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_atgc, 150 Opt_err, 151 }; 152 153 static match_table_t f2fs_tokens = { 154 {Opt_gc_background, "background_gc=%s"}, 155 {Opt_disable_roll_forward, "disable_roll_forward"}, 156 {Opt_norecovery, "norecovery"}, 157 {Opt_discard, "discard"}, 158 {Opt_nodiscard, "nodiscard"}, 159 {Opt_noheap, "no_heap"}, 160 {Opt_heap, "heap"}, 161 {Opt_user_xattr, "user_xattr"}, 162 {Opt_nouser_xattr, "nouser_xattr"}, 163 {Opt_acl, "acl"}, 164 {Opt_noacl, "noacl"}, 165 {Opt_active_logs, "active_logs=%u"}, 166 {Opt_disable_ext_identify, "disable_ext_identify"}, 167 {Opt_inline_xattr, "inline_xattr"}, 168 {Opt_noinline_xattr, "noinline_xattr"}, 169 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 170 {Opt_inline_data, "inline_data"}, 171 {Opt_inline_dentry, "inline_dentry"}, 172 {Opt_noinline_dentry, "noinline_dentry"}, 173 {Opt_flush_merge, "flush_merge"}, 174 {Opt_noflush_merge, "noflush_merge"}, 175 {Opt_nobarrier, "nobarrier"}, 176 {Opt_fastboot, "fastboot"}, 177 {Opt_extent_cache, "extent_cache"}, 178 {Opt_noextent_cache, "noextent_cache"}, 179 {Opt_noinline_data, "noinline_data"}, 180 {Opt_data_flush, "data_flush"}, 181 {Opt_reserve_root, "reserve_root=%u"}, 182 {Opt_resgid, "resgid=%u"}, 183 {Opt_resuid, "resuid=%u"}, 184 {Opt_mode, "mode=%s"}, 185 {Opt_io_size_bits, "io_bits=%u"}, 186 {Opt_fault_injection, "fault_injection=%u"}, 187 {Opt_fault_type, "fault_type=%u"}, 188 {Opt_lazytime, "lazytime"}, 189 {Opt_nolazytime, "nolazytime"}, 190 {Opt_quota, "quota"}, 191 {Opt_noquota, "noquota"}, 192 {Opt_usrquota, "usrquota"}, 193 {Opt_grpquota, "grpquota"}, 194 {Opt_prjquota, "prjquota"}, 195 {Opt_usrjquota, "usrjquota=%s"}, 196 {Opt_grpjquota, "grpjquota=%s"}, 197 {Opt_prjjquota, "prjjquota=%s"}, 198 {Opt_offusrjquota, "usrjquota="}, 199 {Opt_offgrpjquota, "grpjquota="}, 200 {Opt_offprjjquota, "prjjquota="}, 201 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 202 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 203 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 204 {Opt_whint, "whint_mode=%s"}, 205 {Opt_alloc, "alloc_mode=%s"}, 206 {Opt_fsync, "fsync_mode=%s"}, 207 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 208 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 209 {Opt_inlinecrypt, "inlinecrypt"}, 210 {Opt_checkpoint_disable, "checkpoint=disable"}, 211 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 212 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 213 {Opt_checkpoint_enable, "checkpoint=enable"}, 214 {Opt_compress_algorithm, "compress_algorithm=%s"}, 215 {Opt_compress_log_size, "compress_log_size=%u"}, 216 {Opt_compress_extension, "compress_extension=%s"}, 217 {Opt_atgc, "atgc"}, 218 {Opt_err, NULL}, 219 }; 220 221 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 222 { 223 struct va_format vaf; 224 va_list args; 225 int level; 226 227 va_start(args, fmt); 228 229 level = printk_get_level(fmt); 230 vaf.fmt = printk_skip_level(fmt); 231 vaf.va = &args; 232 printk("%c%cF2FS-fs (%s): %pV\n", 233 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 234 235 va_end(args); 236 } 237 238 #ifdef CONFIG_UNICODE 239 static const struct f2fs_sb_encodings { 240 __u16 magic; 241 char *name; 242 char *version; 243 } f2fs_sb_encoding_map[] = { 244 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"}, 245 }; 246 247 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb, 248 const struct f2fs_sb_encodings **encoding, 249 __u16 *flags) 250 { 251 __u16 magic = le16_to_cpu(sb->s_encoding); 252 int i; 253 254 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 255 if (magic == f2fs_sb_encoding_map[i].magic) 256 break; 257 258 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map)) 259 return -EINVAL; 260 261 *encoding = &f2fs_sb_encoding_map[i]; 262 *flags = le16_to_cpu(sb->s_encoding_flags); 263 264 return 0; 265 } 266 #endif 267 268 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 269 { 270 block_t limit = min((sbi->user_block_count << 1) / 1000, 271 sbi->user_block_count - sbi->reserved_blocks); 272 273 /* limit is 0.2% */ 274 if (test_opt(sbi, RESERVE_ROOT) && 275 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 276 F2FS_OPTION(sbi).root_reserved_blocks = limit; 277 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 278 F2FS_OPTION(sbi).root_reserved_blocks); 279 } 280 if (!test_opt(sbi, RESERVE_ROOT) && 281 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 282 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 283 !gid_eq(F2FS_OPTION(sbi).s_resgid, 284 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 285 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 286 from_kuid_munged(&init_user_ns, 287 F2FS_OPTION(sbi).s_resuid), 288 from_kgid_munged(&init_user_ns, 289 F2FS_OPTION(sbi).s_resgid)); 290 } 291 292 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 293 { 294 if (!F2FS_OPTION(sbi).unusable_cap_perc) 295 return; 296 297 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 298 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 299 else 300 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 301 F2FS_OPTION(sbi).unusable_cap_perc; 302 303 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 304 F2FS_OPTION(sbi).unusable_cap, 305 F2FS_OPTION(sbi).unusable_cap_perc); 306 } 307 308 static void init_once(void *foo) 309 { 310 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 311 312 inode_init_once(&fi->vfs_inode); 313 } 314 315 #ifdef CONFIG_QUOTA 316 static const char * const quotatypes[] = INITQFNAMES; 317 #define QTYPE2NAME(t) (quotatypes[t]) 318 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 319 substring_t *args) 320 { 321 struct f2fs_sb_info *sbi = F2FS_SB(sb); 322 char *qname; 323 int ret = -EINVAL; 324 325 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 326 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 327 return -EINVAL; 328 } 329 if (f2fs_sb_has_quota_ino(sbi)) { 330 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 331 return 0; 332 } 333 334 qname = match_strdup(args); 335 if (!qname) { 336 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 337 return -ENOMEM; 338 } 339 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 340 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 341 ret = 0; 342 else 343 f2fs_err(sbi, "%s quota file already specified", 344 QTYPE2NAME(qtype)); 345 goto errout; 346 } 347 if (strchr(qname, '/')) { 348 f2fs_err(sbi, "quotafile must be on filesystem root"); 349 goto errout; 350 } 351 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 352 set_opt(sbi, QUOTA); 353 return 0; 354 errout: 355 kfree(qname); 356 return ret; 357 } 358 359 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 360 { 361 struct f2fs_sb_info *sbi = F2FS_SB(sb); 362 363 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 364 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 365 return -EINVAL; 366 } 367 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 368 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 369 return 0; 370 } 371 372 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 373 { 374 /* 375 * We do the test below only for project quotas. 'usrquota' and 376 * 'grpquota' mount options are allowed even without quota feature 377 * to support legacy quotas in quota files. 378 */ 379 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 380 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 381 return -1; 382 } 383 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 384 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 385 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 386 if (test_opt(sbi, USRQUOTA) && 387 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 388 clear_opt(sbi, USRQUOTA); 389 390 if (test_opt(sbi, GRPQUOTA) && 391 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 392 clear_opt(sbi, GRPQUOTA); 393 394 if (test_opt(sbi, PRJQUOTA) && 395 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 396 clear_opt(sbi, PRJQUOTA); 397 398 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 399 test_opt(sbi, PRJQUOTA)) { 400 f2fs_err(sbi, "old and new quota format mixing"); 401 return -1; 402 } 403 404 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 405 f2fs_err(sbi, "journaled quota format not specified"); 406 return -1; 407 } 408 } 409 410 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 411 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 412 F2FS_OPTION(sbi).s_jquota_fmt = 0; 413 } 414 return 0; 415 } 416 #endif 417 418 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 419 const char *opt, 420 const substring_t *arg, 421 bool is_remount) 422 { 423 struct f2fs_sb_info *sbi = F2FS_SB(sb); 424 #ifdef CONFIG_FS_ENCRYPTION 425 int err; 426 427 if (!f2fs_sb_has_encrypt(sbi)) { 428 f2fs_err(sbi, "Encrypt feature is off"); 429 return -EINVAL; 430 } 431 432 /* 433 * This mount option is just for testing, and it's not worthwhile to 434 * implement the extra complexity (e.g. RCU protection) that would be 435 * needed to allow it to be set or changed during remount. We do allow 436 * it to be specified during remount, but only if there is no change. 437 */ 438 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) { 439 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 440 return -EINVAL; 441 } 442 err = fscrypt_set_test_dummy_encryption( 443 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy); 444 if (err) { 445 if (err == -EEXIST) 446 f2fs_warn(sbi, 447 "Can't change test_dummy_encryption on remount"); 448 else if (err == -EINVAL) 449 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 450 opt); 451 else 452 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 453 opt, err); 454 return -EINVAL; 455 } 456 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 457 #else 458 f2fs_warn(sbi, "Test dummy encryption mount option ignored"); 459 #endif 460 return 0; 461 } 462 463 static int parse_options(struct super_block *sb, char *options, bool is_remount) 464 { 465 struct f2fs_sb_info *sbi = F2FS_SB(sb); 466 substring_t args[MAX_OPT_ARGS]; 467 #ifdef CONFIG_F2FS_FS_COMPRESSION 468 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 469 int ext_cnt; 470 #endif 471 char *p, *name; 472 int arg = 0; 473 kuid_t uid; 474 kgid_t gid; 475 int ret; 476 477 if (!options) 478 return 0; 479 480 while ((p = strsep(&options, ",")) != NULL) { 481 int token; 482 if (!*p) 483 continue; 484 /* 485 * Initialize args struct so we know whether arg was 486 * found; some options take optional arguments. 487 */ 488 args[0].to = args[0].from = NULL; 489 token = match_token(p, f2fs_tokens, args); 490 491 switch (token) { 492 case Opt_gc_background: 493 name = match_strdup(&args[0]); 494 495 if (!name) 496 return -ENOMEM; 497 if (!strcmp(name, "on")) { 498 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 499 } else if (!strcmp(name, "off")) { 500 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 501 } else if (!strcmp(name, "sync")) { 502 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 503 } else { 504 kfree(name); 505 return -EINVAL; 506 } 507 kfree(name); 508 break; 509 case Opt_disable_roll_forward: 510 set_opt(sbi, DISABLE_ROLL_FORWARD); 511 break; 512 case Opt_norecovery: 513 /* this option mounts f2fs with ro */ 514 set_opt(sbi, NORECOVERY); 515 if (!f2fs_readonly(sb)) 516 return -EINVAL; 517 break; 518 case Opt_discard: 519 set_opt(sbi, DISCARD); 520 break; 521 case Opt_nodiscard: 522 if (f2fs_sb_has_blkzoned(sbi)) { 523 f2fs_warn(sbi, "discard is required for zoned block devices"); 524 return -EINVAL; 525 } 526 clear_opt(sbi, DISCARD); 527 break; 528 case Opt_noheap: 529 set_opt(sbi, NOHEAP); 530 break; 531 case Opt_heap: 532 clear_opt(sbi, NOHEAP); 533 break; 534 #ifdef CONFIG_F2FS_FS_XATTR 535 case Opt_user_xattr: 536 set_opt(sbi, XATTR_USER); 537 break; 538 case Opt_nouser_xattr: 539 clear_opt(sbi, XATTR_USER); 540 break; 541 case Opt_inline_xattr: 542 set_opt(sbi, INLINE_XATTR); 543 break; 544 case Opt_noinline_xattr: 545 clear_opt(sbi, INLINE_XATTR); 546 break; 547 case Opt_inline_xattr_size: 548 if (args->from && match_int(args, &arg)) 549 return -EINVAL; 550 set_opt(sbi, INLINE_XATTR_SIZE); 551 F2FS_OPTION(sbi).inline_xattr_size = arg; 552 break; 553 #else 554 case Opt_user_xattr: 555 f2fs_info(sbi, "user_xattr options not supported"); 556 break; 557 case Opt_nouser_xattr: 558 f2fs_info(sbi, "nouser_xattr options not supported"); 559 break; 560 case Opt_inline_xattr: 561 f2fs_info(sbi, "inline_xattr options not supported"); 562 break; 563 case Opt_noinline_xattr: 564 f2fs_info(sbi, "noinline_xattr options not supported"); 565 break; 566 #endif 567 #ifdef CONFIG_F2FS_FS_POSIX_ACL 568 case Opt_acl: 569 set_opt(sbi, POSIX_ACL); 570 break; 571 case Opt_noacl: 572 clear_opt(sbi, POSIX_ACL); 573 break; 574 #else 575 case Opt_acl: 576 f2fs_info(sbi, "acl options not supported"); 577 break; 578 case Opt_noacl: 579 f2fs_info(sbi, "noacl options not supported"); 580 break; 581 #endif 582 case Opt_active_logs: 583 if (args->from && match_int(args, &arg)) 584 return -EINVAL; 585 if (arg != 2 && arg != 4 && 586 arg != NR_CURSEG_PERSIST_TYPE) 587 return -EINVAL; 588 F2FS_OPTION(sbi).active_logs = arg; 589 break; 590 case Opt_disable_ext_identify: 