 } 284 if (!test_opt(sbi, RESERVE_ROOT) && 285 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 286 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 287 !gid_eq(F2FS_OPTION(sbi).s_resgid, 288 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 289 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 290 from_kuid_munged(&init_user_ns, 291 F2FS_OPTION(sbi).s_resuid), 292 from_kgid_munged(&init_user_ns, 293 F2FS_OPTION(sbi).s_resgid)); 294 } 295 296 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 297 { 298 if (!F2FS_OPTION(sbi).unusable_cap_perc) 299 return; 300 301 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 302 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 303 else 304 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 305 F2FS_OPTION(sbi).unusable_cap_perc; 306 307 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 308 F2FS_OPTION(sbi).unusable_cap, 309 F2FS_OPTION(sbi).unusable_cap_perc); 310 } 311 312 static void init_once(void *foo) 313 { 314 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 315 316 inode_init_once(&fi->vfs_inode); 317 } 318 319 #ifdef CONFIG_QUOTA 320 static const char * const quotatypes[] = INITQFNAMES; 321 #define QTYPE2NAME(t) (quotatypes[t]) 322 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 323 substring_t *args) 324 { 325 struct f2fs_sb_info *sbi = F2FS_SB(sb); 326 char *qname; 327 int ret = -EINVAL; 328 329 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 330 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 331 return -EINVAL; 332 } 333 if (f2fs_sb_has_quota_ino(sbi)) { 334 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 335 return 0; 336 } 337 338 qname = match_strdup(args); 339 if (!qname) { 340 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 341 return -ENOMEM; 342 } 343 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 344 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 345 ret = 0; 346 else 347 f2fs_err(sbi, "%s quota file already specified", 348 QTYPE2NAME(qtype)); 349 goto errout; 350 } 351 if (strchr(qname, '/')) { 352 f2fs_err(sbi, "quotafile must be on filesystem root"); 353 goto errout; 354 } 355 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 356 set_opt(sbi, QUOTA); 357 return 0; 358 errout: 359 kfree(qname); 360 return ret; 361 } 362 363 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 364 { 365 struct f2fs_sb_info *sbi = F2FS_SB(sb); 366 367 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 368 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 369 return -EINVAL; 370 } 371 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 372 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 373 return 0; 374 } 375 376 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 377 { 378 /* 379 * We do the test below only for project quotas. 'usrquota' and 380 * 'grpquota' mount options are allowed even without quota feature 381 * to support legacy quotas in quota files. 382 */ 383 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 384 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 385 return -1; 386 } 387 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 388 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 389 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 390 if (test_opt(sbi, USRQUOTA) && 391 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 392 clear_opt(sbi, USRQUOTA); 393 394 if (test_opt(sbi, GRPQUOTA) && 395 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 396 clear_opt(sbi, GRPQUOTA); 397 398 if (test_opt(sbi, PRJQUOTA) && 399 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 400 clear_opt(sbi, PRJQUOTA); 401 402 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 403 test_opt(sbi, PRJQUOTA)) { 404 f2fs_err(sbi, "old and new quota format mixing"); 405 return -1; 406 } 407 408 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 409 f2fs_err(sbi, "journaled quota format not specified"); 410 return -1; 411 } 412 } 413 414 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 415 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 416 F2FS_OPTION(sbi).s_jquota_fmt = 0; 417 } 418 return 0; 419 } 420 #endif 421 422 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 423 const char *opt, 424 const substring_t *arg, 425 bool is_remount) 426 { 427 struct f2fs_sb_info *sbi = F2FS_SB(sb); 428 #ifdef CONFIG_FS_ENCRYPTION 429 int err; 430 431 if (!f2fs_sb_has_encrypt(sbi)) { 432 f2fs_err(sbi, "Encrypt feature is off"); 433 return -EINVAL; 434 } 435 436 /* 437 * This mount option is just for testing, and it's not worthwhile to 438 * implement the extra complexity (e.g. RCU protection) that would be 439 * needed to allow it to be set or changed during remount. We do allow 440 * it to be specified during remount, but only if there is no change. 441 */ 442 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) { 443 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 444 return -EINVAL; 445 } 446 err = fscrypt_set_test_dummy_encryption( 447 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy); 448 if (err) { 449 if (err == -EEXIST) 450 f2fs_warn(sbi, 451 "Can't change test_dummy_encryption on remount"); 452 else if (err == -EINVAL) 453 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 454 opt); 455 else 456 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 457 opt, err); 458 return -EINVAL; 459 } 460 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 461 #else 462 f2fs_warn(sbi, "Test dummy encryption mount option ignored"); 463 #endif 464 return 0; 465 } 466 467 static int parse_options(struct super_block *sb, char *options, bool is_remount) 468 { 469 struct f2fs_sb_info *sbi = F2FS_SB(sb); 470 substring_t args[MAX_OPT_ARGS]; 471 #ifdef CONFIG_F2FS_FS_COMPRESSION 472 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 473 int ext_cnt; 474 #endif 475 char *p, *name; 476 int arg = 0; 477 kuid_t uid; 478 kgid_t gid; 479 int ret; 480 481 if (!options) 482 return 0; 483 484 while ((p = strsep(&options, ",")) != NULL) { 485 int token; 486 if (!*p) 487 continue; 488 /* 489 * Initialize args struct so we know whether arg was 490 * found; some options take optional arguments. 491 */ 492 args[0].to = args[0].from = NULL; 493 token = match_token(p, f2fs_tokens, args); 494 495 switch (token) { 496 case Opt_gc_background: 497 name = match_strdup(&args[0]); 498 499 if (!name) 500 return -ENOMEM; 501 if (!strcmp(name, "on")) { 502 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 503 } else if (!strcmp(name, "off")) { 504 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 505 } else if (!strcmp(name, "sync")) { 506 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 507 } else { 508 kfree(name); 509 return -EINVAL; 510 } 511 kfree(name); 512 break; 513 case Opt_disable_roll_forward: 514 set_opt(sbi, DISABLE_ROLL_FORWARD); 515 break; 516 case Opt_norecovery: 517 /* this option mounts f2fs with ro */ 518 set_opt(sbi, NORECOVERY); 519 if (!f2fs_readonly(sb)) 520 return -EINVAL; 521 break; 522 case Opt_discard: 523 set_opt(sbi, DISCARD); 524 break; 525 case Opt_nodiscard: 526 if (f2fs_sb_has_blkzoned(sbi)) { 527 f2fs_warn(sbi, "discard is required for zoned block devices"); 528 return -EINVAL; 529 } 530 clear_opt(sbi, DISCARD); 531 break; 532 case Opt_noheap: 533 set_opt(sbi, NOHEAP); 534 break; 535 case Opt_heap: 536 clear_opt(sbi, NOHEAP); 537 break; 538 #ifdef CONFIG_F2FS_FS_XATTR 539 case Opt_user_xattr: 540 set_opt(sbi, XATTR_USER); 541 break; 542 case Opt_nouser_xattr: 543 clear_opt(sbi, XATTR_USER); 544 break; 545 case Opt_inline_xattr: 546 set_opt(sbi, INLINE_XATTR); 547 break; 548 case Opt_noinline_xattr: 549 clear_opt(sbi, INLINE_XATTR); 550 break; 551 case Opt_inline_xattr_size: 552 if (args->from && match_int(args, &arg)) 553 return -EINVAL; 554 set_opt(sbi, INLINE_XATTR_SIZE); 555 F2FS_OPTION(sbi).inline_xattr_size = arg; 556 break; 557 #else 558 case Opt_user_xattr: 559 f2fs_info(sbi, "user_xattr options not supported"); 560 break; 561 case Opt_nouser_xattr: 562 f2fs_info(sbi, "nouser_xattr options not supported"); 563 break; 564 case Opt_inline_xattr: 565 f2fs_info(sbi, "inline_xattr options not supported"); 566 break; 567 case Opt_noinline_xattr: 568 f2fs_info(sbi, "noinline_xattr options not supported"); 569 break; 570 #endif 571 #ifdef CONFIG_F2FS_FS_POSIX_ACL 572 case Opt_acl: 573 set_opt(sbi, POSIX_ACL); 574 break; 575 case Opt_noacl: 576 clear_opt(sbi, POSIX_ACL); 577 break; 578 #else 579 case Opt_acl: 580 f2fs_info(sbi, "acl options not supported"); 581 break; 582 case Opt_noacl: 583 f2fs_info(sbi, "noacl options not supported"); 584 break; 585 #endif 586 case Opt_active_logs: 587 if (args->from && match_int(args, &arg)) 588 return -EINVAL; 589 if (arg != 2 && arg != 4 && 590 arg != NR_CURSEG_PERSIST_TYPE) 591 return -EINVAL; 592 F2FS_OPTION(sbi).active_logs = arg; 593 break; 594 case Opt_disable_ext_identify: 595 set_opt(sbi, DISABLE_EXT_IDENTIFY); 596 break; 597 case Opt_inline_data: 598 set_opt(sbi, INLINE_DATA); 599 break; 600 case Opt_inline_dentry: 601 set_opt(sbi, INLINE_DENTRY); 602 break; 603 case Opt_noinline_dentry: 604 clear_opt(sbi, INLINE_DENTRY); 605 break; 606 case Opt_flush_merge: 607 set_opt(sbi, FLUSH_MERGE); 608 break; 609 case Opt_noflush_merge: 610 clear_opt(sbi, FLUSH_MERGE); 611 break; 612 case Opt_nobarrier: 613 set_opt(sbi, NOBARRIER); 614 break; 615 case Opt_fastboot: 616 set_opt(sbi, FASTBOOT); 617 break; 618 case Opt_extent_cache: 619 set_opt(sbi, EXTENT_CACHE); 620 break; 621 case Opt_noextent_cache: 622 clear_opt(sbi, EXTENT_CACHE); 623 break; 624 case Opt_noinline_data: 625 clear_opt(sbi, INLINE_DATA); 626 break; 627 case Opt_data_flush: 628 set_opt(sbi, DATA_FLUSH); 629 break; 630 case Opt_reserve_root: 631 if (args->from && match_int(args, &arg)) 632 return -EINVAL; 633 if (test_opt(sbi, RESERVE_ROOT)) { 634 f2fs_info(sbi, "Preserve previous reserve_root=%u", 635 F2FS_OPTION(sbi).root_reserved_blocks); 636 } else { 637 F2FS_OPTION(sbi).root_reserved_blocks = arg; 638 set_opt(sbi, RESERVE_ROOT); 639 } 640 break; 641 case Opt_resuid: 642 if (args->from && match_int(args, &arg)) 643 return -EINVAL; 644 uid = make_kuid(current_user_ns(), arg); 645 if (!uid_valid(uid)) { 646 f2fs_err(sbi, "Invalid uid value %d", arg); 647 return -EINVAL; 648 } 649 F2FS_OPTION(sbi).s_resuid = uid; 650 break; 651 case Opt_resgid: 652 if (args->from && match_int(args, &arg)) 653 return -EINVAL; 654 gid = make_kgid(current_user_ns(), arg); 655 if (!gid_valid(gid)) { 656 f2fs_err(sbi, "Invalid gid value %d", arg); 657 return -EINVAL; 658 } 659 F2FS_OPTION(sbi).s_resgid = gid; 660 break; 661 case Opt_mode: 662 name = match_strdup(&args[0]); 663 664 if (!name) 665 return -ENOMEM; 666 if (!strcmp(name, "adaptive")) { 667 if (f2fs_sb_has_blkzoned(sbi)) { 668 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 669 kfree(name); 670 return -EINVAL; 671 } 672 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 673 } else if (!strcmp(name, "lfs")) { 674 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 675 } else { 676 kfree(name); 677 return -EINVAL; 678 } 679 kfree(name); 680 break; 681 case Opt_io_size_bits: 682 if (args->from && match_int(args, &arg)) 683 return -EINVAL; 684 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) { 685 f2fs_warn(sbi, "Not support %d, larger than %d", 686 1 << arg, BIO_MAX_PAGES); 687 return -EINVAL; 688 } 689 F2FS_OPTION(sbi).write_io_size_bits = arg; 690 break; 691 #ifdef CONFIG_F2FS_FAULT_INJECTION 692 case Opt_fault_injection: 693 if (args->from && match_int(args, &arg)) 694 return -EINVAL; 695 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 696 set_opt(sbi, FAULT_INJECTION); 697 break; 698 699 case Opt_fault_type: 700 if (args->from && match_int(args, &arg)) 701 return -EINVAL; 702 f2fs_build_fault_attr(sbi, 0, arg); 703 set_opt(sbi, FAULT_INJECTION); 704 break; 705 #else 706 case Opt_fault_injection: 707 f2fs_info(sbi, "fault_injection options not supported"); 708 break; 709 710 case Opt_fault_type: 711 f2fs_info(sbi, "fault_type options not supported"); 712 break; 713 #endif 714 case Opt_lazytime: 715 sb->s_flags |= SB_LAZYTIME; 716 break; 717 case Opt_nolazytime: 718 sb->s_flags &= ~SB_LAZYTIME; 719 break; 720 #ifdef CONFIG_QUOTA 721 case Opt_quota: 722 case Opt_usrquota: 723 set_opt(sbi, USRQUOTA); 724 break; 725 case Opt_grpquota: 726 set_opt(sbi, GRPQUOTA); 727 break; 728 case Opt_prjquota: 729 set_opt(sbi, PRJQUOTA); 730 break; 731 case Opt_usrjquota: 732 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 733 if (ret) 734 return ret; 735 break; 736 case Opt_grpjquota: 737 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 738 if (ret) 739 return ret; 740 break; 741 case Opt_prjjquota: 742 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 743 if (ret) 744 return ret; 745 break; 746 case Opt_offusrjquota: 747 ret = f2fs_clear_qf_name(sb, USRQUOTA); 748 if (ret) 749 return ret; 750 break; 751 case Opt_offgrpjquota: 752 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 753 if (ret) 754 return ret; 755 break; 756 case Opt_offprjjquota: 757 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 758 if (ret) 759 return ret; 760 break; 761 case Opt_jqfmt_vfsold: 762 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 763 break; 764 case Opt_jqfmt_vfsv0: 765 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 766 break; 767 case Opt_jqfmt_vfsv1: 768 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 769 break; 770 case Opt_noquota: 771 clear_opt(sbi, QUOTA); 772 clear_opt(sbi, USRQUOTA); 773 clear_opt(sbi, GRPQUOTA); 774 clear_opt(sbi, PRJQUOTA); 775 break; 776 #else 777 case Opt_quota: 778 case Opt_usrquota: 779 case Opt_grpquota: 780 case Opt_prjquota: 781 case Opt_usrjquota: 782 case Opt_grpjquota: 783 case Opt_prjjquota: 784 case Opt_offusrjquota: 785 case Opt_offgrpjquota: 786 case Opt_offprjjquota: 787 case Opt_jqfmt_vfsold: 788 case Opt_jqfmt_vfsv0: 789 case Opt_jqfmt_vfsv1: 790 case Opt_noquota: 791 f2fs_info(sbi, "quota operations not supported"); 792 break; 793 #endif 794 case Opt_whint: 795 name = match_strdup(&args[0]); 796 if (!name) 797 return -ENOMEM; 798 if (!strcmp(name, "user-based")) { 799 