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_err, 150 }; 151 152 static match_table_t f2fs_tokens = { 153 {Opt_gc_background, "background_gc=%s"}, 154 {Opt_disable_roll_forward, "disable_roll_forward"}, 155 {Opt_norecovery, "norecovery"}, 156 {Opt_discard, "discard"}, 157 {Opt_nodiscard, "nodiscard"}, 158 {Opt_noheap, "no_heap"}, 159 {Opt_heap, "heap"}, 160 {Opt_user_xattr, "user_xattr"}, 161 {Opt_nouser_xattr, "nouser_xattr"}, 162 {Opt_acl, "acl"}, 163 {Opt_noacl, "noacl"}, 164 {Opt_active_logs, "active_logs=%u"}, 165 {Opt_disable_ext_identify, "disable_ext_identify"}, 166 {Opt_inline_xattr, "inline_xattr"}, 167 {Opt_noinline_xattr, "noinline_xattr"}, 168 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 169 {Opt_inline_data, "inline_data"}, 170 {Opt_inline_dentry, "inline_dentry"}, 171 {Opt_noinline_dentry, "noinline_dentry"}, 172 {Opt_flush_merge, "flush_merge"}, 173 {Opt_noflush_merge, "noflush_merge"}, 174 {Opt_nobarrier, "nobarrier"}, 175 {Opt_fastboot, "fastboot"}, 176 {Opt_extent_cache, "extent_cache"}, 177 {Opt_noextent_cache, "noextent_cache"}, 178 {Opt_noinline_data, "noinline_data"}, 179 {Opt_data_flush, "data_flush"}, 180 {Opt_reserve_root, "reserve_root=%u"}, 181 {Opt_resgid, "resgid=%u"}, 182 {Opt_resuid, "resuid=%u"}, 183 {Opt_mode, "mode=%s"}, 184 {Opt_io_size_bits, "io_bits=%u"}, 185 {Opt_fault_injection, "fault_injection=%u"}, 186 {Opt_fault_type, "fault_type=%u"}, 187 {Opt_lazytime, "lazytime"}, 188 {Opt_nolazytime, "nolazytime"}, 189 {Opt_quota, "quota"}, 190 {Opt_noquota, "noquota"}, 191 {Opt_usrquota, "usrquota"}, 192 {Opt_grpquota, "grpquota"}, 193 {Opt_prjquota, "prjquota"}, 194 {Opt_usrjquota, "usrjquota=%s"}, 195 {Opt_grpjquota, "grpjquota=%s"}, 196 {Opt_prjjquota, "prjjquota=%s"}, 197 {Opt_offusrjquota, "usrjquota="}, 198 {Opt_offgrpjquota, "grpjquota="}, 199 {Opt_offprjjquota, "prjjquota="}, 200 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 201 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 202 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 203 {Opt_whint, "whint_mode=%s"}, 204 {Opt_alloc, "alloc_mode=%s"}, 205 {Opt_fsync, "fsync_mode=%s"}, 206 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 207 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 208 {Opt_inlinecrypt, "inlinecrypt"}, 209 {Opt_checkpoint_disable, "checkpoint=disable"}, 210 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 211 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 212 {Opt_checkpoint_enable, "checkpoint=enable"}, 213 {Opt_compress_algorithm, "compress_algorithm=%s"}, 214 {Opt_compress_log_size, "compress_log_size=%u"}, 215 {Opt_compress_extension, "compress_extension=%s"}, 216 {Opt_err, NULL}, 217 }; 218 219 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 220 { 221 struct va_format vaf; 222 va_list args; 223 int level; 224 225 va_start(args, fmt); 226 227 level = printk_get_level(fmt); 228 vaf.fmt = printk_skip_level(fmt); 229 vaf.va = &args; 230 printk("%c%cF2FS-fs (%s): %pV\n", 231 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 232 233 va_end(args); 234 } 235 236 #ifdef CONFIG_UNICODE 237 static const struct f2fs_sb_encodings { 238 __u16 magic; 239 char *name; 240 char *version; 241 } f2fs_sb_encoding_map[] = { 242 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"}, 243 }; 244 245 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb, 246 const struct f2fs_sb_encodings **encoding, 247 __u16 *flags) 248 { 249 __u16 magic = le16_to_cpu(sb->s_encoding); 250 int i; 251 252 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 253 if (magic == f2fs_sb_encoding_map[i].magic) 254 break; 255 256 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map)) 257 return -EINVAL; 258 259 *encoding = &f2fs_sb_encoding_map[i]; 260 *flags = le16_to_cpu(sb->s_encoding_flags); 261 262 return 0; 263 } 264 #endif 265 266 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 267 { 268 block_t limit = min((sbi->user_block_count << 1) / 1000, 269 sbi->user_block_count - sbi->reserved_blocks); 270 271 /* limit is 0.2% */ 272 if (test_opt(sbi, RESERVE_ROOT) && 273 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 274 F2FS_OPTION(sbi).root_reserved_blocks = limit; 275 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 276 F2FS_OPTION(sbi).root_reserved_blocks); 277 } 278 if (!test_opt(sbi, RESERVE_ROOT) && 279 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 280 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 281 !gid_eq(F2FS_OPTION(sbi).s_resgid, 282 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 283 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 284 from_kuid_munged(&init_user_ns, 285 F2FS_OPTION(sbi).s_resuid), 286 from_kgid_munged(&init_user_ns, 287 F2FS_OPTION(sbi).s_resgid)); 288 } 289 290 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 291 { 292 if (!F2FS_OPTION(sbi).unusable_cap_perc) 293 return; 294 295 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 296 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 297 else 298 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 299 F2FS_OPTION(sbi).unusable_cap_perc; 300 301 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 302 F2FS_OPTION(sbi).unusable_cap, 303 F2FS_OPTION(sbi).unusable_cap_perc); 304 } 305 306 static void init_once(void *foo) 307 { 308 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 309 310 inode_init_once(&fi->vfs_inode); 311 } 312 313 #ifdef CONFIG_QUOTA 314 static const char * const quotatypes[] = INITQFNAMES; 315 #define QTYPE2NAME(t) (quotatypes[t]) 316 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 317 substring_t *args) 318 { 319 struct f2fs_sb_info *sbi = F2FS_SB(sb); 320 char *qname; 321 int ret = -EINVAL; 322 323 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 324 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 325 return -EINVAL; 326 } 327 if (f2fs_sb_has_quota_ino(sbi)) { 328 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 329 return 0; 330 } 331 332 qname = match_strdup(args); 333 if (!qname) { 334 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 335 return -ENOMEM; 336 } 337 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 338 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 339 ret = 0; 340 else 341 f2fs_err(sbi, "%s quota file already specified", 342 QTYPE2NAME(qtype)); 343 goto errout; 344 } 345 if (strchr(qname, '/')) { 346 f2fs_err(sbi, "quotafile must be on filesystem root"); 347 goto errout; 348 } 349 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 350 set_opt(sbi, QUOTA); 351 return 0; 352 errout: 353 kfree(qname); 354 return ret; 355 } 356 357 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 358 { 359 struct f2fs_sb_info *sbi = F2FS_SB(sb); 360 361 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 362 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 363 return -EINVAL; 364 } 365 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 366 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 367 return 0; 368 } 369 370 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 371 { 372 /* 373 * We do the test below only for project quotas. 'usrquota' and 374 * 'grpquota' mount options are allowed even without quota feature 375 * to support legacy quotas in quota files. 376 */ 377 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 378 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 379 return -1; 380 } 381 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 382 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 383 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 384 if (test_opt(sbi, USRQUOTA) && 385 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 386 clear_opt(sbi, USRQUOTA); 387 388 if (test_opt(sbi, GRPQUOTA) && 389 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 390 clear_opt(sbi, GRPQUOTA); 391 392 if (test_opt(sbi, PRJQUOTA) && 393 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 394 clear_opt(sbi, PRJQUOTA); 395 396 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 397 test_opt(sbi, PRJQUOTA)) { 398 f2fs_err(sbi, "old and new quota format mixing"); 399 return -1; 400 } 401 402 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 403 f2fs_err(sbi, "journaled quota format not specified"); 404 return -1; 405 } 406 } 407 408 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 409 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 410 F2FS_OPTION(sbi).s_jquota_fmt = 0; 411 } 412 return 0; 413 } 414 #endif 415 416 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 417 const char *opt, 418 const substring_t *arg, 419 bool is_remount) 420 { 421 struct f2fs_sb_info *sbi = F2FS_SB(sb); 422 #ifdef CONFIG_FS_ENCRYPTION 423 int err; 424 425 if (!f2fs_sb_has_encrypt(sbi)) { 426 f2fs_err(sbi, "Encrypt feature is off"); 427 return -EINVAL; 428 } 429 430 /* 431 * This mount option is just for testing, and it's not worthwhile to 432 * implement the extra complexity (e.g. RCU protection) that would be 433 * needed to allow it to be set or changed during remount. We do allow 434 * it to be specified during remount, but only if there is no change. 435 */ 436 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) { 437 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 438 return -EINVAL; 439 } 440 err = fscrypt_set_test_dummy_encryption( 441 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy); 442 if (err) { 443 if (err == -EEXIST) 444 f2fs_warn(sbi, 445 "Can't change test_dummy_encryption on remount"); 446 else if (err == -EINVAL) 447 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 448 opt); 449 else 450 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 451 opt, err); 452 return -EINVAL; 453 } 454 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 455 #else 456 f2fs_warn(sbi, "Test dummy encryption mount option ignored"); 457 #endif 458 return 0; 459 } 460 461 static int parse_options(struct super_block *sb, char *options, bool is_remount) 462 { 463 struct f2fs_sb_info *sbi = F2FS_SB(sb); 464 substring_t args[MAX_OPT_ARGS]; 465 #ifdef CONFIG_F2FS_FS_COMPRESSION 466 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 467 int ext_cnt; 468 #endif 469 char *p, *name; 470 int arg = 0; 471 kuid_t uid; 472 kgid_t gid; 473 int ret; 474 475 if (!options) 476 return 0; 477 478 while ((p = strsep(&options, ",")) != NULL) { 479 int token; 480 if (!*p) 481 continue; 482 /* 483 * Initialize args struct so we know whether arg was 484 * found; some options take optional arguments. 485 */ 486 args[0].to = args[0].from = NULL; 487 token = match_token(p, f2fs_tokens, args); 488 489 switch (token) { 490 case Opt_gc_background: 491 name = match_strdup(&args[0]); 492 493 if (!name) 494 return -ENOMEM; 495 if (!strcmp(name, "on")) { 496 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 497 } else if (!strcmp(name, "off")) { 498 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 499 } else if (!strcmp(name, "sync")) { 500 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 501 } else { 502 kfree(name); 503 return -EINVAL; 504 } 505 kfree(name); 506 break; 507 case Opt_disable_roll_forward: 508 set_opt(sbi, DISABLE_ROLL_FORWARD); 509 break; 510 case Opt_norecovery: 511 /* this option mounts f2fs with ro */ 512 set_opt(sbi, NORECOVERY); 513 if (!f2fs_readonly(sb)) 514 return -EINVAL; 515 break; 516 case Opt_discard: 517 set_opt(sbi, DISCARD); 518 break; 519 case Opt_nodiscard: 520 if (f2fs_sb_has_blkzoned(sbi)) { 521 f2fs_warn(sbi, "discard is required for zoned block devices"); 522 return -EINVAL; 523 } 524 clear_opt(sbi, DISCARD); 525 break; 526 case Opt_noheap: 527 set_opt(sbi, NOHEAP); 528 break; 529 case Opt_heap: 530 clear_opt(sbi, NOHEAP); 531 break; 532 #ifdef CONFIG_F2FS_FS_XATTR 533 case Opt_user_xattr: 534 set_opt(sbi, XATTR_USER); 535 break; 536 case Opt_nouser_xattr: 537 clear_opt(sbi, XATTR_USER); 538 break; 539 case Opt_inline_xattr: 540 set_opt(sbi, INLINE_XATTR); 541 break; 542 case Opt_noinline_xattr: 543 clear_opt(sbi, INLINE_XATTR); 544 break; 545 case Opt_inline_xattr_size: 546 if (args->from && match_int(args, &arg)) 547 return -EINVAL; 548 set_opt(sbi, INLINE_XATTR_SIZE); 549 F2FS_OPTION(sbi).inline_xattr_size = arg; 550 break; 551 #else 552 case Opt_user_xattr: 553 f2fs_info(sbi, "user_xattr options not supported"); 554 break; 555 case Opt_nouser_xattr: 556 f2fs_info(sbi, "nouser_xattr options not supported"); 557 break; 558 case Opt_inline_xattr: 559 f2fs_info(sbi, "inline_xattr options not supported"); 560 break; 561 case Opt_noinline_xattr: 562 f2fs_info(sbi, "noinline_xattr options not supported"); 563 break; 564 #endif 565 #ifdef CONFIG_F2FS_FS_POSIX_ACL 566 case Opt_acl: 567 set_opt(sbi, POSIX_ACL); 568 break; 569 case Opt_noacl: 570 clear_opt(sbi, POSIX_ACL); 571 break; 572 #else 573 case Opt_acl: 574 f2fs_info(sbi, "acl options not supported"); 575 break; 576 case Opt_noacl: 577 f2fs_info(sbi, "noacl options not supported"); 578 break; 579 #endif 580 case Opt_active_logs: 581 if (args->from && match_int(args, &arg)) 582 return -EINVAL; 583 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE) 584 return -EINVAL; 585 F2FS_OPTION(sbi).active_logs = arg; 586 break; 587 case Opt_disable_ext_identify: 588 set_opt(sbi, DISABLE_EXT_IDENTIFY); 589 break; 590 case Opt_inline_data: 591 set_opt(sbi, INLINE_DATA); 592 break; 593 case Opt_inline_dentry: 594 set_opt(sbi, INLINE_DENTRY); 595 break; 596 case Opt_noinline_dentry: 597 clear_opt(sbi, INLINE_DENTRY); 598 break; 599 case Opt_flush_merge: 600 set_opt(sbi, FLUSH_MERGE); 601 break; 602 case Opt_noflush_merge: 603 clear_opt(sbi, FLUSH_MERGE); 604 break; 605 case Opt_nobarrier: 606 set_opt(sbi, NOBARRIER); 607 break; 608 case Opt_fastboot: 609 set_opt(sbi, FASTBOOT); 610 break; 611 case Opt_extent_cache: 612 set_opt(sbi, EXTENT_CACHE); 613 break; 614 case Opt_noextent_cache: 615 clear_opt(sbi, EXTENT_CACHE); 616 break; 617 case Opt_noinline_data: 618 clear_opt(sbi, INLINE_DATA); 619 break; 620 case Opt_data_flush: 621 set_opt(sbi, DATA_FLUSH); 622 break; 623 case Opt_reserve_root: 624 if (args->from && match_int(args, &arg)) 625 return -EINVAL; 626 if (test_opt(sbi, RESERVE_ROOT)) { 627 f2fs_info(sbi, "Preserve previous reserve_root=%u", 628 F2FS_OPTION(sbi).root_reserved_blocks); 629 } else { 630 F2FS_OPTION(sbi).root_reserved_blocks = arg; 631 set_opt(sbi, RESERVE_ROOT); 632 } 633 break; 634 case Opt_resuid: 635 if (args->from && match_int(args, &arg)) 636 return -EINVAL; 637 uid = make_kuid(current_user_ns(), arg); 638 if (!uid_valid(uid)) { 639 f2fs_err(sbi, "Invalid uid value %d", arg); 640 return -EINVAL; 641 } 642 F2FS_OPTION(sbi).s_resuid = uid; 643 break; 644 case Opt_resgid: 645 if (args->from && match_int(args, &arg)) 646 return -EINVAL; 647 gid = make_kgid(current_user_ns(), arg); 648 if (!gid_valid(gid)) { 649 f2fs_err(sbi, "Invalid gid value %d", arg); 650 return -EINVAL; 651 } 652 F2FS_OPTION(sbi).s_resgid = gid; 653 break; 654 case Opt_mode: 655 name = match_strdup(&args[0]); 656 657 if (!name) 658 return -ENOMEM; 659 if (!strcmp(name, "adaptive")) { 660 if (f2fs_sb_has_blkzoned(sbi)) { 661 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 662 kfree(name); 663 return -EINVAL; 664 } 665 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 666 } else if (!strcmp(name, "lfs")) { 667 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 668 } else { 669 kfree(name); 670 return -EINVAL; 671 } 672 kfree(name); 673 break; 674 case Opt_io_size_bits: 675 if (args->from && match_int(args, &arg)) 676 return -EINVAL; 677 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) { 678 f2fs_warn(sbi, "Not support %d, larger than %d", 679 1 << arg, BIO_MAX_PAGES); 680 return -EINVAL; 681 } 682 F2FS_OPTION(sbi).write_io_size_bits = arg; 683 break; 684 #ifdef CONFIG_F2FS_FAULT_INJECTION 685 case Opt_fault_injection: 686 if (args->from && match_int(args, &arg)) 687 return -EINVAL; 688 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 689 set_opt(sbi, FAULT_INJECTION); 690 break; 691 692 case Opt_fault_type: 693 if (args->from && match_int(args, &arg)) 694 return -EINVAL; 695 f2fs_build_fault_attr(sbi, 0, arg); 696 set_opt(sbi, FAULT_INJECTION); 697 break; 698 #else 699 case Opt_fault_injection: 700 f2fs_info(sbi, "fault_injection options not supported"); 701 break; 702 703 case Opt_fault_type: 704 f2fs_info(sbi, "fault_type options not supported"); 705 break; 706 #endif 707 case Opt_lazytime: 708 sb->s_flags |= SB_LAZYTIME; 709 break; 710 case Opt_nolazytime: 711 sb->s_flags &= ~SB_LAZYTIME; 712 break; 713 #ifdef CONFIG_QUOTA 714 case Opt_quota: 715 case Opt_usrquota: 716 set_opt(sbi, USRQUOTA); 717 break; 718 case Opt_grpquota: 719 set_opt(sbi, GRPQUOTA); 720 break; 721 case Opt_prjquota: 722 set_opt(sbi, PRJQUOTA); 723 break; 724 case Opt_usrjquota: 725 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 726 if (ret) 727 return ret; 728 break; 729 case Opt_grpjquota: 730 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 731 if (ret) 732 return ret; 733 break; 734 case Opt_prjjquota: 735 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 736 if (ret) 737 return ret; 738 break; 739 case Opt_offusrjquota: 740 ret = f2fs_clear_qf_name(sb, USRQUOTA); 741 if (ret) 742 return ret; 743 break; 744 case Opt_offgrpjquota: 745 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 746 if (ret) 747 return ret; 748 break; 749 case Opt_offprjjquota: 750 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 751 if (ret) 752 return ret; 753 break; 754 case Opt_jqfmt_vfsold: 755 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 756 break; 757 case Opt_jqfmt_vfsv0: 758 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 759 break; 760 case Opt_jqfmt_vfsv1: 761 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 762 break; 763 case Opt_noquota: 764 clear_opt(sbi, QUOTA); 765 clear_opt(sbi, USRQUOTA); 766 clear_opt(sbi, GRPQUOTA); 767 clear_opt(sbi, PRJQUOTA); 768 break; 769 #else 770 case Opt_quota: 771 case Opt_usrquota: 772 case Opt_grpquota: 773 case Opt_prjquota: 774 case Opt_usrjquota: 775 case Opt_grpjquota: 776 case Opt_prjjquota: 777 case Opt_offusrjquota: 778 case Opt_offgrpjquota: 779 case Opt_offprjjquota: 780 case Opt_jqfmt_vfsold: 781 case Opt_jqfmt_vfsv0: 782 case Opt_jqfmt_vfsv1: 783 case Opt_noquota: 784 f2fs_info(sbi, "quota operations not supported"); 785 break; 786 #endif 787 case Opt_whint: 788 name = match_strdup(&args[0]); 789 if (!name) 790 return -ENOMEM; 791 if (!strcmp(name, "user-based")) { 792 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER; 793 } else if (!strcmp(name, "off")) { 794 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 795 } else if (!strcmp(name, "fs-based")) { 796 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS; 797 } else { 798 kfree(name); 799 return -EINVAL; 800 } 801 kfree(name); 802 break; 803 case Opt_alloc: 804 name = match_strdup(&args[0]); 805 if (!name) 806 return -ENOMEM; 807 808 if (!strcmp(name, "default")) { 809 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 810 } else if (!strcmp(name, "reuse")) { 811 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 812 } else { 813 kfree(name); 814 return -EINVAL; 815 } 816 kfree(name); 817 break; 818 case Opt_fsync: 819 name = match_strdup(&args[0]); 820 if (!name) 821 return -ENOMEM; 822 if (!strcmp(name, "posix")) { 823 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 824 } else if (!strcmp(name, "strict")) { 825 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 826 } else if (!strcmp(name, "nobarrier")) { 827 F2FS_OPTION(sbi).fsync_mode = 828 FSYNC_MODE_NOBARRIER; 829 } else { 830 kfree(name); 831 return -EINVAL; 832 } 833 kfree(name); 834 break; 835 case Opt_test_dummy_encryption: 836 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 837 is_remount); 838 if (ret) 839 return ret; 840 break; 841 case Opt_inlinecrypt: 842 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 843 sb->s_flags |= SB_INLINECRYPT; 844 #else 845 f2fs_info(sbi, "inline encryption not supported"); 846 #endif 847 break; 848 case Opt_checkpoint_disable_cap_perc: 849 if (args->from && match_int(args, &arg)) 850 return -EINVAL; 851 if (arg < 0 || arg > 100) 852 return -EINVAL; 853 F2FS_OPTION(sbi).unusable_cap_perc = arg; 854 set_opt(sbi, DISABLE_CHECKPOINT); 855 break; 856 case Opt_checkpoint_disable_cap: 857 if (args->from && match_int(args, &arg)) 858 return -EINVAL; 859 F2FS_OPTION(sbi).unusable_cap = arg; 860 set_opt(sbi, DISABLE_CHECKPOINT); 861 break; 862 case Opt_checkpoint_disable: 863 set_opt(sbi, DISABLE_CHECKPOINT); 864 break; 865 case Opt_checkpoint_enable: 866 clear_opt(sbi, DISABLE_CHECKPOINT); 867 break; 868 #ifdef CONFIG_F2FS_FS_COMPRESSION 869 case Opt_compress_algorithm: 870 if (!f2fs_sb_has_compression(sbi)) { 871 f2fs_err(sbi, "Compression feature if off"); 872 return -EINVAL; 873 } 874 name = match_strdup(&args[0]); 875 if (!name) 876 return -ENOMEM; 877 if (!strcmp(name, "lzo")) { 878 F2FS_OPTION(sbi).compress_algorithm = 879 COMPRESS_LZO; 880 } else if (!strcmp(name, "lz4")) { 881 F2FS_OPTION(sbi).compress_algorithm = 882 COMPRESS_LZ4; 883 } else if (!strcmp(name, "zstd")) { 884 F2FS_OPTION(sbi).compress_algorithm = 885 COMPRESS_ZSTD; 886 } else if (!strcmp(name, "lzo-rle")) { 887 F2FS_OPTION(sbi).compress_algorithm = 888 COMPRESS_LZORLE; 889 } else { 890 kfree(name); 891 return -EINVAL; 892 } 893 kfree(name); 894 break; 895 case Opt_compress_log_size: 896 if (!f2fs_sb_has_compression(sbi)) { 897 f2fs_err(sbi, "Compression feature is off"); 898 return -EINVAL; 899 } 900 if (args->from && match_int(args, &arg)) 901 return -EINVAL; 902 if (arg < MIN_COMPRESS_LOG_SIZE || 903 arg > MAX_COMPRESS_LOG_SIZE) { 904 f2fs_err(sbi, 905 "Compress cluster log size is out of range"); 906 return -EINVAL; 907 } 908 F2FS_OPTION(sbi).compress_log_size = arg; 909 break; 910 case Opt_compress_extension: 911 if (!f2fs_sb_has_compression(sbi)) { 912 f2fs_err(sbi, "Compression feature is off"); 913 return -EINVAL; 914 } 915 name = match_strdup(&args[0]); 916 if (!name) 917 return -ENOMEM; 918 919 ext = F2FS_OPTION(sbi).extensions; 920 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 921 922 if (strlen(name) >= F2FS_EXTENSION_LEN || 923 ext_cnt >= COMPRESS_EXT_NUM) { 924 f2fs_err(sbi, 925 "invalid extension length/number"); 926 kfree(name); 927 return -EINVAL; 928 } 929 930 strcpy(ext[ext_cnt], name); 931 F2FS_OPTION(sbi).compress_ext_cnt++; 932 kfree(name); 933 break; 934 #else 935 case Opt_compress_algorithm: 936 case Opt_compress_log_size: 937 case Opt_compress_extension: 938 f2fs_info(sbi, "compression options not supported"); 939 break; 940 #endif 941 default: 942 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 943 p); 944 return -EINVAL; 945 } 946 } 947 #ifdef CONFIG_QUOTA 948 if (f2fs_check_quota_options(sbi)) 949 return -EINVAL; 950 #else 951 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 952 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 953 return -EINVAL; 954 } 955 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 956 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 957 return -EINVAL; 958 } 959 #endif 960 #ifndef CONFIG_UNICODE 961 if (f2fs_sb_has_casefold(sbi)) { 962 f2fs_err(sbi, 963 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 964 return -EINVAL; 965 } 966 #endif 967 968 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 969 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO", 970 F2FS_IO_SIZE_KB(sbi)); 971 return -EINVAL; 972 } 973 974 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 975 int min_size, max_size; 976 977 if (!f2fs_sb_has_extra_attr(sbi) || 978 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 979 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 980 return -EINVAL; 981 } 982 if (!test_opt(sbi, INLINE_XATTR)) { 983 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 984 return -EINVAL; 985 } 986 987 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 988 max_size = MAX_INLINE_XATTR_SIZE; 989 990 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 991 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 992 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 993 min_size, max_size); 994 return -EINVAL; 995 } 996 } 997 998 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 999 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n"); 1000 return -EINVAL; 1001 } 1002 1003 /* Not pass down write hints if the number of active logs is lesser 1004 * than NR_CURSEG_TYPE. 1005 */ 1006 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE) 1007 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1008 return 0; 1009 } 1010 1011 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1012 { 1013 struct f2fs_inode_info *fi; 1014 1015 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); 1016 if (!fi) 1017 return NULL; 1018 1019 init_once((void *) fi); 1020 1021 /* Initialize f2fs-specific inode info */ 1022 atomic_set(&fi->dirty_pages, 0); 1023 init_rwsem(&fi->i_sem); 1024 spin_lock_init(&fi->i_size_lock); 1025 INIT_LIST_HEAD(&fi->dirty_list); 1026 INIT_LIST_HEAD(&fi->gdirty_list); 1027 INIT_LIST_HEAD(&fi->inmem_ilist); 1028 INIT_LIST_HEAD(&fi->inmem_pages); 1029 mutex_init(&fi->inmem_lock); 1030 init_rwsem(&fi->i_gc_rwsem[READ]); 1031 init_rwsem(&fi->i_gc_rwsem[WRITE]); 1032 init_rwsem(&fi->i_mmap_sem); 1033 init_rwsem(&fi->i_xattr_sem); 1034 1035 /* Will be used by directory only */ 1036 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1037 1038 fi->ra_offset = -1; 1039 1040 return &fi->vfs_inode; 1041 } 1042 1043 static int f2fs_drop_inode(struct inode *inode) 1044 { 1045 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1046 int ret; 1047 1048 /* 1049 * during filesystem shutdown, if checkpoint is disabled, 1050 * drop useless meta/node dirty pages. 1051 */ 1052 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1053 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1054 inode->i_ino == F2FS_META_INO(sbi)) { 1055 trace_f2fs_drop_inode(inode, 1); 1056 return 1; 1057 } 1058 } 1059 1060 /* 1061 * This is to avoid a deadlock condition like below. 1062 * writeback_single_inode(inode) 1063 * - f2fs_write_data_page 1064 * - f2fs_gc -> iput -> evict 1065 * - inode_wait_for_writeback(inode) 1066 */ 1067 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1068 if (!inode->i_nlink && !is_bad_inode(inode)) { 1069 /* to avoid evict_inode call simultaneously */ 1070 atomic_inc(&inode->i_count); 1071 spin_unlock(&inode->i_lock); 1072 1073 /* some remained atomic pages should discarded */ 1074 if (f2fs_is_atomic_file(inode)) 1075 f2fs_drop_inmem_pages(inode); 1076 1077 /* should remain fi->extent_tree for writepage */ 1078 f2fs_destroy_extent_node(inode); 1079 1080 sb_start_intwrite(inode->i_sb); 1081 f2fs_i_size_write(inode, 0); 1082 1083 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1084 inode, NULL, 0, DATA); 1085 truncate_inode_pages_final(inode->i_mapping); 1086 1087 if (F2FS_HAS_BLOCKS(inode)) 1088 f2fs_truncate(inode); 1089 1090 sb_end_intwrite(inode->i_sb); 1091 1092 spin_lock(&inode->i_lock); 1093 atomic_dec(&inode->i_count); 1094 } 1095 trace_f2fs_drop_inode(inode, 0); 1096 return 0; 1097 } 1098 ret = generic_drop_inode(inode); 1099 if (!ret) 1100 ret = fscrypt_drop_inode(inode); 1101 trace_f2fs_drop_inode(inode, ret); 1102 return ret; 1103 } 1104 1105 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1106 { 1107 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1108 int ret = 0; 1109 1110 spin_lock(&sbi->inode_lock[DIRTY_META]); 1111 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1112 ret = 1; 1113 } else { 1114 set_inode_flag(inode, FI_DIRTY_INODE); 1115 stat_inc_dirty_inode(sbi, DIRTY_META); 1116 } 1117 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1118 list_add_tail(&F2FS_I(inode)->gdirty_list, 1119 &sbi->inode_list[DIRTY_META]); 1120 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1121 } 1122 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1123 return ret; 1124 } 1125 1126 void f2fs_inode_synced(struct inode *inode) 1127 { 1128 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1129 1130 spin_lock(&sbi->inode_lock[DIRTY_META]); 1131 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1132 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1133 return; 1134 } 1135 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1136 list_del_init(&F2FS_I(inode)->gdirty_list); 1137 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1138 } 1139 clear_inode_flag(inode, FI_DIRTY_INODE); 1140 clear_inode_flag(inode, FI_AUTO_RECOVER); 1141 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1142 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1143 } 1144 1145 /* 1146 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1147 * 1148 * We should call set_dirty_inode to write the dirty inode through write_inode. 