591 set_opt(sbi, DISABLE_EXT_IDENTIFY); 592 break; 593 case Opt_inline_data: 594 set_opt(sbi, INLINE_DATA); 595 break; 596 case Opt_inline_dentry: 597 set_opt(sbi, INLINE_DENTRY); 598 break; 599 case Opt_noinline_dentry: 600 clear_opt(sbi, INLINE_DENTRY); 601 break; 602 case Opt_flush_merge: 603 set_opt(sbi, FLUSH_MERGE); 604 break; 605 case Opt_noflush_merge: 606 clear_opt(sbi, FLUSH_MERGE); 607 break; 608 case Opt_nobarrier: 609 set_opt(sbi, NOBARRIER); 610 break; 611 case Opt_fastboot: 612 set_opt(sbi, FASTBOOT); 613 break; 614 case Opt_extent_cache: 615 set_opt(sbi, EXTENT_CACHE); 616 break; 617 case Opt_noextent_cache: 618 clear_opt(sbi, EXTENT_CACHE); 619 break; 620 case Opt_noinline_data: 621 clear_opt(sbi, INLINE_DATA); 622 break; 623 case Opt_data_flush: 624 set_opt(sbi, DATA_FLUSH); 625 break; 626 case Opt_reserve_root: 627 if (args->from && match_int(args, &arg)) 628 return -EINVAL; 629 if (test_opt(sbi, RESERVE_ROOT)) { 630 f2fs_info(sbi, "Preserve previous reserve_root=%u", 631 F2FS_OPTION(sbi).root_reserved_blocks); 632 } else { 633 F2FS_OPTION(sbi).root_reserved_blocks = arg; 634 set_opt(sbi, RESERVE_ROOT); 635 } 636 break; 637 case Opt_resuid: 638 if (args->from && match_int(args, &arg)) 639 return -EINVAL; 640 uid = make_kuid(current_user_ns(), arg); 641 if (!uid_valid(uid)) { 642 f2fs_err(sbi, "Invalid uid value %d", arg); 643 return -EINVAL; 644 } 645 F2FS_OPTION(sbi).s_resuid = uid; 646 break; 647 case Opt_resgid: 648 if (args->from && match_int(args, &arg)) 649 return -EINVAL; 650 gid = make_kgid(current_user_ns(), arg); 651 if (!gid_valid(gid)) { 652 f2fs_err(sbi, "Invalid gid value %d", arg); 653 return -EINVAL; 654 } 655 F2FS_OPTION(sbi).s_resgid = gid; 656 break; 657 case Opt_mode: 658 name = match_strdup(&args[0]); 659 660 if (!name) 661 return -ENOMEM; 662 if (!strcmp(name, "adaptive")) { 663 if (f2fs_sb_has_blkzoned(sbi)) { 664 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 665 kfree(name); 666 return -EINVAL; 667 } 668 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 669 } else if (!strcmp(name, "lfs")) { 670 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 671 } else { 672 kfree(name); 673 return -EINVAL; 674 } 675 kfree(name); 676 break; 677 case Opt_io_size_bits: 678 if (args->from && match_int(args, &arg)) 679 return -EINVAL; 680 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) { 681 f2fs_warn(sbi, "Not support %d, larger than %d", 682 1 << arg, BIO_MAX_PAGES); 683 return -EINVAL; 684 } 685 F2FS_OPTION(sbi).write_io_size_bits = arg; 686 break; 687 #ifdef CONFIG_F2FS_FAULT_INJECTION 688 case Opt_fault_injection: 689 if (args->from && match_int(args, &arg)) 690 return -EINVAL; 691 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 692 set_opt(sbi, FAULT_INJECTION); 693 break; 694 695 case Opt_fault_type: 696 if (args->from && match_int(args, &arg)) 697 return -EINVAL; 698 f2fs_build_fault_attr(sbi, 0, arg); 699 set_opt(sbi, FAULT_INJECTION); 700 break; 701 #else 702 case Opt_fault_injection: 703 f2fs_info(sbi, "fault_injection options not supported"); 704 break; 705 706 case Opt_fault_type: 707 f2fs_info(sbi, "fault_type options not supported"); 708 break; 709 #endif 710 case Opt_lazytime: 711 sb->s_flags |= SB_LAZYTIME; 712 break; 713 case Opt_nolazytime: 714 sb->s_flags &= ~SB_LAZYTIME; 715 break; 716 #ifdef CONFIG_QUOTA 717 case Opt_quota: 718 case Opt_usrquota: 719 set_opt(sbi, USRQUOTA); 720 break; 721 case Opt_grpquota: 722 set_opt(sbi, GRPQUOTA); 723 break; 724 case Opt_prjquota: 725 set_opt(sbi, PRJQUOTA); 726 break; 727 case Opt_usrjquota: 728 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 729 if (ret) 730 return ret; 731 break; 732 case Opt_grpjquota: 733 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 734 if (ret) 735 return ret; 736 break; 737 case Opt_prjjquota: 738 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 739 if (ret) 740 return ret; 741 break; 742 case Opt_offusrjquota: 743 ret = f2fs_clear_qf_name(sb, USRQUOTA); 744 if (ret) 745 return ret; 746 break; 747 case Opt_offgrpjquota: 748 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 749 if (ret) 750 return ret; 751 break; 752 case Opt_offprjjquota: 753 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 754 if (ret) 755 return ret; 756 break; 757 case Opt_jqfmt_vfsold: 758 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 759 break; 760 case Opt_jqfmt_vfsv0: 761 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 762 break; 763 case Opt_jqfmt_vfsv1: 764 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 765 break; 766 case Opt_noquota: 767 clear_opt(sbi, QUOTA); 768 clear_opt(sbi, USRQUOTA); 769 clear_opt(sbi, GRPQUOTA); 770 clear_opt(sbi, PRJQUOTA); 771 break; 772 #else 773 case Opt_quota: 774 case Opt_usrquota: 775 case Opt_grpquota: 776 case Opt_prjquota: 777 case Opt_usrjquota: 778 case Opt_grpjquota: 779 case Opt_prjjquota: 780 case Opt_offusrjquota: 781 case Opt_offgrpjquota: 782 case Opt_offprjjquota: 783 case Opt_jqfmt_vfsold: 784 case Opt_jqfmt_vfsv0: 785 case Opt_jqfmt_vfsv1: 786 case Opt_noquota: 787 f2fs_info(sbi, "quota operations not supported"); 788 break; 789 #endif 790 case Opt_whint: 791 name = match_strdup(&args[0]); 792 if (!name) 793 return -ENOMEM; 794 if (!strcmp(name, "user-based")) { 795 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER; 796 } else if (!strcmp(name, "off")) { 797 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 798 } else if (!strcmp(name, "fs-based")) { 799 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS; 800 } else { 801 kfree(name); 802 return -EINVAL; 803 } 804 kfree(name); 805 break; 806 case Opt_alloc: 807 name = match_strdup(&args[0]); 808 if (!name) 809 return -ENOMEM; 810 811 if (!strcmp(name, "default")) { 812 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 813 } else if (!strcmp(name, "reuse")) { 814 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 815 } else { 816 kfree(name); 817 return -EINVAL; 818 } 819 kfree(name); 820 break; 821 case Opt_fsync: 822 name = match_strdup(&args[0]); 823 if (!name) 824 return -ENOMEM; 825 if (!strcmp(name, "posix")) { 826 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 827 } else if (!strcmp(name, "strict")) { 828 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 829 } else if (!strcmp(name, "nobarrier")) { 830 F2FS_OPTION(sbi).fsync_mode = 831 FSYNC_MODE_NOBARRIER; 832 } else { 833 kfree(name); 834 return -EINVAL; 835 } 836 kfree(name); 837 break; 838 case Opt_test_dummy_encryption: 839 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 840 is_remount); 841 if (ret) 842 return ret; 843 break; 844 case Opt_inlinecrypt: 845 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 846 sb->s_flags |= SB_INLINECRYPT; 847 #else 848 f2fs_info(sbi, "inline encryption not supported"); 849 #endif 850 break; 851 case Opt_checkpoint_disable_cap_perc: 852 if (args->from && match_int(args, &arg)) 853 return -EINVAL; 854 if (arg < 0 || arg > 100) 855 return -EINVAL; 856 F2FS_OPTION(sbi).unusable_cap_perc = arg; 857 set_opt(sbi, DISABLE_CHECKPOINT); 858 break; 859 case Opt_checkpoint_disable_cap: 860 if (args->from && match_int(args, &arg)) 861 return -EINVAL; 862 F2FS_OPTION(sbi).unusable_cap = arg; 863 set_opt(sbi, DISABLE_CHECKPOINT); 864 break; 865 case Opt_checkpoint_disable: 866 set_opt(sbi, DISABLE_CHECKPOINT); 867 break; 868 case Opt_checkpoint_enable: 869 clear_opt(sbi, DISABLE_CHECKPOINT); 870 break; 871 #ifdef CONFIG_F2FS_FS_COMPRESSION 872 case Opt_compress_algorithm: 873 if (!f2fs_sb_has_compression(sbi)) { 874 f2fs_info(sbi, "Image doesn't support compression"); 875 break; 876 } 877 name = match_strdup(&args[0]); 878 if (!name) 879 return -ENOMEM; 880 if (!strcmp(name, "lzo")) { 881 F2FS_OPTION(sbi).compress_algorithm = 882 COMPRESS_LZO; 883 } else if (!strcmp(name, "lz4")) { 884 F2FS_OPTION(sbi).compress_algorithm = 885 COMPRESS_LZ4; 886 } else if (!strcmp(name, "zstd")) { 887 F2FS_OPTION(sbi).compress_algorithm = 888 COMPRESS_ZSTD; 889 } else if (!strcmp(name, "lzo-rle")) { 890 F2FS_OPTION(sbi).compress_algorithm = 891 COMPRESS_LZORLE; 892 } else { 893 kfree(name); 894 return -EINVAL; 895 } 896 kfree(name); 897 break; 898 case Opt_compress_log_size: 899 if (!f2fs_sb_has_compression(sbi)) { 900 f2fs_info(sbi, "Image doesn't support compression"); 901 break; 902 } 903 if (args->from && match_int(args, &arg)) 904 return -EINVAL; 905 if (arg < MIN_COMPRESS_LOG_SIZE || 906 arg > MAX_COMPRESS_LOG_SIZE) { 907 f2fs_err(sbi, 908 "Compress cluster log size is out of range"); 909 return -EINVAL; 910 } 911 F2FS_OPTION(sbi).compress_log_size = arg; 912 break; 913 case Opt_compress_extension: 914 if (!f2fs_sb_has_compression(sbi)) { 915 f2fs_info(sbi, "Image doesn't support compression"); 916 break; 917 } 918 name = match_strdup(&args[0]); 919 if (!name) 920 return -ENOMEM; 921 922 ext = F2FS_OPTION(sbi).extensions; 923 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 924 925 if (strlen(name) >= F2FS_EXTENSION_LEN || 926 ext_cnt >= COMPRESS_EXT_NUM) { 927 f2fs_err(sbi, 928 "invalid extension length/number"); 929 kfree(name); 930 return -EINVAL; 931 } 932 933 strcpy(ext[ext_cnt], name); 934 F2FS_OPTION(sbi).compress_ext_cnt++; 935 kfree(name); 936 break; 937 #else 938 case Opt_compress_algorithm: 939 case Opt_compress_log_size: 940 case Opt_compress_extension: 941 f2fs_info(sbi, "compression options not supported"); 942 break; 943 #endif 944 case Opt_atgc: 945 set_opt(sbi, ATGC); 946 break; 947 default: 948 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 949 p); 950 return -EINVAL; 951 } 952 } 953 #ifdef CONFIG_QUOTA 954 if (f2fs_check_quota_options(sbi)) 955 return -EINVAL; 956 #else 957 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 958 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 959 return -EINVAL; 960 } 961 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 962 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 963 return -EINVAL; 964 } 965 #endif 966 #ifndef CONFIG_UNICODE 967 if (f2fs_sb_has_casefold(sbi)) { 968 f2fs_err(sbi, 969 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 970 return -EINVAL; 971 } 972 #endif 973 /* 974 * The BLKZONED feature indicates that the drive was formatted with 975 * zone alignment optimization. This is optional for host-aware 976 * devices, but mandatory for host-managed zoned block devices. 977 */ 978 #ifndef CONFIG_BLK_DEV_ZONED 979 if (f2fs_sb_has_blkzoned(sbi)) { 980 f2fs_err(sbi, "Zoned block device support is not enabled"); 981 return -EINVAL; 982 } 983 #endif 984 985 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 986 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO", 987 F2FS_IO_SIZE_KB(sbi)); 988 return -EINVAL; 989 } 990 991 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 992 int min_size, max_size; 993 994 if (!f2fs_sb_has_extra_attr(sbi) || 995 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 996 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 997 return -EINVAL; 998 } 999 if (!test_opt(sbi, INLINE_XATTR)) { 1000 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1001 return -EINVAL; 1002 } 1003 1004 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 1005 max_size = MAX_INLINE_XATTR_SIZE; 1006 1007 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1008 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1009 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1010 min_size, max_size); 1011 return -EINVAL; 1012 } 1013 } 1014 1015 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1016 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n"); 1017 return -EINVAL; 1018 } 1019 1020 /* Not pass down write hints if the number of active logs is lesser 1021 * than NR_CURSEG_PERSIST_TYPE. 1022 */ 1023 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE) 1024 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1025 return 0; 1026 } 1027 1028 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1029 { 1030 struct f2fs_inode_info *fi; 1031 1032 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); 1033 if (!fi) 1034 return NULL; 1035 1036 init_once((void *) fi); 1037 1038 /* Initialize f2fs-specific inode info */ 1039 atomic_set(&fi->dirty_pages, 0); 1040 atomic_set(&fi->i_compr_blocks, 0); 1041 init_rwsem(&fi->i_sem); 1042 spin_lock_init(&fi->i_size_lock); 1043 INIT_LIST_HEAD(&fi->dirty_list); 1044 INIT_LIST_HEAD(&fi->gdirty_list); 1045 INIT_LIST_HEAD(&fi->inmem_ilist); 1046 INIT_LIST_HEAD(&fi->inmem_pages); 1047 mutex_init(&fi->inmem_lock); 1048 init_rwsem(&fi->i_gc_rwsem[READ]); 1049 init_rwsem(&fi->i_gc_rwsem[WRITE]); 1050 init_rwsem(&fi->i_mmap_sem); 1051 init_rwsem(&fi->i_xattr_sem); 1052 1053 /* Will be used by directory only */ 1054 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1055 1056 fi->ra_offset = -1; 1057 1058 return &fi->vfs_inode; 1059 } 1060 1061 static int f2fs_drop_inode(struct inode *inode) 1062 { 1063 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1064 int ret; 1065 1066 /* 1067 * during filesystem shutdown, if checkpoint is disabled, 1068 * drop useless meta/node dirty pages. 1069 */ 1070 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1071 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1072 inode->i_ino == F2FS_META_INO(sbi)) { 1073 trace_f2fs_drop_inode(inode, 1); 1074 return 1; 1075 } 1076 } 1077 1078 /* 1079 * This is to avoid a deadlock condition like below. 