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER; 800 } else if (!strcmp(name, "off")) { 801 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 802 } else if (!strcmp(name, "fs-based")) { 803 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS; 804 } else { 805 kfree(name); 806 return -EINVAL; 807 } 808 kfree(name); 809 break; 810 case Opt_alloc: 811 name = match_strdup(&args[0]); 812 if (!name) 813 return -ENOMEM; 814 815 if (!strcmp(name, "default")) { 816 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 817 } else if (!strcmp(name, "reuse")) { 818 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 819 } else { 820 kfree(name); 821 return -EINVAL; 822 } 823 kfree(name); 824 break; 825 case Opt_fsync: 826 name = match_strdup(&args[0]); 827 if (!name) 828 return -ENOMEM; 829 if (!strcmp(name, "posix")) { 830 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 831 } else if (!strcmp(name, "strict")) { 832 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 833 } else if (!strcmp(name, "nobarrier")) { 834 F2FS_OPTION(sbi).fsync_mode = 835 FSYNC_MODE_NOBARRIER; 836 } else { 837 kfree(name); 838 return -EINVAL; 839 } 840 kfree(name); 841 break; 842 case Opt_test_dummy_encryption: 843 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 844 is_remount); 845 if (ret) 846 return ret; 847 break; 848 case Opt_inlinecrypt: 849 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 850 sb->s_flags |= SB_INLINECRYPT; 851 #else 852 f2fs_info(sbi, "inline encryption not supported"); 853 #endif 854 break; 855 case Opt_checkpoint_disable_cap_perc: 856 if (args->from && match_int(args, &arg)) 857 return -EINVAL; 858 if (arg < 0 || arg > 100) 859 return -EINVAL; 860 F2FS_OPTION(sbi).unusable_cap_perc = arg; 861 set_opt(sbi, DISABLE_CHECKPOINT); 862 break; 863 case Opt_checkpoint_disable_cap: 864 if (args->from && match_int(args, &arg)) 865 return -EINVAL; 866 F2FS_OPTION(sbi).unusable_cap = arg; 867 set_opt(sbi, DISABLE_CHECKPOINT); 868 break; 869 case Opt_checkpoint_disable: 870 set_opt(sbi, DISABLE_CHECKPOINT); 871 break; 872 case Opt_checkpoint_enable: 873 clear_opt(sbi, DISABLE_CHECKPOINT); 874 break; 875 #ifdef CONFIG_F2FS_FS_COMPRESSION 876 case Opt_compress_algorithm: 877 if (!f2fs_sb_has_compression(sbi)) { 878 f2fs_info(sbi, "Image doesn't support compression"); 879 break; 880 } 881 name = match_strdup(&args[0]); 882 if (!name) 883 return -ENOMEM; 884 if (!strcmp(name, "lzo")) { 885 F2FS_OPTION(sbi).compress_algorithm = 886 COMPRESS_LZO; 887 } else if (!strcmp(name, "lz4")) { 888 F2FS_OPTION(sbi).compress_algorithm = 889 COMPRESS_LZ4; 890 } else if (!strcmp(name, "zstd")) { 891 F2FS_OPTION(sbi).compress_algorithm = 892 COMPRESS_ZSTD; 893 } else if (!strcmp(name, "lzo-rle")) { 894 F2FS_OPTION(sbi).compress_algorithm = 895 COMPRESS_LZORLE; 896 } else { 897 kfree(name); 898 return -EINVAL; 899 } 900 kfree(name); 901 break; 902 case Opt_compress_log_size: 903 if (!f2fs_sb_has_compression(sbi)) { 904 f2fs_info(sbi, "Image doesn't support compression"); 905 break; 906 } 907 if (args->from && match_int(args, &arg)) 908 return -EINVAL; 909 if (arg < MIN_COMPRESS_LOG_SIZE || 910 arg > MAX_COMPRESS_LOG_SIZE) { 911 f2fs_err(sbi, 912 "Compress cluster log size is out of range"); 913 return -EINVAL; 914 } 915 F2FS_OPTION(sbi).compress_log_size = arg; 916 break; 917 case Opt_compress_extension: 918 if (!f2fs_sb_has_compression(sbi)) { 919 f2fs_info(sbi, "Image doesn't support compression"); 920 break; 921 } 922 name = match_strdup(&args[0]); 923 if (!name) 924 return -ENOMEM; 925 926 ext = F2FS_OPTION(sbi).extensions; 927 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 928 929 if (strlen(name) >= F2FS_EXTENSION_LEN || 930 ext_cnt >= COMPRESS_EXT_NUM) { 931 f2fs_err(sbi, 932 "invalid extension length/number"); 933 kfree(name); 934 return -EINVAL; 935 } 936 937 strcpy(ext[ext_cnt], name); 938 F2FS_OPTION(sbi).compress_ext_cnt++; 939 kfree(name); 940 break; 941 case Opt_compress_chksum: 942 F2FS_OPTION(sbi).compress_chksum = true; 943 break; 944 case Opt_compress_mode: 945 name = match_strdup(&args[0]); 946 if (!name) 947 return -ENOMEM; 948 if (!strcmp(name, "fs")) { 949 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 950 } else if (!strcmp(name, "user")) { 951 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 952 } else { 953 kfree(name); 954 return -EINVAL; 955 } 956 kfree(name); 957 break; 958 #else 959 case Opt_compress_algorithm: 960 case Opt_compress_log_size: 961 case Opt_compress_extension: 962 case Opt_compress_chksum: 963 case Opt_compress_mode: 964 f2fs_info(sbi, "compression options not supported"); 965 break; 966 #endif 967 case Opt_atgc: 968 set_opt(sbi, ATGC); 969 break; 970 default: 971 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 972 p); 973 return -EINVAL; 974 } 975 } 976 #ifdef CONFIG_QUOTA 977 if (f2fs_check_quota_options(sbi)) 978 return -EINVAL; 979 #else 980 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 981 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 982 return -EINVAL; 983 } 984 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 985 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 986 return -EINVAL; 987 } 988 #endif 989 #ifndef CONFIG_UNICODE 990 if (f2fs_sb_has_casefold(sbi)) { 991 f2fs_err(sbi, 992 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 993 return -EINVAL; 994 } 995 #endif 996 /* 997 * The BLKZONED feature indicates that the drive was formatted with 998 * zone alignment optimization. This is optional for host-aware 999 * devices, but mandatory for host-managed zoned block devices. 1000 */ 1001 #ifndef CONFIG_BLK_DEV_ZONED 1002 if (f2fs_sb_has_blkzoned(sbi)) { 1003 f2fs_err(sbi, "Zoned block device support is not enabled"); 1004 return -EINVAL; 1005 } 1006 #endif 1007 1008 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 1009 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO", 1010 F2FS_IO_SIZE_KB(sbi)); 1011 return -EINVAL; 1012 } 1013 1014 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1015 int min_size, max_size; 1016 1017 if (!f2fs_sb_has_extra_attr(sbi) || 1018 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1019 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1020 return -EINVAL; 1021 } 1022 if (!test_opt(sbi, INLINE_XATTR)) { 1023 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1024 return -EINVAL; 1025 } 1026 1027 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 1028 max_size = MAX_INLINE_XATTR_SIZE; 1029 1030 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1031 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1032 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1033 min_size, max_size); 1034 return -EINVAL; 1035 } 1036 } 1037 1038 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1039 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n"); 1040 return -EINVAL; 1041 } 1042 1043 /* Not pass down write hints if the number of active logs is lesser 1044 * than NR_CURSEG_PERSIST_TYPE. 1045 */ 1046 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE) 1047 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1048 return 0; 1049 } 1050 1051 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1052 { 1053 struct f2fs_inode_info *fi; 1054 1055 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); 1056 if (!fi) 1057 return NULL; 1058 1059 init_once((void *) fi); 1060 1061 /* Initialize f2fs-specific inode info */ 1062 atomic_set(&fi->dirty_pages, 0); 1063 atomic_set(&fi->i_compr_blocks, 0); 1064 init_rwsem(&fi->i_sem); 1065 spin_lock_init(&fi->i_size_lock); 1066 INIT_LIST_HEAD(&fi->dirty_list); 1067 INIT_LIST_HEAD(&fi->gdirty_list); 1068 INIT_LIST_HEAD(&fi->inmem_ilist); 1069 INIT_LIST_HEAD(&fi->inmem_pages); 1070 mutex_init(&fi->inmem_lock); 1071 init_rwsem(&fi->i_gc_rwsem[READ]); 1072 init_rwsem(&fi->i_gc_rwsem[WRITE]); 1073 init_rwsem(&fi->i_mmap_sem); 1074 init_rwsem(&fi->i_xattr_sem); 1075 1076 /* Will be used by directory only */ 1077 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1078 1079 fi->ra_offset = -1; 1080 1081 return &fi->vfs_inode; 1082 } 1083 1084 static int f2fs_drop_inode(struct inode *inode) 1085 { 1086 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1087 int ret; 1088 1089 /* 1090 * during filesystem shutdown, if checkpoint is disabled, 1091 * drop useless meta/node dirty pages. 1092 */ 1093 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1094 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1095 inode->i_ino == F2FS_META_INO(sbi)) { 1096 trace_f2fs_drop_inode(inode, 1); 1097 return 1; 1098 } 1099 } 1100 1101 /* 1102 * This is to avoid a deadlock condition like below. 1103 * writeback_single_inode(inode) 1104 * - f2fs_write_data_page 1105 * - f2fs_gc -> iput -> evict 1106 * - inode_wait_for_writeback(inode) 1107 */ 1108 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1109 if (!inode->i_nlink && !is_bad_inode(inode)) { 1110 /* to avoid evict_inode call simultaneously */ 1111 atomic_inc(&inode->i_count); 1112 spin_unlock(&inode->i_lock); 1113 1114 /* some remained atomic pages should discarded */ 1115 if (f2fs_is_atomic_file(inode)) 1116 f2fs_drop_inmem_pages(inode); 1117 1118 /* should remain fi->extent_tree for writepage */ 1119 f2fs_destroy_extent_node(inode); 1120 1121 sb_start_intwrite(inode->i_sb); 1122 f2fs_i_size_write(inode, 0); 1123 1124 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1125 inode, NULL, 0, DATA); 1126 truncate_inode_pages_final(inode->i_mapping); 1127 1128 if (F2FS_HAS_BLOCKS(inode)) 1129 f2fs_truncate(inode); 1130 1131 sb_end_intwrite(inode->i_sb); 1132 1133 spin_lock(&inode->i_lock); 1134 atomic_dec(&inode->i_count); 1135 } 1136 trace_f2fs_drop_inode(inode, 0); 1137 return 0; 1138 } 1139 ret = generic_drop_inode(inode); 1140 if (!ret) 1141 ret = fscrypt_drop_inode(inode); 1142 trace_f2fs_drop_inode(inode, ret); 1143 return ret; 1144 } 1145 1146 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1147 { 1148 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1149 int ret = 0; 1150 1151 spin_lock(&sbi->inode_lock[DIRTY_META]); 1152 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1153 ret = 1; 1154 } else { 1155 set_inode_flag(inode, FI_DIRTY_INODE); 1156 stat_inc_dirty_inode(sbi, DIRTY_META); 1157 } 1158 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1159 list_add_tail(&F2FS_I(inode)->gdirty_list, 1160 &sbi->inode_list[DIRTY_META]); 1161 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1162 } 1163 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1164 return ret; 1165 } 1166 1167 void f2fs_inode_synced(struct inode *inode) 1168 { 1169 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1170 1171 spin_lock(&sbi->inode_lock[DIRTY_META]); 1172 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1173 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1174 return; 1175 } 1176 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1177 list_del_init(&F2FS_I(inode)->gdirty_list); 1178 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1179 } 1180 clear_inode_flag(inode, FI_DIRTY_INODE); 1181 clear_inode_flag(inode, FI_AUTO_RECOVER); 1182 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1183 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1184 } 1185 1186 /* 1187 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1188 * 1189 * We should call set_dirty_inode to write the dirty inode through write_inode. 1190 */ 1191 static void f2fs_dirty_inode(struct inode *inode, int flags) 1192 { 1193 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1194 1195 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1196 inode->i_ino == F2FS_META_INO(sbi)) 1197 return; 1198 1199 if (flags == I_DIRTY_TIME) 1200 return; 1201 1202 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1203 clear_inode_flag(inode, FI_AUTO_RECOVER); 1204 1205 f2fs_inode_dirtied(inode, false); 1206 } 1207 1208 static void f2fs_free_inode(struct inode *inode) 1209 { 1210 fscrypt_free_inode(inode); 1211 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1212 } 1213 1214 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1215 { 1216 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1217 percpu_counter_destroy(&sbi->total_valid_inode_count); 1218 } 1219 1220 static void destroy_device_list(struct f2fs_sb_info *sbi) 1221 { 1222 int i; 1223 1224 for (i = 0; i < sbi->s_ndevs; i++) { 1225 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1226 #ifdef CONFIG_BLK_DEV_ZONED 1227 kvfree(FDEV(i).blkz_seq); 1228 kfree(FDEV(i).zone_capacity_blocks); 1229 #endif 1230 } 1231 kvfree(sbi->devs); 1232 } 1233 1234 static void f2fs_put_super(struct super_block *sb) 1235 { 1236 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1237 int i; 1238 bool dropped; 1239 1240 /* unregister procfs/sysfs entries in advance to avoid race case */ 1241 f2fs_unregister_sysfs(sbi); 1242 1243 f2fs_quota_off_umount(sb); 1244 1245 /* prevent remaining shrinker jobs */ 1246 mutex_lock(&sbi->umount_mutex); 1247 1248 /* 1249 * We don't need to do checkpoint when superblock is clean. 1250 * But, the previous checkpoint was not done by umount, it needs to do 1251 * clean checkpoint again. 1252 */ 1253 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1254 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1255 struct cp_control cpc = { 1256 .reason = CP_UMOUNT, 1257 }; 1258 f2fs_write_checkpoint(sbi, &cpc); 1259 } 1260 1261 /* be sure to wait for any on-going discard commands */ 1262 dropped = f2fs_issue_discard_timeout(sbi); 1263 1264 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && 1265 !sbi->discard_blks && !dropped) { 1266 struct cp_control cpc = { 1267 .reason = CP_UMOUNT | CP_TRIMMED, 1268 }; 1269 f2fs_write_checkpoint(sbi, &cpc); 1270 } 1271 1272 /* 1273 * normally superblock is clean, so we need to release this. 1274 * In addition, EIO will skip do checkpoint, we need this as well. 1275 */ 1276 f2fs_release_ino_entry(sbi, true); 1277 1278 f2fs_leave_shrinker(sbi); 1279 mutex_unlock(&sbi->umount_mutex); 1280 1281 /* our cp_error case, we can wait for any writeback page */ 1282 f2fs_flush_merged_writes(sbi); 1283 1284 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1285 1286 f2fs_bug_on(sbi, sbi->fsync_node_num); 1287 1288 iput(sbi->node_inode); 1289 sbi->node_inode = NULL; 1290 1291 iput(sbi->meta_inode); 1292 sbi->meta_inode = NULL; 1293 1294 /* 1295 * iput() can update stat information, if f2fs_write_checkpoint() 1296 * above failed with error. 