1149 */ 1150 static void f2fs_dirty_inode(struct inode *inode, int flags) 1151 { 1152 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1153 1154 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1155 inode->i_ino == F2FS_META_INO(sbi)) 1156 return; 1157 1158 if (flags == I_DIRTY_TIME) 1159 return; 1160 1161 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1162 clear_inode_flag(inode, FI_AUTO_RECOVER); 1163 1164 f2fs_inode_dirtied(inode, false); 1165 } 1166 1167 static void f2fs_free_inode(struct inode *inode) 1168 { 1169 fscrypt_free_inode(inode); 1170 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1171 } 1172 1173 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1174 { 1175 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1176 percpu_counter_destroy(&sbi->total_valid_inode_count); 1177 } 1178 1179 static void destroy_device_list(struct f2fs_sb_info *sbi) 1180 { 1181 int i; 1182 1183 for (i = 0; i < sbi->s_ndevs; i++) { 1184 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1185 #ifdef CONFIG_BLK_DEV_ZONED 1186 kvfree(FDEV(i).blkz_seq); 1187 #endif 1188 } 1189 kvfree(sbi->devs); 1190 } 1191 1192 static void f2fs_put_super(struct super_block *sb) 1193 { 1194 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1195 int i; 1196 bool dropped; 1197 1198 /* unregister procfs/sysfs entries in advance to avoid race case */ 1199 f2fs_unregister_sysfs(sbi); 1200 1201 f2fs_quota_off_umount(sb); 1202 1203 /* prevent remaining shrinker jobs */ 1204 mutex_lock(&sbi->umount_mutex); 1205 1206 /* 1207 * We don't need to do checkpoint when superblock is clean. 1208 * But, the previous checkpoint was not done by umount, it needs to do 1209 * clean checkpoint again. 1210 */ 1211 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1212 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1213 struct cp_control cpc = { 1214 .reason = CP_UMOUNT, 1215 }; 1216 f2fs_write_checkpoint(sbi, &cpc); 1217 } 1218 1219 /* be sure to wait for any on-going discard commands */ 1220 dropped = f2fs_issue_discard_timeout(sbi); 1221 1222 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && 1223 !sbi->discard_blks && !dropped) { 1224 struct cp_control cpc = { 1225 .reason = CP_UMOUNT | CP_TRIMMED, 1226 }; 1227 f2fs_write_checkpoint(sbi, &cpc); 1228 } 1229 1230 /* 1231 * normally superblock is clean, so we need to release this. 1232 * In addition, EIO will skip do checkpoint, we need this as well. 1233 */ 1234 f2fs_release_ino_entry(sbi, true); 1235 1236 f2fs_leave_shrinker(sbi); 1237 mutex_unlock(&sbi->umount_mutex); 1238 1239 /* our cp_error case, we can wait for any writeback page */ 1240 f2fs_flush_merged_writes(sbi); 1241 1242 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1243 1244 f2fs_bug_on(sbi, sbi->fsync_node_num); 1245 1246 iput(sbi->node_inode); 1247 sbi->node_inode = NULL; 1248 1249 iput(sbi->meta_inode); 1250 sbi->meta_inode = NULL; 1251 1252 /* 1253 * iput() can update stat information, if f2fs_write_checkpoint() 1254 * above failed with error. 1255 */ 1256 f2fs_destroy_stats(sbi); 1257 1258 /* destroy f2fs internal modules */ 1259 f2fs_destroy_node_manager(sbi); 1260 f2fs_destroy_segment_manager(sbi); 1261 1262 f2fs_destroy_post_read_wq(sbi); 1263 1264 kvfree(sbi->ckpt); 1265 1266 sb->s_fs_info = NULL; 1267 if (sbi->s_chksum_driver) 1268 crypto_free_shash(sbi->s_chksum_driver); 1269 kfree(sbi->raw_super); 1270 1271 destroy_device_list(sbi); 1272 f2fs_destroy_xattr_caches(sbi); 1273 mempool_destroy(sbi->write_io_dummy); 1274 #ifdef CONFIG_QUOTA 1275 for (i = 0; i < MAXQUOTAS; i++) 1276 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1277 #endif 1278 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1279 destroy_percpu_info(sbi); 1280 for (i = 0; i < NR_PAGE_TYPE; i++) 1281 kvfree(sbi->write_io[i]); 1282 #ifdef CONFIG_UNICODE 1283 utf8_unload(sbi->s_encoding); 1284 #endif 1285 kfree(sbi); 1286 } 1287 1288 int f2fs_sync_fs(struct super_block *sb, int sync) 1289 { 1290 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1291 int err = 0; 1292 1293 if (unlikely(f2fs_cp_error(sbi))) 1294 return 0; 1295 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1296 return 0; 1297 1298 trace_f2fs_sync_fs(sb, sync); 1299 1300 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1301 return -EAGAIN; 1302 1303 if (sync) { 1304 struct cp_control cpc; 1305 1306 cpc.reason = __get_cp_reason(sbi); 1307 1308 down_write(&sbi->gc_lock); 1309 err = f2fs_write_checkpoint(sbi, &cpc); 1310 up_write(&sbi->gc_lock); 1311 } 1312 f2fs_trace_ios(NULL, 1); 1313 1314 return err; 1315 } 1316 1317 static int f2fs_freeze(struct super_block *sb) 1318 { 1319 if (f2fs_readonly(sb)) 1320 return 0; 1321 1322 /* IO error happened before */ 1323 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1324 return -EIO; 1325 1326 /* must be clean, since sync_filesystem() was already called */ 1327 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1328 return -EINVAL; 1329 return 0; 1330 } 1331 1332 static int f2fs_unfreeze(struct super_block *sb) 1333 { 1334 return 0; 1335 } 1336 1337 #ifdef CONFIG_QUOTA 1338 static int f2fs_statfs_project(struct super_block *sb, 1339 kprojid_t projid, struct kstatfs *buf) 1340 { 1341 struct kqid qid; 1342 struct dquot *dquot; 1343 u64 limit; 1344 u64 curblock; 1345 1346 qid = make_kqid_projid(projid); 1347 dquot = dqget(sb, qid); 1348 if (IS_ERR(dquot)) 1349 return PTR_ERR(dquot); 1350 spin_lock(&dquot->dq_dqb_lock); 1351 1352 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1353 dquot->dq_dqb.dqb_bhardlimit); 1354 if (limit) 1355 limit >>= sb->s_blocksize_bits; 1356 1357 if (limit && buf->f_blocks > limit) { 1358 curblock = (dquot->dq_dqb.dqb_curspace + 1359 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1360 buf->f_blocks = limit; 1361 buf->f_bfree = buf->f_bavail = 1362 (buf->f_blocks > curblock) ? 1363 (buf->f_blocks - curblock) : 0; 1364 } 1365 1366 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1367 dquot->dq_dqb.dqb_ihardlimit); 1368 1369 if (limit && buf->f_files > limit) { 1370 buf->f_files = limit; 1371 buf->f_ffree = 1372 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1373 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1374 } 1375 1376 spin_unlock(&dquot->dq_dqb_lock); 1377 dqput(dquot); 1378 return 0; 1379 } 1380 #endif 1381 1382 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1383 { 1384 struct super_block *sb = dentry->d_sb; 1385 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1386 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1387 block_t total_count, user_block_count, start_count; 1388 u64 avail_node_count; 1389 1390 total_count = le64_to_cpu(sbi->raw_super->block_count); 1391 user_block_count = sbi->user_block_count; 1392 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1393 buf->f_type = F2FS_SUPER_MAGIC; 1394 buf->f_bsize = sbi->blocksize; 1395 1396 buf->f_blocks = total_count - start_count; 1397 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1398 sbi->current_reserved_blocks; 1399 1400 spin_lock(&sbi->stat_lock); 1401 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1402 buf->f_bfree = 0; 1403 else 1404 buf->f_bfree -= sbi->unusable_block_count; 1405 spin_unlock(&sbi->stat_lock); 1406 1407 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1408 buf->f_bavail = buf->f_bfree - 1409 F2FS_OPTION(sbi).root_reserved_blocks; 1410 else 1411 buf->f_bavail = 0; 1412 1413 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1414 1415 if (avail_node_count > user_block_count) { 1416 buf->f_files = user_block_count; 1417 buf->f_ffree = buf->f_bavail; 1418 } else { 1419 buf->f_files = avail_node_count; 1420 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1421 buf->f_bavail); 1422 } 1423 1424 buf->f_namelen = F2FS_NAME_LEN; 1425 buf->f_fsid.val[0] = (u32)id; 1426 buf->f_fsid.val[1] = (u32)(id >> 32); 1427 1428 #ifdef CONFIG_QUOTA 1429 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1430 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1431 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1432 } 1433 #endif 1434 return 0; 1435 } 1436 1437 static inline void f2fs_show_quota_options(struct seq_file *seq, 1438 struct super_block *sb) 1439 { 1440 #ifdef CONFIG_QUOTA 1441 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1442 1443 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1444 char *fmtname = ""; 1445 1446 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1447 case QFMT_VFS_OLD: 1448 fmtname = "vfsold"; 1449 break; 1450 case QFMT_VFS_V0: 1451 fmtname = "vfsv0"; 1452 break; 1453 case QFMT_VFS_V1: 1454 fmtname = "vfsv1"; 1455 break; 1456 } 1457 seq_printf(seq, ",jqfmt=%s", fmtname); 1458 } 1459 1460 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1461 seq_show_option(seq, "usrjquota", 1462 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1463 1464 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1465 seq_show_option(seq, "grpjquota", 1466 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1467 1468 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1469 seq_show_option(seq, "prjjquota", 1470 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1471 #endif 1472 } 1473 1474 static inline void f2fs_show_compress_options(struct seq_file *seq, 1475 struct super_block *sb) 1476 { 1477 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1478 char *algtype = ""; 1479 int i; 1480 1481 if (!f2fs_sb_has_compression(sbi)) 1482 return; 1483 1484 switch (F2FS_OPTION(sbi).compress_algorithm) { 1485 case COMPRESS_LZO: 1486 algtype = "lzo"; 1487 break; 1488 case COMPRESS_LZ4: 1489 algtype = "lz4"; 1490 break; 1491 case COMPRESS_ZSTD: 1492 algtype = "zstd"; 1493 break; 1494 case COMPRESS_LZORLE: 1495 algtype = "lzo-rle"; 1496 break; 1497 } 1498 seq_printf(seq, ",compress_algorithm=%s", algtype); 1499 1500 seq_printf(seq, ",compress_log_size=%u", 1501 F2FS_OPTION(sbi).compress_log_size); 1502 1503 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1504 seq_printf(seq, ",compress_extension=%s", 1505 F2FS_OPTION(sbi).extensions[i]); 1506 } 1507 } 1508 1509 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1510 { 1511 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1512 1513 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1514 seq_printf(seq, ",background_gc=%s", "sync"); 1515 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1516 seq_printf(seq, ",background_gc=%s", "on"); 1517 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1518 seq_printf(seq, ",background_gc=%s", "off"); 1519 1520 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1521 seq_puts(seq, ",disable_roll_forward"); 1522 if (test_opt(sbi, NORECOVERY)) 1523 seq_puts(seq, ",norecovery"); 1524 if (test_opt(sbi, DISCARD)) 1525 seq_puts(seq, ",discard"); 1526 else 1527 seq_puts(seq, ",nodiscard"); 1528 if (test_opt(sbi, NOHEAP)) 1529 seq_puts(seq, ",no_heap"); 1530 else 1531 seq_puts(seq, ",heap"); 1532 #ifdef CONFIG_F2FS_FS_XATTR 1533 if (test_opt(sbi, XATTR_USER)) 1534 seq_puts(seq, ",user_xattr"); 1535 else 1536 seq_puts(seq, ",nouser_xattr"); 1537 if (test_opt(sbi, INLINE_XATTR)) 1538 seq_puts(seq, ",inline_xattr"); 1539 else 1540 seq_puts(seq, ",noinline_xattr"); 1541 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1542 seq_printf(seq, ",inline_xattr_size=%u", 1543 F2FS_OPTION(sbi).inline_xattr_size); 1544 #endif 1545 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1546 if (test_opt(sbi, POSIX_ACL)) 1547 seq_puts(seq, ",acl"); 1548 else 1549 seq_puts(seq, ",noacl"); 1550 #endif 1551 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1552 seq_puts(seq, ",disable_ext_identify"); 1553 if (test_opt(sbi, INLINE_DATA)) 1554 seq_puts(seq, ",inline_data"); 1555 else 1556 seq_puts(seq, ",noinline_data"); 1557 if (test_opt(sbi, INLINE_DENTRY)) 1558 seq_puts(seq, ",inline_dentry"); 1559 else 1560 seq_puts(seq, ",noinline_dentry"); 1561 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1562 seq_puts(seq, ",flush_merge"); 1563 if (test_opt(sbi, NOBARRIER)) 1564 seq_puts(seq, ",nobarrier"); 1565 if (test_opt(sbi, FASTBOOT)) 1566 seq_puts(seq, ",fastboot"); 1567 if (test_opt(sbi, EXTENT_CACHE)) 1568 seq_puts(seq, ",extent_cache"); 1569 else 1570 seq_puts(seq, ",noextent_cache"); 1571 if (test_opt(sbi, DATA_FLUSH)) 1572 seq_puts(seq, ",data_flush"); 1573 1574 seq_puts(seq, ",mode="); 1575 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1576 seq_puts(seq, "adaptive"); 1577 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1578 seq_puts(seq, "lfs"); 1579 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1580 if (test_opt(sbi, RESERVE_ROOT)) 1581 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1582 F2FS_OPTION(sbi).root_reserved_blocks, 1583 from_kuid_munged(&init_user_ns, 1584 F2FS_OPTION(sbi).s_resuid), 1585 from_kgid_munged(&init_user_ns, 1586 F2FS_OPTION(sbi).s_resgid)); 1587 if (F2FS_IO_SIZE_BITS(sbi)) 1588 seq_printf(seq, ",io_bits=%u", 1589 F2FS_OPTION(sbi).write_io_size_bits); 1590 #ifdef CONFIG_F2FS_FAULT_INJECTION 1591 if (test_opt(sbi, FAULT_INJECTION)) { 1592 seq_printf(seq, ",fault_injection=%u", 1593 F2FS_OPTION(sbi).fault_info.inject_rate); 1594 seq_printf(seq, ",fault_type=%u", 1595 F2FS_OPTION(sbi).fault_info.inject_type); 1596 } 1597 #endif 1598 #ifdef CONFIG_QUOTA 1599 if (test_opt(sbi, QUOTA)) 1600 seq_puts(seq, ",quota"); 1601 if (test_opt(sbi, USRQUOTA)) 1602 seq_puts(seq, ",usrquota"); 1603 if (test_opt(sbi, GRPQUOTA)) 1604 seq_puts(seq, ",grpquota"); 1605 if (test_opt(sbi, PRJQUOTA)) 1606 seq_puts(seq, ",prjquota"); 1607 #endif 1608 f2fs_show_quota_options(seq, sbi->sb); 1609 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1610 seq_printf(seq, ",whint_mode=%s", "user-based"); 1611 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1612 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1613 1614 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1615 1616 if (sbi->sb->s_flags & SB_INLINECRYPT) 1617 seq_puts(seq, ",inlinecrypt"); 1618 1619 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1620 seq_printf(seq, ",alloc_mode=%s", "default"); 1621 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1622 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1623 1624 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1625 seq_printf(seq, ",checkpoint=disable:%u", 1626 F2FS_OPTION(sbi).unusable_cap); 