1080 * writeback_single_inode(inode) 1081 * - f2fs_write_data_page 1082 * - f2fs_gc -> iput -> evict 1083 * - inode_wait_for_writeback(inode) 1084 */ 1085 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1086 if (!inode->i_nlink && !is_bad_inode(inode)) { 1087 /* to avoid evict_inode call simultaneously */ 1088 atomic_inc(&inode->i_count); 1089 spin_unlock(&inode->i_lock); 1090 1091 /* some remained atomic pages should discarded */ 1092 if (f2fs_is_atomic_file(inode)) 1093 f2fs_drop_inmem_pages(inode); 1094 1095 /* should remain fi->extent_tree for writepage */ 1096 f2fs_destroy_extent_node(inode); 1097 1098 sb_start_intwrite(inode->i_sb); 1099 f2fs_i_size_write(inode, 0); 1100 1101 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1102 inode, NULL, 0, DATA); 1103 truncate_inode_pages_final(inode->i_mapping); 1104 1105 if (F2FS_HAS_BLOCKS(inode)) 1106 f2fs_truncate(inode); 1107 1108 sb_end_intwrite(inode->i_sb); 1109 1110 spin_lock(&inode->i_lock); 1111 atomic_dec(&inode->i_count); 1112 } 1113 trace_f2fs_drop_inode(inode, 0); 1114 return 0; 1115 } 1116 ret = generic_drop_inode(inode); 1117 if (!ret) 1118 ret = fscrypt_drop_inode(inode); 1119 trace_f2fs_drop_inode(inode, ret); 1120 return ret; 1121 } 1122 1123 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1124 { 1125 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1126 int ret = 0; 1127 1128 spin_lock(&sbi->inode_lock[DIRTY_META]); 1129 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1130 ret = 1; 1131 } else { 1132 set_inode_flag(inode, FI_DIRTY_INODE); 1133 stat_inc_dirty_inode(sbi, DIRTY_META); 1134 } 1135 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1136 list_add_tail(&F2FS_I(inode)->gdirty_list, 1137 &sbi->inode_list[DIRTY_META]); 1138 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1139 } 1140 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1141 return ret; 1142 } 1143 1144 void f2fs_inode_synced(struct inode *inode) 1145 { 1146 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1147 1148 spin_lock(&sbi->inode_lock[DIRTY_META]); 1149 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1150 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1151 return; 1152 } 1153 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1154 list_del_init(&F2FS_I(inode)->gdirty_list); 1155 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1156 } 1157 clear_inode_flag(inode, FI_DIRTY_INODE); 1158 clear_inode_flag(inode, FI_AUTO_RECOVER); 1159 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1160 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1161 } 1162 1163 /* 1164 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1165 * 1166 * We should call set_dirty_inode to write the dirty inode through write_inode. 1167 */ 1168 static void f2fs_dirty_inode(struct inode *inode, int flags) 1169 { 1170 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1171 1172 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1173 inode->i_ino == F2FS_META_INO(sbi)) 1174 return; 1175 1176 if (flags == I_DIRTY_TIME) 1177 return; 1178 1179 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1180 clear_inode_flag(inode, FI_AUTO_RECOVER); 1181 1182 f2fs_inode_dirtied(inode, false); 1183 } 1184 1185 static void f2fs_free_inode(struct inode *inode) 1186 { 1187 fscrypt_free_inode(inode); 1188 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1189 } 1190 1191 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1192 { 1193 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1194 percpu_counter_destroy(&sbi->total_valid_inode_count); 1195 } 1196 1197 static void destroy_device_list(struct f2fs_sb_info *sbi) 1198 { 1199 int i; 1200 1201 for (i = 0; i < sbi->s_ndevs; i++) { 1202 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1203 #ifdef CONFIG_BLK_DEV_ZONED 1204 kvfree(FDEV(i).blkz_seq); 1205 kfree(FDEV(i).zone_capacity_blocks); 1206 #endif 1207 } 1208 kvfree(sbi->devs); 1209 } 1210 1211 static void f2fs_put_super(struct super_block *sb) 1212 { 1213 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1214 int i; 1215 bool dropped; 1216 1217 /* unregister procfs/sysfs entries in advance to avoid race case */ 1218 f2fs_unregister_sysfs(sbi); 1219 1220 f2fs_quota_off_umount(sb); 1221 1222 /* prevent remaining shrinker jobs */ 1223 mutex_lock(&sbi->umount_mutex); 1224 1225 /* 1226 * We don't need to do checkpoint when superblock is clean. 1227 * But, the previous checkpoint was not done by umount, it needs to do 1228 * clean checkpoint again. 1229 */ 1230 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1231 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1232 struct cp_control cpc = { 1233 .reason = CP_UMOUNT, 1234 }; 1235 f2fs_write_checkpoint(sbi, &cpc); 1236 } 1237 1238 /* be sure to wait for any on-going discard commands */ 1239 dropped = f2fs_issue_discard_timeout(sbi); 1240 1241 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && 1242 !sbi->discard_blks && !dropped) { 1243 struct cp_control cpc = { 1244 .reason = CP_UMOUNT | CP_TRIMMED, 1245 }; 1246 f2fs_write_checkpoint(sbi, &cpc); 1247 } 1248 1249 /* 1250 * normally superblock is clean, so we need to release this. 1251 * In addition, EIO will skip do checkpoint, we need this as well. 1252 */ 1253 f2fs_release_ino_entry(sbi, true); 1254 1255 f2fs_leave_shrinker(sbi); 1256 mutex_unlock(&sbi->umount_mutex); 1257 1258 /* our cp_error case, we can wait for any writeback page */ 1259 f2fs_flush_merged_writes(sbi); 1260 1261 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1262 1263 f2fs_bug_on(sbi, sbi->fsync_node_num); 1264 1265 iput(sbi->node_inode); 1266 sbi->node_inode = NULL; 1267 1268 iput(sbi->meta_inode); 1269 sbi->meta_inode = NULL; 1270 1271 /* 1272 * iput() can update stat information, if f2fs_write_checkpoint() 1273 * above failed with error. 1274 */ 1275 f2fs_destroy_stats(sbi); 1276 1277 /* destroy f2fs internal modules */ 1278 f2fs_destroy_node_manager(sbi); 1279 f2fs_destroy_segment_manager(sbi); 1280 1281 f2fs_destroy_post_read_wq(sbi); 1282 1283 kvfree(sbi->ckpt); 1284 1285 sb->s_fs_info = NULL; 1286 if (sbi->s_chksum_driver) 1287 crypto_free_shash(sbi->s_chksum_driver); 1288 kfree(sbi->raw_super); 1289 1290 destroy_device_list(sbi); 1291 f2fs_destroy_page_array_cache(sbi); 1292 f2fs_destroy_xattr_caches(sbi); 1293 mempool_destroy(sbi->write_io_dummy); 1294 #ifdef CONFIG_QUOTA 1295 for (i = 0; i < MAXQUOTAS; i++) 1296 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1297 #endif 1298 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1299 destroy_percpu_info(sbi); 1300 for (i = 0; i < NR_PAGE_TYPE; i++) 1301 kvfree(sbi->write_io[i]); 1302 #ifdef CONFIG_UNICODE 1303 utf8_unload(sb->s_encoding); 1304 #endif 1305 kfree(sbi); 1306 } 1307 1308 int f2fs_sync_fs(struct super_block *sb, int sync) 1309 { 1310 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1311 int err = 0; 1312 1313 if (unlikely(f2fs_cp_error(sbi))) 1314 return 0; 1315 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1316 return 0; 1317 1318 trace_f2fs_sync_fs(sb, sync); 1319 1320 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1321 return -EAGAIN; 1322 1323 if (sync) { 1324 struct cp_control cpc; 1325 1326 cpc.reason = __get_cp_reason(sbi); 1327 1328 down_write(&sbi->gc_lock); 1329 err = f2fs_write_checkpoint(sbi, &cpc); 1330 up_write(&sbi->gc_lock); 1331 } 1332 f2fs_trace_ios(NULL, 1); 1333 1334 return err; 1335 } 1336 1337 static int f2fs_freeze(struct super_block *sb) 1338 { 1339 if (f2fs_readonly(sb)) 1340 return 0; 1341 1342 /* IO error happened before */ 1343 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1344 return -EIO; 1345 1346 /* must be clean, since sync_filesystem() was already called */ 1347 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1348 return -EINVAL; 1349 return 0; 1350 } 1351 1352 static int f2fs_unfreeze(struct super_block *sb) 1353 { 1354 return 0; 1355 } 1356 1357 #ifdef CONFIG_QUOTA 1358 static int f2fs_statfs_project(struct super_block *sb, 1359 kprojid_t projid, struct kstatfs *buf) 1360 { 1361 struct kqid qid; 1362 struct dquot *dquot; 1363 u64 limit; 1364 u64 curblock; 1365 1366 qid = make_kqid_projid(projid); 1367 dquot = dqget(sb, qid); 1368 if (IS_ERR(dquot)) 1369 return PTR_ERR(dquot); 1370 spin_lock(&dquot->dq_dqb_lock); 1371 1372 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1373 dquot->dq_dqb.dqb_bhardlimit); 1374 if (limit) 1375 limit >>= sb->s_blocksize_bits; 1376 1377 if (limit && buf->f_blocks > limit) { 1378 curblock = (dquot->dq_dqb.dqb_curspace + 1379 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1380 buf->f_blocks = limit; 1381 buf->f_bfree = buf->f_bavail = 1382 (buf->f_blocks > curblock) ? 1383 (buf->f_blocks - curblock) : 0; 1384 } 1385 1386 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1387 dquot->dq_dqb.dqb_ihardlimit); 1388 1389 if (limit && buf->f_files > limit) { 1390 buf->f_files = limit; 1391 buf->f_ffree = 1392 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1393 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1394 } 1395 1396 spin_unlock(&dquot->dq_dqb_lock); 1397 dqput(dquot); 1398 return 0; 1399 } 1400 #endif 1401 1402 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1403 { 1404 struct super_block *sb = dentry->d_sb; 1405 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1406 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1407 block_t total_count, user_block_count, start_count; 1408 u64 avail_node_count; 1409 1410 total_count = le64_to_cpu(sbi->raw_super->block_count); 1411 user_block_count = sbi->user_block_count; 1412 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1413 buf->f_type = F2FS_SUPER_MAGIC; 1414 buf->f_bsize = sbi->blocksize; 1415 1416 buf->f_blocks = total_count - start_count; 1417 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1418 sbi->current_reserved_blocks; 1419 1420 spin_lock(&sbi->stat_lock); 1421 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1422 buf->f_bfree = 0; 1423 else 1424 buf->f_bfree -= sbi->unusable_block_count; 1425 spin_unlock(&sbi->stat_lock); 1426 1427 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1428 buf->f_bavail = buf->f_bfree - 1429 F2FS_OPTION(sbi).root_reserved_blocks; 1430 else 1431 buf->f_bavail = 0; 1432 1433 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1434 1435 if (avail_node_count > user_block_count) { 1436 buf->f_files = user_block_count; 1437 buf->f_ffree = buf->f_bavail; 1438 } else { 1439 buf->f_files = avail_node_count; 1440 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1441 buf->f_bavail); 1442 } 1443 1444 buf->f_namelen = F2FS_NAME_LEN; 1445 buf->f_fsid = u64_to_fsid(id); 1446 1447 #ifdef CONFIG_QUOTA 1448 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1449 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1450 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1451 } 1452 #endif 1453 return 0; 1454 } 1455 1456 static inline void f2fs_show_quota_options(struct seq_file *seq, 1457 struct super_block *sb) 1458 { 1459 #ifdef CONFIG_QUOTA 1460 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1461 1462 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1463 char *fmtname = ""; 1464 1465 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1466 case QFMT_VFS_OLD: 1467 fmtname = "vfsold"; 1468 break; 1469 case QFMT_VFS_V0: 1470 fmtname = "vfsv0"; 1471 break; 1472 case QFMT_VFS_V1: 1473 fmtname = "vfsv1"; 1474 break; 1475 } 1476 seq_printf(seq, ",jqfmt=%s", fmtname); 1477 } 1478 1479 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1480 seq_show_option(seq, "usrjquota", 1481 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1482 1483 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1484 seq_show_option(seq, "grpjquota", 1485 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1486 1487 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1488 seq_show_option(seq, "prjjquota", 1489 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1490 #endif 1491 } 1492 1493 static inline void f2fs_show_compress_options(struct seq_file *seq, 1494 struct super_block *sb) 1495 { 1496 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1497 char *algtype = ""; 1498 int i; 1499 1500 if (!f2fs_sb_has_compression(sbi)) 1501 return; 1502 1503 switch (F2FS_OPTION(sbi).compress_algorithm) { 1504 case COMPRESS_LZO: 1505 algtype = "lzo"; 1506 break; 1507 case COMPRESS_LZ4: 1508 algtype = "lz4"; 1509 break; 1510 case COMPRESS_ZSTD: 1511 algtype = "zstd"; 1512 break; 1513 case COMPRESS_LZORLE: 1514 algtype = "lzo-rle"; 1515 break; 1516 } 1517 seq_printf(seq, ",compress_algorithm=%s", algtype); 1518 1519 seq_printf(seq, ",compress_log_size=%u", 1520 F2FS_OPTION(sbi).compress_log_size); 1521 1522 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1523 seq_printf(seq, ",compress_extension=%s", 1524 F2FS_OPTION(sbi).extensions[i]); 1525 } 1526 } 1527 1528 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1529 { 1530 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1531 1532 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1533 seq_printf(seq, ",background_gc=%s", "sync"); 