1297 */ 1298 f2fs_destroy_stats(sbi); 1299 1300 /* destroy f2fs internal modules */ 1301 f2fs_destroy_node_manager(sbi); 1302 f2fs_destroy_segment_manager(sbi); 1303 1304 f2fs_destroy_post_read_wq(sbi); 1305 1306 kvfree(sbi->ckpt); 1307 1308 sb->s_fs_info = NULL; 1309 if (sbi->s_chksum_driver) 1310 crypto_free_shash(sbi->s_chksum_driver); 1311 kfree(sbi->raw_super); 1312 1313 destroy_device_list(sbi); 1314 f2fs_destroy_page_array_cache(sbi); 1315 f2fs_destroy_xattr_caches(sbi); 1316 mempool_destroy(sbi->write_io_dummy); 1317 #ifdef CONFIG_QUOTA 1318 for (i = 0; i < MAXQUOTAS; i++) 1319 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1320 #endif 1321 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1322 destroy_percpu_info(sbi); 1323 for (i = 0; i < NR_PAGE_TYPE; i++) 1324 kvfree(sbi->write_io[i]); 1325 #ifdef CONFIG_UNICODE 1326 utf8_unload(sb->s_encoding); 1327 #endif 1328 kfree(sbi); 1329 } 1330 1331 int f2fs_sync_fs(struct super_block *sb, int sync) 1332 { 1333 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1334 int err = 0; 1335 1336 if (unlikely(f2fs_cp_error(sbi))) 1337 return 0; 1338 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1339 return 0; 1340 1341 trace_f2fs_sync_fs(sb, sync); 1342 1343 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1344 return -EAGAIN; 1345 1346 if (sync) { 1347 struct cp_control cpc; 1348 1349 cpc.reason = __get_cp_reason(sbi); 1350 1351 down_write(&sbi->gc_lock); 1352 err = f2fs_write_checkpoint(sbi, &cpc); 1353 up_write(&sbi->gc_lock); 1354 } 1355 f2fs_trace_ios(NULL, 1); 1356 1357 return err; 1358 } 1359 1360 static int f2fs_freeze(struct super_block *sb) 1361 { 1362 if (f2fs_readonly(sb)) 1363 return 0; 1364 1365 /* IO error happened before */ 1366 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1367 return -EIO; 1368 1369 /* must be clean, since sync_filesystem() was already called */ 1370 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1371 return -EINVAL; 1372 return 0; 1373 } 1374 1375 static int f2fs_unfreeze(struct super_block *sb) 1376 { 1377 return 0; 1378 } 1379 1380 #ifdef CONFIG_QUOTA 1381 static int f2fs_statfs_project(struct super_block *sb, 1382 kprojid_t projid, struct kstatfs *buf) 1383 { 1384 struct kqid qid; 1385 struct dquot *dquot; 1386 u64 limit; 1387 u64 curblock; 1388 1389 qid = make_kqid_projid(projid); 1390 dquot = dqget(sb, qid); 1391 if (IS_ERR(dquot)) 1392 return PTR_ERR(dquot); 1393 spin_lock(&dquot->dq_dqb_lock); 1394 1395 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1396 dquot->dq_dqb.dqb_bhardlimit); 1397 if (limit) 1398 limit >>= sb->s_blocksize_bits; 1399 1400 if (limit && buf->f_blocks > limit) { 1401 curblock = (dquot->dq_dqb.dqb_curspace + 1402 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1403 buf->f_blocks = limit; 1404 buf->f_bfree = buf->f_bavail = 1405 (buf->f_blocks > curblock) ? 1406 (buf->f_blocks - curblock) : 0; 1407 } 1408 1409 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1410 dquot->dq_dqb.dqb_ihardlimit); 1411 1412 if (limit && buf->f_files > limit) { 1413 buf->f_files = limit; 1414 buf->f_ffree = 1415 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1416 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1417 } 1418 1419 spin_unlock(&dquot->dq_dqb_lock); 1420 dqput(dquot); 1421 return 0; 1422 } 1423 #endif 1424 1425 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1426 { 1427 struct super_block *sb = dentry->d_sb; 1428 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1429 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1430 block_t total_count, user_block_count, start_count; 1431 u64 avail_node_count; 1432 1433 total_count = le64_to_cpu(sbi->raw_super->block_count); 1434 user_block_count = sbi->user_block_count; 1435 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1436 buf->f_type = F2FS_SUPER_MAGIC; 1437 buf->f_bsize = sbi->blocksize; 1438 1439 buf->f_blocks = total_count - start_count; 1440 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1441 sbi->current_reserved_blocks; 1442 1443 spin_lock(&sbi->stat_lock); 1444 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1445 buf->f_bfree = 0; 1446 else 1447 buf->f_bfree -= sbi->unusable_block_count; 1448 spin_unlock(&sbi->stat_lock); 1449 1450 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1451 buf->f_bavail = buf->f_bfree - 1452 F2FS_OPTION(sbi).root_reserved_blocks; 1453 else 1454 buf->f_bavail = 0; 1455 1456 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1457 1458 if (avail_node_count > user_block_count) { 1459 buf->f_files = user_block_count; 1460 buf->f_ffree = buf->f_bavail; 1461 } else { 1462 buf->f_files = avail_node_count; 1463 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1464 buf->f_bavail); 1465 } 1466 1467 buf->f_namelen = F2FS_NAME_LEN; 1468 buf->f_fsid = u64_to_fsid(id); 1469 1470 #ifdef CONFIG_QUOTA 1471 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1472 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1473 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1474 } 1475 #endif 1476 return 0; 1477 } 1478 1479 static inline void f2fs_show_quota_options(struct seq_file *seq, 1480 struct super_block *sb) 1481 { 1482 #ifdef CONFIG_QUOTA 1483 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1484 1485 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1486 char *fmtname = ""; 1487 1488 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1489 case QFMT_VFS_OLD: 1490 fmtname = "vfsold"; 1491 break; 1492 case QFMT_VFS_V0: 1493 fmtname = "vfsv0"; 1494 break; 1495 case QFMT_VFS_V1: 1496 fmtname = "vfsv1"; 1497 break; 1498 } 1499 seq_printf(seq, ",jqfmt=%s", fmtname); 1500 } 1501 1502 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1503 seq_show_option(seq, "usrjquota", 1504 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1505 1506 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1507 seq_show_option(seq, "grpjquota", 1508 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1509 1510 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1511 seq_show_option(seq, "prjjquota", 1512 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1513 #endif 1514 } 1515 1516 static inline void f2fs_show_compress_options(struct seq_file *seq, 1517 struct super_block *sb) 1518 { 1519 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1520 char *algtype = ""; 1521 int i; 1522 1523 if (!f2fs_sb_has_compression(sbi)) 1524 return; 1525 1526 switch (F2FS_OPTION(sbi).compress_algorithm) { 1527 case COMPRESS_LZO: 1528 algtype = "lzo"; 1529 break; 1530 case COMPRESS_LZ4: 1531 algtype = "lz4"; 1532 break; 1533 case COMPRESS_ZSTD: 1534 algtype = "zstd"; 1535 break; 1536 case COMPRESS_LZORLE: 1537 algtype = "lzo-rle"; 1538 break; 1539 } 1540 seq_printf(seq, ",compress_algorithm=%s", algtype); 1541 1542 seq_printf(seq, ",compress_log_size=%u", 1543 F2FS_OPTION(sbi).compress_log_size); 1544 1545 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1546 seq_printf(seq, ",compress_extension=%s", 1547 F2FS_OPTION(sbi).extensions[i]); 1548 } 1549 1550 if (F2FS_OPTION(sbi).compress_chksum) 1551 seq_puts(seq, ",compress_chksum"); 1552 1553 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1554 seq_printf(seq, ",compress_mode=%s", "fs"); 1555 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1556 seq_printf(seq, ",compress_mode=%s", "user"); 1557 } 1558 1559 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1560 { 1561 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1562 1563 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1564 seq_printf(seq, ",background_gc=%s", "sync"); 1565 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1566 seq_printf(seq, ",background_gc=%s", "on"); 1567 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1568 seq_printf(seq, ",background_gc=%s", "off"); 1569 1570 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1571 seq_puts(seq, ",disable_roll_forward"); 1572 if (test_opt(sbi, NORECOVERY)) 1573 seq_puts(seq, ",norecovery"); 1574 if (test_opt(sbi, DISCARD)) 1575 seq_puts(seq, ",discard"); 1576 else 1577 seq_puts(seq, ",nodiscard"); 1578 if (test_opt(sbi, NOHEAP)) 1579 seq_puts(seq, ",no_heap"); 1580 else 1581 seq_puts(seq, ",heap"); 1582 #ifdef CONFIG_F2FS_FS_XATTR 1583 if (test_opt(sbi, XATTR_USER)) 1584 seq_puts(seq, ",user_xattr"); 1585 else 1586 seq_puts(seq, ",nouser_xattr"); 1587 if (test_opt(sbi, INLINE_XATTR)) 1588 seq_puts(seq, ",inline_xattr"); 1589 else 1590 seq_puts(seq, ",noinline_xattr"); 1591 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1592 seq_printf(seq, ",inline_xattr_size=%u", 1593 F2FS_OPTION(sbi).inline_xattr_size); 1594 #endif 1595 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1596 if (test_opt(sbi, POSIX_ACL)) 1597 seq_puts(seq, ",acl"); 1598 else 1599 seq_puts(seq, ",noacl"); 1600 #endif 1601 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1602 seq_puts(seq, ",disable_ext_identify"); 1603 if (test_opt(sbi, INLINE_DATA)) 1604 seq_puts(seq, ",inline_data"); 1605 else 1606 seq_puts(seq, ",noinline_data"); 1607 if (test_opt(sbi, INLINE_DENTRY)) 1608 seq_puts(seq, ",inline_dentry"); 1609 else 1610 seq_puts(seq, ",noinline_dentry"); 1611 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1612 seq_puts(seq, ",flush_merge"); 1613 if (test_opt(sbi, NOBARRIER)) 1614 seq_puts(seq, ",nobarrier"); 1615 if (test_opt(sbi, FASTBOOT)) 1616 seq_puts(seq, ",fastboot"); 1617 if (test_opt(sbi, EXTENT_CACHE)) 1618 seq_puts(seq, ",extent_cache"); 1619 else 1620 seq_puts(seq, ",noextent_cache"); 1621 if (test_opt(sbi, DATA_FLUSH)) 1622 seq_puts(seq, ",data_flush"); 1623 1624 seq_puts(seq, ",mode="); 1625 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1626 seq_puts(seq, "adaptive"); 1627 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1628 seq_puts(seq, "lfs"); 1629 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1630 if (test_opt(sbi, RESERVE_ROOT)) 1631 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1632 F2FS_OPTION(sbi).root_reserved_blocks, 1633 from_kuid_munged(&init_user_ns, 1634 F2FS_OPTION(sbi).s_resuid), 1635 from_kgid_munged(&init_user_ns, 1636 F2FS_OPTION(sbi).s_resgid)); 1637 if (F2FS_IO_SIZE_BITS(sbi)) 1638 seq_printf(seq, ",io_bits=%u", 1639 F2FS_OPTION(sbi).write_io_size_bits); 1640 #ifdef CONFIG_F2FS_FAULT_INJECTION 1641 if (test_opt(sbi, FAULT_INJECTION)) { 1642 seq_printf(seq, ",fault_injection=%u", 1643 F2FS_OPTION(sbi).fault_info.inject_rate); 1644 seq_printf(seq, ",fault_type=%u", 1645 F2FS_OPTION(sbi).fault_info.inject_type); 1646 } 1647 #endif 1648 #ifdef CONFIG_QUOTA 1649 if (test_opt(sbi, QUOTA)) 1650 seq_puts(seq, ",quota"); 1651 if (test_opt(sbi, USRQUOTA)) 1652 seq_puts(seq, ",usrquota"); 1653 if (test_opt(sbi, GRPQUOTA)) 1654 seq_puts(seq, ",grpquota"); 1655 if (test_opt(sbi, PRJQUOTA)) 1656 seq_puts(seq, ",prjquota"); 1657 #endif 1658 f2fs_show_quota_options(seq, sbi->sb); 1659 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1660 seq_printf(seq, ",whint_mode=%s", "user-based"); 1661 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1662 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1663 1664 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1665 1666 if (sbi->sb->s_flags & SB_INLINECRYPT) 1667 seq_puts(seq, ",inlinecrypt"); 1668 1669 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1670 seq_printf(seq, ",alloc_mode=%s", "default"); 1671 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1672 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1673 1674 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1675 seq_printf(seq, ",checkpoint=disable:%u", 1676 F2FS_OPTION(sbi).unusable_cap); 1677 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1678 seq_printf(seq, ",fsync_mode=%s", "posix"); 1679 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1680 seq_printf(seq, ",fsync_mode=%s", "strict"); 1681 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1682 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 1683 1684 #ifdef CONFIG_F2FS_FS_COMPRESSION 1685 f2fs_show_compress_options(seq, sbi->sb); 1686 #endif 1687 1688 if (test_opt(sbi, ATGC)) 1689 seq_puts(seq, ",atgc"); 1690 return 0; 1691 } 1692 1693 static void default_options(struct f2fs_sb_info *sbi) 1694 { 1695 /* init some FS parameters */ 1696 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 1697 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 1698 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1699 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1700 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1701 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 1702 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 1703 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 1704 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 1705 F2FS_OPTION(sbi).compress_ext_cnt = 0; 1706 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1707 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 1708 1709 sbi->sb->s_flags &= ~SB_INLINECRYPT; 1710 1711 set_opt(sbi, INLINE_XATTR); 1712 set_opt(sbi, INLINE_DATA); 1713 set_opt(sbi, INLINE_DENTRY); 1714 set_opt(sbi, EXTENT_CACHE); 1715 set_opt(sbi, NOHEAP); 1716 clear_opt(sbi, DISABLE_CHECKPOINT); 1717 F2FS_OPTION(sbi).unusable_cap = 0; 1718 sbi->sb->s_flags |= SB_LAZYTIME; 1719 set_opt(sbi, FLUSH_MERGE); 1720 set_opt(sbi, DISCARD); 1721 if (f2fs_sb_has_blkzoned(sbi)) 1722 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 1723 else 1724 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 1725 1726 #ifdef CONFIG_F2FS_FS_XATTR 1727 set_opt(sbi, XATTR_USER); 1728 #endif 1729 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1730 set_opt(sbi, POSIX_ACL); 1731 #endif 1732 1733 f2fs_build_fault_attr(sbi, 0, 0); 1734 } 1735 1736 #ifdef CONFIG_QUOTA 1737 static int f2fs_enable_quotas(struct super_block *sb); 1738 #endif 1739 1740 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 1741 { 1742 