1627 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1628 seq_printf(seq, ",fsync_mode=%s", "posix"); 1629 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1630 seq_printf(seq, ",fsync_mode=%s", "strict"); 1631 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1632 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 1633 1634 #ifdef CONFIG_F2FS_FS_COMPRESSION 1635 f2fs_show_compress_options(seq, sbi->sb); 1636 #endif 1637 return 0; 1638 } 1639 1640 static void default_options(struct f2fs_sb_info *sbi) 1641 { 1642 /* init some FS parameters */ 1643 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE; 1644 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 1645 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1646 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1647 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1648 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 1649 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 1650 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 1651 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 1652 F2FS_OPTION(sbi).compress_ext_cnt = 0; 1653 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 1654 1655 sbi->sb->s_flags &= ~SB_INLINECRYPT; 1656 1657 set_opt(sbi, INLINE_XATTR); 1658 set_opt(sbi, INLINE_DATA); 1659 set_opt(sbi, INLINE_DENTRY); 1660 set_opt(sbi, EXTENT_CACHE); 1661 set_opt(sbi, NOHEAP); 1662 clear_opt(sbi, DISABLE_CHECKPOINT); 1663 F2FS_OPTION(sbi).unusable_cap = 0; 1664 sbi->sb->s_flags |= SB_LAZYTIME; 1665 set_opt(sbi, FLUSH_MERGE); 1666 set_opt(sbi, DISCARD); 1667 if (f2fs_sb_has_blkzoned(sbi)) 1668 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 1669 else 1670 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 1671 1672 #ifdef CONFIG_F2FS_FS_XATTR 1673 set_opt(sbi, XATTR_USER); 1674 #endif 1675 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1676 set_opt(sbi, POSIX_ACL); 1677 #endif 1678 1679 f2fs_build_fault_attr(sbi, 0, 0); 1680 } 1681 1682 #ifdef CONFIG_QUOTA 1683 static int f2fs_enable_quotas(struct super_block *sb); 1684 #endif 1685 1686 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 1687 { 1688 unsigned int s_flags = sbi->sb->s_flags; 1689 struct cp_control cpc; 1690 int err = 0; 1691 int ret; 1692 block_t unusable; 1693 1694 if (s_flags & SB_RDONLY) { 1695 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 1696 return -EINVAL; 1697 } 1698 sbi->sb->s_flags |= SB_ACTIVE; 1699 1700 f2fs_update_time(sbi, DISABLE_TIME); 1701 1702 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 1703 down_write(&sbi->gc_lock); 1704 err = f2fs_gc(sbi, true, false, NULL_SEGNO); 1705 if (err == -ENODATA) { 1706 err = 0; 1707 break; 1708 } 1709 if (err && err != -EAGAIN) 1710 break; 1711 } 1712 1713 ret = sync_filesystem(sbi->sb); 1714 if (ret || err) { 1715 err = ret ? ret: err; 1716 goto restore_flag; 1717 } 1718 1719 unusable = f2fs_get_unusable_blocks(sbi); 1720 if (f2fs_disable_cp_again(sbi, unusable)) { 1721 err = -EAGAIN; 1722 goto restore_flag; 1723 } 1724 1725 down_write(&sbi->gc_lock); 1726 cpc.reason = CP_PAUSE; 1727 set_sbi_flag(sbi, SBI_CP_DISABLED); 1728 err = f2fs_write_checkpoint(sbi, &cpc); 1729 if (err) 1730 goto out_unlock; 1731 1732 spin_lock(&sbi->stat_lock); 1733 sbi->unusable_block_count = unusable; 1734 spin_unlock(&sbi->stat_lock); 1735 1736 out_unlock: 1737 up_write(&sbi->gc_lock); 1738 restore_flag: 1739 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 1740 return err; 1741 } 1742 1743 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 1744 { 1745 down_write(&sbi->gc_lock); 1746 f2fs_dirty_to_prefree(sbi); 1747 1748 clear_sbi_flag(sbi, SBI_CP_DISABLED); 1749 set_sbi_flag(sbi, SBI_IS_DIRTY); 1750 up_write(&sbi->gc_lock); 1751 1752 f2fs_sync_fs(sbi->sb, 1); 1753 } 1754 1755 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 1756 { 1757 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1758 struct f2fs_mount_info org_mount_opt; 1759 unsigned long old_sb_flags; 1760 int err; 1761 bool need_restart_gc = false; 1762 bool need_stop_gc = false; 1763 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 1764 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT); 1765 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 1766 bool checkpoint_changed; 1767 #ifdef CONFIG_QUOTA 1768 int i, j; 1769 #endif 1770 1771 /* 1772 * Save the old mount options in case we 1773 * need to restore them. 1774 */ 1775 org_mount_opt = sbi->mount_opt; 1776 old_sb_flags = sb->s_flags; 1777 1778 #ifdef CONFIG_QUOTA 1779 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 1780 for (i = 0; i < MAXQUOTAS; i++) { 1781 if (F2FS_OPTION(sbi).s_qf_names[i]) { 1782 org_mount_opt.s_qf_names[i] = 1783 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 1784 GFP_KERNEL); 1785 if (!org_mount_opt.s_qf_names[i]) { 1786 for (j = 0; j < i; j++) 1787 kfree(org_mount_opt.s_qf_names[j]); 1788 return -ENOMEM; 1789 } 1790 } else { 1791 org_mount_opt.s_qf_names[i] = NULL; 1792 } 1793 } 1794 #endif 1795 1796 /* recover superblocks we couldn't write due to previous RO mount */ 1797 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 1798 err = f2fs_commit_super(sbi, false); 1799 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 1800 err); 1801 if (!err) 1802 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1803 } 1804 1805 default_options(sbi); 1806 1807 /* parse mount options */ 1808 err = parse_options(sb, data, true); 1809 if (err) 1810 goto restore_opts; 1811 checkpoint_changed = 1812 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT); 1813 1814 /* 1815 * Previous and new state of filesystem is RO, 1816 * so skip checking GC and FLUSH_MERGE conditions. 1817 */ 1818 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 1819 goto skip; 1820 1821 #ifdef CONFIG_QUOTA 1822 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 1823 err = dquot_suspend(sb, -1); 1824 if (err < 0) 1825 goto restore_opts; 1826 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 1827 /* dquot_resume needs RW */ 1828 sb->s_flags &= ~SB_RDONLY; 1829 if (sb_any_quota_suspended(sb)) { 1830 dquot_resume(sb, -1); 1831 } else if (f2fs_sb_has_quota_ino(sbi)) { 1832 err = f2fs_enable_quotas(sb); 1833 if (err) 1834 goto restore_opts; 1835 } 1836 } 1837 #endif 1838 /* disallow enable/disable extent_cache dynamically */ 1839 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 1840 err = -EINVAL; 1841 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 1842 goto restore_opts; 1843 } 1844 1845 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 1846 err = -EINVAL; 1847 f2fs_warn(sbi, "switch io_bits option is not allowed"); 1848 goto restore_opts; 1849 } 1850 1851 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 1852 err = -EINVAL; 1853 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 1854 goto restore_opts; 1855 } 1856 1857 /* 1858 * We stop the GC thread if FS is mounted as RO 1859 * or if background_gc = off is passed in mount 1860 * option. Also sync the filesystem. 1861 */ 1862 if ((*flags & SB_RDONLY) || 1863 F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) { 1864 if (sbi->gc_thread) { 1865 f2fs_stop_gc_thread(sbi); 1866 need_restart_gc = true; 1867 } 1868 } else if (!sbi->gc_thread) { 1869 err = f2fs_start_gc_thread(sbi); 1870 if (err) 1871 goto restore_opts; 1872 need_stop_gc = true; 1873 } 1874 1875 if (*flags & SB_RDONLY || 1876 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 1877 writeback_inodes_sb(sb, WB_REASON_SYNC); 1878 sync_inodes_sb(sb); 1879 1880 set_sbi_flag(sbi, SBI_IS_DIRTY); 1881 set_sbi_flag(sbi, SBI_IS_CLOSE); 1882 f2fs_sync_fs(sb, 1); 1883 clear_sbi_flag(sbi, SBI_IS_CLOSE); 1884 } 1885 1886 if (checkpoint_changed) { 1887 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 1888 err = f2fs_disable_checkpoint(sbi); 1889 if (err) 1890 goto restore_gc; 1891 } else { 1892 f2fs_enable_checkpoint(sbi); 1893 } 1894 } 1895 1896 /* 1897 * We stop issue flush thread if FS is mounted as RO 1898 * or if flush_merge is not passed in mount option. 1899 */ 1900 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 1901 clear_opt(sbi, FLUSH_MERGE); 1902 f2fs_destroy_flush_cmd_control(sbi, false); 1903 } else { 1904 err = f2fs_create_flush_cmd_control(sbi); 1905 if (err) 1906 goto restore_gc; 1907 } 1908 skip: 1909 #ifdef CONFIG_QUOTA 1910 /* Release old quota file names */ 1911 for (i = 0; i < MAXQUOTAS; i++) 1912 kfree(org_mount_opt.s_qf_names[i]); 1913 #endif 1914 /* Update the POSIXACL Flag */ 1915 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 1916 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 1917 1918 limit_reserve_root(sbi); 1919 adjust_unusable_cap_perc(sbi); 1920 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 1921 return 0; 1922 restore_gc: 1923 if (need_restart_gc) { 1924 if (f2fs_start_gc_thread(sbi)) 1925 f2fs_warn(sbi, "background gc thread has stopped"); 1926 } else if (need_stop_gc) { 1927 f2fs_stop_gc_thread(sbi); 1928 } 1929 restore_opts: 1930 #ifdef CONFIG_QUOTA 1931 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 1932 for (i = 0; i < MAXQUOTAS; i++) { 1933 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1934 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 1935 } 1936 #endif 1937 sbi->mount_opt = org_mount_opt; 1938 sb->s_flags = old_sb_flags; 1939 return err; 1940 } 1941 1942 #ifdef CONFIG_QUOTA 1943 /* Read data from quotafile */ 1944 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 1945 size_t len, loff_t off) 1946 { 1947 struct inode *inode = sb_dqopt(sb)->files[type]; 1948 struct address_space *mapping = inode->i_mapping; 1949 block_t blkidx = F2FS_BYTES_TO_BLK(off); 1950 int offset = off & (sb->s_blocksize - 1); 1951 int tocopy; 1952 size_t toread; 1953 loff_t i_size = i_size_read(inode); 1954 struct page *page; 1955 char *kaddr; 1956 1957 if (off > i_size) 1958 return 0; 1959 1960 if (off + len > i_size) 1961 len = i_size - off; 1962 toread = len; 1963 while (toread > 0) { 1964 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 1965 repeat: 1966 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 1967 if (IS_ERR(page)) { 1968 if (PTR_ERR(page) == -ENOMEM) { 1969 congestion_wait(BLK_RW_ASYNC, 1970 DEFAULT_IO_TIMEOUT); 1971 goto repeat; 1972 } 1973 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 1974 return PTR_ERR(page); 1975 } 1976 1977 lock_page(page); 1978 1979 if (unlikely(page->mapping != mapping)) { 1980 f2fs_put_page(page, 1); 1981 goto repeat; 1982 } 1983 if (unlikely(!PageUptodate(page))) { 1984 f2fs_put_page(page, 1); 1985 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 1986 return -EIO; 1987 } 1988 1989 kaddr = kmap_atomic(page); 1990 memcpy(data, kaddr + offset, tocopy); 1991 kunmap_atomic(kaddr); 1992 f2fs_put_page(page, 1); 1993 1994 offset = 0; 1995 toread -= tocopy; 1996 data += tocopy; 1997 blkidx++; 1998 } 1999 return len; 2000 } 2001 2002 /* Write to quotafile */ 2003 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2004 const char *data, size_t len, loff_t off) 2005 { 2006 struct inode *inode = sb_dqopt(sb)->files[type]; 2007 struct address_space *mapping = inode->i_mapping; 2008 const struct address_space_operations *a_ops = mapping->a_ops; 2009 int offset = off & (sb->s_blocksize - 1); 2010 size_t towrite = len; 2011 struct page *page; 2012 void *fsdata = NULL; 2013 char *kaddr; 2014 int err = 0; 2015 int tocopy; 2016 2017 while (towrite > 0) { 2018 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2019 towrite); 2020 retry: 2021 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2022 &page, &fsdata); 2023 if (unlikely(err)) { 2024 if (err == -ENOMEM) { 2025 congestion_wait(BLK_RW_ASYNC, 2026 DEFAULT_IO_TIMEOUT); 2027 goto retry; 2028 } 2029 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2030 break; 2031 } 2032 2033 kaddr = kmap_atomic(page); 2034 memcpy(kaddr + offset, data, tocopy); 2035 kunmap_atomic(kaddr); 2036 flush_dcache_page(page); 2037 2038 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2039 page, fsdata); 2040 offset = 0; 2041 towrite -= tocopy; 2042 off += tocopy; 2043 data += tocopy; 2044 cond_resched(); 2045 } 2046 2047 if (len == towrite) 2048 return err; 2049 inode->i_mtime = inode->i_ctime = current_time(inode); 2050 f2fs_mark_inode_dirty_sync(inode, false); 2051 return len - towrite; 2052 } 2053 2054 static struct dquot **f2fs_get_dquots(struct inode *inode) 2055 { 2056 return F2FS_I(inode)->i_dquot; 2057 } 2058 2059 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2060 { 2061 return &F2FS_I(inode)->i_reserved_quota; 2062 } 2063 2064 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2065 { 2066 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2067 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2068 return 0; 2069 } 2070 2071 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2072 F2FS_OPTION(sbi).s_jquota_fmt, type); 2073 } 2074 2075 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2076 { 2077 int enabled = 0; 2078 int i, err; 2079 2080 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2081 err = f2fs_enable_quotas(sbi->sb); 2082 if (err) { 2083 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2084 return 0; 2085 } 2086 return 1; 2087 } 2088 2089 for (i = 0; i < MAXQUOTAS; i++) { 2090 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2091 err = f2fs_quota_on_mount(sbi, i); 2092 if (!err) { 2093 enabled = 1; 2094 continue; 2095 } 2096 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2097 err, i); 2098 } 2099 } 2100 return enabled; 2101 } 2102 2103 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2104 unsigned int flags) 2105 { 2106 struct inode *qf_inode; 2107 unsigned long qf_inum; 2108 int err; 2109 2110 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2111 2112 qf_inum = f2fs_qf_ino(sb, type); 2113 if (!qf_inum) 2114 return -EPERM; 2115 2116 qf_inode = f2fs_iget(sb, qf_inum); 2117 if (IS_ERR(qf_inode)) { 2118 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2119 return PTR_ERR(qf_inode); 2120 } 2121 2122 /* Don't account quota for quota files to avoid recursion */ 2123 qf_inode->i_flags |= S_NOQUOTA; 2124 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2125 iput(qf_inode); 2126 return err; 2127 } 2128 2129 static int f2fs_enable_quotas(struct super_block *sb) 2130 { 2131 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2132 int type, err = 0; 2133 unsigned long qf_inum; 2134 bool quota_mopt[MAXQUOTAS] = { 2135 test_opt(sbi, USRQUOTA), 2136 test_opt(sbi, GRPQUOTA), 2137 test_opt(sbi, PRJQUOTA), 2138 }; 2139 2140 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2141 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2142 return 0; 2143 } 2144 2145 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2146 2147 for (type = 0; type < MAXQUOTAS; type++) { 2148 qf_inum = f2fs_qf_ino(sb, type); 2149 if (qf_inum) { 2150 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2151 DQUOT_USAGE_ENABLED | 2152 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2153 if (err) { 2154 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2155 type, err); 2156 for (type--; type >= 0; type--) 2157 dquot_quota_off(sb, type); 2158 set_sbi_flag(F2FS_SB(sb), 2159 SBI_QUOTA_NEED_REPAIR); 2160 return err; 2161 } 2162 } 2163 } 2164 return 0; 2165 } 2166 2167 int f2fs_quota_sync(struct super_block *sb, int type) 2168 { 2169 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2170 struct quota_info *dqopt = sb_dqopt(sb); 2171 int cnt; 2172 int ret; 2173 2174 /* 2175 * do_quotactl 2176 * f2fs_quota_sync 2177 * down_read(quota_sem) 2178 * dquot_writeback_dquots() 2179 * f2fs_dquot_commit 2180 * block_operation 2181 * down_read(quota_sem) 2182 */ 2183 f2fs_lock_op(sbi); 2184 2185 down_read(&sbi->quota_sem); 2186 ret = dquot_writeback_dquots(sb, type); 2187 if (ret) 2188 goto out; 2189 2190 /* 2191 * Now when everything is written we can discard the pagecache so 2192 * that userspace sees the changes. 