1534 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1535 seq_printf(seq, ",background_gc=%s", "on"); 1536 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1537 seq_printf(seq, ",background_gc=%s", "off"); 1538 1539 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1540 seq_puts(seq, ",disable_roll_forward"); 1541 if (test_opt(sbi, NORECOVERY)) 1542 seq_puts(seq, ",norecovery"); 1543 if (test_opt(sbi, DISCARD)) 1544 seq_puts(seq, ",discard"); 1545 else 1546 seq_puts(seq, ",nodiscard"); 1547 if (test_opt(sbi, NOHEAP)) 1548 seq_puts(seq, ",no_heap"); 1549 else 1550 seq_puts(seq, ",heap"); 1551 #ifdef CONFIG_F2FS_FS_XATTR 1552 if (test_opt(sbi, XATTR_USER)) 1553 seq_puts(seq, ",user_xattr"); 1554 else 1555 seq_puts(seq, ",nouser_xattr"); 1556 if (test_opt(sbi, INLINE_XATTR)) 1557 seq_puts(seq, ",inline_xattr"); 1558 else 1559 seq_puts(seq, ",noinline_xattr"); 1560 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1561 seq_printf(seq, ",inline_xattr_size=%u", 1562 F2FS_OPTION(sbi).inline_xattr_size); 1563 #endif 1564 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1565 if (test_opt(sbi, POSIX_ACL)) 1566 seq_puts(seq, ",acl"); 1567 else 1568 seq_puts(seq, ",noacl"); 1569 #endif 1570 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1571 seq_puts(seq, ",disable_ext_identify"); 1572 if (test_opt(sbi, INLINE_DATA)) 1573 seq_puts(seq, ",inline_data"); 1574 else 1575 seq_puts(seq, ",noinline_data"); 1576 if (test_opt(sbi, INLINE_DENTRY)) 1577 seq_puts(seq, ",inline_dentry"); 1578 else 1579 seq_puts(seq, ",noinline_dentry"); 1580 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1581 seq_puts(seq, ",flush_merge"); 1582 if (test_opt(sbi, NOBARRIER)) 1583 seq_puts(seq, ",nobarrier"); 1584 if (test_opt(sbi, FASTBOOT)) 1585 seq_puts(seq, ",fastboot"); 1586 if (test_opt(sbi, EXTENT_CACHE)) 1587 seq_puts(seq, ",extent_cache"); 1588 else 1589 seq_puts(seq, ",noextent_cache"); 1590 if (test_opt(sbi, DATA_FLUSH)) 1591 seq_puts(seq, ",data_flush"); 1592 1593 seq_puts(seq, ",mode="); 1594 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1595 seq_puts(seq, "adaptive"); 1596 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1597 seq_puts(seq, "lfs"); 1598 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1599 if (test_opt(sbi, RESERVE_ROOT)) 1600 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1601 F2FS_OPTION(sbi).root_reserved_blocks, 1602 from_kuid_munged(&init_user_ns, 1603 F2FS_OPTION(sbi).s_resuid), 1604 from_kgid_munged(&init_user_ns, 1605 F2FS_OPTION(sbi).s_resgid)); 1606 if (F2FS_IO_SIZE_BITS(sbi)) 1607 seq_printf(seq, ",io_bits=%u", 1608 F2FS_OPTION(sbi).write_io_size_bits); 1609 #ifdef CONFIG_F2FS_FAULT_INJECTION 1610 if (test_opt(sbi, FAULT_INJECTION)) { 1611 seq_printf(seq, ",fault_injection=%u", 1612 F2FS_OPTION(sbi).fault_info.inject_rate); 1613 seq_printf(seq, ",fault_type=%u", 1614 F2FS_OPTION(sbi).fault_info.inject_type); 1615 } 1616 #endif 1617 #ifdef CONFIG_QUOTA 1618 if (test_opt(sbi, QUOTA)) 1619 seq_puts(seq, ",quota"); 1620 if (test_opt(sbi, USRQUOTA)) 1621 seq_puts(seq, ",usrquota"); 1622 if (test_opt(sbi, GRPQUOTA)) 1623 seq_puts(seq, ",grpquota"); 1624 if (test_opt(sbi, PRJQUOTA)) 1625 seq_puts(seq, ",prjquota"); 1626 #endif 1627 f2fs_show_quota_options(seq, sbi->sb); 1628 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1629 seq_printf(seq, ",whint_mode=%s", "user-based"); 1630 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1631 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1632 1633 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1634 1635 if (sbi->sb->s_flags & SB_INLINECRYPT) 1636 seq_puts(seq, ",inlinecrypt"); 1637 1638 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1639 seq_printf(seq, ",alloc_mode=%s", "default"); 1640 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1641 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1642 1643 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1644 seq_printf(seq, ",checkpoint=disable:%u", 1645 F2FS_OPTION(sbi).unusable_cap); 1646 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1647 seq_printf(seq, ",fsync_mode=%s", "posix"); 1648 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1649 seq_printf(seq, ",fsync_mode=%s", "strict"); 1650 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1651 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 1652 1653 #ifdef CONFIG_F2FS_FS_COMPRESSION 1654 f2fs_show_compress_options(seq, sbi->sb); 1655 #endif 1656 1657 if (test_opt(sbi, ATGC)) 1658 seq_puts(seq, ",atgc"); 1659 return 0; 1660 } 1661 1662 static void default_options(struct f2fs_sb_info *sbi) 1663 { 1664 /* init some FS parameters */ 1665 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 1666 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 1667 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1668 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1669 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1670 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 1671 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 1672 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 1673 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 1674 F2FS_OPTION(sbi).compress_ext_cnt = 0; 1675 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 1676 1677 sbi->sb->s_flags &= ~SB_INLINECRYPT; 1678 1679 set_opt(sbi, INLINE_XATTR); 1680 set_opt(sbi, INLINE_DATA); 1681 set_opt(sbi, INLINE_DENTRY); 1682 set_opt(sbi, EXTENT_CACHE); 1683 set_opt(sbi, NOHEAP); 1684 clear_opt(sbi, DISABLE_CHECKPOINT); 1685 F2FS_OPTION(sbi).unusable_cap = 0; 1686 sbi->sb->s_flags |= SB_LAZYTIME; 1687 set_opt(sbi, FLUSH_MERGE); 1688 set_opt(sbi, DISCARD); 1689 if (f2fs_sb_has_blkzoned(sbi)) 1690 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 1691 else 1692 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 1693 1694 #ifdef CONFIG_F2FS_FS_XATTR 1695 set_opt(sbi, XATTR_USER); 1696 #endif 1697 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1698 set_opt(sbi, POSIX_ACL); 1699 #endif 1700 1701 f2fs_build_fault_attr(sbi, 0, 0); 1702 } 1703 1704 #ifdef CONFIG_QUOTA 1705 static int f2fs_enable_quotas(struct super_block *sb); 1706 #endif 1707 1708 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 1709 { 1710 unsigned int s_flags = sbi->sb->s_flags; 1711 struct cp_control cpc; 1712 int err = 0; 1713 int ret; 1714 block_t unusable; 1715 1716 if (s_flags & SB_RDONLY) { 1717 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 1718 return -EINVAL; 1719 } 1720 sbi->sb->s_flags |= SB_ACTIVE; 1721 1722 f2fs_update_time(sbi, DISABLE_TIME); 1723 1724 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 1725 down_write(&sbi->gc_lock); 1726 err = f2fs_gc(sbi, true, false, NULL_SEGNO); 1727 if (err == -ENODATA) { 1728 err = 0; 1729 break; 1730 } 1731 if (err && err != -EAGAIN) 1732 break; 1733 } 1734 1735 ret = sync_filesystem(sbi->sb); 1736 if (ret || err) { 1737 err = ret ? ret: err; 1738 goto restore_flag; 1739 } 1740 1741 unusable = f2fs_get_unusable_blocks(sbi); 1742 if (f2fs_disable_cp_again(sbi, unusable)) { 1743 err = -EAGAIN; 1744 goto restore_flag; 1745 } 1746 1747 down_write(&sbi->gc_lock); 1748 cpc.reason = CP_PAUSE; 1749 set_sbi_flag(sbi, SBI_CP_DISABLED); 1750 err = f2fs_write_checkpoint(sbi, &cpc); 1751 if (err) 1752 goto out_unlock; 1753 1754 spin_lock(&sbi->stat_lock); 1755 sbi->unusable_block_count = unusable; 1756 spin_unlock(&sbi->stat_lock); 1757 1758 out_unlock: 1759 up_write(&sbi->gc_lock); 1760 restore_flag: 1761 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 1762 return err; 1763 } 1764 1765 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 1766 { 1767 down_write(&sbi->gc_lock); 1768 f2fs_dirty_to_prefree(sbi); 1769 1770 clear_sbi_flag(sbi, SBI_CP_DISABLED); 1771 set_sbi_flag(sbi, SBI_IS_DIRTY); 1772 up_write(&sbi->gc_lock); 1773 1774 f2fs_sync_fs(sbi->sb, 1); 1775 } 1776 1777 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 1778 { 1779 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1780 struct f2fs_mount_info org_mount_opt; 1781 unsigned long old_sb_flags; 1782 int err; 1783 bool need_restart_gc = false; 1784 bool need_stop_gc = false; 1785 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 1786 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT); 1787 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 1788 bool no_atgc = !test_opt(sbi, ATGC); 1789 bool checkpoint_changed; 1790 #ifdef CONFIG_QUOTA 1791 int i, j; 1792 #endif 1793 1794 /* 1795 * Save the old mount options in case we 1796 * need to restore them. 1797 */ 1798 org_mount_opt = sbi->mount_opt; 1799 old_sb_flags = sb->s_flags; 1800 1801 #ifdef CONFIG_QUOTA 1802 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 1803 for (i = 0; i < MAXQUOTAS; i++) { 1804 if (F2FS_OPTION(sbi).s_qf_names[i]) { 1805 org_mount_opt.s_qf_names[i] = 1806 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 1807 GFP_KERNEL); 1808 if (!org_mount_opt.s_qf_names[i]) { 1809 for (j = 0; j < i; j++) 1810 kfree(org_mount_opt.s_qf_names[j]); 1811 return -ENOMEM; 1812 } 1813 } else { 1814 org_mount_opt.s_qf_names[i] = NULL; 1815 } 1816 } 1817 #endif 1818 1819 /* recover superblocks we couldn't write due to previous RO mount */ 1820 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 1821 err = f2fs_commit_super(sbi, false); 1822 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 1823 err); 1824 if (!err) 1825 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1826 } 1827 1828 default_options(sbi); 1829 1830 /* parse mount options */ 1831 err = parse_options(sb, data, true); 1832 if (err) 1833 goto restore_opts; 1834 checkpoint_changed = 1835 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT); 1836 1837 /* 1838 * Previous and new state of filesystem is RO, 1839 * so skip checking GC and FLUSH_MERGE conditions. 1840 */ 1841 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 1842 goto skip; 1843 1844 #ifdef CONFIG_QUOTA 1845 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 1846 err = dquot_suspend(sb, -1); 1847 if (err < 0) 1848 goto restore_opts; 1849 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 1850 /* dquot_resume needs RW */ 1851 sb->s_flags &= ~SB_RDONLY; 1852 if (sb_any_quota_suspended(sb)) { 1853 dquot_resume(sb, -1); 1854 } else if (f2fs_sb_has_quota_ino(sbi)) { 1855 err = f2fs_enable_quotas(sb); 1856 if (err) 1857 goto restore_opts; 1858 } 1859 } 1860 #endif 1861 /* disallow enable atgc dynamically */ 1862 if (no_atgc == !!test_opt(sbi, ATGC)) { 1863 err = -EINVAL; 1864 f2fs_warn(sbi, "switch atgc option is not allowed"); 1865 goto restore_opts; 1866 } 1867 1868 /* disallow enable/disable extent_cache dynamically */ 1869 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 1870 err = -EINVAL; 1871 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 1872 goto restore_opts; 1873 } 1874 1875 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 1876 err = -EINVAL; 1877 f2fs_warn(sbi, "switch io_bits option is not allowed"); 1878 goto restore_opts; 1879 } 1880 1881 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 1882 err = -EINVAL; 1883 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 1884 goto restore_opts; 1885 } 1886 1887 /* 1888 * We stop the GC thread if FS is mounted as RO 1889 * or if background_gc = off is passed in mount 1890 * option. Also sync the filesystem. 1891 */ 1892 if ((*flags & SB_RDONLY) || 1893 F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) { 1894 if (sbi->gc_thread) { 1895 f2fs_stop_gc_thread(sbi); 1896 need_restart_gc = true; 1897 } 1898 } else if (!sbi->gc_thread) { 1899 err = f2fs_start_gc_thread(sbi); 1900 if (err) 1901 goto restore_opts; 1902 need_stop_gc = true; 1903 } 1904 1905 if (*flags & SB_RDONLY || 1906 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 1907 writeback_inodes_sb(sb, WB_REASON_SYNC); 1908 sync_inodes_sb(sb); 1909 1910 set_sbi_flag(sbi, SBI_IS_DIRTY); 1911 set_sbi_flag(sbi, SBI_IS_CLOSE); 1912 f2fs_sync_fs(sb, 1); 1913 clear_sbi_flag(sbi, SBI_IS_CLOSE); 1914 } 1915 1916 if (checkpoint_changed) { 1917 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 1918 err = f2fs_disable_checkpoint(sbi); 1919 if (err) 1920 goto restore_gc; 1921 } else { 1922 f2fs_enable_checkpoint(sbi); 1923 } 1924 } 1925 1926 /* 1927 * We stop issue flush thread if FS is mounted as RO 1928 * or if flush_merge is not passed in mount option. 