unsigned int s_flags = sbi->sb->s_flags; 1743 struct cp_control cpc; 1744 int err = 0; 1745 int ret; 1746 block_t unusable; 1747 1748 if (s_flags & SB_RDONLY) { 1749 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 1750 return -EINVAL; 1751 } 1752 sbi->sb->s_flags |= SB_ACTIVE; 1753 1754 f2fs_update_time(sbi, DISABLE_TIME); 1755 1756 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 1757 down_write(&sbi->gc_lock); 1758 err = f2fs_gc(sbi, true, false, NULL_SEGNO); 1759 if (err == -ENODATA) { 1760 err = 0; 1761 break; 1762 } 1763 if (err && err != -EAGAIN) 1764 break; 1765 } 1766 1767 ret = sync_filesystem(sbi->sb); 1768 if (ret || err) { 1769 err = ret ? ret: err; 1770 goto restore_flag; 1771 } 1772 1773 unusable = f2fs_get_unusable_blocks(sbi); 1774 if (f2fs_disable_cp_again(sbi, unusable)) { 1775 err = -EAGAIN; 1776 goto restore_flag; 1777 } 1778 1779 down_write(&sbi->gc_lock); 1780 cpc.reason = CP_PAUSE; 1781 set_sbi_flag(sbi, SBI_CP_DISABLED); 1782 err = f2fs_write_checkpoint(sbi, &cpc); 1783 if (err) 1784 goto out_unlock; 1785 1786 spin_lock(&sbi->stat_lock); 1787 sbi->unusable_block_count = unusable; 1788 spin_unlock(&sbi->stat_lock); 1789 1790 out_unlock: 1791 up_write(&sbi->gc_lock); 1792 restore_flag: 1793 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 1794 return err; 1795 } 1796 1797 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 1798 { 1799 down_write(&sbi->gc_lock); 1800 f2fs_dirty_to_prefree(sbi); 1801 1802 clear_sbi_flag(sbi, SBI_CP_DISABLED); 1803 set_sbi_flag(sbi, SBI_IS_DIRTY); 1804 up_write(&sbi->gc_lock); 1805 1806 f2fs_sync_fs(sbi->sb, 1); 1807 } 1808 1809 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 1810 { 1811 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1812 struct f2fs_mount_info org_mount_opt; 1813 unsigned long old_sb_flags; 1814 int err; 1815 bool need_restart_gc = false; 1816 bool need_stop_gc = false; 1817 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 1818 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT); 1819 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 1820 bool no_atgc = !test_opt(sbi, ATGC); 1821 bool checkpoint_changed; 1822 #ifdef CONFIG_QUOTA 1823 int i, j; 1824 #endif 1825 1826 /* 1827 * Save the old mount options in case we 1828 * need to restore them. 1829 */ 1830 org_mount_opt = sbi->mount_opt; 1831 old_sb_flags = sb->s_flags; 1832 1833 #ifdef CONFIG_QUOTA 1834 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 1835 for (i = 0; i < MAXQUOTAS; i++) { 1836 if (F2FS_OPTION(sbi).s_qf_names[i]) { 1837 org_mount_opt.s_qf_names[i] = 1838 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 1839 GFP_KERNEL); 1840 if (!org_mount_opt.s_qf_names[i]) { 1841 for (j = 0; j < i; j++) 1842 kfree(org_mount_opt.s_qf_names[j]); 1843 return -ENOMEM; 1844 } 1845 } else { 1846 org_mount_opt.s_qf_names[i] = NULL; 1847 } 1848 } 1849 #endif 1850 1851 /* recover superblocks we couldn't write due to previous RO mount */ 1852 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 1853 err = f2fs_commit_super(sbi, false); 1854 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 1855 err); 1856 if (!err) 1857 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1858 } 1859 1860 default_options(sbi); 1861 1862 /* parse mount options */ 1863 err = parse_options(sb, data, true); 1864 if (err) 1865 goto restore_opts; 1866 checkpoint_changed = 1867 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT); 1868 1869 /* 1870 * Previous and new state of filesystem is RO, 1871 * so skip checking GC and FLUSH_MERGE conditions. 1872 */ 1873 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 1874 goto skip; 1875 1876 #ifdef CONFIG_QUOTA 1877 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 1878 err = dquot_suspend(sb, -1); 1879 if (err < 0) 1880 goto restore_opts; 1881 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 1882 /* dquot_resume needs RW */ 1883 sb->s_flags &= ~SB_RDONLY; 1884 if (sb_any_quota_suspended(sb)) { 1885 dquot_resume(sb, -1); 1886 } else if (f2fs_sb_has_quota_ino(sbi)) { 1887 err = f2fs_enable_quotas(sb); 1888 if (err) 1889 goto restore_opts; 1890 } 1891 } 1892 #endif 1893 /* disallow enable atgc dynamically */ 1894 if (no_atgc == !!test_opt(sbi, ATGC)) { 1895 err = -EINVAL; 1896 f2fs_warn(sbi, "switch atgc option is not allowed"); 1897 goto restore_opts; 1898 } 1899 1900 /* disallow enable/disable extent_cache dynamically */ 1901 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 1902 err = -EINVAL; 1903 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 1904 goto restore_opts; 1905 } 1906 1907 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 1908 err = -EINVAL; 1909 f2fs_warn(sbi, "switch io_bits option is not allowed"); 1910 goto restore_opts; 1911 } 1912 1913 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 1914 err = -EINVAL; 1915 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 1916 goto restore_opts; 1917 } 1918 1919 /* 1920 * We stop the GC thread if FS is mounted as RO 1921 * or if background_gc = off is passed in mount 1922 * option. Also sync the filesystem. 1923 */ 1924 if ((*flags & SB_RDONLY) || 1925 F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) { 1926 if (sbi->gc_thread) { 1927 f2fs_stop_gc_thread(sbi); 1928 need_restart_gc = true; 1929 } 1930 } else if (!sbi->gc_thread) { 1931 err = f2fs_start_gc_thread(sbi); 1932 if (err) 1933 goto restore_opts; 1934 need_stop_gc = true; 1935 } 1936 1937 if (*flags & SB_RDONLY || 1938 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 1939 sync_inodes_sb(sb); 1940 1941 set_sbi_flag(sbi, SBI_IS_DIRTY); 1942 set_sbi_flag(sbi, SBI_IS_CLOSE); 1943 f2fs_sync_fs(sb, 1); 1944 clear_sbi_flag(sbi, SBI_IS_CLOSE); 1945 } 1946 1947 if (checkpoint_changed) { 1948 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 1949 err = f2fs_disable_checkpoint(sbi); 1950 if (err) 1951 goto restore_gc; 1952 } else { 1953 f2fs_enable_checkpoint(sbi); 1954 } 1955 } 1956 1957 /* 1958 * We stop issue flush thread if FS is mounted as RO 1959 * or if flush_merge is not passed in mount option. 1960 */ 1961 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 1962 clear_opt(sbi, FLUSH_MERGE); 1963 f2fs_destroy_flush_cmd_control(sbi, false); 1964 } else { 1965 err = f2fs_create_flush_cmd_control(sbi); 1966 if (err) 1967 goto restore_gc; 1968 } 1969 skip: 1970 #ifdef CONFIG_QUOTA 1971 /* Release old quota file names */ 1972 for (i = 0; i < MAXQUOTAS; i++) 1973 kfree(org_mount_opt.s_qf_names[i]); 1974 #endif 1975 /* Update the POSIXACL Flag */ 1976 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 1977 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 1978 1979 limit_reserve_root(sbi); 1980 adjust_unusable_cap_perc(sbi); 1981 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 1982 return 0; 1983 restore_gc: 1984 if (need_restart_gc) { 1985 if (f2fs_start_gc_thread(sbi)) 1986 f2fs_warn(sbi, "background gc thread has stopped"); 1987 } else if (need_stop_gc) { 1988 f2fs_stop_gc_thread(sbi); 1989 } 1990 restore_opts: 1991 #ifdef CONFIG_QUOTA 1992 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 1993 for (i = 0; i < MAXQUOTAS; i++) { 1994 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1995 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 1996 } 1997 #endif 1998 sbi->mount_opt = org_mount_opt; 1999 sb->s_flags = old_sb_flags; 2000 return err; 2001 } 2002 2003 #ifdef CONFIG_QUOTA 2004 /* Read data from quotafile */ 2005 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2006 size_t len, loff_t off) 2007 { 2008 struct inode *inode = sb_dqopt(sb)->files[type]; 2009 struct address_space *mapping = inode->i_mapping; 2010 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2011 int offset = off & (sb->s_blocksize - 1); 2012 int tocopy; 2013 size_t toread; 2014 loff_t i_size = i_size_read(inode); 2015 struct page *page; 2016 char *kaddr; 2017 2018 if (off > i_size) 2019 return 0; 2020 2021 if (off + len > i_size) 2022 len = i_size - off; 2023 toread = len; 2024 while (toread > 0) { 2025 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2026 repeat: 2027 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2028 if (IS_ERR(page)) { 2029 if (PTR_ERR(page) == -ENOMEM) { 2030 congestion_wait(BLK_RW_ASYNC, 2031 DEFAULT_IO_TIMEOUT); 2032 goto repeat; 2033 } 2034 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2035 return PTR_ERR(page); 2036 } 2037 2038 lock_page(page); 2039 2040 if (unlikely(page->mapping != mapping)) { 2041 f2fs_put_page(page, 1); 2042 goto repeat; 2043 } 2044 if (unlikely(!PageUptodate(page))) { 2045 f2fs_put_page(page, 1); 2046 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2047 return -EIO; 2048 } 2049 2050 kaddr = kmap_atomic(page); 2051 memcpy(data, kaddr + offset, tocopy); 2052 kunmap_atomic(kaddr); 2053 f2fs_put_page(page, 1); 2054 2055 offset = 0; 2056 toread -= tocopy; 2057 data += tocopy; 2058 blkidx++; 2059 } 2060 return len; 2061 } 2062 2063 /* Write to quotafile */ 2064 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2065 const char *data, size_t len, loff_t off) 2066 { 2067 struct inode *inode = sb_dqopt(sb)->files[type]; 2068 struct address_space *mapping = inode->i_mapping; 2069 const struct address_space_operations *a_ops = mapping->a_ops; 2070 int offset = off & (sb->s_blocksize - 1); 2071 size_t towrite = len; 2072 struct page *page; 2073 void *fsdata = NULL; 2074 char *kaddr; 2075 int err = 0; 2076 int tocopy; 2077 2078 while (towrite > 0) { 2079 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2080 towrite); 2081 retry: 2082 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2083 &page, &fsdata); 2084 if (unlikely(err)) { 2085 if (err == -ENOMEM) { 2086 congestion_wait(BLK_RW_ASYNC, 2087 DEFAULT_IO_TIMEOUT); 2088 goto retry; 2089 } 2090 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2091 break; 2092 } 2093 2094 kaddr = kmap_atomic(page); 2095 memcpy(kaddr + offset, data, tocopy); 2096 kunmap_atomic(kaddr); 2097 flush_dcache_page(page); 2098 2099 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2100 page, fsdata); 2101 offset = 0; 2102 towrite -= tocopy; 2103 off += tocopy; 2104 data += tocopy; 2105 cond_resched(); 2106 } 2107 2108 if (len == towrite) 2109 return err; 2110 inode->i_mtime = inode->i_ctime = current_time(inode); 2111 f2fs_mark_inode_dirty_sync(inode, false); 2112 return len - towrite; 2113 } 2114 2115 static struct dquot **f2fs_get_dquots(struct inode *inode) 2116 { 2117 return F2FS_I(inode)->i_dquot; 2118 } 2119 2120 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2121 { 2122 return &F2FS_I(inode)->i_reserved_quota; 2123 } 2124 2125 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2126 { 2127 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2128 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2129 return 0; 2130 } 2131 2132 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2133 F2FS_OPTION(sbi).s_jquota_fmt, type); 2134 } 2135 2136 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2137 { 2138 int enabled = 0; 2139 int i, err; 2140 2141 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2142 err = f2fs_enable_quotas(sbi->sb); 2143 if (err) { 2144 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2145 return 0; 2146 } 2147 return 1; 2148 } 2149 2150 for (i = 0; i < MAXQUOTAS; i++) { 2151 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2152 err = f2fs_quota_on_mount(sbi, i); 2153 if (!err) { 2154 enabled = 1; 2155 continue; 2156 } 2157 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2158 err, i); 2159 } 2160 } 2161 return enabled; 2162 } 2163 2164 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2165 unsigned int flags) 2166 { 2167 struct inode *qf_inode; 2168 unsigned long qf_inum; 2169 int err; 2170 2171 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2172 2173 qf_inum = f2fs_qf_ino(sb, type); 2174 if (!qf_inum) 2175 return -EPERM; 2176 2177 qf_inode = f2fs_iget(sb, qf_inum); 2178 if (IS_ERR(qf_inode)) { 2179 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2180 return PTR_ERR(qf_inode); 2181 } 2182 2183 /* Don't account quota for quota files to avoid recursion */ 2184 qf_inode->i_flags |= S_NOQUOTA; 2185 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2186 iput(qf_inode); 2187 return err; 2188 } 2189 2190 static int f2fs_enable_quotas(struct super_block *sb) 2191 { 2192 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2193 int type, err = 0; 2194 unsigned long qf_inum; 2195 bool quota_mopt[MAXQUOTAS] = { 2196 test_opt(sbi, USRQUOTA), 2197 test_opt(sbi, GRPQUOTA), 2198 test_opt(sbi, PRJQUOTA), 2199 }; 2200 2201 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2202 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2203 return 0; 2204 } 2205 2206 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2207 2208 for (type = 0; type < MAXQUOTAS; type++) { 2209 qf_inum = f2fs_qf_ino(sb, type); 2210 if (qf_inum) { 2211 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2212 DQUOT_USAGE_ENABLED | 2213 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2214 if (err) { 2215 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2216 type, err); 2217 for (type--; type >= 0; type--) 2218 dquot_quota_off(sb, type); 2219 set_sbi_flag(F2FS_SB(sb), 2220 SBI_QUOTA_NEED_REPAIR); 2221 return err; 2222 } 2223 } 2224 } 2225 return 0; 2226 } 2227 2228 int f2fs_quota_sync(struct super_block *sb, int type) 2229 { 2230 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2231 struct quota_info *dqopt = sb_dqopt(sb); 2232 int cnt; 2233 int ret; 2234 2235 /* 2236 * do_quotactl 2237 * f2fs_quota_sync 2238 * down_read(quota_sem) 2239 * dquot_writeback_dquots() 2240 * f2fs_dquot_commit 2241 * block_operation 2242 * down_read(quota_sem) 2243 */ 2244 f2fs_lock_op(sbi); 2245 2246 down_read(&sbi->quota_sem); 2247 ret = dquot_writeback_dquots(sb, type); 2248 if (ret) 2249 goto out; 2250 2251 /* 2252 * Now when everything is written we can discard the pagecache so 2253 * that userspace sees the changes. 