2193 */ 2194 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2195 struct address_space *mapping; 2196 2197 if (type != -1 && cnt != type) 2198 continue; 2199 if (!sb_has_quota_active(sb, cnt)) 2200 continue; 2201 2202 mapping = dqopt->files[cnt]->i_mapping; 2203 2204 ret = filemap_fdatawrite(mapping); 2205 if (ret) 2206 goto out; 2207 2208 /* if we are using journalled quota */ 2209 if (is_journalled_quota(sbi)) 2210 continue; 2211 2212 ret = filemap_fdatawait(mapping); 2213 if (ret) 2214 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2215 2216 inode_lock(dqopt->files[cnt]); 2217 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0); 2218 inode_unlock(dqopt->files[cnt]); 2219 } 2220 out: 2221 if (ret) 2222 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2223 up_read(&sbi->quota_sem); 2224 f2fs_unlock_op(sbi); 2225 return ret; 2226 } 2227 2228 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2229 const struct path *path) 2230 { 2231 struct inode *inode; 2232 int err; 2233 2234 /* if quota sysfile exists, deny enabling quota with specific file */ 2235 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2236 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2237 return -EBUSY; 2238 } 2239 2240 err = f2fs_quota_sync(sb, type); 2241 if (err) 2242 return err; 2243 2244 err = dquot_quota_on(sb, type, format_id, path); 2245 if (err) 2246 return err; 2247 2248 inode = d_inode(path->dentry); 2249 2250 inode_lock(inode); 2251 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2252 f2fs_set_inode_flags(inode); 2253 inode_unlock(inode); 2254 f2fs_mark_inode_dirty_sync(inode, false); 2255 2256 return 0; 2257 } 2258 2259 static int __f2fs_quota_off(struct super_block *sb, int type) 2260 { 2261 struct inode *inode = sb_dqopt(sb)->files[type]; 2262 int err; 2263 2264 if (!inode || !igrab(inode)) 2265 return dquot_quota_off(sb, type); 2266 2267 err = f2fs_quota_sync(sb, type); 2268 if (err) 2269 goto out_put; 2270 2271 err = dquot_quota_off(sb, type); 2272 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2273 goto out_put; 2274 2275 inode_lock(inode); 2276 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2277 f2fs_set_inode_flags(inode); 2278 inode_unlock(inode); 2279 f2fs_mark_inode_dirty_sync(inode, false); 2280 out_put: 2281 iput(inode); 2282 return err; 2283 } 2284 2285 static int f2fs_quota_off(struct super_block *sb, int type) 2286 { 2287 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2288 int err; 2289 2290 err = __f2fs_quota_off(sb, type); 2291 2292 /* 2293 * quotactl can shutdown journalled quota, result in inconsistence 2294 * between quota record and fs data by following updates, tag the 2295 * flag to let fsck be aware of it. 2296 */ 2297 if (is_journalled_quota(sbi)) 2298 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2299 return err; 2300 } 2301 2302 void f2fs_quota_off_umount(struct super_block *sb) 2303 { 2304 int type; 2305 int err; 2306 2307 for (type = 0; type < MAXQUOTAS; type++) { 2308 err = __f2fs_quota_off(sb, type); 2309 if (err) { 2310 int ret = dquot_quota_off(sb, type); 2311 2312 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2313 type, err, ret); 2314 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2315 } 2316 } 2317 /* 2318 * In case of checkpoint=disable, we must flush quota blocks. 2319 * This can cause NULL exception for node_inode in end_io, since 2320 * put_super already dropped it. 2321 */ 2322 sync_filesystem(sb); 2323 } 2324 2325 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2326 { 2327 struct quota_info *dqopt = sb_dqopt(sb); 2328 int type; 2329 2330 for (type = 0; type < MAXQUOTAS; type++) { 2331 if (!dqopt->files[type]) 2332 continue; 2333 f2fs_inode_synced(dqopt->files[type]); 2334 } 2335 } 2336 2337 static int f2fs_dquot_commit(struct dquot *dquot) 2338 { 2339 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2340 int ret; 2341 2342 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2343 ret = dquot_commit(dquot); 2344 if (ret < 0) 2345 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2346 up_read(&sbi->quota_sem); 2347 return ret; 2348 } 2349 2350 static int f2fs_dquot_acquire(struct dquot *dquot) 2351 { 2352 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2353 int ret; 2354 2355 down_read(&sbi->quota_sem); 2356 ret = dquot_acquire(dquot); 2357 if (ret < 0) 2358 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2359 up_read(&sbi->quota_sem); 2360 return ret; 2361 } 2362 2363 static int f2fs_dquot_release(struct dquot *dquot) 2364 { 2365 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2366 int ret = dquot_release(dquot); 2367 2368 if (ret < 0) 2369 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2370 return ret; 2371 } 2372 2373 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2374 { 2375 struct super_block *sb = dquot->dq_sb; 2376 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2377 int ret = dquot_mark_dquot_dirty(dquot); 2378 2379 /* if we are using journalled quota */ 2380 if (is_journalled_quota(sbi)) 2381 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2382 2383 return ret; 2384 } 2385 2386 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2387 { 2388 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2389 int ret = dquot_commit_info(sb, type); 2390 2391 if (ret < 0) 2392 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2393 return ret; 2394 } 2395 2396 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2397 { 2398 *projid = F2FS_I(inode)->i_projid; 2399 return 0; 2400 } 2401 2402 static const struct dquot_operations f2fs_quota_operations = { 2403 .get_reserved_space = f2fs_get_reserved_space, 2404 .write_dquot = f2fs_dquot_commit, 2405 .acquire_dquot = f2fs_dquot_acquire, 2406 .release_dquot = f2fs_dquot_release, 2407 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2408 .write_info = f2fs_dquot_commit_info, 2409 .alloc_dquot = dquot_alloc, 2410 .destroy_dquot = dquot_destroy, 2411 .get_projid = f2fs_get_projid, 2412 .get_next_id = dquot_get_next_id, 2413 }; 2414 2415 static const struct quotactl_ops f2fs_quotactl_ops = { 2416 .quota_on = f2fs_quota_on, 2417 .quota_off = f2fs_quota_off, 2418 .quota_sync = f2fs_quota_sync, 2419 .get_state = dquot_get_state, 2420 .set_info = dquot_set_dqinfo, 2421 .get_dqblk = dquot_get_dqblk, 2422 .set_dqblk = dquot_set_dqblk, 2423 .get_nextdqblk = dquot_get_next_dqblk, 2424 }; 2425 #else 2426 int f2fs_quota_sync(struct super_block *sb, int type) 2427 { 2428 return 0; 2429 } 2430 2431 void f2fs_quota_off_umount(struct super_block *sb) 2432 { 2433 } 2434 #endif 2435 2436 static const struct super_operations f2fs_sops = { 2437 .alloc_inode = f2fs_alloc_inode, 2438 .free_inode = f2fs_free_inode, 2439 .drop_inode = f2fs_drop_inode, 2440 .write_inode = f2fs_write_inode, 2441 .dirty_inode = f2fs_dirty_inode, 2442 .show_options = f2fs_show_options, 2443 #ifdef CONFIG_QUOTA 2444 .quota_read = f2fs_quota_read, 2445 .quota_write = f2fs_quota_write, 2446 .get_dquots = f2fs_get_dquots, 2447 #endif 2448 .evict_inode = f2fs_evict_inode, 2449 .put_super = f2fs_put_super, 2450 .sync_fs = f2fs_sync_fs, 2451 .freeze_fs = f2fs_freeze, 2452 .unfreeze_fs = f2fs_unfreeze, 2453 .statfs = f2fs_statfs, 2454 .remount_fs = f2fs_remount, 2455 }; 2456 2457 #ifdef CONFIG_FS_ENCRYPTION 2458 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2459 { 2460 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2461 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2462 ctx, len, NULL); 2463 } 2464 2465 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2466 void *fs_data) 2467 { 2468 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2469 2470 /* 2471 * Encrypting the root directory is not allowed because fsck 2472 * expects lost+found directory to exist and remain unencrypted 2473 * if LOST_FOUND feature is enabled. 2474 * 2475 */ 2476 if (f2fs_sb_has_lost_found(sbi) && 2477 inode->i_ino == F2FS_ROOT_INO(sbi)) 2478 return -EPERM; 2479 2480 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2481 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2482 ctx, len, fs_data, XATTR_CREATE); 2483 } 2484 2485 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 2486 { 2487 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 2488 } 2489 2490 static bool f2fs_has_stable_inodes(struct super_block *sb) 2491 { 2492 return true; 2493 } 2494 2495 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 2496 int *ino_bits_ret, int *lblk_bits_ret) 2497 { 2498 *ino_bits_ret = 8 * sizeof(nid_t); 2499 *lblk_bits_ret = 8 * sizeof(block_t); 2500 } 2501 2502 static int f2fs_get_num_devices(struct super_block *sb) 2503 { 2504 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2505 2506 if (f2fs_is_multi_device(sbi)) 2507 return sbi->s_ndevs; 2508 return 1; 2509 } 2510 2511 static void f2fs_get_devices(struct super_block *sb, 2512 struct request_queue **devs) 2513 { 2514 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2515 int i; 2516 2517 for (i = 0; i < sbi->s_ndevs; i++) 2518 devs[i] = bdev_get_queue(FDEV(i).bdev); 2519 } 2520 2521 static const struct fscrypt_operations f2fs_cryptops = { 2522 .key_prefix = "f2fs:", 2523 .get_context = f2fs_get_context, 2524 .set_context = f2fs_set_context, 2525 .get_dummy_policy = f2fs_get_dummy_policy, 2526 .empty_dir = f2fs_empty_dir, 2527 .max_namelen = F2FS_NAME_LEN, 2528 .has_stable_inodes = f2fs_has_stable_inodes, 2529 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 2530 .get_num_devices = f2fs_get_num_devices, 2531 .get_devices = f2fs_get_devices, 2532 }; 2533 #endif 2534 2535 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 2536 u64 ino, u32 generation) 2537 { 2538 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2539 struct inode *inode; 2540 2541 if (f2fs_check_nid_range(sbi, ino)) 2542 return ERR_PTR(-ESTALE); 2543 2544 /* 2545 * f2fs_iget isn't quite right if the inode is currently unallocated! 2546 * However f2fs_iget currently does appropriate checks to handle stale 2547 * inodes so everything is OK. 2548 */ 2549 inode = f2fs_iget(sb, ino); 2550 if (IS_ERR(inode)) 2551 return ERR_CAST(inode); 2552 if (unlikely(generation && inode->i_generation != generation)) { 2553 /* we didn't find the right inode.. */ 2554 iput(inode); 2555 return ERR_PTR(-ESTALE); 2556 } 2557 return inode; 2558 } 2559 2560 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 2561 int fh_len, int fh_type) 2562 { 2563 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 2564 f2fs_nfs_get_inode); 2565 } 2566 2567 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 2568 int fh_len, int fh_type) 2569 { 2570 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 2571 f2fs_nfs_get_inode); 2572 } 2573 2574 static const struct export_operations f2fs_export_ops = { 2575 .fh_to_dentry = f2fs_fh_to_dentry, 2576 .fh_to_parent = f2fs_fh_to_parent, 2577 .get_parent = f2fs_get_parent, 2578 }; 2579 2580 static loff_t max_file_blocks(void) 2581 { 2582 loff_t result = 0; 2583 loff_t leaf_count = DEF_ADDRS_PER_BLOCK; 2584 2585 /* 2586 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 2587 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 2588 * space in inode.i_addr, it will be more safe to reassign 2589 * result as zero. 2590 */ 2591 2592 /* two direct node blocks */ 2593 result += (leaf_count * 2); 2594 2595 /* two indirect node blocks */ 2596 leaf_count *= NIDS_PER_BLOCK; 2597 result += (leaf_count * 2); 2598 2599 /* one double indirect node block */ 2600 leaf_count *= NIDS_PER_BLOCK; 2601 result += leaf_count; 2602 2603 return result; 2604 } 2605 2606 static int __f2fs_commit_super(struct buffer_head *bh, 2607 struct f2fs_super_block *super) 2608 { 2609 lock_buffer(bh); 2610 if (super) 2611 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 2612 set_buffer_dirty(bh); 2613 unlock_buffer(bh); 2614 2615 /* it's rare case, we can do fua all the time */ 2616 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 2617 } 2618 2619 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 2620 struct buffer_head *bh) 2621 { 2622 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2623 (bh->b_data + F2FS_SUPER_OFFSET); 2624 struct super_block *sb = sbi->sb; 2625 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 2626 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 2627 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 2628 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 2629 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 2630 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 2631 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 2632 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 2633 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 2634 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 2635 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2636 u32 