1929 */ 1930 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 1931 clear_opt(sbi, FLUSH_MERGE); 1932 f2fs_destroy_flush_cmd_control(sbi, false); 1933 } else { 1934 err = f2fs_create_flush_cmd_control(sbi); 1935 if (err) 1936 goto restore_gc; 1937 } 1938 skip: 1939 #ifdef CONFIG_QUOTA 1940 /* Release old quota file names */ 1941 for (i = 0; i < MAXQUOTAS; i++) 1942 kfree(org_mount_opt.s_qf_names[i]); 1943 #endif 1944 /* Update the POSIXACL Flag */ 1945 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 1946 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 1947 1948 limit_reserve_root(sbi); 1949 adjust_unusable_cap_perc(sbi); 1950 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 1951 return 0; 1952 restore_gc: 1953 if (need_restart_gc) { 1954 if (f2fs_start_gc_thread(sbi)) 1955 f2fs_warn(sbi, "background gc thread has stopped"); 1956 } else if (need_stop_gc) { 1957 f2fs_stop_gc_thread(sbi); 1958 } 1959 restore_opts: 1960 #ifdef CONFIG_QUOTA 1961 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 1962 for (i = 0; i < MAXQUOTAS; i++) { 1963 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1964 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 1965 } 1966 #endif 1967 sbi->mount_opt = org_mount_opt; 1968 sb->s_flags = old_sb_flags; 1969 return err; 1970 } 1971 1972 #ifdef CONFIG_QUOTA 1973 /* Read data from quotafile */ 1974 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 1975 size_t len, loff_t off) 1976 { 1977 struct inode *inode = sb_dqopt(sb)->files[type]; 1978 struct address_space *mapping = inode->i_mapping; 1979 block_t blkidx = F2FS_BYTES_TO_BLK(off); 1980 int offset = off & (sb->s_blocksize - 1); 1981 int tocopy; 1982 size_t toread; 1983 loff_t i_size = i_size_read(inode); 1984 struct page *page; 1985 char *kaddr; 1986 1987 if (off > i_size) 1988 return 0; 1989 1990 if (off + len > i_size) 1991 len = i_size - off; 1992 toread = len; 1993 while (toread > 0) { 1994 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 1995 repeat: 1996 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 1997 if (IS_ERR(page)) { 1998 if (PTR_ERR(page) == -ENOMEM) { 1999 congestion_wait(BLK_RW_ASYNC, 2000 DEFAULT_IO_TIMEOUT); 2001 goto repeat; 2002 } 2003 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2004 return PTR_ERR(page); 2005 } 2006 2007 lock_page(page); 2008 2009 if (unlikely(page->mapping != mapping)) { 2010 f2fs_put_page(page, 1); 2011 goto repeat; 2012 } 2013 if (unlikely(!PageUptodate(page))) { 2014 f2fs_put_page(page, 1); 2015 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2016 return -EIO; 2017 } 2018 2019 kaddr = kmap_atomic(page); 2020 memcpy(data, kaddr + offset, tocopy); 2021 kunmap_atomic(kaddr); 2022 f2fs_put_page(page, 1); 2023 2024 offset = 0; 2025 toread -= tocopy; 2026 data += tocopy; 2027 blkidx++; 2028 } 2029 return len; 2030 } 2031 2032 /* Write to quotafile */ 2033 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2034 const char *data, size_t len, loff_t off) 2035 { 2036 struct inode *inode = sb_dqopt(sb)->files[type]; 2037 struct address_space *mapping = inode->i_mapping; 2038 const struct address_space_operations *a_ops = mapping->a_ops; 2039 int offset = off & (sb->s_blocksize - 1); 2040 size_t towrite = len; 2041 struct page *page; 2042 void *fsdata = NULL; 2043 char *kaddr; 2044 int err = 0; 2045 int tocopy; 2046 2047 while (towrite > 0) { 2048 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2049 towrite); 2050 retry: 2051 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2052 &page, &fsdata); 2053 if (unlikely(err)) { 2054 if (err == -ENOMEM) { 2055 congestion_wait(BLK_RW_ASYNC, 2056 DEFAULT_IO_TIMEOUT); 2057 goto retry; 2058 } 2059 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2060 break; 2061 } 2062 2063 kaddr = kmap_atomic(page); 2064 memcpy(kaddr + offset, data, tocopy); 2065 kunmap_atomic(kaddr); 2066 flush_dcache_page(page); 2067 2068 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2069 page, fsdata); 2070 offset = 0; 2071 towrite -= tocopy; 2072 off += tocopy; 2073 data += tocopy; 2074 cond_resched(); 2075 } 2076 2077 if (len == towrite) 2078 return err; 2079 inode->i_mtime = inode->i_ctime = current_time(inode); 2080 f2fs_mark_inode_dirty_sync(inode, false); 2081 return len - towrite; 2082 } 2083 2084 static struct dquot **f2fs_get_dquots(struct inode *inode) 2085 { 2086 return F2FS_I(inode)->i_dquot; 2087 } 2088 2089 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2090 { 2091 return &F2FS_I(inode)->i_reserved_quota; 2092 } 2093 2094 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2095 { 2096 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2097 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2098 return 0; 2099 } 2100 2101 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2102 F2FS_OPTION(sbi).s_jquota_fmt, type); 2103 } 2104 2105 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2106 { 2107 int enabled = 0; 2108 int i, err; 2109 2110 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2111 err = f2fs_enable_quotas(sbi->sb); 2112 if (err) { 2113 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2114 return 0; 2115 } 2116 return 1; 2117 } 2118 2119 for (i = 0; i < MAXQUOTAS; i++) { 2120 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2121 err = f2fs_quota_on_mount(sbi, i); 2122 if (!err) { 2123 enabled = 1; 2124 continue; 2125 } 2126 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2127 err, i); 2128 } 2129 } 2130 return enabled; 2131 } 2132 2133 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2134 unsigned int flags) 2135 { 2136 struct inode *qf_inode; 2137 unsigned long qf_inum; 2138 int err; 2139 2140 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2141 2142 qf_inum = f2fs_qf_ino(sb, type); 2143 if (!qf_inum) 2144 return -EPERM; 2145 2146 qf_inode = f2fs_iget(sb, qf_inum); 2147 if (IS_ERR(qf_inode)) { 2148 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2149 return PTR_ERR(qf_inode); 2150 } 2151 2152 /* Don't account quota for quota files to avoid recursion */ 2153 qf_inode->i_flags |= S_NOQUOTA; 2154 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2155 iput(qf_inode); 2156 return err; 2157 } 2158 2159 static int f2fs_enable_quotas(struct super_block *sb) 2160 { 2161 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2162 int type, err = 0; 2163 unsigned long qf_inum; 2164 bool quota_mopt[MAXQUOTAS] = { 2165 test_opt(sbi, USRQUOTA), 2166 test_opt(sbi, GRPQUOTA), 2167 test_opt(sbi, PRJQUOTA), 2168 }; 2169 2170 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2171 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2172 return 0; 2173 } 2174 2175 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2176 2177 for (type = 0; type < MAXQUOTAS; type++) { 2178 qf_inum = f2fs_qf_ino(sb, type); 2179 if (qf_inum) { 2180 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2181 DQUOT_USAGE_ENABLED | 2182 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2183 if (err) { 2184 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2185 type, err); 2186 for (type--; type >= 0; type--) 2187 dquot_quota_off(sb, type); 2188 set_sbi_flag(F2FS_SB(sb), 2189 SBI_QUOTA_NEED_REPAIR); 2190 return err; 2191 } 2192 } 2193 } 2194 return 0; 2195 } 2196 2197 int f2fs_quota_sync(struct super_block *sb, int type) 2198 { 2199 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2200 struct quota_info *dqopt = sb_dqopt(sb); 2201 int cnt; 2202 int ret; 2203 2204 /* 2205 * do_quotactl 2206 * f2fs_quota_sync 2207 * down_read(quota_sem) 2208 * dquot_writeback_dquots() 2209 * f2fs_dquot_commit 2210 * block_operation 2211 * down_read(quota_sem) 2212 */ 2213 f2fs_lock_op(sbi); 2214 2215 down_read(&sbi->quota_sem); 2216 ret = dquot_writeback_dquots(sb, type); 2217 if (ret) 2218 goto out; 2219 2220 /* 2221 * Now when everything is written we can discard the pagecache so 2222 * that userspace sees the changes. 2223 */ 2224 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2225 struct address_space *mapping; 2226 2227 if (type != -1 && cnt != type) 2228 continue; 2229 if (!sb_has_quota_active(sb, cnt)) 2230 continue; 2231 2232 mapping = dqopt->files[cnt]->i_mapping; 2233 2234 ret = filemap_fdatawrite(mapping); 2235 if (ret) 2236 goto out; 2237 2238 /* if we are using journalled quota */ 2239 if (is_journalled_quota(sbi)) 2240 continue; 2241 2242 ret = filemap_fdatawait(mapping); 2243 if (ret) 2244 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2245 2246 inode_lock(dqopt->files[cnt]); 2247 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0); 2248 inode_unlock(dqopt->files[cnt]); 2249 } 2250 out: 2251 if (ret) 2252 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2253 up_read(&sbi->quota_sem); 2254 f2fs_unlock_op(sbi); 2255 return ret; 2256 } 2257 2258 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2259 const struct path *path) 2260 { 2261 struct inode *inode; 2262 int err; 2263 2264 /* if quota sysfile exists, deny enabling quota with specific file */ 2265 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2266 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2267 return -EBUSY; 2268 } 2269 2270 err = f2fs_quota_sync(sb, type); 2271 if (err) 2272 return err; 2273 2274 err = dquot_quota_on(sb, type, format_id, path); 2275 if (err) 2276 return err; 2277 2278 inode = d_inode(path->dentry); 2279 2280 inode_lock(inode); 2281 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2282 f2fs_set_inode_flags(inode); 2283 inode_unlock(inode); 2284 f2fs_mark_inode_dirty_sync(inode, false); 2285 2286 return 0; 2287 } 2288 2289 static int __f2fs_quota_off(struct super_block *sb, int type) 2290 { 2291 struct inode *inode = sb_dqopt(sb)->files[type]; 2292 int err; 2293 2294 if (!inode || !igrab(inode)) 2295 return dquot_quota_off(sb, type); 2296 2297 err = f2fs_quota_sync(sb, type); 2298 if (err) 2299 goto out_put; 2300 2301 err = dquot_quota_off(sb, type); 2302 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2303 goto out_put; 2304 2305 inode_lock(inode); 2306 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2307 f2fs_set_inode_flags(inode); 2308 inode_unlock(inode); 2309 f2fs_mark_inode_dirty_sync(inode, false); 2310 out_put: 2311 iput(inode); 2312 return err; 2313 } 2314 2315 static int f2fs_quota_off(struct super_block *sb, int type) 2316 { 2317 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2318 int err; 2319 2320 err = __f2fs_quota_off(sb, type); 2321 2322 /* 2323 * quotactl can shutdown journalled quota, result in inconsistence 2324 * between quota record and fs data by following updates, tag the 2325 * flag to let fsck be aware of it. 2326 */ 2327 if (is_journalled_quota(sbi)) 2328 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2329 return err; 2330 } 2331 2332 void f2fs_quota_off_umount(struct super_block *sb) 2333 { 2334 int type; 2335 int err; 2336 2337 for (type = 0; type < MAXQUOTAS; type++) { 2338 err = __f2fs_quota_off(sb, type); 2339 if (err) { 2340 int ret = dquot_quota_off(sb, type); 2341 2342 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2343 type, err, ret); 2344 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2345 } 2346 } 2347 /* 2348 * In case of checkpoint=disable, we must flush quota blocks. 2349 * This can cause NULL exception for node_inode in end_io, since 2350 * put_super already dropped it. 2351 */ 2352 sync_filesystem(sb); 2353 } 2354 2355 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2356 { 2357 struct quota_info *dqopt = sb_dqopt(sb); 2358 int type; 2359 2360 for (type = 0; type < MAXQUOTAS; type++) { 2361 if (!dqopt->files[type]) 2362 continue; 2363 f2fs_inode_synced(dqopt->files[type]); 2364 } 2365 } 2366 2367 static int f2fs_dquot_commit(struct dquot *dquot) 2368 { 2369 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2370 int ret; 2371 2372 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2373 ret = dquot_commit(dquot); 2374 if (ret < 0) 2375 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2376 up_read(&sbi->quota_sem); 2377 return ret; 2378 } 2379 2380 static int f2fs_dquot_acquire(struct dquot *dquot) 2381 { 2382 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2383 int ret; 2384 2385 down_read(&sbi->quota_sem); 2386 ret = dquot_acquire(dquot); 2387 if (ret < 0) 2388 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2389 up_read(&sbi->quota_sem); 2390 return ret; 2391 } 2392 2393 static int f2fs_dquot_release(struct dquot *dquot) 2394 { 2395 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2396 int ret = dquot_release(dquot); 2397 2398 if (ret < 0) 2399 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2400 return ret; 2401 } 2402 2403 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2404 { 2405 struct super_block *sb = dquot->dq_sb; 2406 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2407 int ret = dquot_mark_dquot_dirty(dquot); 2408 2409 /* if we are using journalled quota */ 2410 if (is_journalled_quota(sbi)) 2411 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2412 2413 return ret; 2414 } 2415 2416 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2417 { 2418 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2419 int ret = dquot_commit_info(sb, type); 2420 2421 if (ret < 0) 2422 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2423 return ret; 2424 } 2425 2426 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2427 { 2428 *projid = F2FS_I(inode)->i_projid; 2429 return 0; 2430 } 2431 2432 static const struct dquot_operations f2fs_quota_operations = { 2433 .get_reserved_space = f2fs_get_reserved_space, 2434 .write_dquot = f2fs_dquot_commit, 2435 .acquire_dquot = f2fs_dquot_acquire, 2436 .release_dquot = f2fs_dquot_release, 2437 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2438 .write_info = f2fs_dquot_commit_info, 2439 .alloc_dquot = dquot_alloc, 2440 .destroy_dquot = dquot_destroy, 2441 .get_projid = f2fs_get_projid, 2442 .get_next_id = dquot_get_next_id, 2443 }; 2444 2445 static const struct quotactl_ops f2fs_quotactl_ops = { 2446 .quota_on = f2fs_quota_on, 2447 .quota_off = f2fs_quota_off, 2448 .quota_sync = f2fs_quota_sync, 2449 .get_state = dquot_get_state, 2450 .set_info = dquot_set_dqinfo, 2451 .get_dqblk = dquot_get_dqblk, 2452 .set_dqblk = dquot_set_dqblk, 2453 .get_nextdqblk = dquot_get_next_dqblk, 2454 }; 2455 #else 2456 int f2fs_quota_sync(struct super_block *sb, int type) 2457 { 2458 return 0; 2459 } 2460 2461 void f2fs_quota_off_umount(struct super_block *sb) 2462 { 2463 } 2464 #endif 2465 2466 static const struct super_operations f2fs_sops = { 2467 .alloc_inode = f2fs_alloc_inode, 2468 .free_inode = f2fs_free_inode, 2469 .drop_inode = f2fs_drop_inode, 2470 .write_inode = f2fs_write_inode, 2471 .dirty_inode = f2fs_dirty_inode, 2472 .show_options = f2fs_show_options, 2473 #ifdef CONFIG_QUOTA 2474 .quota_read = f2fs_quota_read, 2475 .quota_write = f2fs_quota_write, 2476 .get_dquots = f2fs_get_dquots, 2477 #endif 2478 .evict_inode = f2fs_evict_inode, 2479 .put_super = f2fs_put_super, 2480 .sync_fs = f2fs_sync_fs, 2481 .freeze_fs = f2fs_freeze, 2482 .unfreeze_fs = f2fs_unfreeze, 2483 .statfs = f2fs_statfs, 2484 .remount_fs = f2fs_remount, 2485 }; 2486 2487 #ifdef CONFIG_FS_ENCRYPTION 2488 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2489 { 2490 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2491 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2492 ctx, len, NULL); 2493 } 2494 2495 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2496 void *fs_data) 2497 { 2498 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2499 2500 /* 2501 * Encrypting the root directory is not allowed because fsck 2502 * expects lost+found directory to exist and remain unencrypted 2503 * if LOST_FOUND feature is enabled. 