2254 */ 2255 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2256 struct address_space *mapping; 2257 2258 if (type != -1 && cnt != type) 2259 continue; 2260 if (!sb_has_quota_active(sb, cnt)) 2261 continue; 2262 2263 mapping = dqopt->files[cnt]->i_mapping; 2264 2265 ret = filemap_fdatawrite(mapping); 2266 if (ret) 2267 goto out; 2268 2269 /* if we are using journalled quota */ 2270 if (is_journalled_quota(sbi)) 2271 continue; 2272 2273 ret = filemap_fdatawait(mapping); 2274 if (ret) 2275 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2276 2277 inode_lock(dqopt->files[cnt]); 2278 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0); 2279 inode_unlock(dqopt->files[cnt]); 2280 } 2281 out: 2282 if (ret) 2283 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2284 up_read(&sbi->quota_sem); 2285 f2fs_unlock_op(sbi); 2286 return ret; 2287 } 2288 2289 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2290 const struct path *path) 2291 { 2292 struct inode *inode; 2293 int err; 2294 2295 /* if quota sysfile exists, deny enabling quota with specific file */ 2296 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2297 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2298 return -EBUSY; 2299 } 2300 2301 err = f2fs_quota_sync(sb, type); 2302 if (err) 2303 return err; 2304 2305 err = dquot_quota_on(sb, type, format_id, path); 2306 if (err) 2307 return err; 2308 2309 inode = d_inode(path->dentry); 2310 2311 inode_lock(inode); 2312 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2313 f2fs_set_inode_flags(inode); 2314 inode_unlock(inode); 2315 f2fs_mark_inode_dirty_sync(inode, false); 2316 2317 return 0; 2318 } 2319 2320 static int __f2fs_quota_off(struct super_block *sb, int type) 2321 { 2322 struct inode *inode = sb_dqopt(sb)->files[type]; 2323 int err; 2324 2325 if (!inode || !igrab(inode)) 2326 return dquot_quota_off(sb, type); 2327 2328 err = f2fs_quota_sync(sb, type); 2329 if (err) 2330 goto out_put; 2331 2332 err = dquot_quota_off(sb, type); 2333 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2334 goto out_put; 2335 2336 inode_lock(inode); 2337 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2338 f2fs_set_inode_flags(inode); 2339 inode_unlock(inode); 2340 f2fs_mark_inode_dirty_sync(inode, false); 2341 out_put: 2342 iput(inode); 2343 return err; 2344 } 2345 2346 static int f2fs_quota_off(struct super_block *sb, int type) 2347 { 2348 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2349 int err; 2350 2351 err = __f2fs_quota_off(sb, type); 2352 2353 /* 2354 * quotactl can shutdown journalled quota, result in inconsistence 2355 * between quota record and fs data by following updates, tag the 2356 * flag to let fsck be aware of it. 2357 */ 2358 if (is_journalled_quota(sbi)) 2359 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2360 return err; 2361 } 2362 2363 void f2fs_quota_off_umount(struct super_block *sb) 2364 { 2365 int type; 2366 int err; 2367 2368 for (type = 0; type < MAXQUOTAS; type++) { 2369 err = __f2fs_quota_off(sb, type); 2370 if (err) { 2371 int ret = dquot_quota_off(sb, type); 2372 2373 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2374 type, err, ret); 2375 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2376 } 2377 } 2378 /* 2379 * In case of checkpoint=disable, we must flush quota blocks. 2380 * This can cause NULL exception for node_inode in end_io, since 2381 * put_super already dropped it. 2382 */ 2383 sync_filesystem(sb); 2384 } 2385 2386 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2387 { 2388 struct quota_info *dqopt = sb_dqopt(sb); 2389 int type; 2390 2391 for (type = 0; type < MAXQUOTAS; type++) { 2392 if (!dqopt->files[type]) 2393 continue; 2394 f2fs_inode_synced(dqopt->files[type]); 2395 } 2396 } 2397 2398 static int f2fs_dquot_commit(struct dquot *dquot) 2399 { 2400 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2401 int ret; 2402 2403 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2404 ret = dquot_commit(dquot); 2405 if (ret < 0) 2406 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2407 up_read(&sbi->quota_sem); 2408 return ret; 2409 } 2410 2411 static int f2fs_dquot_acquire(struct dquot *dquot) 2412 { 2413 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2414 int ret; 2415 2416 down_read(&sbi->quota_sem); 2417 ret = dquot_acquire(dquot); 2418 if (ret < 0) 2419 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2420 up_read(&sbi->quota_sem); 2421 return ret; 2422 } 2423 2424 static int f2fs_dquot_release(struct dquot *dquot) 2425 { 2426 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2427 int ret = dquot_release(dquot); 2428 2429 if (ret < 0) 2430 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2431 return ret; 2432 } 2433 2434 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2435 { 2436 struct super_block *sb = dquot->dq_sb; 2437 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2438 int ret = dquot_mark_dquot_dirty(dquot); 2439 2440 /* if we are using journalled quota */ 2441 if (is_journalled_quota(sbi)) 2442 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2443 2444 return ret; 2445 } 2446 2447 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2448 { 2449 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2450 int ret = dquot_commit_info(sb, type); 2451 2452 if (ret < 0) 2453 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2454 return ret; 2455 } 2456 2457 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2458 { 2459 *projid = F2FS_I(inode)->i_projid; 2460 return 0; 2461 } 2462 2463 static const struct dquot_operations f2fs_quota_operations = { 2464 .get_reserved_space = f2fs_get_reserved_space, 2465 .write_dquot = f2fs_dquot_commit, 2466 .acquire_dquot = f2fs_dquot_acquire, 2467 .release_dquot = f2fs_dquot_release, 2468 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2469 .write_info = f2fs_dquot_commit_info, 2470 .alloc_dquot = dquot_alloc, 2471 .destroy_dquot = dquot_destroy, 2472 .get_projid = f2fs_get_projid, 2473 .get_next_id = dquot_get_next_id, 2474 }; 2475 2476 static const struct quotactl_ops f2fs_quotactl_ops = { 2477 .quota_on = f2fs_quota_on, 2478 .quota_off = f2fs_quota_off, 2479 .quota_sync = f2fs_quota_sync, 2480 .get_state = dquot_get_state, 2481 .set_info = dquot_set_dqinfo, 2482 .get_dqblk = dquot_get_dqblk, 2483 .set_dqblk = dquot_set_dqblk, 2484 .get_nextdqblk = dquot_get_next_dqblk, 2485 }; 2486 #else 2487 int f2fs_quota_sync(struct super_block *sb, int type) 2488 { 2489 return 0; 2490 } 2491 2492 void f2fs_quota_off_umount(struct super_block *sb) 2493 { 2494 } 2495 #endif 2496 2497 static const struct super_operations f2fs_sops = { 2498 .alloc_inode = f2fs_alloc_inode, 2499 .free_inode = f2fs_free_inode, 2500 .drop_inode = f2fs_drop_inode, 2501 .write_inode = f2fs_write_inode, 2502 .dirty_inode = f2fs_dirty_inode, 2503 .show_options = f2fs_show_options, 2504 #ifdef CONFIG_QUOTA 2505 .quota_read = f2fs_quota_read, 2506 .quota_write = f2fs_quota_write, 2507 .get_dquots = f2fs_get_dquots, 2508 #endif 2509 .evict_inode = f2fs_evict_inode, 2510 .put_super = f2fs_put_super, 2511 .sync_fs = f2fs_sync_fs, 2512 .freeze_fs = f2fs_freeze, 2513 .unfreeze_fs = f2fs_unfreeze, 2514 .statfs = f2fs_statfs, 2515 .remount_fs = f2fs_remount, 2516 }; 2517 2518 #ifdef CONFIG_FS_ENCRYPTION 2519 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2520 { 2521 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2522 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2523 ctx, len, NULL); 2524 } 2525 2526 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2527 void *fs_data) 2528 { 2529 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2530 2531 /* 2532 * Encrypting the root directory is not allowed because fsck 2533 * expects lost+found directory to exist and remain unencrypted 2534 * if LOST_FOUND feature is enabled. 2535 * 2536 */ 2537 if (f2fs_sb_has_lost_found(sbi) && 2538 inode->i_ino == F2FS_ROOT_INO(sbi)) 2539 return -EPERM; 2540 2541 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2542 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2543 ctx, len, fs_data, XATTR_CREATE); 2544 } 2545 2546 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 2547 { 2548 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 2549 } 2550 2551 static bool f2fs_has_stable_inodes(struct super_block *sb) 2552 { 2553 return true; 2554 } 2555 2556 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 2557 int *ino_bits_ret, int *lblk_bits_ret) 2558 { 2559 *ino_bits_ret = 8 * sizeof(nid_t); 2560 *lblk_bits_ret = 8 * sizeof(block_t); 2561 } 2562 2563 static int f2fs_get_num_devices(struct super_block *sb) 2564 { 2565 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2566 2567 if (f2fs_is_multi_device(sbi)) 2568 return sbi->s_ndevs; 2569 return 1; 2570 } 2571 2572 static void f2fs_get_devices(struct super_block *sb, 2573 struct request_queue **devs) 2574 { 2575 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2576 int i; 2577 2578 for (i = 0; i < sbi->s_ndevs; i++) 2579 devs[i] = bdev_get_queue(FDEV(i).bdev); 2580 } 2581 2582 static const struct fscrypt_operations f2fs_cryptops = { 2583 .key_prefix = "f2fs:", 2584 .get_context = f2fs_get_context, 2585 .set_context = f2fs_set_context, 2586 .get_dummy_policy = f2fs_get_dummy_policy, 2587 .empty_dir = f2fs_empty_dir, 2588 .max_namelen = F2FS_NAME_LEN, 2589 .has_stable_inodes = f2fs_has_stable_inodes, 2590 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 2591 .get_num_devices = f2fs_get_num_devices, 2592 .get_devices = f2fs_get_devices, 2593 }; 2594 #endif 2595 2596 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 2597 u64 ino, u32 generation) 2598 { 2599 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2600 struct inode *inode; 2601 2602 if (f2fs_check_nid_range(sbi, ino)) 2603 return ERR_PTR(-ESTALE); 2604 2605 /* 2606 * f2fs_iget isn't quite right if the inode is currently unallocated! 2607 * However f2fs_iget currently does appropriate checks to handle stale 2608 * inodes so everything is OK. 2609 */ 2610 inode = f2fs_iget(sb, ino); 2611 if (IS_ERR(inode)) 2612 return ERR_CAST(inode); 2613 if (unlikely(generation && inode->i_generation != generation)) { 2614 /* we didn't find the right inode.. */ 2615 iput(inode); 2616 return ERR_PTR(-ESTALE); 2617 } 2618 return inode; 2619 } 2620 2621 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 2622 int fh_len, int fh_type) 2623 { 2624 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 2625 f2fs_nfs_get_inode); 2626 } 2627 2628 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 2629 int fh_len, int fh_type) 2630 { 2631 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 2632 f2fs_nfs_get_inode); 2633 } 2634 2635 static const struct export_operations f2fs_export_ops = { 2636 .fh_to_dentry = f2fs_fh_to_dentry, 2637 .fh_to_parent = f2fs_fh_to_parent, 2638 .get_parent = f2fs_get_parent, 2639 }; 2640 2641 static loff_t max_file_blocks(void) 2642 { 2643 loff_t result = 0; 2644 loff_t leaf_count = DEF_ADDRS_PER_BLOCK; 2645 2646 /* 2647 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 2648 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 2649 * space in inode.i_addr, it will be more safe to reassign 2650 * result as zero. 2651 */ 2652 2653 /* two direct node blocks */ 2654 result += (leaf_count * 2); 2655 2656 /* two indirect node blocks */ 2657 leaf_count *= NIDS_PER_BLOCK; 2658 result += (leaf_count * 2); 2659 2660 /* one double indirect node block */ 2661 leaf_count *= NIDS_PER_BLOCK; 2662 result += leaf_count; 2663 2664 return result; 2665 } 2666 2667 static int __f2fs_commit_super(struct buffer_head *bh, 2668 struct f2fs_super_block *super) 2669 { 2670 lock_buffer(bh); 2671 if (super) 2672 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 2673 set_buffer_dirty(bh); 2674 unlock_buffer(bh); 2675 2676 /* it's rare case, we can do fua all the time */ 2677 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 2678 } 2679 2680 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 2681 struct buffer_head *bh) 2682 { 2683 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2684 (bh->b_data + F2FS_SUPER_OFFSET); 2685 struct super_block *sb = sbi->sb; 2686 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 2687 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 2688 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 2689 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 2690 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 2691 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 2692 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 2693 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 2694 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 2695 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 2696 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2697 u32 segment_count = le32_to_cpu(raw_super->segment_count); 2698 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2699 u64 main_end_blkaddr = main_blkaddr + 2700 (segment_count_main << log_blocks_per_seg); 2701 u64 seg_end_blkaddr = segment0_blkaddr + 2702 (segment_count << log_blocks_per_seg); 2703 2704 if (segment0_blkaddr != cp_blkaddr) { 2705 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 2706 segment0_blkaddr, cp_blkaddr); 2707 return true; 2708 } 2709 2710 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 2711 sit_blkaddr) { 2712 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 2713 cp_blkaddr, sit_blkaddr, 2714 segment_count_ckpt << log_blocks_per_seg); 2715 return true; 2716 } 2717 2718 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 2719 nat_blkaddr) { 2720 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 2721 sit_blkaddr, nat_blkaddr, 2722 segment_count_sit << log_blocks_per_seg); 2723 return true; 2724 } 2725 2726 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 2727 ssa_blkaddr) { 2728 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 2729 nat_blkaddr, ssa_blkaddr, 2730 segment_count_nat << log_blocks_per_seg); 2731 return true; 2732 } 2733 2734 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 2735 main_blkaddr) { 2736 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 2737 ssa_blkaddr, main_blkaddr, 2738 segment_count_ssa << log_blocks_per_seg); 2739 return true; 2740 } 2741 2742 if (main_end_blkaddr > seg_end_blkaddr) { 2743 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 