segment_count = le32_to_cpu(raw_super->segment_count); 2637 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2638 u64 main_end_blkaddr = main_blkaddr + 2639 (segment_count_main << log_blocks_per_seg); 2640 u64 seg_end_blkaddr = segment0_blkaddr + 2641 (segment_count << log_blocks_per_seg); 2642 2643 if (segment0_blkaddr != cp_blkaddr) { 2644 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 2645 segment0_blkaddr, cp_blkaddr); 2646 return true; 2647 } 2648 2649 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 2650 sit_blkaddr) { 2651 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 2652 cp_blkaddr, sit_blkaddr, 2653 segment_count_ckpt << log_blocks_per_seg); 2654 return true; 2655 } 2656 2657 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 2658 nat_blkaddr) { 2659 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 2660 sit_blkaddr, nat_blkaddr, 2661 segment_count_sit << log_blocks_per_seg); 2662 return true; 2663 } 2664 2665 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 2666 ssa_blkaddr) { 2667 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 2668 nat_blkaddr, ssa_blkaddr, 2669 segment_count_nat << log_blocks_per_seg); 2670 return true; 2671 } 2672 2673 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 2674 main_blkaddr) { 2675 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 2676 ssa_blkaddr, main_blkaddr, 2677 segment_count_ssa << log_blocks_per_seg); 2678 return true; 2679 } 2680 2681 if (main_end_blkaddr > seg_end_blkaddr) { 2682 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)", 2683 main_blkaddr, 2684 segment0_blkaddr + 2685 (segment_count << log_blocks_per_seg), 2686 segment_count_main << log_blocks_per_seg); 2687 return true; 2688 } else if (main_end_blkaddr < seg_end_blkaddr) { 2689 int err = 0; 2690 char *res; 2691 2692 /* fix in-memory information all the time */ 2693 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 2694 segment0_blkaddr) >> log_blocks_per_seg); 2695 2696 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 2697 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2698 res = "internally"; 2699 } else { 2700 err = __f2fs_commit_super(bh, NULL); 2701 res = err ? "failed" : "done"; 2702 } 2703 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)", 2704 res, main_blkaddr, 2705 segment0_blkaddr + 2706 (segment_count << log_blocks_per_seg), 2707 segment_count_main << log_blocks_per_seg); 2708 if (err) 2709 return true; 2710 } 2711 return false; 2712 } 2713 2714 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 2715 struct buffer_head *bh) 2716 { 2717 block_t segment_count, segs_per_sec, secs_per_zone; 2718 block_t total_sections, blocks_per_seg; 2719 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2720 (bh->b_data + F2FS_SUPER_OFFSET); 2721 unsigned int blocksize; 2722 size_t crc_offset = 0; 2723 __u32 crc = 0; 2724 2725 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 2726 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 2727 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 2728 return -EINVAL; 2729 } 2730 2731 /* Check checksum_offset and crc in superblock */ 2732 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 2733 crc_offset = le32_to_cpu(raw_super->checksum_offset); 2734 if (crc_offset != 2735 offsetof(struct f2fs_super_block, crc)) { 2736 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 2737 crc_offset); 2738 return -EFSCORRUPTED; 2739 } 2740 crc = le32_to_cpu(raw_super->crc); 2741 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 2742 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 2743 return -EFSCORRUPTED; 2744 } 2745 } 2746 2747 /* Currently, support only 4KB page cache size */ 2748 if (F2FS_BLKSIZE != PAGE_SIZE) { 2749 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB", 2750 PAGE_SIZE); 2751 return -EFSCORRUPTED; 2752 } 2753 2754 /* Currently, support only 4KB block size */ 2755 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); 2756 if (blocksize != F2FS_BLKSIZE) { 2757 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB", 2758 blocksize); 2759 return -EFSCORRUPTED; 2760 } 2761 2762 /* check log blocks per segment */ 2763 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 2764 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 2765 le32_to_cpu(raw_super->log_blocks_per_seg)); 2766 return -EFSCORRUPTED; 2767 } 2768 2769 /* Currently, support 512/1024/2048/4096 bytes sector size */ 2770 if (le32_to_cpu(raw_super->log_sectorsize) > 2771 F2FS_MAX_LOG_SECTOR_SIZE || 2772 le32_to_cpu(raw_super->log_sectorsize) < 2773 F2FS_MIN_LOG_SECTOR_SIZE) { 2774 f2fs_info(sbi, "Invalid log sectorsize (%u)", 2775 le32_to_cpu(raw_super->log_sectorsize)); 2776 return -EFSCORRUPTED; 2777 } 2778 if (le32_to_cpu(raw_super->log_sectors_per_block) + 2779 le32_to_cpu(raw_super->log_sectorsize) != 2780 F2FS_MAX_LOG_SECTOR_SIZE) { 2781 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 2782 le32_to_cpu(raw_super->log_sectors_per_block), 2783 le32_to_cpu(raw_super->log_sectorsize)); 2784 return -EFSCORRUPTED; 2785 } 2786 2787 segment_count = le32_to_cpu(raw_super->segment_count); 2788 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 2789 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 2790 total_sections = le32_to_cpu(raw_super->section_count); 2791 2792 /* blocks_per_seg should be 512, given the above check */ 2793 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 2794 2795 if (segment_count > F2FS_MAX_SEGMENT || 2796 segment_count < F2FS_MIN_SEGMENTS) { 2797 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 2798 return -EFSCORRUPTED; 2799 } 2800 2801 if (total_sections > segment_count || 2802 total_sections < F2FS_MIN_SEGMENTS || 2803 segs_per_sec > segment_count || !segs_per_sec) { 2804 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 2805 segment_count, total_sections, segs_per_sec); 2806 return -EFSCORRUPTED; 2807 } 2808 2809 if ((segment_count / segs_per_sec) < total_sections) { 2810 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 2811 segment_count, segs_per_sec, total_sections); 2812 return -EFSCORRUPTED; 2813 } 2814 2815 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 2816 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 2817 segment_count, le64_to_cpu(raw_super->block_count)); 2818 return -EFSCORRUPTED; 2819 } 2820 2821 if (RDEV(0).path[0]) { 2822 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 2823 int i = 1; 2824 2825 while (i < MAX_DEVICES && RDEV(i).path[0]) { 2826 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 2827 i++; 2828 } 2829 if (segment_count != dev_seg_count) { 2830 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 2831 segment_count, dev_seg_count); 2832 return -EFSCORRUPTED; 2833 } 2834 } 2835 2836 if (secs_per_zone > total_sections || !secs_per_zone) { 2837 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 2838 secs_per_zone, total_sections); 2839 return -EFSCORRUPTED; 2840 } 2841 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 2842 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 2843 (le32_to_cpu(raw_super->extension_count) + 2844 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 2845 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 2846 le32_to_cpu(raw_super->extension_count), 2847 raw_super->hot_ext_count, 2848 F2FS_MAX_EXTENSION); 2849 return -EFSCORRUPTED; 2850 } 2851 2852 if (le32_to_cpu(raw_super->cp_payload) > 2853 (blocks_per_seg - F2FS_CP_PACKS)) { 2854 f2fs_info(sbi, "Insane cp_payload (%u > %u)", 2855 le32_to_cpu(raw_super->cp_payload), 2856 blocks_per_seg - F2FS_CP_PACKS); 2857 return -EFSCORRUPTED; 2858 } 2859 2860 /* check reserved ino info */ 2861 if (le32_to_cpu(raw_super->node_ino) != 1 || 2862 le32_to_cpu(raw_super->meta_ino) != 2 || 2863 le32_to_cpu(raw_super->root_ino) != 3) { 2864 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 2865 le32_to_cpu(raw_super->node_ino), 2866 le32_to_cpu(raw_super->meta_ino), 2867 le32_to_cpu(raw_super->root_ino)); 2868 return -EFSCORRUPTED; 2869 } 2870 2871 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 2872 if (sanity_check_area_boundary(sbi, bh)) 2873 return -EFSCORRUPTED; 2874 2875 return 0; 2876 } 2877 2878 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 2879 { 2880 unsigned int total, fsmeta; 2881 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 2882 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 2883 unsigned int ovp_segments, reserved_segments; 2884 unsigned int main_segs, blocks_per_seg; 2885 unsigned int sit_segs, nat_segs; 2886 unsigned int sit_bitmap_size, nat_bitmap_size; 2887 unsigned int log_blocks_per_seg; 2888 unsigned int segment_count_main; 2889 unsigned int cp_pack_start_sum, cp_payload; 2890 block_t user_block_count, valid_user_blocks; 2891 block_t avail_node_count, valid_node_count; 2892 int i, j; 2893 2894 total = le32_to_cpu(raw_super->segment_count); 2895 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 2896 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 2897 fsmeta += sit_segs; 2898 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 2899 fsmeta += nat_segs; 2900 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 2901 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 2902 2903 if (unlikely(fsmeta >= total)) 2904 return 1; 2905 2906 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 2907 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 2908 2909 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS || 2910 ovp_segments == 0 || reserved_segments == 0)) { 2911 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 2912 return 1; 2913 } 2914 2915 user_block_count = le64_to_cpu(ckpt->user_block_count); 2916 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2917 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2918 if (!user_block_count || user_block_count >= 2919 segment_count_main << log_blocks_per_seg) { 2920 f2fs_err(sbi, "Wrong user_block_count: %u", 2921 user_block_count); 2922 return 1; 2923 } 2924 2925 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 2926 if (valid_user_blocks > user_block_count) { 2927 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 2928 valid_user_blocks, user_block_count); 2929 return 1; 2930 } 2931 2932 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 2933 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 2934 if (valid_node_count > avail_node_count) { 2935 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 2936 valid_node_count, avail_node_count); 2937 return 1; 2938 } 2939 2940 main_segs = le32_to_cpu(raw_super->segment_count_main); 2941 blocks_per_seg = sbi->blocks_per_seg; 2942 2943 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 2944 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 2945 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 2946 return 1; 2947 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 2948 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 2949 le32_to_cpu(ckpt->cur_node_segno[j])) { 2950 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 2951 i, j, 2952 le32_to_cpu(ckpt->cur_node_segno[i])); 2953 return 1; 2954 } 2955 } 2956 } 2957 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 2958 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 2959 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 2960 return 1; 2961 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 2962 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 2963 le32_to_cpu(ckpt->cur_data_segno[j])) { 2964 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 2965 i, j, 2966 le32_to_cpu(ckpt->cur_data_segno[i])); 2967 return 1; 2968 } 2969 } 2970 } 2971 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 2972 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 2973 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 2974 le32_to_cpu(ckpt->cur_data_segno[j])) { 2975 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 2976 i, j, 2977 le32_to_cpu(ckpt->cur_node_segno[i])); 2978 return 1; 2979 } 2980 } 2981 } 2982 2983 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 2984 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 2985 2986 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 2987 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 2988 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 2989 sit_bitmap_size, nat_bitmap_size); 2990 return 1; 2991 } 2992 2993 cp_pack_start_sum = __start_sum_addr(sbi); 2994 cp_payload = __cp_payload(sbi); 2995 if (cp_pack_start_sum < cp_payload + 1 || 2996 cp_pack_start_sum > blocks_per_seg - 1 - 2997 NR_CURSEG_TYPE) { 2998 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 2999 cp_pack_start_sum); 3000 return 1; 3001 } 3002 3003 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3004 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3005 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3006 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3007 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3008 le32_to_cpu(ckpt->checksum_offset)); 3009 return 1; 3010 } 3011 3012 if (unlikely(f2fs_cp_error(sbi))) { 3013 f2fs_err(sbi, "A bug case: need to run fsck"); 3014 return 1; 3015 } 3016 return 0; 3017 } 3018 3019 static void init_sb_info(struct f2fs_sb_info *sbi) 3020 { 3021 struct f2fs_super_block *raw_super = sbi->raw_super; 3022 int i; 3023 3024 sbi->log_sectors_per_block = 3025 le32_to_cpu(raw_super->log_sectors_per_block); 3026 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3027 sbi->blocksize = 1 << sbi->log_blocksize; 3028 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3029 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3030 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3031 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3032 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3033 sbi->total_node_count = 3034 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3035 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3036 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); 3037 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); 3038 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); 3039 sbi->cur_victim_sec = NULL_SECNO; 3040 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3041 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3042 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3043 sbi->migration_granularity = sbi->segs_per_sec; 3044 3045 sbi->dir_level = DEF_DIR_LEVEL; 3046 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3047 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3048 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3049 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3050 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3051 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3052 DEF_UMOUNT_DISCARD_TIMEOUT; 3053 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3054 3055 for (i = 0; i < NR_COUNT_TYPE; i++) 3056 atomic_set(&sbi->nr_pages[i], 0); 3057 3058 for (i = 0; i < META; i++) 3059 atomic_set(&sbi->wb_sync_req[i], 0); 3060 3061 INIT_LIST_HEAD(&sbi->s_list); 3062 mutex_init(&sbi->umount_mutex); 3063 init_rwsem(&sbi->io_order_lock); 3064 spin_lock_init(&sbi->cp_lock); 3065 3066 sbi->dirty_device = 0; 3067 spin_lock_init(&sbi->dev_lock); 3068 3069 init_rwsem(&sbi->sb_lock); 3070 init_rwsem(&sbi->pin_sem); 3071 } 3072 3073 static int init_percpu_info(struct f2fs_sb_info *sbi) 3074 { 3075 int err; 3076 3077 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3078 if (err) 3079 return err; 3080 3081 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3082 GFP_KERNEL); 3083 if (err) 3084 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3085 3086 return err; 3087 } 3088 3089 #ifdef CONFIG_BLK_DEV_ZONED 3090 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3091 void *data) 3092 { 3093 struct f2fs_dev_info *dev = data; 3094 3095 if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL) 3096 set_bit(idx, dev->blkz_seq); 3097 return 0; 3098 } 3099 3100 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3101 { 3102 struct block_device *bdev = FDEV(devi).bdev; 3103 sector_t nr_sectors = bdev->bd_part->nr_sects; 3104 int ret; 3105 3106 if (!f2fs_sb_has_blkzoned(sbi)) 3107 return 0; 3108 3109 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3110 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 3111 return -EINVAL; 3112 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 3113 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3114 __ilog2_u32(sbi->blocks_per_blkz)) 3115 return -EINVAL; 3116 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3117 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3118 sbi->log_blocks_per_blkz; 3119 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 3120 FDEV(devi).nr_blkz++; 3121 3122 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3123 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3124 * sizeof(unsigned long), 3125 GFP_KERNEL); 3126 if (!FDEV(devi).blkz_seq) 3127 return -ENOMEM; 3128 3129 /* Get block zones type */ 3130 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3131 &FDEV(devi)); 3132 if (ret < 0) 3133 return ret; 3134 3135 return 0; 3136 } 3137 #endif 3138 3139 /* 3140 * Read f2fs raw super block. 3141 * Because we have two copies of super block, so read both of them 3142 * to get the first valid one. If any one of them is broken, we pass 3143 * them recovery flag back to the caller. 3144 */ 3145 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3146 struct f2fs_super_block **raw_super, 3147 int *valid_super_block, int *recovery) 3148 { 3149 struct super_block *sb = sbi->sb; 3150 int block; 3151 struct buffer_head *bh; 3152 struct f2fs_super_block *super; 3153 int err = 0; 3154 3155 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3156 if (!super) 3157 return -ENOMEM; 3158 3159 for (block = 0; block < 2; block++) { 3160 bh = sb_bread(sb, block); 3161 if (!bh) { 3162 f2fs_err(sbi, "Unable to read %dth superblock", 3163 block + 1); 3164 err = -EIO; 3165 *recovery = 1; 3166 continue; 3167 } 3168 3169 /* sanity checking of raw super */ 3170 err = sanity_check_raw_super(sbi, bh); 3171 if (err) { 3172 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3173 block + 1); 3174 brelse(bh); 3175 *recovery = 1; 3176 continue; 3177 } 3178 3179 if (!*raw_super) { 3180 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3181 sizeof(*super)); 3182 *valid_super_block = block; 3183 *raw_super = super; 3184 } 3185 brelse(bh); 3186 } 3187 3188 /* No valid superblock */ 3189 if (!*raw_super) 3190 kfree(super); 3191 else 3192 err = 0; 3193 3194 return err; 3195 } 3196 3197 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3198 { 3199 struct buffer_head *bh; 3200 __u32 crc = 0; 3201 int err; 3202 3203 if ((recover && f2fs_readonly(sbi->sb)) || 3204 bdev_read_only(sbi->sb->s_bdev)) { 3205 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3206 return -EROFS; 3207 } 3208 3209 /* we should update superblock crc here */ 3210 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3211 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3212 offsetof(struct f2fs_super_block, crc)); 3213 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3214 } 3215 3216 /* write back-up superblock first */ 3217 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3218 if (!bh) 3219 return -EIO; 3220 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3221 brelse(bh); 3222 3223 /* if we are in recovery path, skip writing valid superblock */ 3224 if (recover || err) 3225 return err; 3226 3227 /* write current valid superblock */ 3228 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3229 if (!bh) 3230 return -EIO; 3231 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3232 brelse(bh); 3233 return err; 3234 } 3235 3236 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3237 { 3238 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3239 unsigned int max_devices = MAX_DEVICES; 3240 int i; 3241 3242 /* Initialize single device information */ 3243 if (!RDEV(0).path[0]) { 3244 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3245 return 0; 3246 max_devices = 1; 3247 } 3248 3249 /* 3250 * Initialize multiple devices information, or single 3251 * zoned block device information. 3252 */ 3253 sbi->devs = f2fs_kzalloc(sbi, 3254 array_size(max_devices, 3255 sizeof(struct f2fs_dev_info)), 3256 GFP_KERNEL); 3257 if (!sbi->devs) 3258 return -ENOMEM; 3259 3260 for (i = 0; i < max_devices; i++) { 3261 3262 if (i > 0 && !RDEV(i).path[0]) 3263 break; 3264 3265 if (max_devices == 1) { 3266 /* Single zoned block device mount */ 3267 FDEV(0).bdev = 3268 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3269 sbi->sb->s_mode, sbi->sb->s_type); 3270 } else { 3271 /* Multi-device mount */ 3272 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3273 FDEV(i).total_segments = 3274 le32_to_cpu(RDEV(i).total_segments); 3275 if (i == 0) { 3276 FDEV(i).start_blk = 0; 3277 FDEV(i).end_blk = FDEV(i).start_blk + 3278 (FDEV(i).total_segments << 3279 sbi->log_blocks_per_seg) - 1 + 3280 le32_to_cpu(raw_super->segment0_blkaddr); 3281 } else { 3282 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3283 FDEV(i).end_blk = FDEV(i).start_blk + 3284 (FDEV(i).total_segments << 3285 sbi->log_blocks_per_seg) - 1; 3286 } 3287 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3288 sbi->sb->s_mode, sbi->sb->s_type); 3289 } 3290 if (IS_ERR(FDEV(i).bdev)) 3291 return PTR_ERR(FDEV(i).bdev); 3292 3293 /* to release errored devices */ 3294 sbi->s_ndevs = i + 1; 3295 3296 #ifdef CONFIG_BLK_DEV_ZONED 3297 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3298 !f2fs_sb_has_blkzoned(sbi)) { 3299 f2fs_err(sbi, "Zoned block device feature not enabled\n"); 3300 return -EINVAL; 3301 } 3302 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3303 if (init_blkz_info(sbi, i)) { 3304 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3305 return -EINVAL; 3306 } 3307 if (max_devices == 1) 3308 break; 3309 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3310 i, FDEV(i).path, 3311 FDEV(i).total_segments, 3312 FDEV(i).start_blk, FDEV(i).end_blk, 3313 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3314 "Host-aware" : "Host-managed"); 3315 continue; 3316 } 3317 #endif 3318 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3319 i, FDEV(i).path, 3320 FDEV(i).total_segments, 3321 FDEV(i).start_blk, FDEV(i).end_blk); 3322 } 3323 f2fs_info(sbi, 3324 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3325 return 0; 3326 } 3327 3328 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3329 { 3330 #ifdef CONFIG_UNICODE 3331 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) { 3332 const struct f2fs_sb_encodings *encoding_info; 3333 struct unicode_map *encoding; 3334 __u16 encoding_flags; 3335 3336 if (f2fs_sb_has_encrypt(sbi)) { 3337 f2fs_err(sbi, 3338 "Can't mount with encoding and encryption"); 3339 return -EINVAL; 3340 } 3341 3342 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info, 3343 &encoding_flags)) { 3344 f2fs_err(sbi, 3345 "Encoding requested by superblock is unknown"); 3346 return -EINVAL; 3347 } 3348 3349 encoding = utf8_load(encoding_info->version); 3350 if (IS_ERR(encoding)) { 3351 f2fs_err(sbi, 3352 "can't mount with superblock charset: %s-%s " 3353 "not supported by the kernel. flags: 0x%x.", 3354 encoding_info->name, encoding_info->version, 3355 encoding_flags); 3356 return PTR_ERR(encoding); 3357 } 3358 f2fs_info(sbi, "Using encoding defined by superblock: " 3359 "%s-%s with flags 0x%hx", encoding_info->name, 3360 encoding_info->version?:"\b", encoding_flags); 3361 3362 sbi->s_encoding = encoding; 3363 sbi->s_encoding_flags = encoding_flags; 3364 sbi->sb->s_d_op = &f2fs_dentry_ops; 3365 } 3366 #else 3367 if (f2fs_sb_has_casefold(sbi)) { 3368 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3369 return -EINVAL; 3370 } 3371 #endif 3372 return 0; 3373 } 3374 3375 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3376 { 3377 struct f2fs_sm_info *sm_i = SM_I(sbi); 3378 3379 /* adjust parameters according to the volume size */ 3380 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3381 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3382 sm_i->dcc_info->discard_granularity = 1; 3383 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 3384 } 3385 3386 sbi->readdir_ra = 1; 3387 } 3388 3389 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3390 { 3391 struct f2fs_sb_info *sbi; 3392 struct f2fs_super_block *raw_super; 3393 struct inode *root; 3394 int err; 3395 bool skip_recovery = false, need_fsck = false; 3396 char *options = NULL; 3397 int recovery, i, valid_super_block; 3398 struct curseg_info *seg_i; 3399 int retry_cnt = 1; 3400 3401 try_onemore: 3402 err = -EINVAL; 3403 raw_super = NULL; 3404 valid_super_block = -1; 3405 recovery = 0; 3406 3407 /* allocate memory for f2fs-specific super block info */ 3408 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 3409 if (!sbi) 3410 return -ENOMEM; 3411 3412 sbi->sb = sb; 3413 3414 /* Load the checksum driver */ 3415 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 3416 if (IS_ERR(sbi->s_chksum_driver)) { 3417 f2fs_err(sbi, "Cannot load crc32 driver."); 3418 err = PTR_ERR(sbi->s_chksum_driver); 3419 sbi->s_chksum_driver = NULL; 3420 goto free_sbi; 3421 } 3422 3423 /* set a block size */ 3424 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 3425 f2fs_err(sbi, "unable to set blocksize"); 3426 goto free_sbi; 3427 } 3428 3429 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 3430 &recovery); 3431 if (err) 3432 goto free_sbi; 3433 3434 sb->s_fs_info = sbi; 3435 sbi->raw_super = raw_super; 3436 3437 /* precompute checksum seed for metadata */ 3438 if (f2fs_sb_has_inode_chksum(sbi)) 3439 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 3440 sizeof(raw_super->uuid)); 3441 3442 /* 3443 * The BLKZONED feature indicates that the drive was formatted with 3444 * zone alignment optimization. This is optional for host-aware 3445 * devices, but mandatory for host-managed zoned block devices. 3446 */ 3447 #ifndef CONFIG_BLK_DEV_ZONED 3448 if (f2fs_sb_has_blkzoned(sbi)) { 3449 f2fs_err(sbi, "Zoned block device support is not enabled"); 3450 err = -EOPNOTSUPP; 3451 goto free_sb_buf; 3452 } 3453 #endif 3454 default_options(sbi); 3455 /* parse mount options */ 3456 options = kstrdup((const char *)data, GFP_KERNEL); 3457 if (data && !options) { 3458 err = -ENOMEM; 3459 goto free_sb_buf; 3460 } 3461 3462 err = parse_options(sb, options, false); 3463 if (err) 3464 goto free_options; 3465 3466 sbi->max_file_blocks = max_file_blocks(); 3467 sb->s_maxbytes = sbi->max_file_blocks << 3468 le32_to_cpu(raw_super->log_blocksize); 3469 sb->s_max_links = F2FS_LINK_MAX; 3470 3471 err = f2fs_setup_casefold(sbi); 3472 if (err) 3473 goto free_options; 3474 3475 #ifdef CONFIG_QUOTA 3476 sb->dq_op = &f2fs_quota_operations; 3477 sb->s_qcop = &f2fs_quotactl_ops; 3478 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 3479 3480 if (f2fs_sb_has_quota_ino(sbi)) { 3481 for (i = 0; i < MAXQUOTAS; i++) { 3482 if (f2fs_qf_ino(sbi->sb, i)) 3483 sbi->nquota_files++; 3484 } 3485 } 3486 #endif 3487 3488 sb->s_op = &f2fs_sops; 3489 #ifdef CONFIG_FS_ENCRYPTION 3490 sb->s_cop = &f2fs_cryptops; 3491 #endif 3492 #ifdef CONFIG_FS_VERITY 3493 sb->s_vop = &f2fs_verityops; 3494 #endif 3495 sb->s_xattr = f2fs_xattr_handlers; 3496 sb->s_export_op = &f2fs_export_ops; 3497 sb->s_magic = F2FS_SUPER_MAGIC; 3498 sb->s_time_gran = 1; 3499 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3500 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 3501 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 3502 sb->s_iflags |= SB_I_CGROUPWB; 3503 3504 /* init f2fs-specific super block info */ 3505 sbi->valid_super_block = valid_super_block; 3506 init_rwsem(&sbi->gc_lock); 3507 mutex_init(&sbi->writepages); 3508 mutex_init(&sbi->cp_mutex); 3509 init_rwsem(&sbi->node_write); 3510 init_rwsem(&sbi->node_change); 3511 3512 /* disallow all the data/node/meta page writes */ 3513 set_sbi_flag(sbi, SBI_POR_DOING); 3514 spin_lock_init(&sbi->stat_lock); 3515 3516 /* init iostat info */ 3517 spin_lock_init(&sbi->iostat_lock); 3518 sbi->iostat_enable = false; 3519 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS; 3520 3521 for (i = 0; i < NR_PAGE_TYPE; i++) { 3522 int n = (i == META) ? 