2504 * 2505 */ 2506 if (f2fs_sb_has_lost_found(sbi) && 2507 inode->i_ino == F2FS_ROOT_INO(sbi)) 2508 return -EPERM; 2509 2510 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2511 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2512 ctx, len, fs_data, XATTR_CREATE); 2513 } 2514 2515 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 2516 { 2517 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 2518 } 2519 2520 static bool f2fs_has_stable_inodes(struct super_block *sb) 2521 { 2522 return true; 2523 } 2524 2525 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 2526 int *ino_bits_ret, int *lblk_bits_ret) 2527 { 2528 *ino_bits_ret = 8 * sizeof(nid_t); 2529 *lblk_bits_ret = 8 * sizeof(block_t); 2530 } 2531 2532 static int f2fs_get_num_devices(struct super_block *sb) 2533 { 2534 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2535 2536 if (f2fs_is_multi_device(sbi)) 2537 return sbi->s_ndevs; 2538 return 1; 2539 } 2540 2541 static void f2fs_get_devices(struct super_block *sb, 2542 struct request_queue **devs) 2543 { 2544 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2545 int i; 2546 2547 for (i = 0; i < sbi->s_ndevs; i++) 2548 devs[i] = bdev_get_queue(FDEV(i).bdev); 2549 } 2550 2551 static const struct fscrypt_operations f2fs_cryptops = { 2552 .key_prefix = "f2fs:", 2553 .get_context = f2fs_get_context, 2554 .set_context = f2fs_set_context, 2555 .get_dummy_policy = f2fs_get_dummy_policy, 2556 .empty_dir = f2fs_empty_dir, 2557 .max_namelen = F2FS_NAME_LEN, 2558 .has_stable_inodes = f2fs_has_stable_inodes, 2559 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 2560 .get_num_devices = f2fs_get_num_devices, 2561 .get_devices = f2fs_get_devices, 2562 }; 2563 #endif 2564 2565 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 2566 u64 ino, u32 generation) 2567 { 2568 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2569 struct inode *inode; 2570 2571 if (f2fs_check_nid_range(sbi, ino)) 2572 return ERR_PTR(-ESTALE); 2573 2574 /* 2575 * f2fs_iget isn't quite right if the inode is currently unallocated! 2576 * However f2fs_iget currently does appropriate checks to handle stale 2577 * inodes so everything is OK. 2578 */ 2579 inode = f2fs_iget(sb, ino); 2580 if (IS_ERR(inode)) 2581 return ERR_CAST(inode); 2582 if (unlikely(generation && inode->i_generation != generation)) { 2583 /* we didn't find the right inode.. */ 2584 iput(inode); 2585 return ERR_PTR(-ESTALE); 2586 } 2587 return inode; 2588 } 2589 2590 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 2591 int fh_len, int fh_type) 2592 { 2593 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 2594 f2fs_nfs_get_inode); 2595 } 2596 2597 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 2598 int fh_len, int fh_type) 2599 { 2600 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 2601 f2fs_nfs_get_inode); 2602 } 2603 2604 static const struct export_operations f2fs_export_ops = { 2605 .fh_to_dentry = f2fs_fh_to_dentry, 2606 .fh_to_parent = f2fs_fh_to_parent, 2607 .get_parent = f2fs_get_parent, 2608 }; 2609 2610 static loff_t max_file_blocks(void) 2611 { 2612 loff_t result = 0; 2613 loff_t leaf_count = DEF_ADDRS_PER_BLOCK; 2614 2615 /* 2616 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 2617 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 2618 * space in inode.i_addr, it will be more safe to reassign 2619 * result as zero. 2620 */ 2621 2622 /* two direct node blocks */ 2623 result += (leaf_count * 2); 2624 2625 /* two indirect node blocks */ 2626 leaf_count *= NIDS_PER_BLOCK; 2627 result += (leaf_count * 2); 2628 2629 /* one double indirect node block */ 2630 leaf_count *= NIDS_PER_BLOCK; 2631 result += leaf_count; 2632 2633 return result; 2634 } 2635 2636 static int __f2fs_commit_super(struct buffer_head *bh, 2637 struct f2fs_super_block *super) 2638 { 2639 lock_buffer(bh); 2640 if (super) 2641 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 2642 set_buffer_dirty(bh); 2643 unlock_buffer(bh); 2644 2645 /* it's rare case, we can do fua all the time */ 2646 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 2647 } 2648 2649 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 2650 struct buffer_head *bh) 2651 { 2652 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2653 (bh->b_data + F2FS_SUPER_OFFSET); 2654 struct super_block *sb = sbi->sb; 2655 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 2656 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 2657 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 2658 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 2659 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 2660 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 2661 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 2662 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 2663 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 2664 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 2665 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2666 u32 segment_count = le32_to_cpu(raw_super->segment_count); 2667 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2668 u64 main_end_blkaddr = main_blkaddr + 2669 (segment_count_main << log_blocks_per_seg); 2670 u64 seg_end_blkaddr = segment0_blkaddr + 2671 (segment_count << log_blocks_per_seg); 2672 2673 if (segment0_blkaddr != cp_blkaddr) { 2674 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 2675 segment0_blkaddr, cp_blkaddr); 2676 return true; 2677 } 2678 2679 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 2680 sit_blkaddr) { 2681 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 2682 cp_blkaddr, sit_blkaddr, 2683 segment_count_ckpt << log_blocks_per_seg); 2684 return true; 2685 } 2686 2687 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 2688 nat_blkaddr) { 2689 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 2690 sit_blkaddr, nat_blkaddr, 2691 segment_count_sit << log_blocks_per_seg); 2692 return true; 2693 } 2694 2695 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 2696 ssa_blkaddr) { 2697 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 2698 nat_blkaddr, ssa_blkaddr, 2699 segment_count_nat << log_blocks_per_seg); 2700 return true; 2701 } 2702 2703 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 2704 main_blkaddr) { 2705 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 2706 ssa_blkaddr, main_blkaddr, 2707 segment_count_ssa << log_blocks_per_seg); 2708 return true; 2709 } 2710 2711 if (main_end_blkaddr > seg_end_blkaddr) { 2712 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 2713 main_blkaddr, seg_end_blkaddr, 2714 segment_count_main << log_blocks_per_seg); 2715 return true; 2716 } else if (main_end_blkaddr < seg_end_blkaddr) { 2717 int err = 0; 2718 char *res; 2719 2720 /* fix in-memory information all the time */ 2721 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 2722 segment0_blkaddr) >> log_blocks_per_seg); 2723 2724 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 2725 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2726 res = "internally"; 2727 } else { 2728 err = __f2fs_commit_super(bh, NULL); 2729 res = err ? "failed" : "done"; 2730 } 2731 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 2732 res, main_blkaddr, seg_end_blkaddr, 2733 segment_count_main << log_blocks_per_seg); 2734 if (err) 2735 return true; 2736 } 2737 return false; 2738 } 2739 2740 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 2741 struct buffer_head *bh) 2742 { 2743 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 2744 block_t total_sections, blocks_per_seg; 2745 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2746 (bh->b_data + F2FS_SUPER_OFFSET); 2747 unsigned int blocksize; 2748 size_t crc_offset = 0; 2749 __u32 crc = 0; 2750 2751 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 2752 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 2753 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 2754 return -EINVAL; 2755 } 2756 2757 /* Check checksum_offset and crc in superblock */ 2758 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 2759 crc_offset = le32_to_cpu(raw_super->checksum_offset); 2760 if (crc_offset != 2761 offsetof(struct f2fs_super_block, crc)) { 2762 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 2763 crc_offset); 2764 return -EFSCORRUPTED; 2765 } 2766 crc = le32_to_cpu(raw_super->crc); 2767 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 2768 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 2769 return -EFSCORRUPTED; 2770 } 2771 } 2772 2773 /* Currently, support only 4KB page cache size */ 2774 if (F2FS_BLKSIZE != PAGE_SIZE) { 2775 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB", 2776 PAGE_SIZE); 2777 return -EFSCORRUPTED; 2778 } 2779 2780 /* Currently, support only 4KB block size */ 2781 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); 2782 if (blocksize != F2FS_BLKSIZE) { 2783 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB", 2784 blocksize); 2785 return -EFSCORRUPTED; 2786 } 2787 2788 /* check log blocks per segment */ 2789 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 2790 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 2791 le32_to_cpu(raw_super->log_blocks_per_seg)); 2792 return -EFSCORRUPTED; 2793 } 2794 2795 /* Currently, support 512/1024/2048/4096 bytes sector size */ 2796 if (le32_to_cpu(raw_super->log_sectorsize) > 2797 F2FS_MAX_LOG_SECTOR_SIZE || 2798 le32_to_cpu(raw_super->log_sectorsize) < 2799 F2FS_MIN_LOG_SECTOR_SIZE) { 2800 f2fs_info(sbi, "Invalid log sectorsize (%u)", 2801 le32_to_cpu(raw_super->log_sectorsize)); 2802 return -EFSCORRUPTED; 2803 } 2804 if (le32_to_cpu(raw_super->log_sectors_per_block) + 2805 le32_to_cpu(raw_super->log_sectorsize) != 2806 F2FS_MAX_LOG_SECTOR_SIZE) { 2807 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 2808 le32_to_cpu(raw_super->log_sectors_per_block), 2809 le32_to_cpu(raw_super->log_sectorsize)); 2810 return -EFSCORRUPTED; 2811 } 2812 2813 segment_count = le32_to_cpu(raw_super->segment_count); 2814 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2815 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 2816 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 2817 total_sections = le32_to_cpu(raw_super->section_count); 2818 2819 /* blocks_per_seg should be 512, given the above check */ 2820 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 2821 2822 if (segment_count > F2FS_MAX_SEGMENT || 2823 segment_count < F2FS_MIN_SEGMENTS) { 2824 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 2825 return -EFSCORRUPTED; 2826 } 2827 2828 if (total_sections > segment_count_main || total_sections < 1 || 2829 segs_per_sec > segment_count || !segs_per_sec) { 2830 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 2831 segment_count, total_sections, segs_per_sec); 2832 return -EFSCORRUPTED; 2833 } 2834 2835 if (segment_count_main != total_sections * segs_per_sec) { 2836 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 2837 segment_count_main, total_sections, segs_per_sec); 2838 return -EFSCORRUPTED; 2839 } 2840 2841 if ((segment_count / segs_per_sec) < total_sections) { 2842 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 2843 segment_count, segs_per_sec, total_sections); 2844 return -EFSCORRUPTED; 2845 } 2846 2847 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 2848 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 2849 segment_count, le64_to_cpu(raw_super->block_count)); 2850 return -EFSCORRUPTED; 2851 } 2852 2853 if (RDEV(0).path[0]) { 2854 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 2855 int i = 1; 2856 2857 while (i < MAX_DEVICES && RDEV(i).path[0]) { 2858 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 2859 i++; 2860 } 2861 if (segment_count != dev_seg_count) { 2862 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 2863 segment_count, dev_seg_count); 2864 return -EFSCORRUPTED; 2865 } 2866 } else { 2867 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 2868 !bdev_is_zoned(sbi->sb->s_bdev)) { 2869 f2fs_info(sbi, "Zoned block device path is missing"); 2870 return -EFSCORRUPTED; 2871 } 2872 } 2873 2874 if (secs_per_zone > total_sections || !secs_per_zone) { 2875 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 2876 secs_per_zone, total_sections); 2877 return -EFSCORRUPTED; 2878 } 2879 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 2880 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 2881 (le32_to_cpu(raw_super->extension_count) + 2882 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 2883 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 2884 le32_to_cpu(raw_super->extension_count), 2885 raw_super->hot_ext_count, 2886 F2FS_MAX_EXTENSION); 2887 return -EFSCORRUPTED; 2888 } 2889 2890 if (le32_to_cpu(raw_super->cp_payload) > 2891 (blocks_per_seg - F2FS_CP_PACKS)) { 2892 f2fs_info(sbi, "Insane cp_payload (%u > %u)", 2893 le32_to_cpu(raw_super->cp_payload), 2894 blocks_per_seg - F2FS_CP_PACKS); 2895 return -EFSCORRUPTED; 2896 } 2897 2898 /* check reserved ino info */ 2899 if (le32_to_cpu(raw_super->node_ino) != 1 || 2900 le32_to_cpu(raw_super->meta_ino) != 2 || 2901 le32_to_cpu(raw_super->root_ino) != 3) { 2902 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 2903 le32_to_cpu(raw_super->node_ino), 2904 le32_to_cpu(raw_super->meta_ino), 2905 le32_to_cpu(raw_super->root_ino)); 2906 return -EFSCORRUPTED; 2907 } 2908 2909 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 2910 if (sanity_check_area_boundary(sbi, bh)) 2911 return -EFSCORRUPTED; 2912 2913 return 0; 2914 } 2915 2916 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 2917 { 2918 unsigned int total, fsmeta; 2919 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 2920 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 2921 unsigned int ovp_segments, reserved_segments; 2922 unsigned int main_segs, blocks_per_seg; 2923 unsigned int sit_segs, nat_segs; 2924 unsigned int sit_bitmap_size, nat_bitmap_size; 2925 unsigned int log_blocks_per_seg; 2926 unsigned int segment_count_main; 2927 unsigned int cp_pack_start_sum, cp_payload; 2928 block_t user_block_count, valid_user_blocks; 2929 block_t avail_node_count, valid_node_count; 2930 int i, j; 2931 2932 total = le32_to_cpu(raw_super->segment_count); 2933 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 2934 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 2935 fsmeta += sit_segs; 2936 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 2937 fsmeta += nat_segs; 2938 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 2939 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 2940 2941 if (unlikely(fsmeta >= total)) 2942 return 1; 2943 2944 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 2945 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 2946 2947 if (unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 2948 ovp_segments == 0 || reserved_segments == 0)) { 2949 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 2950 return 1; 2951 } 2952 2953 user_block_count = le64_to_cpu(ckpt->user_block_count); 2954 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2955 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2956 if (!user_block_count || user_block_count >= 2957 segment_count_main << log_blocks_per_seg) { 2958 f2fs_err(sbi, "Wrong user_block_count: %u", 2959 user_block_count); 2960 return 1; 2961 } 2962 2963 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 2964 if (valid_user_blocks > user_block_count) { 2965 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 2966 valid_user_blocks, user_block_count); 2967 return 1; 2968 } 2969 2970 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 2971 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 2972 if (valid_node_count > avail_node_count) { 2973 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 2974 valid_node_count, avail_node_count); 2975 return 1; 2976 } 2977 2978 main_segs = le32_to_cpu(raw_super->segment_count_main); 2979 blocks_per_seg = sbi->blocks_per_seg; 2980 2981 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 2982 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 2983 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 2984 return 1; 2985 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 2986 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 2987 le32_to_cpu(ckpt->cur_node_segno[j])) { 2988 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 2989 i, j, 2990 le32_to_cpu(ckpt->cur_node_segno[i])); 2991 return 1; 2992 } 2993 } 2994 } 2995 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 2996 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 2997 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 2998 return 1; 2999 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3000 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3001 le32_to_cpu(ckpt->cur_data_segno[j])) { 3002 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3003 i, j, 3004 le32_to_cpu(ckpt->cur_data_segno[i])); 3005 return 1; 3006 } 3007 } 3008 } 3009 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3010 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3011 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3012 le32_to_cpu(ckpt->cur_data_segno[j])) { 3013 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3014 i, j, 3015 le32_to_cpu(ckpt->cur_node_segno[i])); 3016 return 1; 3017 } 3018 } 3019 } 3020 3021 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3022 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3023 3024 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3025 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3026 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3027 sit_bitmap_size, nat_bitmap_size); 3028 return 1; 3029 } 3030 3031 cp_pack_start_sum = __start_sum_addr(sbi); 3032 cp_payload = __cp_payload(sbi); 3033 if (cp_pack_start_sum < cp_payload + 1 || 3034 cp_pack_start_sum > blocks_per_seg - 1 - 3035 NR_CURSEG_PERSIST_TYPE) { 3036 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3037 cp_pack_start_sum); 3038 return 1; 3039 } 3040 3041 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3042 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3043 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3044 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3045 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3046 le32_to_cpu(ckpt->checksum_offset)); 3047 return 1; 3048 } 3049 3050 if (unlikely(f2fs_cp_error(sbi))) { 3051 f2fs_err(sbi, "A bug case: need to run fsck"); 3052 return 1; 3053 } 3054 return 0; 3055 } 3056 3057 static void init_sb_info(struct f2fs_sb_info *sbi) 3058 { 3059 struct f2fs_super_block *raw_super = sbi->raw_super; 3060 int i; 3061 3062 sbi->log_sectors_per_block = 3063 le32_to_cpu(raw_super->log_sectors_per_block); 3064 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3065 sbi->blocksize = 1 << sbi->log_blocksize; 3066 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3067 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3068 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3069 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3070 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3071 sbi->total_node_count = 3072 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3073 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3074 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); 3075 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); 3076 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); 3077 sbi->cur_victim_sec = NULL_SECNO; 3078 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3079 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3080 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3081 sbi->migration_granularity = sbi->segs_per_sec; 3082 3083 sbi->dir_level = DEF_DIR_LEVEL; 3084 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3085 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3086 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3087 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3088 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3089 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3090 DEF_UMOUNT_DISCARD_TIMEOUT; 3091 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3092 3093 for (i = 0; i < NR_COUNT_TYPE; i++) 3094 atomic_set(&sbi->nr_pages[i], 0); 3095 3096 for (i = 0; i < META; i++) 3097 atomic_set(&sbi->wb_sync_req[i], 0); 3098 3099 INIT_LIST_HEAD(&sbi->s_list); 3100 mutex_init(&sbi->umount_mutex); 3101 init_rwsem(&sbi->io_order_lock); 3102 spin_lock_init(&sbi->cp_lock); 3103 3104 sbi->dirty_device = 0; 3105 spin_lock_init(&sbi->dev_lock); 3106 3107 init_rwsem(&sbi->sb_lock); 3108 init_rwsem(&sbi->pin_sem); 3109 } 3110 3111 static int init_percpu_info(struct f2fs_sb_info *sbi) 3112 { 3113 int err; 3114 3115 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3116 if (err) 3117 return err; 3118 3119 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3120 GFP_KERNEL); 3121 if (err) 3122 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3123 3124 return err; 3125 } 3126 3127 #ifdef CONFIG_BLK_DEV_ZONED 3128 3129 struct f2fs_report_zones_args { 3130 struct f2fs_dev_info *dev; 3131 bool zone_cap_mismatch; 3132 }; 3133 3134 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3135 void *data) 3136 { 3137 struct f2fs_report_zones_args *rz_args = data; 3138 3139 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3140 return 0; 3141 3142 set_bit(idx, rz_args->dev->blkz_seq); 3143 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >> 3144 F2FS_LOG_SECTORS_PER_BLOCK; 3145 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch) 3146 rz_args->zone_cap_mismatch = true; 3147 3148 return 0; 3149 } 3150 3151 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3152 { 3153 struct block_device *bdev = FDEV(devi).bdev; 3154 sector_t nr_sectors = bdev->bd_part->nr_sects; 3155 struct f2fs_report_zones_args rep_zone_arg; 3156 int ret; 3157 3158 if (!f2fs_sb_has_blkzoned(sbi)) 3159 return 0; 3160 3161 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3162 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 3163 return -EINVAL; 3164 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 3165 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3166 __ilog2_u32(sbi->blocks_per_blkz)) 3167 return -EINVAL; 3168 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3169 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3170 sbi->log_blocks_per_blkz; 3171 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 3172 FDEV(devi).nr_blkz++; 3173 3174 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3175 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3176 * sizeof(unsigned long), 3177 GFP_KERNEL); 3178 if (!FDEV(devi).blkz_seq) 3179 return -ENOMEM; 3180 3181 /* Get block zones type and zone-capacity */ 3182 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi, 3183 FDEV(devi).nr_blkz * sizeof(block_t), 3184 GFP_KERNEL); 3185 if (!FDEV(devi).zone_capacity_blocks) 3186 return -ENOMEM; 3187 3188 rep_zone_arg.dev = &FDEV(devi); 3189 rep_zone_arg.zone_cap_mismatch = false; 3190 3191 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3192 &rep_zone_arg); 3193 if (ret < 0) 3194 return ret; 3195 3196 if (!rep_zone_arg.zone_cap_mismatch) { 3197 kfree(FDEV(devi).zone_capacity_blocks); 3198 FDEV(devi).zone_capacity_blocks = NULL; 3199 } 3200 3201 return 0; 3202 } 3203 #endif 3204 3205 /* 3206 * Read f2fs raw super block. 3207 * Because we have two copies of super block, so read both of them 3208 * to get the first valid one. If any one of them is broken, we pass 3209 * them recovery flag back to the caller. 3210 */ 3211 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3212 struct f2fs_super_block **raw_super, 3213 int *valid_super_block, int *recovery) 3214 { 3215 struct super_block *sb = sbi->sb; 3216 int block; 3217 struct buffer_head *bh; 3218 struct f2fs_super_block *super; 3219 int err = 0; 3220 3221 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3222 if (!super) 3223 return -ENOMEM; 3224 3225 for (block = 0; block < 2; block++) { 3226 bh = sb_bread(sb, block); 3227 if (!bh) { 3228 f2fs_err(sbi, "Unable to read %dth superblock", 3229 block + 1); 3230 err = -EIO; 3231 *recovery = 1; 3232 continue; 3233 } 3234 3235 /* sanity checking of raw super */ 3236 err = sanity_check_raw_super(sbi, bh); 3237 if (err) { 3238 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3239 block + 1); 3240 brelse(bh); 3241 *recovery = 1; 3242 continue; 3243 } 3244 3245 if (!*raw_super) { 3246 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3247 sizeof(*super)); 3248 *valid_super_block = block; 3249 *raw_super = super; 3250 } 3251 brelse(bh); 3252 } 3253 3254 /* No valid superblock */ 3255 if (!