2744 main_blkaddr, seg_end_blkaddr, 2745 segment_count_main << log_blocks_per_seg); 2746 return true; 2747 } else if (main_end_blkaddr < seg_end_blkaddr) { 2748 int err = 0; 2749 char *res; 2750 2751 /* fix in-memory information all the time */ 2752 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 2753 segment0_blkaddr) >> log_blocks_per_seg); 2754 2755 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 2756 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2757 res = "internally"; 2758 } else { 2759 err = __f2fs_commit_super(bh, NULL); 2760 res = err ? "failed" : "done"; 2761 } 2762 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 2763 res, main_blkaddr, seg_end_blkaddr, 2764 segment_count_main << log_blocks_per_seg); 2765 if (err) 2766 return true; 2767 } 2768 return false; 2769 } 2770 2771 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 2772 struct buffer_head *bh) 2773 { 2774 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 2775 block_t total_sections, blocks_per_seg; 2776 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2777 (bh->b_data + F2FS_SUPER_OFFSET); 2778 size_t crc_offset = 0; 2779 __u32 crc = 0; 2780 2781 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 2782 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 2783 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 2784 return -EINVAL; 2785 } 2786 2787 /* Check checksum_offset and crc in superblock */ 2788 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 2789 crc_offset = le32_to_cpu(raw_super->checksum_offset); 2790 if (crc_offset != 2791 offsetof(struct f2fs_super_block, crc)) { 2792 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 2793 crc_offset); 2794 return -EFSCORRUPTED; 2795 } 2796 crc = le32_to_cpu(raw_super->crc); 2797 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 2798 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 2799 return -EFSCORRUPTED; 2800 } 2801 } 2802 2803 /* Currently, support only 4KB block size */ 2804 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 2805 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 2806 le32_to_cpu(raw_super->log_blocksize), 2807 F2FS_BLKSIZE_BITS); 2808 return -EFSCORRUPTED; 2809 } 2810 2811 /* check log blocks per segment */ 2812 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 2813 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 2814 le32_to_cpu(raw_super->log_blocks_per_seg)); 2815 return -EFSCORRUPTED; 2816 } 2817 2818 /* Currently, support 512/1024/2048/4096 bytes sector size */ 2819 if (le32_to_cpu(raw_super->log_sectorsize) > 2820 F2FS_MAX_LOG_SECTOR_SIZE || 2821 le32_to_cpu(raw_super->log_sectorsize) < 2822 F2FS_MIN_LOG_SECTOR_SIZE) { 2823 f2fs_info(sbi, "Invalid log sectorsize (%u)", 2824 le32_to_cpu(raw_super->log_sectorsize)); 2825 return -EFSCORRUPTED; 2826 } 2827 if (le32_to_cpu(raw_super->log_sectors_per_block) + 2828 le32_to_cpu(raw_super->log_sectorsize) != 2829 F2FS_MAX_LOG_SECTOR_SIZE) { 2830 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 2831 le32_to_cpu(raw_super->log_sectors_per_block), 2832 le32_to_cpu(raw_super->log_sectorsize)); 2833 return -EFSCORRUPTED; 2834 } 2835 2836 segment_count = le32_to_cpu(raw_super->segment_count); 2837 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2838 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 2839 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 2840 total_sections = le32_to_cpu(raw_super->section_count); 2841 2842 /* blocks_per_seg should be 512, given the above check */ 2843 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 2844 2845 if (segment_count > F2FS_MAX_SEGMENT || 2846 segment_count < F2FS_MIN_SEGMENTS) { 2847 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 2848 return -EFSCORRUPTED; 2849 } 2850 2851 if (total_sections > segment_count_main || total_sections < 1 || 2852 segs_per_sec > segment_count || !segs_per_sec) { 2853 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 2854 segment_count, total_sections, segs_per_sec); 2855 return -EFSCORRUPTED; 2856 } 2857 2858 if (segment_count_main != total_sections * segs_per_sec) { 2859 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 2860 segment_count_main, total_sections, segs_per_sec); 2861 return -EFSCORRUPTED; 2862 } 2863 2864 if ((segment_count / segs_per_sec) < total_sections) { 2865 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 2866 segment_count, segs_per_sec, total_sections); 2867 return -EFSCORRUPTED; 2868 } 2869 2870 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 2871 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 2872 segment_count, le64_to_cpu(raw_super->block_count)); 2873 return -EFSCORRUPTED; 2874 } 2875 2876 if (RDEV(0).path[0]) { 2877 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 2878 int i = 1; 2879 2880 while (i < MAX_DEVICES && RDEV(i).path[0]) { 2881 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 2882 i++; 2883 } 2884 if (segment_count != dev_seg_count) { 2885 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 2886 segment_count, dev_seg_count); 2887 return -EFSCORRUPTED; 2888 } 2889 } else { 2890 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 2891 !bdev_is_zoned(sbi->sb->s_bdev)) { 2892 f2fs_info(sbi, "Zoned block device path is missing"); 2893 return -EFSCORRUPTED; 2894 } 2895 } 2896 2897 if (secs_per_zone > total_sections || !secs_per_zone) { 2898 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 2899 secs_per_zone, total_sections); 2900 return -EFSCORRUPTED; 2901 } 2902 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 2903 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 2904 (le32_to_cpu(raw_super->extension_count) + 2905 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 2906 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 2907 le32_to_cpu(raw_super->extension_count), 2908 raw_super->hot_ext_count, 2909 F2FS_MAX_EXTENSION); 2910 return -EFSCORRUPTED; 2911 } 2912 2913 if (le32_to_cpu(raw_super->cp_payload) > 2914 (blocks_per_seg - F2FS_CP_PACKS)) { 2915 f2fs_info(sbi, "Insane cp_payload (%u > %u)", 2916 le32_to_cpu(raw_super->cp_payload), 2917 blocks_per_seg - F2FS_CP_PACKS); 2918 return -EFSCORRUPTED; 2919 } 2920 2921 /* check reserved ino info */ 2922 if (le32_to_cpu(raw_super->node_ino) != 1 || 2923 le32_to_cpu(raw_super->meta_ino) != 2 || 2924 le32_to_cpu(raw_super->root_ino) != 3) { 2925 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 2926 le32_to_cpu(raw_super->node_ino), 2927 le32_to_cpu(raw_super->meta_ino), 2928 le32_to_cpu(raw_super->root_ino)); 2929 return -EFSCORRUPTED; 2930 } 2931 2932 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 2933 if (sanity_check_area_boundary(sbi, bh)) 2934 return -EFSCORRUPTED; 2935 2936 return 0; 2937 } 2938 2939 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 2940 { 2941 unsigned int total, fsmeta; 2942 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 2943 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 2944 unsigned int ovp_segments, reserved_segments; 2945 unsigned int main_segs, blocks_per_seg; 2946 unsigned int sit_segs, nat_segs; 2947 unsigned int sit_bitmap_size, nat_bitmap_size; 2948 unsigned int log_blocks_per_seg; 2949 unsigned int segment_count_main; 2950 unsigned int cp_pack_start_sum, cp_payload; 2951 block_t user_block_count, valid_user_blocks; 2952 block_t avail_node_count, valid_node_count; 2953 int i, j; 2954 2955 total = le32_to_cpu(raw_super->segment_count); 2956 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 2957 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 2958 fsmeta += sit_segs; 2959 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 2960 fsmeta += nat_segs; 2961 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 2962 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 2963 2964 if (unlikely(fsmeta >= total)) 2965 return 1; 2966 2967 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 2968 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 2969 2970 if (unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 2971 ovp_segments == 0 || reserved_segments == 0)) { 2972 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 2973 return 1; 2974 } 2975 2976 user_block_count = le64_to_cpu(ckpt->user_block_count); 2977 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2978 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2979 if (!user_block_count || user_block_count >= 2980 segment_count_main << log_blocks_per_seg) { 2981 f2fs_err(sbi, "Wrong user_block_count: %u", 2982 user_block_count); 2983 return 1; 2984 } 2985 2986 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 2987 if (valid_user_blocks > user_block_count) { 2988 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 2989 valid_user_blocks, user_block_count); 2990 return 1; 2991 } 2992 2993 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 2994 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 2995 if (valid_node_count > avail_node_count) { 2996 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 2997 valid_node_count, avail_node_count); 2998 return 1; 2999 } 3000 3001 main_segs = le32_to_cpu(raw_super->segment_count_main); 3002 blocks_per_seg = sbi->blocks_per_seg; 3003 3004 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3005 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3006 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3007 return 1; 3008 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3009 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3010 le32_to_cpu(ckpt->cur_node_segno[j])) { 3011 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3012 i, j, 3013 le32_to_cpu(ckpt->cur_node_segno[i])); 3014 return 1; 3015 } 3016 } 3017 } 3018 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3019 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3020 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3021 return 1; 3022 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3023 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3024 le32_to_cpu(ckpt->cur_data_segno[j])) { 3025 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3026 i, j, 3027 le32_to_cpu(ckpt->cur_data_segno[i])); 3028 return 1; 3029 } 3030 } 3031 } 3032 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3033 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3034 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3035 le32_to_cpu(ckpt->cur_data_segno[j])) { 3036 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3037 i, j, 3038 le32_to_cpu(ckpt->cur_node_segno[i])); 3039 return 1; 3040 } 3041 } 3042 } 3043 3044 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3045 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3046 3047 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3048 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3049 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3050 sit_bitmap_size, nat_bitmap_size); 3051 return 1; 3052 } 3053 3054 cp_pack_start_sum = __start_sum_addr(sbi); 3055 cp_payload = __cp_payload(sbi); 3056 if (cp_pack_start_sum < cp_payload + 1 || 3057 cp_pack_start_sum > blocks_per_seg - 1 - 3058 NR_CURSEG_PERSIST_TYPE) { 3059 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3060 cp_pack_start_sum); 3061 return 1; 3062 } 3063 3064 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3065 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3066 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3067 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3068 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3069 le32_to_cpu(ckpt->checksum_offset)); 3070 return 1; 3071 } 3072 3073 if (unlikely(f2fs_cp_error(sbi))) { 3074 f2fs_err(sbi, "A bug case: need to run fsck"); 3075 return 1; 3076 } 3077 return 0; 3078 } 3079 3080 static void init_sb_info(struct f2fs_sb_info *sbi) 3081 { 3082 struct f2fs_super_block *raw_super = sbi->raw_super; 3083 int i; 3084 3085 sbi->log_sectors_per_block = 3086 le32_to_cpu(raw_super->log_sectors_per_block); 3087 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3088 sbi->blocksize = 1 << sbi->log_blocksize; 3089 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3090 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3091 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3092 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3093 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3094 sbi->total_node_count = 3095 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3096 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3097 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3098 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3099 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3100 sbi->cur_victim_sec = NULL_SECNO; 3101 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3102 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3103 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3104 sbi->migration_granularity = sbi->segs_per_sec; 3105 3106 sbi->dir_level = DEF_DIR_LEVEL; 3107 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3108 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3109 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3110 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3111 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3112 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3113 DEF_UMOUNT_DISCARD_TIMEOUT; 3114 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3115 3116 for (i = 0; i < NR_COUNT_TYPE; i++) 3117 atomic_set(&sbi->nr_pages[i], 0); 3118 3119 for (i = 0; i < META; i++) 3120 atomic_set(&sbi->wb_sync_req[i], 0); 3121 3122 INIT_LIST_HEAD(&sbi->s_list); 3123 mutex_init(&sbi->umount_mutex); 3124 init_rwsem(&sbi->io_order_lock); 3125 spin_lock_init(&sbi->cp_lock); 3126 3127 sbi->dirty_device = 0; 3128 spin_lock_init(&sbi->dev_lock); 3129 3130 init_rwsem(&sbi->sb_lock); 3131 init_rwsem(&sbi->pin_sem); 3132 } 3133 3134 static int init_percpu_info(struct f2fs_sb_info *sbi) 3135 { 3136 int err; 3137 3138 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3139 if (err) 3140 return err; 3141 3142 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3143 GFP_KERNEL); 3144 