1: NR_TEMP_TYPE; 3523 int j; 3524 3525 sbi->write_io[i] = 3526 f2fs_kmalloc(sbi, 3527 array_size(n, 3528 sizeof(struct f2fs_bio_info)), 3529 GFP_KERNEL); 3530 if (!sbi->write_io[i]) { 3531 err = -ENOMEM; 3532 goto free_bio_info; 3533 } 3534 3535 for (j = HOT; j < n; j++) { 3536 init_rwsem(&sbi->write_io[i][j].io_rwsem); 3537 sbi->write_io[i][j].sbi = sbi; 3538 sbi->write_io[i][j].bio = NULL; 3539 spin_lock_init(&sbi->write_io[i][j].io_lock); 3540 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 3541 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); 3542 init_rwsem(&sbi->write_io[i][j].bio_list_lock); 3543 } 3544 } 3545 3546 init_rwsem(&sbi->cp_rwsem); 3547 init_rwsem(&sbi->quota_sem); 3548 init_waitqueue_head(&sbi->cp_wait); 3549 init_sb_info(sbi); 3550 3551 err = init_percpu_info(sbi); 3552 if (err) 3553 goto free_bio_info; 3554 3555 if (F2FS_IO_ALIGNED(sbi)) { 3556 sbi->write_io_dummy = 3557 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 3558 if (!sbi->write_io_dummy) { 3559 err = -ENOMEM; 3560 goto free_percpu; 3561 } 3562 } 3563 3564 /* init per sbi slab cache */ 3565 err = f2fs_init_xattr_caches(sbi); 3566 if (err) 3567 goto free_io_dummy; 3568 3569 /* get an inode for meta space */ 3570 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 3571 if (IS_ERR(sbi->meta_inode)) { 3572 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 3573 err = PTR_ERR(sbi->meta_inode); 3574 goto free_xattr_cache; 3575 } 3576 3577 err = f2fs_get_valid_checkpoint(sbi); 3578 if (err) { 3579 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 3580 goto free_meta_inode; 3581 } 3582 3583 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 3584 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3585 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 3586 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3587 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 3588 } 3589 3590 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 3591 set_sbi_flag(sbi, SBI_NEED_FSCK); 3592 3593 /* Initialize device list */ 3594 err = f2fs_scan_devices(sbi); 3595 if (err) { 3596 f2fs_err(sbi, "Failed to find devices"); 3597 goto free_devices; 3598 } 3599 3600 err = f2fs_init_post_read_wq(sbi); 3601 if (err) { 3602 f2fs_err(sbi, "Failed to initialize post read workqueue"); 3603 goto free_devices; 3604 } 3605 3606 sbi->total_valid_node_count = 3607 le32_to_cpu(sbi->ckpt->valid_node_count); 3608 percpu_counter_set(&sbi->total_valid_inode_count, 3609 le32_to_cpu(sbi->ckpt->valid_inode_count)); 3610 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 3611 sbi->total_valid_block_count = 3612 le64_to_cpu(sbi->ckpt->valid_block_count); 3613 sbi->last_valid_block_count = sbi->total_valid_block_count; 3614 sbi->reserved_blocks = 0; 3615 sbi->current_reserved_blocks = 0; 3616 limit_reserve_root(sbi); 3617 adjust_unusable_cap_perc(sbi); 3618 3619 for (i = 0; i < NR_INODE_TYPE; i++) { 3620 INIT_LIST_HEAD(&sbi->inode_list[i]); 3621 spin_lock_init(&sbi->inode_lock[i]); 3622 } 3623 mutex_init(&sbi->flush_lock); 3624 3625 f2fs_init_extent_cache_info(sbi); 3626 3627 f2fs_init_ino_entry_info(sbi); 3628 3629 f2fs_init_fsync_node_info(sbi); 3630 3631 /* setup f2fs internal modules */ 3632 err = f2fs_build_segment_manager(sbi); 3633 if (err) { 3634 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 3635 err); 3636 goto free_sm; 3637 } 3638 err = f2fs_build_node_manager(sbi); 3639 if (err) { 3640 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 3641 err); 3642 goto free_nm; 3643 } 3644 3645 /* For write statistics */ 3646 if (sb->s_bdev->bd_part) 3647 sbi->sectors_written_start = 3648 (u64)part_stat_read(sb->s_bdev->bd_part, 3649 sectors[STAT_WRITE]); 3650 3651 /* Read accumulated write IO statistics if exists */ 3652 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 3653 if (__exist_node_summaries(sbi)) 3654 sbi->kbytes_written = 3655 le64_to_cpu(seg_i->journal->info.kbytes_written); 3656 3657 f2fs_build_gc_manager(sbi); 3658 3659 err = f2fs_build_stats(sbi); 3660 if (err) 3661 goto free_nm; 3662 3663 /* get an inode for node space */ 3664 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 3665 if (IS_ERR(sbi->node_inode)) { 3666 f2fs_err(sbi, "Failed to read node inode"); 3667 err = PTR_ERR(sbi->node_inode); 3668 goto free_stats; 3669 } 3670 3671 /* read root inode and dentry */ 3672 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 3673 if (IS_ERR(root)) { 3674 f2fs_err(sbi, "Failed to read root inode"); 3675 err = PTR_ERR(root); 3676 goto free_node_inode; 3677 } 3678 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 3679 !root->i_size || !root->i_nlink) { 3680 iput(root); 3681 err = -EINVAL; 3682 goto free_node_inode; 3683 } 3684 3685 sb->s_root = d_make_root(root); /* allocate root dentry */ 3686 if (!sb->s_root) { 3687 err = -ENOMEM; 3688 goto free_node_inode; 3689 } 3690 3691 err = f2fs_register_sysfs(sbi); 3692 if (err) 3693 goto free_root_inode; 3694 3695 #ifdef CONFIG_QUOTA 3696 /* Enable quota usage during mount */ 3697 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 3698 err = f2fs_enable_quotas(sb); 3699 if (err) 3700 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 3701 } 3702 #endif 3703 /* if there are any orphan inodes, free them */ 3704 err = f2fs_recover_orphan_inodes(sbi); 3705 if (err) 3706 goto free_meta; 3707 3708 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 3709 goto reset_checkpoint; 3710 3711 /* recover fsynced data */ 3712 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 3713 !test_opt(sbi, NORECOVERY)) { 3714 /* 3715 * mount should be failed, when device has readonly mode, and 3716 * previous checkpoint was not done by clean system shutdown. 3717 */ 3718 if (f2fs_hw_is_readonly(sbi)) { 3719 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3720 err = -EROFS; 3721 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable"); 3722 goto free_meta; 3723 } 3724 f2fs_info(sbi, "write access unavailable, skipping recovery"); 3725 goto reset_checkpoint; 3726 } 3727 3728 if (need_fsck) 3729 set_sbi_flag(sbi, SBI_NEED_FSCK); 3730 3731 if (skip_recovery) 3732 goto reset_checkpoint; 3733 3734 err = f2fs_recover_fsync_data(sbi, false); 3735 if (err < 0) { 3736 if (err != -ENOMEM) 3737 skip_recovery = true; 3738 need_fsck = true; 3739 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 3740 err); 3741 goto free_meta; 3742 } 3743 } else { 3744 err = f2fs_recover_fsync_data(sbi, true); 3745 3746 if (!f2fs_readonly(sb) && err > 0) { 3747 err = -EINVAL; 3748 f2fs_err(sbi, "Need to recover fsync data"); 3749 goto free_meta; 3750 } 3751 } 3752 3753 /* 3754 * If the f2fs is not readonly and fsync data recovery succeeds, 3755 * check zoned block devices' write pointer consistency. 3756 */ 3757 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 3758 err = f2fs_check_write_pointer(sbi); 3759 if (err) 3760 goto free_meta; 3761 } 3762 3763 reset_checkpoint: 3764 /* f2fs_recover_fsync_data() cleared this already */ 3765 clear_sbi_flag(sbi, SBI_POR_DOING); 3766 3767 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 3768 err = f2fs_disable_checkpoint(sbi); 3769 if (err) 3770 goto sync_free_meta; 3771 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 3772 f2fs_enable_checkpoint(sbi); 3773 } 3774 3775 /* 3776 * If filesystem is not mounted as read-only then 3777 * do start the gc_thread. 3778 */ 3779 if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) { 3780 /* After POR, we can run background GC thread.*/ 3781 err = f2fs_start_gc_thread(sbi); 3782 if (err) 3783 goto sync_free_meta; 3784 } 3785 kvfree(options); 3786 3787 /* recover broken superblock */ 3788 if (recovery) { 3789 err = f2fs_commit_super(sbi, true); 3790 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 3791 sbi->valid_super_block ? 1 : 2, err); 3792 } 3793 3794 f2fs_join_shrinker(sbi); 3795 3796 f2fs_tuning_parameters(sbi); 3797 3798 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 3799 cur_cp_version(F2FS_CKPT(sbi))); 3800 f2fs_update_time(sbi, CP_TIME); 3801 f2fs_update_time(sbi, REQ_TIME); 3802 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3803 return 0; 3804 3805 sync_free_meta: 3806 /* safe to flush all the data */ 3807 sync_filesystem(sbi->sb); 3808 retry_cnt = 0; 3809 3810 free_meta: 3811 #ifdef CONFIG_QUOTA 3812 f2fs_truncate_quota_inode_pages(sb); 3813 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 3814 f2fs_quota_off_umount(sbi->sb); 3815 #endif 3816 /* 3817 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 3818 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 3819 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 3820 * falls into an infinite loop in f2fs_sync_meta_pages(). 3821 */ 3822 truncate_inode_pages_final(META_MAPPING(sbi)); 3823 /* evict some inodes being cached by GC */ 3824 evict_inodes(sb); 3825 f2fs_unregister_sysfs(sbi); 3826 free_root_inode: 3827 dput(sb->s_root); 3828 sb->s_root = NULL; 3829 free_node_inode: 3830 f2fs_release_ino_entry(sbi, true); 3831 truncate_inode_pages_final(NODE_MAPPING(sbi)); 3832 iput(sbi->node_inode); 3833 sbi->node_inode = NULL; 3834 free_stats: 3835 f2fs_destroy_stats(sbi); 3836 free_nm: 3837 f2fs_destroy_node_manager(sbi); 3838 free_sm: 3839 f2fs_destroy_segment_manager(sbi); 3840 f2fs_destroy_post_read_wq(sbi); 3841 free_devices: 3842 destroy_device_list(sbi); 3843 kvfree(sbi->ckpt); 3844 free_meta_inode: 3845 make_bad_inode(sbi->meta_inode); 3846 iput(sbi->meta_inode); 3847 sbi->meta_inode = NULL; 3848 free_xattr_cache: 3849 f2fs_destroy_xattr_caches(sbi); 3850 free_io_dummy: 3851 mempool_destroy(sbi->write_io_dummy); 3852 free_percpu: 3853 destroy_percpu_info(sbi); 3854 free_bio_info: 3855 for (i = 0; i < NR_PAGE_TYPE; i++) 3856 kvfree(sbi->write_io[i]); 3857 3858 #ifdef CONFIG_UNICODE 3859 utf8_unload(sbi->s_encoding); 3860 #endif 3861 free_options: 3862 #ifdef CONFIG_QUOTA 3863 for (i = 0; i < MAXQUOTAS; i++) 3864 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 3865 #endif 3866 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 3867 kvfree(options); 3868 free_sb_buf: 3869 kfree(raw_super); 3870 free_sbi: 3871 if (sbi->s_chksum_driver) 3872 crypto_free_shash(sbi->s_chksum_driver); 3873 kfree(sbi); 3874 3875 /* give only one another chance */ 3876 if (retry_cnt > 0 && skip_recovery) { 3877 retry_cnt--; 3878 shrink_dcache_sb(sb); 3879 goto try_onemore; 3880 } 3881 return err; 3882 } 3883 3884 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 3885 const char *dev_name, void *data) 3886 { 3887 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 3888 } 3889 3890 static void kill_f2fs_super(struct super_block *sb) 3891 { 3892 if (sb->s_root) { 3893 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3894 3895 set_sbi_flag(sbi, SBI_IS_CLOSE); 3896 f2fs_stop_gc_thread(sbi); 3897 f2fs_stop_discard_thread(sbi); 3898 3899 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 3900 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3901 struct cp_control cpc = { 3902 .reason = CP_UMOUNT, 3903 }; 3904 f2fs_write_checkpoint(sbi, &cpc); 3905 } 3906 3907 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 3908 sb->s_flags &= ~SB_RDONLY; 3909 } 3910 kill_block_super(sb); 3911 } 3912 3913 static struct file_system_type f2fs_fs_type = { 3914 .owner = THIS_MODULE, 3915 .name = "f2fs", 3916 .mount = f2fs_mount, 3917 .kill_sb = kill_f2fs_super, 3918 .fs_flags = FS_REQUIRES_DEV, 3919 }; 3920 MODULE_ALIAS_FS("f2fs"); 3921 3922 static int __init init_inodecache(void) 3923 { 3924 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 3925 sizeof(struct f2fs_inode_info), 0, 3926 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 3927 if (!f2fs_inode_cachep) 3928 return -ENOMEM; 3929 return 0; 3930 } 3931 3932 static void destroy_inodecache(void) 3933 { 3934 /* 3935 * Make sure all delayed rcu free inodes are flushed before we 3936 * destroy cache. 3937 */ 3938 rcu_barrier(); 3939 kmem_cache_destroy(f2fs_inode_cachep); 3940 } 3941 3942 static int __init init_f2fs_fs(void) 3943 { 3944 int err; 3945 3946 if (PAGE_SIZE != F2FS_BLKSIZE) { 3947 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 3948 PAGE_SIZE, F2FS_BLKSIZE); 3949 return -EINVAL; 3950 } 3951 3952 f2fs_build_trace_ios(); 3953 3954 err = init_inodecache(); 3955 if (err) 3956 goto fail; 3957 err = f2fs_create_node_manager_caches(); 3958 if (err) 3959 goto free_inodecache; 3960 err = f2fs_create_segment_manager_caches(); 3961 if (err) 3962 goto free_node_manager_caches; 3963 err = f2fs_create_checkpoint_caches(); 3964 if (err) 3965 goto free_segment_manager_caches; 3966 err = f2fs_create_extent_cache(); 3967 if (err) 3968 goto free_checkpoint_caches; 3969 err = f2fs_init_sysfs(); 3970 if (err) 3971 goto free_extent_cache; 3972 err = register_shrinker(&f2fs_shrinker_info); 3973 if (err) 3974 goto free_sysfs; 3975 err = register_filesystem(&f2fs_fs_type); 3976 if (err) 3977 goto free_shrinker; 3978 f2fs_create_root_stats(); 3979 err = f2fs_init_post_read_processing(); 3980 if (err) 3981 goto free_root_stats; 3982 err = f2fs_init_bio_entry_cache(); 3983 if (err) 3984 goto free_post_read; 3985 err = f2fs_init_bioset(); 3986 if (err) 3987 goto free_bio_enrty_cache; 3988 err = f2fs_init_compress_mempool(); 3989 if (err) 3990 goto free_bioset; 3991 return 0; 3992 free_bioset: 3993 f2fs_destroy_bioset(); 3994 free_bio_enrty_cache: 3995 f2fs_destroy_bio_entry_cache(); 3996 free_post_read: 3997 f2fs_destroy_post_read_processing(); 3998 free_root_stats: 3999 f2fs_destroy_root_stats(); 4000 unregister_filesystem(&f2fs_fs_type); 4001 free_shrinker: 4002 unregister_shrinker(&f2fs_shrinker_info); 4003 free_sysfs: 4004 f2fs_exit_sysfs(); 4005 free_extent_cache: 4006 f2fs_destroy_extent_cache(); 4007 free_checkpoint_caches: 4008 f2fs_destroy_checkpoint_caches(); 4009 free_segment_manager_caches: 4010 f2fs_destroy_segment_manager_caches(); 4011 free_node_manager_caches: 4012 f2fs_destroy_node_manager_caches(); 4013 free_inodecache: 4014 destroy_inodecache(); 4015 fail: 4016 return err; 4017 } 4018 4019 static void __exit exit_f2fs_fs(void) 4020 { 4021 f2fs_destroy_compress_mempool(); 4022 f2fs_destroy_bioset(); 4023 f2fs_destroy_bio_entry_cache(); 4024 f2fs_destroy_post_read_processing(); 4025 f2fs_destroy_root_stats(); 4026 unregister_filesystem(&f2fs_fs_type); 4027 unregister_shrinker(&f2fs_shrinker_info); 4028 f2fs_exit_sysfs(); 4029 f2fs_destroy_extent_cache(); 4030 f2fs_destroy_checkpoint_caches(); 4031 f2fs_destroy_segment_manager_caches(); 4032 f2fs_destroy_node_manager_caches(); 4033 destroy_inodecache(); 4034 f2fs_destroy_trace_ios(); 4035 } 4036 4037 module_init(init_f2fs_fs) 4038 module_exit(exit_f2fs_fs) 4039 4040 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4041 MODULE_DESCRIPTION("Flash Friendly File System"); 4042 MODULE_LICENSE("GPL"); 4043 4044