if (err) 3145 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3146 3147 return err; 3148 } 3149 3150 #ifdef CONFIG_BLK_DEV_ZONED 3151 3152 struct f2fs_report_zones_args { 3153 struct f2fs_dev_info *dev; 3154 bool zone_cap_mismatch; 3155 }; 3156 3157 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3158 void *data) 3159 { 3160 struct f2fs_report_zones_args *rz_args = data; 3161 3162 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3163 return 0; 3164 3165 set_bit(idx, rz_args->dev->blkz_seq); 3166 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >> 3167 F2FS_LOG_SECTORS_PER_BLOCK; 3168 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch) 3169 rz_args->zone_cap_mismatch = true; 3170 3171 return 0; 3172 } 3173 3174 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3175 { 3176 struct block_device *bdev = FDEV(devi).bdev; 3177 sector_t nr_sectors = bdev_nr_sectors(bdev); 3178 struct f2fs_report_zones_args rep_zone_arg; 3179 int ret; 3180 3181 if (!f2fs_sb_has_blkzoned(sbi)) 3182 return 0; 3183 3184 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3185 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 3186 return -EINVAL; 3187 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 3188 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3189 __ilog2_u32(sbi->blocks_per_blkz)) 3190 return -EINVAL; 3191 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3192 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3193 sbi->log_blocks_per_blkz; 3194 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 3195 FDEV(devi).nr_blkz++; 3196 3197 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3198 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3199 * sizeof(unsigned long), 3200 GFP_KERNEL); 3201 if (!FDEV(devi).blkz_seq) 3202 return -ENOMEM; 3203 3204 /* Get block zones type and zone-capacity */ 3205 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi, 3206 FDEV(devi).nr_blkz * sizeof(block_t), 3207 GFP_KERNEL); 3208 if (!FDEV(devi).zone_capacity_blocks) 3209 return -ENOMEM; 3210 3211 rep_zone_arg.dev = &FDEV(devi); 3212 rep_zone_arg.zone_cap_mismatch = false; 3213 3214 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3215 &rep_zone_arg); 3216 if (ret < 0) 3217 return ret; 3218 3219 if (!rep_zone_arg.zone_cap_mismatch) { 3220 kfree(FDEV(devi).zone_capacity_blocks); 3221 FDEV(devi).zone_capacity_blocks = NULL; 3222 } 3223 3224 return 0; 3225 } 3226 #endif 3227 3228 /* 3229 * Read f2fs raw super block. 3230 * Because we have two copies of super block, so read both of them 3231 * to get the first valid one. If any one of them is broken, we pass 3232 * them recovery flag back to the caller. 3233 */ 3234 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3235 struct f2fs_super_block **raw_super, 3236 int *valid_super_block, int *recovery) 3237 { 3238 struct super_block *sb = sbi->sb; 3239 int block; 3240 struct buffer_head *bh; 3241 struct f2fs_super_block *super; 3242 int err = 0; 3243 3244 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3245 if (!super) 3246 return -ENOMEM; 3247 3248 for (block = 0; block < 2; block++) { 3249 bh = sb_bread(sb, block); 3250 if (!bh) { 3251 f2fs_err(sbi, "Unable to read %dth superblock", 3252 block + 1); 3253 err = -EIO; 3254 *recovery = 1; 3255 continue; 3256 } 3257 3258 /* sanity checking of raw super */ 3259 err = sanity_check_raw_super(sbi, bh); 3260 if (err) { 3261 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3262 block + 1); 3263 brelse(bh); 3264 *recovery = 1; 3265 continue; 3266 } 3267 3268 if (!*raw_super) { 3269 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3270 sizeof(*super)); 3271 *valid_super_block = block; 3272 *raw_super = super; 3273 } 3274 brelse(bh); 3275 } 3276 3277 /* No valid superblock */ 3278 if (!*raw_super) 3279 kfree(super); 3280 else 3281 err = 0; 3282 3283 return err; 3284 } 3285 3286 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3287 { 3288 struct buffer_head *bh; 3289 __u32 crc = 0; 3290 int err; 3291 3292 if ((recover && f2fs_readonly(sbi->sb)) || 3293 bdev_read_only(sbi->sb->s_bdev)) { 3294 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3295 return -EROFS; 3296 } 3297 3298 /* we should update superblock crc here */ 3299 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3300 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3301 offsetof(struct f2fs_super_block, crc)); 3302 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3303 } 3304 3305 /* write back-up superblock first */ 3306 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3307 if (!bh) 3308 return -EIO; 3309 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3310 brelse(bh); 3311 3312 /* if we are in recovery path, skip writing valid superblock */ 3313 if (recover || err) 3314 return err; 3315 3316 /* write current valid superblock */ 3317 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3318 if (!bh) 3319 return -EIO; 3320 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3321 brelse(bh); 3322 return err; 3323 } 3324 3325 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3326 { 3327 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3328 unsigned int max_devices = MAX_DEVICES; 3329 int i; 3330 3331 /* Initialize single device information */ 3332 if (!RDEV(0).path[0]) { 3333 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3334 return 0; 3335 max_devices = 1; 3336 } 3337 3338 /* 3339 * Initialize multiple devices information, or single 3340 * zoned block device information. 3341 */ 3342 sbi->devs = f2fs_kzalloc(sbi, 3343 array_size(max_devices, 3344 sizeof(struct f2fs_dev_info)), 3345 GFP_KERNEL); 3346 if (!sbi->devs) 3347 return -ENOMEM; 3348 3349 for (i = 0; i < max_devices; i++) { 3350 3351 if (i > 0 && !RDEV(i).path[0]) 3352 break; 3353 3354 if (max_devices == 1) { 3355 /* Single zoned block device mount */ 3356 FDEV(0).bdev = 3357 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3358 sbi->sb->s_mode, sbi->sb->s_type); 3359 } else { 3360 /* Multi-device mount */ 3361 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3362 FDEV(i).total_segments = 3363 le32_to_cpu(RDEV(i).total_segments); 3364 if (i == 0) { 3365 FDEV(i).start_blk = 0; 3366 FDEV(i).end_blk = FDEV(i).start_blk + 3367 (FDEV(i).total_segments << 3368 sbi->log_blocks_per_seg) - 1 + 3369 le32_to_cpu(raw_super->segment0_blkaddr); 3370 } else { 3371 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3372 FDEV(i).end_blk = FDEV(i).start_blk + 3373 (FDEV(i).total_segments << 3374 sbi->log_blocks_per_seg) - 1; 3375 } 3376 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3377 sbi->sb->s_mode, sbi->sb->s_type); 3378 } 3379 if (IS_ERR(FDEV(i).bdev)) 3380 return PTR_ERR(FDEV(i).bdev); 3381 3382 /* to release errored devices */ 3383 sbi->s_ndevs = i + 1; 3384 3385 #ifdef CONFIG_BLK_DEV_ZONED 3386 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3387 !f2fs_sb_has_blkzoned(sbi)) { 3388 f2fs_err(sbi, "Zoned block device feature not enabled\n"); 3389 return -EINVAL; 3390 } 3391 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3392 if (init_blkz_info(sbi, i)) { 3393 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3394 return -EINVAL; 3395 } 3396 if (max_devices == 1) 3397 break; 3398 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3399 i, FDEV(i).path, 3400 FDEV(i).total_segments, 3401 FDEV(i).start_blk, FDEV(i).end_blk, 3402 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3403 "Host-aware" : "Host-managed"); 3404 continue; 3405 } 3406 #endif 3407 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3408 i, FDEV(i).path, 3409 FDEV(i).total_segments, 3410 FDEV(i).start_blk, FDEV(i).end_blk); 3411 } 3412 f2fs_info(sbi, 3413 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3414 return 0; 3415 } 3416 3417 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3418 { 3419 #ifdef CONFIG_UNICODE 3420 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 3421 const struct f2fs_sb_encodings *encoding_info; 3422 struct unicode_map *encoding; 3423 __u16 encoding_flags; 3424 3425 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info, 3426 &encoding_flags)) { 3427 f2fs_err(sbi, 3428 "Encoding requested by superblock is unknown"); 3429 return -EINVAL; 3430 } 3431 3432 encoding = utf8_load(encoding_info->version); 3433 if (IS_ERR(encoding)) { 3434 f2fs_err(sbi, 3435 "can't mount with superblock charset: %s-%s " 3436 "not supported by the kernel. flags: 0x%x.", 3437 encoding_info->name, encoding_info->version, 3438 encoding_flags); 3439 return PTR_ERR(encoding); 3440 } 3441 f2fs_info(sbi, "Using encoding defined by superblock: " 3442 "%s-%s with flags 0x%hx", encoding_info->name, 3443 encoding_info->version?:"\b", encoding_flags); 3444 3445 sbi->sb->s_encoding = encoding; 3446 sbi->sb->s_encoding_flags = encoding_flags; 3447 } 3448 #else 3449 if (f2fs_sb_has_casefold(sbi)) { 3450 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3451 return -EINVAL; 3452 } 3453 #endif 3454 return 0; 3455 } 3456 3457 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3458 { 3459 struct f2fs_sm_info *sm_i = SM_I(sbi); 3460 3461 /* adjust parameters according to the volume size */ 3462 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3463 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3464 sm_i->dcc_info->discard_granularity = 1; 3465 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 3466 } 3467 3468 sbi->readdir_ra = 1; 3469 } 3470 3471 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3472 { 3473 struct f2fs_sb_info *sbi; 3474 struct f2fs_super_block *raw_super; 3475 struct inode *root; 3476 int err; 3477 bool skip_recovery = false, need_fsck = false; 3478 char *options = NULL; 3479 int recovery, i, valid_super_block; 3480 struct curseg_info *seg_i; 3481 int retry_cnt = 1; 3482 3483 try_onemore: 3484 err = -EINVAL; 3485 raw_super = NULL; 3486 valid_super_block = -1; 3487 recovery = 0; 3488 3489 /* allocate memory for f2fs-specific super block info */ 3490 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 3491 if (!sbi) 3492 return -ENOMEM; 3493 3494 sbi->sb = sb; 3495 3496 /* Load the checksum driver */ 3497 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 3498 if (IS_ERR(sbi->s_chksum_driver)) { 3499 f2fs_err(sbi, "Cannot load crc32 driver."); 3500 err = PTR_ERR(sbi->s_chksum_driver); 3501 sbi->s_chksum_driver = NULL; 3502 goto free_sbi; 3503 } 3504 3505 /* set a block size */ 3506 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 3507 f2fs_err(sbi, "unable to set blocksize"); 3508 goto free_sbi; 3509 } 3510 3511 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 3512 &recovery); 3513 if (err) 3514 goto free_sbi; 3515 3516 sb->s_fs_info = sbi; 3517 sbi->raw_super = raw_super; 3518 3519 /* precompute checksum seed for metadata */ 3520 if (f2fs_sb_has_inode_chksum(sbi)) 3521 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 3522 sizeof(raw_super->uuid)); 3523 3524 default_options(sbi); 3525 /* parse mount options */ 3526 options = kstrdup((const char *)data, GFP_KERNEL); 3527 if (data && !options) { 3528 err = -ENOMEM; 3529 goto free_sb_buf; 3530 } 3531 3532 err = parse_options(sb, options, false); 3533 if (err) 3534 goto free_options; 3535 3536 sbi->max_file_blocks = max_file_blocks(); 3537 sb->s_maxbytes = sbi->max_file_blocks << 3538 le32_to_cpu(raw_super->log_blocksize); 3539 sb->s_max_links = F2FS_LINK_MAX; 3540 3541 err = f2fs_setup_casefold(sbi); 3542 if (err) 3543 goto free_options; 3544 3545 #ifdef CONFIG_QUOTA 3546 sb->dq_op = &f2fs_quota_operations; 3547 sb->s_qcop = &f2fs_quotactl_ops; 3548 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 3549 3550 if (f2fs_sb_has_quota_ino(sbi)) { 3551 for (i = 0; i < MAXQUOTAS; i++) { 3552 if (f2fs_qf_ino(sbi->sb, i)) 3553 sbi->nquota_files++; 3554 } 3555 } 3556 #endif 3557 3558 sb->s_op = &f2fs_sops; 3559 #ifdef CONFIG_FS_ENCRYPTION 3560 sb->s_cop = &f2fs_cryptops; 3561 #endif 3562 #ifdef CONFIG_FS_VERITY 3563 sb->s_vop = &f2fs_verityops; 3564 #endif 3565 sb->s_xattr = f2fs_xattr_handlers; 3566 sb->s_export_op = &f2fs_export_ops; 3567 sb->s_magic = F2FS_SUPER_MAGIC; 3568 sb->s_time_gran = 1; 3569 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3570 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 3571 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 3572 sb->s_iflags |= SB_I_CGROUPWB; 3573 3574 /* init f2fs-specific super block info */ 3575 sbi->valid_super_block = valid_super_block; 3576 init_rwsem(&sbi->gc_lock); 3577 mutex_init(&sbi->writepages); 3578 init_rwsem(&sbi->cp_global_sem); 3579 init_rwsem(&sbi->node_write); 3580 init_rwsem(&sbi->node_change); 3581 3582 /* disallow all the data/node/meta page writes */ 3583 set_sbi_flag(sbi, SBI_POR_DOING); 3584 spin_lock_init(&sbi->stat_lock); 3585 3586 /* init iostat info */ 3587 spin_lock_init(&sbi->iostat_lock); 3588 sbi->iostat_enable = false; 3589 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS; 3590 3591 for (i = 0; i < NR_PAGE_TYPE; i++) { 3592 int n = (i == META) ? 1: NR_TEMP_TYPE; 3593 int j; 3594 3595 sbi->write_io[i] = 3596 f2fs_kmalloc(sbi, 3597 array_size(n, 3598 sizeof(struct f2fs_bio_info)), 3599 GFP_KERNEL); 3600 if (!sbi->write_io[i]) { 3601 err = -ENOMEM; 3602 goto free_bio_info; 3603 } 3604 3605 for (j = HOT; j < n; j++) { 3606 init_rwsem(&sbi->write_io[i][j].io_rwsem); 3607 sbi->write_io[i][j].sbi = sbi; 3608 sbi->write_io[i][j].bio = NULL; 3609 spin_lock_init(&sbi->write_io[i][j].io_lock); 3610 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 3611 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); 3612 init_rwsem(&sbi->write_io[i][j].bio_list_lock); 3613 } 3614 } 3615 3616 init_rwsem(&sbi->cp_rwsem); 3617 init_rwsem(&sbi->quota_sem); 3618 init_waitqueue_head(&sbi->cp_wait); 3619 init_sb_info(sbi); 3620 3621 err = init_percpu_info(sbi); 3622 if (err) 3623 goto free_bio_info; 3624 3625 if (F2FS_IO_ALIGNED(sbi)) { 3626 sbi->write_io_dummy = 3627 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 3628 if (!sbi->write_io_dummy) { 3629 err = -ENOMEM; 3630 goto free_percpu; 3631 } 3632 } 3633 3634 /* init per sbi slab cache */ 3635 err = f2fs_init_xattr_caches(sbi); 3636 if (err) 3637 goto free_io_dummy; 3638 err = f2fs_init_page_array_cache(sbi); 3639 if (err) 3640 goto free_xattr_cache; 3641 3642 /* get an inode for meta space */ 3643 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 3644 if (IS_ERR(sbi->meta_inode)) { 3645 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 3646 err = PTR_ERR(sbi->meta_inode); 3647 goto free_page_array_cache; 3648 } 3649 3650 err = f2fs_get_valid_checkpoint(sbi); 3651 if (err) { 3652 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 3653 goto free_meta_inode; 3654 } 3655 3656 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 3657 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3658 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 3659 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3660 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 3661 } 3662 3663 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 3664 set_sbi_flag(sbi, SBI_NEED_FSCK); 3665 3666 /* Initialize device list */ 3667 err = f2fs_scan_devices(sbi); 3668 if (err) { 3669 f2fs_err(sbi, "Failed to find devices"); 3670 goto free_devices; 3671 } 3672 3673 err = f2fs_init_post_read_wq(sbi); 3674 if (err) { 3675 f2fs_err(sbi, "Failed to initialize post read workqueue"); 3676 goto free_devices; 3677 } 3678 3679 sbi->total_valid_node_count = 3680 le32_to_cpu(sbi->ckpt->valid_node_count); 3681 percpu_counter_set(&sbi->total_valid_inode_count, 3682 le32_to_cpu(sbi->ckpt->valid_inode_count)); 3683 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 3684 sbi->total_valid_block_count = 3685 le64_to_cpu(sbi->ckpt->valid_block_count); 3686 sbi->last_valid_block_count = sbi->total_valid_block_count; 3687 sbi->reserved_blocks = 0; 3688 sbi->current_reserved_blocks = 0; 3689 limit_reserve_root(sbi); 3690 adjust_unusable_cap_perc(sbi); 3691 3692 for (i = 0; i < NR_INODE_TYPE; i++) { 3693 INIT_LIST_HEAD(&sbi->inode_list[i]); 3694 spin_lock_init(&sbi->inode_lock[i]); 3695 } 3696 mutex_init(&sbi->flush_lock); 3697 3698 f2fs_init_extent_cache_info(sbi); 3699 3700 f2fs_init_ino_entry_info(sbi); 3701 3702 f2fs_init_fsync_node_info(sbi); 3703 3704 /* setup f2fs internal modules */ 3705 err = f2fs_build_segment_manager(sbi); 3706 if (err) { 3707 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 3708 err); 3709 goto free_sm; 3710 } 3711 err = f2fs_build_node_manager(sbi); 3712 if (err) { 3713 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 3714 err); 3715 goto free_nm; 3716 } 3717 3718 /* For write statistics */ 3719 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 3720 3721 /* Read accumulated write IO statistics if exists */ 3722 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 3723 if (__exist_node_summaries(sbi)) 3724 sbi->kbytes_written = 3725 le64_to_cpu(seg_i->journal->info.kbytes_written); 3726 3727 f2fs_build_gc_manager(sbi); 3728 3729 err = f2fs_build_stats(sbi); 3730 if (err) 3731 goto free_nm; 3732 3733 /* get an inode for node space */ 3734 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 3735 if (IS_ERR(sbi->node_inode)) { 3736 f2fs_err(sbi, "Failed to read node inode"); 3737 err = PTR_ERR(sbi->node_inode); 3738 goto free_stats; 3739 } 3740 3741 /* read root inode and dentry */ 3742 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 3743 if (IS_ERR(root)) { 3744 f2fs_err(sbi, "Failed to read root inode"); 3745 err = PTR_ERR(root); 3746 goto free_node_inode; 3747 } 3748 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 3749 !root->i_size || !root->i_nlink) { 3750 iput(root); 3751 err = -EINVAL; 3752 goto free_node_inode; 3753 } 3754 3755 sb->s_root = d_make_root(root); /* allocate root dentry */ 3756 if (!sb->s_root) { 3757 err = -ENOMEM; 3758 goto free_node_inode; 3759 } 3760 3761 err = f2fs_register_sysfs(sbi); 3762 if (err) 3763 goto free_root_inode; 3764 3765 #ifdef CONFIG_QUOTA 3766 /* Enable quota usage during mount */ 3767 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 3768 err = f2fs_enable_quotas(sb); 3769 if (err) 3770 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 3771 } 3772 #endif 3773 /* if there are any orphan inodes, free them */ 3774 err = f2fs_recover_orphan_inodes(sbi); 3775 if (err) 3776 goto free_meta; 3777 3778 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 3779 goto reset_checkpoint; 3780 3781 /* recover fsynced data */ 3782 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 3783 !test_opt(sbi, NORECOVERY)) { 3784 /* 3785 * mount should be failed, when device has readonly mode, and 3786 * previous checkpoint was not done by clean system shutdown. 3787 */ 3788 if (f2fs_hw_is_readonly(sbi)) { 3789 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3790 err = -EROFS; 3791 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable"); 3792 goto free_meta; 3793 } 3794 f2fs_info(sbi, "write access unavailable, skipping recovery"); 3795 goto reset_checkpoint; 3796 } 3797 3798 if (need_fsck) 3799 set_sbi_flag(sbi, SBI_NEED_FSCK); 3800 3801 if (skip_recovery) 3802 goto reset_checkpoint; 3803 3804 err = f2fs_recover_fsync_data(sbi, false); 3805 if (err < 0) { 3806 if (err != -ENOMEM) 3807 skip_recovery = true; 3808 need_fsck = true; 3809 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 3810 err); 3811 goto free_meta; 3812 } 3813 } else { 3814 err = f2fs_recover_fsync_data(sbi, true); 3815 3816 if (!f2fs_readonly(sb) && err > 0) { 3817 err = -EINVAL; 3818 f2fs_err(sbi, "Need to recover fsync data"); 3819 goto free_meta; 3820 } 3821 } 3822 3823 /* 3824 * If the f2fs is not readonly and fsync data recovery succeeds, 3825 * check zoned block devices' write pointer consistency. 3826 */ 3827 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 3828 err = f2fs_check_write_pointer(sbi); 3829 if (err) 3830 goto free_meta; 3831 } 3832 3833 reset_checkpoint: 3834 f2fs_init_inmem_curseg(sbi); 3835 3836 /* f2fs_recover_fsync_data() cleared this already */ 3837 clear_sbi_flag(sbi, SBI_POR_DOING); 3838 3839 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 3840 err = f2fs_disable_checkpoint(sbi); 3841 if (err) 3842 goto sync_free_meta; 3843 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 3844 f2fs_enable_checkpoint(sbi); 3845 } 3846 3847 /* 3848 * If filesystem is not mounted as read-only then 3849 * do start the gc_thread. 3850 */ 3851 if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) { 3852 /* After POR, we can run background GC thread.*/ 3853 err = f2fs_start_gc_thread(sbi); 3854 if (err) 3855 goto sync_free_meta; 3856 } 3857 kvfree(options); 3858 3859 /* recover broken superblock */ 3860 if (recovery) { 3861 err = f2fs_commit_super(sbi, true); 3862 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 3863 sbi->valid_super_block ? 1 : 2, err); 3864 } 3865 3866 f2fs_join_shrinker(sbi); 3867 3868 f2fs_tuning_parameters(sbi); 3869 3870 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 3871 cur_cp_version(F2FS_CKPT(sbi))); 3872 f2fs_update_time(sbi, CP_TIME); 3873 f2fs_update_time(sbi, REQ_TIME); 3874 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3875 return 0; 3876 3877 sync_free_meta: 3878 /* safe to flush all the data */ 3879 sync_filesystem(sbi->sb); 3880 retry_cnt = 0; 3881 3882 free_meta: 3883 #ifdef CONFIG_QUOTA 3884 f2fs_truncate_quota_inode_pages(sb); 3885 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 3886 f2fs_quota_off_umount(sbi->sb); 3887 #endif 3888 /* 3889 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 3890 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 3891 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 3892 * falls into an infinite loop in f2fs_sync_meta_pages(). 3893 */ 3894 truncate_inode_pages_final(META_MAPPING(sbi)); 3895 /* evict some inodes being cached by GC */ 3896 evict_inodes(sb); 3897 f2fs_unregister_sysfs(sbi); 3898 free_root_inode: 3899 dput(sb->s_root); 3900 sb->s_root = NULL; 3901 free_node_inode: 3902 f2fs_release_ino_entry(sbi, true); 3903 truncate_inode_pages_final(NODE_MAPPING(sbi)); 3904 iput(sbi->node_inode); 3905 sbi->node_inode = NULL; 3906 free_stats: 3907 f2fs_destroy_stats(sbi); 3908 free_nm: 3909 f2fs_destroy_node_manager(sbi); 3910 free_sm: 3911 f2fs_destroy_segment_manager(sbi); 3912 f2fs_destroy_post_read_wq(sbi); 3913 free_devices: 3914 destroy_device_list(sbi); 3915 kvfree(sbi->ckpt); 3916 free_meta_inode: 3917 make_bad_inode(sbi->meta_inode); 3918 iput(sbi->meta_inode); 3919 sbi->meta_inode = NULL; 3920 free_page_array_cache: 3921 f2fs_destroy_page_array_cache(sbi); 3922 free_xattr_cache: 3923 f2fs_destroy_xattr_caches(sbi); 3924 free_io_dummy: 3925 mempool_destroy(sbi->write_io_dummy); 3926 free_percpu: 3927 destroy_percpu_info(sbi); 3928 free_bio_info: 3929 for (i = 0; i < NR_PAGE_TYPE; i++) 3930 kvfree(sbi->write_io[i]); 3931 3932 #ifdef CONFIG_UNICODE 3933 utf8_unload(sb->s_encoding); 3934 sb->s_encoding = NULL; 3935 #endif 3936 free_options: 3937 #ifdef CONFIG_QUOTA 3938 for (i = 0; i < MAXQUOTAS; i++) 3939 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 3940 #endif 3941 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 3942 kvfree(options); 3943 free_sb_buf: 3944 kfree(raw_super); 3945 free_sbi: 3946 if (sbi->s_chksum_driver) 3947 crypto_free_shash(sbi->s_chksum_driver); 3948 kfree(sbi); 3949 3950 /* give only one another chance */ 3951 if (retry_cnt > 0 && skip_recovery) { 3952 retry_cnt--; 3953 shrink_dcache_sb(sb); 3954 goto try_onemore; 3955 } 3956 return err; 3957 } 3958 3959 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 3960 const char *dev_name, void *data) 3961 { 3962 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 3963 } 3964 3965 static void kill_f2fs_super(struct super_block *sb) 3966 { 3967 if (sb->s_root) { 3968 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3969 3970 set_sbi_flag(sbi, SBI_IS_CLOSE); 3971 f2fs_stop_gc_thread(sbi); 3972 f2fs_stop_discard_thread(sbi); 3973 3974 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 3975 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3976 struct cp_control cpc = { 3977 .reason = CP_UMOUNT, 3978 }; 3979 f2fs_write_checkpoint(sbi, &cpc); 3980 } 3981 3982 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 3983 sb->s_flags &= ~SB_RDONLY; 3984 } 3985 kill_block_super(sb); 3986 } 3987 3988 static struct file_system_type f2fs_fs_type = { 3989 .owner = THIS_MODULE, 3990 .name = "f2fs", 3991 .mount = f2fs_mount, 3992 .kill_sb = kill_f2fs_super, 3993 .fs_flags = FS_REQUIRES_DEV, 3994 }; 3995 MODULE_ALIAS_FS("f2fs"); 3996 3997 static int __init init_inodecache(void) 3998 { 3999 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4000 sizeof(struct f2fs_inode_info), 0, 4001 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4002 if (!f2fs_inode_cachep) 4003 return -ENOMEM; 4004 return 0; 4005 } 4006 4007 static void destroy_inodecache(void) 4008 { 4009 /* 4010 * Make sure all delayed rcu free inodes are flushed before we 4011 * destroy cache. 4012 */ 4013 rcu_barrier(); 4014 kmem_cache_destroy(f2fs_inode_cachep); 4015 } 4016 4017 static int __init init_f2fs_fs(void) 4018 { 4019 int err; 4020 4021 if (PAGE_SIZE != F2FS_BLKSIZE) { 4022 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4023 PAGE_SIZE, F2FS_BLKSIZE); 4024 return -EINVAL; 4025 } 4026 4027 f2fs_build_trace_ios(); 4028 4029 err = init_inodecache(); 4030 if (err) 4031 goto fail; 4032 err = f2fs_create_node_manager_caches(); 4033 if (err) 4034 goto free_inodecache; 4035 err = f2fs_create_segment_manager_caches(); 4036 if (err) 4037 goto free_node_manager_caches; 4038 err = f2fs_create_checkpoint_caches(); 4039 if (err) 4040 goto free_segment_manager_caches; 4041 err = f2fs_create_extent_cache(); 4042 if (err) 4043 goto free_checkpoint_caches; 4044 err = f2fs_create_garbage_collection_cache(); 4045 if (err) 4046 goto free_extent_cache; 4047 err = f2fs_init_sysfs(); 4048 if (err) 4049 goto free_garbage_collection_cache; 4050 err = register_shrinker(&f2fs_shrinker_info); 4051 if (err) 4052 goto free_sysfs; 4053 err = register_filesystem(&f2fs_fs_type); 4054 if (err) 4055 goto free_shrinker; 4056 f2fs_create_root_stats(); 4057 err = f2fs_init_post_read_processing(); 4058 if (err) 4059 goto free_root_stats; 4060 err = f2fs_init_bio_entry_cache(); 4061 if (err) 4062 goto free_post_read; 4063 err = f2fs_init_bioset(); 4064 if (err) 4065 goto free_bio_enrty_cache; 4066 err = f2fs_init_compress_mempool(); 4067 if (err) 4068 goto free_bioset; 4069 err = f2fs_init_compress_cache(); 4070 if (err) 4071 goto free_compress_mempool; 4072 return 0; 4073 free_compress_mempool: 4074 f2fs_destroy_compress_mempool(); 4075 free_bioset: 4076 f2fs_destroy_bioset(); 4077 free_bio_enrty_cache: 4078 f2fs_destroy_bio_entry_cache(); 4079 free_post_read: 4080 f2fs_destroy_post_read_processing(); 4081 free_root_stats: 4082 f2fs_destroy_root_stats(); 4083 unregister_filesystem(&f2fs_fs_type); 4084 free_shrinker: 4085 unregister_shrinker(&f2fs_shrinker_info); 4086 free_sysfs: 4087 f2fs_exit_sysfs(); 4088 free_garbage_collection_cache: 4089 f2fs_destroy_garbage_collection_cache(); 4090 free_extent_cache: 4091 f2fs_destroy_extent_cache(); 4092 free_checkpoint_caches: 4093 f2fs_destroy_checkpoint_caches(); 4094 free_segment_manager_caches: 4095 f2fs_destroy_segment_manager_caches(); 4096 free_node_manager_caches: 4097 f2fs_destroy_node_manager_caches(); 4098 free_inodecache: 4099 destroy_inodecache(); 4100 fail: 4101 return err; 4102 } 4103 4104 static void __exit exit_f2fs_fs(void) 4105 { 4106 f2fs_destroy_compress_cache(); 4107 f2fs_destroy_compress_mempool(); 4108 f2fs_destroy_bioset(); 4109 f2fs_destroy_bio_entry_cache(); 4110 f2fs_destroy_post_read_processing(); 4111 f2fs_destroy_root_stats(); 4112 unregister_filesystem(&f2fs_fs_type); 4113 unregister_shrinker(&f2fs_shrinker_info); 4114 f2fs_exit_sysfs(); 4115 f2fs_destroy_garbage_collection_cache(); 4116 f2fs_destroy_extent_cache(); 4117 f2fs_destroy_checkpoint_caches(); 4118 f2fs_destroy_segment_manager_caches(); 4119 f2fs_destroy_node_manager_caches(); 4120 destroy_inodecache(); 4121 f2fs_destroy_trace_ios(); 4122 } 4123 4124 module_init(init_f2fs_fs) 4125 module_exit(exit_f2fs_fs) 4126 4127 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4128 MODULE_DESCRIPTION("Flash Friendly File System"); 4129 MODULE_LICENSE("GPL"); 4130 4131