*raw_super) 3256 kfree(super); 3257 else 3258 err = 0; 3259 3260 return err; 3261 } 3262 3263 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3264 { 3265 struct buffer_head *bh; 3266 __u32 crc = 0; 3267 int err; 3268 3269 if ((recover && f2fs_readonly(sbi->sb)) || 3270 bdev_read_only(sbi->sb->s_bdev)) { 3271 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3272 return -EROFS; 3273 } 3274 3275 /* we should update superblock crc here */ 3276 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3277 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3278 offsetof(struct f2fs_super_block, crc)); 3279 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3280 } 3281 3282 /* write back-up superblock first */ 3283 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3284 if (!bh) 3285 return -EIO; 3286 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3287 brelse(bh); 3288 3289 /* if we are in recovery path, skip writing valid superblock */ 3290 if (recover || err) 3291 return err; 3292 3293 /* write current valid superblock */ 3294 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3295 if (!bh) 3296 return -EIO; 3297 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3298 brelse(bh); 3299 return err; 3300 } 3301 3302 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3303 { 3304 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3305 unsigned int max_devices = MAX_DEVICES; 3306 int i; 3307 3308 /* Initialize single device information */ 3309 if (!RDEV(0).path[0]) { 3310 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3311 return 0; 3312 max_devices = 1; 3313 } 3314 3315 /* 3316 * Initialize multiple devices information, or single 3317 * zoned block device information. 3318 */ 3319 sbi->devs = f2fs_kzalloc(sbi, 3320 array_size(max_devices, 3321 sizeof(struct f2fs_dev_info)), 3322 GFP_KERNEL); 3323 if (!sbi->devs) 3324 return -ENOMEM; 3325 3326 for (i = 0; i < max_devices; i++) { 3327 3328 if (i > 0 && !RDEV(i).path[0]) 3329 break; 3330 3331 if (max_devices == 1) { 3332 /* Single zoned block device mount */ 3333 FDEV(0).bdev = 3334 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3335 sbi->sb->s_mode, sbi->sb->s_type); 3336 } else { 3337 /* Multi-device mount */ 3338 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3339 FDEV(i).total_segments = 3340 le32_to_cpu(RDEV(i).total_segments); 3341 if (i == 0) { 3342 FDEV(i).start_blk = 0; 3343 FDEV(i).end_blk = FDEV(i).start_blk + 3344 (FDEV(i).total_segments << 3345 sbi->log_blocks_per_seg) - 1 + 3346 le32_to_cpu(raw_super->segment0_blkaddr); 3347 } else { 3348 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3349 FDEV(i).end_blk = FDEV(i).start_blk + 3350 (FDEV(i).total_segments << 3351 sbi->log_blocks_per_seg) - 1; 3352 } 3353 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3354 sbi->sb->s_mode, sbi->sb->s_type); 3355 } 3356 if (IS_ERR(FDEV(i).bdev)) 3357 return PTR_ERR(FDEV(i).bdev); 3358 3359 /* to release errored devices */ 3360 sbi->s_ndevs = i + 1; 3361 3362 #ifdef CONFIG_BLK_DEV_ZONED 3363 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3364 !f2fs_sb_has_blkzoned(sbi)) { 3365 f2fs_err(sbi, "Zoned block device feature not enabled\n"); 3366 return -EINVAL; 3367 } 3368 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3369 if (init_blkz_info(sbi, i)) { 3370 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3371 return -EINVAL; 3372 } 3373 if (max_devices == 1) 3374 break; 3375 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3376 i, FDEV(i).path, 3377 FDEV(i).total_segments, 3378 FDEV(i).start_blk, FDEV(i).end_blk, 3379 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3380 "Host-aware" : "Host-managed"); 3381 continue; 3382 } 3383 #endif 3384 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3385 i, FDEV(i).path, 3386 FDEV(i).total_segments, 3387 FDEV(i).start_blk, FDEV(i).end_blk); 3388 } 3389 f2fs_info(sbi, 3390 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3391 return 0; 3392 } 3393 3394 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3395 { 3396 #ifdef CONFIG_UNICODE 3397 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 3398 const struct f2fs_sb_encodings *encoding_info; 3399 struct unicode_map *encoding; 3400 __u16 encoding_flags; 3401 3402 if (f2fs_sb_has_encrypt(sbi)) { 3403 f2fs_err(sbi, 3404 "Can't mount with encoding and encryption"); 3405 return -EINVAL; 3406 } 3407 3408 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info, 3409 &encoding_flags)) { 3410 f2fs_err(sbi, 3411 "Encoding requested by superblock is unknown"); 3412 return -EINVAL; 3413 } 3414 3415 encoding = utf8_load(encoding_info->version); 3416 if (IS_ERR(encoding)) { 3417 f2fs_err(sbi, 3418 "can't mount with superblock charset: %s-%s " 3419 "not supported by the kernel. flags: 0x%x.", 3420 encoding_info->name, encoding_info->version, 3421 encoding_flags); 3422 return PTR_ERR(encoding); 3423 } 3424 f2fs_info(sbi, "Using encoding defined by superblock: " 3425 "%s-%s with flags 0x%hx", encoding_info->name, 3426 encoding_info->version?:"\b", encoding_flags); 3427 3428 sbi->sb->s_encoding = encoding; 3429 sbi->sb->s_encoding_flags = encoding_flags; 3430 sbi->sb->s_d_op = &f2fs_dentry_ops; 3431 } 3432 #else 3433 if (f2fs_sb_has_casefold(sbi)) { 3434 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3435 return -EINVAL; 3436 } 3437 #endif 3438 return 0; 3439 } 3440 3441 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3442 { 3443 struct f2fs_sm_info *sm_i = SM_I(sbi); 3444 3445 /* adjust parameters according to the volume size */ 3446 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3447 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3448 sm_i->dcc_info->discard_granularity = 1; 3449 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 3450 } 3451 3452 sbi->readdir_ra = 1; 3453 } 3454 3455 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3456 { 3457 struct f2fs_sb_info *sbi; 3458 struct f2fs_super_block *raw_super; 3459 struct inode *root; 3460 int err; 3461 bool skip_recovery = false, need_fsck = false; 3462 char *options = NULL; 3463 int recovery, i, valid_super_block; 3464 struct curseg_info *seg_i; 3465 int retry_cnt = 1; 3466 3467 try_onemore: 3468 err = -EINVAL; 3469 raw_super = NULL; 3470 valid_super_block = -1; 3471 recovery = 0; 3472 3473 /* allocate memory for f2fs-specific super block info */ 3474 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 3475 if (!sbi) 3476 return -ENOMEM; 3477 3478 sbi->sb = sb; 3479 3480 /* Load the checksum driver */ 3481 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 3482 if (IS_ERR(sbi->s_chksum_driver)) { 3483 f2fs_err(sbi, "Cannot load crc32 driver."); 3484 err = PTR_ERR(sbi->s_chksum_driver); 3485 sbi->s_chksum_driver = NULL; 3486 goto free_sbi; 3487 } 3488 3489 /* set a block size */ 3490 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 3491 f2fs_err(sbi, "unable to set blocksize"); 3492 goto free_sbi; 3493 } 3494 3495 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 3496 &recovery); 3497 if (err) 3498 goto free_sbi; 3499 3500 sb->s_fs_info = sbi; 3501 sbi->raw_super = raw_super; 3502 3503 /* precompute checksum seed for metadata */ 3504 if (f2fs_sb_has_inode_chksum(sbi)) 3505 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 3506 sizeof(raw_super->uuid)); 3507 3508 default_options(sbi); 3509 /* parse mount options */ 3510 options = kstrdup((const char *)data, GFP_KERNEL); 3511 if (data && !options) { 3512 err = -ENOMEM; 3513 goto free_sb_buf; 3514 } 3515 3516 err = parse_options(sb, options, false); 3517 if (err) 3518 goto free_options; 3519 3520 sbi->max_file_blocks = max_file_blocks(); 3521 sb->s_maxbytes = sbi->max_file_blocks << 3522 le32_to_cpu(raw_super->log_blocksize); 3523 sb->s_max_links = F2FS_LINK_MAX; 3524 3525 err = f2fs_setup_casefold(sbi); 3526 if (err) 3527 goto free_options; 3528 3529 #ifdef CONFIG_QUOTA 3530 sb->dq_op = &f2fs_quota_operations; 3531 sb->s_qcop = &f2fs_quotactl_ops; 3532 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 3533 3534 if (f2fs_sb_has_quota_ino(sbi)) { 3535 for (i = 0; i < MAXQUOTAS; i++) { 3536 if (f2fs_qf_ino(sbi->sb, i)) 3537 sbi->nquota_files++; 3538 } 3539 } 3540 #endif 3541 3542 sb->s_op = &f2fs_sops; 3543 #ifdef CONFIG_FS_ENCRYPTION 3544 sb->s_cop = &f2fs_cryptops; 3545 #endif 3546 #ifdef CONFIG_FS_VERITY 3547 sb->s_vop = &f2fs_verityops; 3548 #endif 3549 sb->s_xattr = f2fs_xattr_handlers; 3550 sb->s_export_op = &f2fs_export_ops; 3551 sb->s_magic = F2FS_SUPER_MAGIC; 3552 sb->s_time_gran = 1; 3553 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3554 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 3555 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 3556 sb->s_iflags |= SB_I_CGROUPWB; 3557 3558 /* init f2fs-specific super block info */ 3559 sbi->valid_super_block = valid_super_block; 3560 init_rwsem(&sbi->gc_lock); 3561 mutex_init(&sbi->writepages); 3562 mutex_init(&sbi->cp_mutex); 3563 init_rwsem(&sbi->node_write); 3564 init_rwsem(&sbi->node_change); 3565 3566 /* disallow all the data/node/meta page writes */ 3567 set_sbi_flag(sbi, SBI_POR_DOING); 3568 spin_lock_init(&sbi->stat_lock); 3569 3570 /* init iostat info */ 3571 spin_lock_init(&sbi->iostat_lock); 3572 sbi->iostat_enable = false; 3573 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS; 3574 3575 for (i = 0; i < NR_PAGE_TYPE; i++) { 3576 int n = (i == META) ? 1: NR_TEMP_TYPE; 3577 int j; 3578 3579 sbi->write_io[i] = 3580 f2fs_kmalloc(sbi, 3581 array_size(n, 3582 sizeof(struct f2fs_bio_info)), 3583 GFP_KERNEL); 3584 if (!sbi->write_io[i]) { 3585 err = -ENOMEM; 3586 goto free_bio_info; 3587 } 3588 3589 for (j = HOT; j < n; j++) { 3590 init_rwsem(&sbi->write_io[i][j].io_rwsem); 3591 sbi->write_io[i][j].sbi = sbi; 3592 sbi->write_io[i][j].bio = NULL; 3593 spin_lock_init(&sbi->write_io[i][j].io_lock); 3594 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 3595 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); 3596 init_rwsem(&sbi->write_io[i][j].bio_list_lock); 3597 } 3598 } 3599 3600 init_rwsem(&sbi->cp_rwsem); 3601 init_rwsem(&sbi->quota_sem); 3602 init_waitqueue_head(&sbi->cp_wait); 3603 init_sb_info(sbi); 3604 3605 err = init_percpu_info(sbi); 3606 if (err) 3607 goto free_bio_info; 3608 3609 if (F2FS_IO_ALIGNED(sbi)) { 3610 sbi->write_io_dummy = 3611 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 3612 if (!sbi->write_io_dummy) { 3613 err = -ENOMEM; 3614 goto free_percpu; 3615 } 3616 } 3617 3618 /* init per sbi slab cache */ 3619 err = f2fs_init_xattr_caches(sbi); 3620 if (err) 3621 goto free_io_dummy; 3622 err = f2fs_init_page_array_cache(sbi); 3623 if (err) 3624 goto free_xattr_cache; 3625 3626 /* get an inode for meta space */ 3627 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 3628 if (IS_ERR(sbi->meta_inode)) { 3629 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 3630 err = PTR_ERR(sbi->meta_inode); 3631 goto free_page_array_cache; 3632 } 3633 3634 err = f2fs_get_valid_checkpoint(sbi); 3635 if (err) { 3636 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 3637 goto free_meta_inode; 3638 } 3639 3640 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 3641 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3642 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 3643 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3644 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 3645 } 3646 3647 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 3648 set_sbi_flag(sbi, SBI_NEED_FSCK); 3649 3650 /* Initialize device list */ 3651 err = f2fs_scan_devices(sbi); 3652 if (err) { 3653 f2fs_err(sbi, "Failed to find devices"); 3654 goto free_devices; 3655 } 3656 3657 err = f2fs_init_post_read_wq(sbi); 3658 if (err) { 3659 f2fs_err(sbi, "Failed to initialize post read workqueue"); 3660 goto free_devices; 3661 } 3662 3663 sbi->total_valid_node_count = 3664 le32_to_cpu(sbi->ckpt->valid_node_count); 3665 percpu_counter_set(&sbi->total_valid_inode_count, 3666 le32_to_cpu(sbi->ckpt->valid_inode_count)); 3667 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 3668 sbi->total_valid_block_count = 3669 le64_to_cpu(sbi->ckpt->valid_block_count); 3670 sbi->last_valid_block_count = sbi->total_valid_block_count; 3671 sbi->reserved_blocks = 0; 3672 sbi->current_reserved_blocks = 0; 3673 limit_reserve_root(sbi); 3674 adjust_unusable_cap_perc(sbi); 3675 3676 for (i = 0; i < NR_INODE_TYPE; i++) { 3677 INIT_LIST_HEAD(&sbi->inode_list[i]); 3678 spin_lock_init(&sbi->inode_lock[i]); 3679 } 3680 mutex_init(&sbi->flush_lock); 3681 3682 f2fs_init_extent_cache_info(sbi); 3683 3684 f2fs_init_ino_entry_info(sbi); 3685 3686 f2fs_init_fsync_node_info(sbi); 3687 3688 /* setup f2fs internal modules */ 3689 err = f2fs_build_segment_manager(sbi); 3690 if (err) { 3691 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 3692 err); 3693 goto free_sm; 3694 } 3695 err = f2fs_build_node_manager(sbi); 3696 if (err) { 3697 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 3698 err); 3699 goto free_nm; 3700 } 3701 3702 /* For write statistics */ 3703 if (sb->s_bdev->bd_part) 3704 sbi->sectors_written_start = 3705 (u64)part_stat_read(sb->s_bdev->bd_part, 3706 sectors[STAT_WRITE]); 3707 3708 /* Read accumulated write IO statistics if exists */ 3709 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 3710 if (__exist_node_summaries(sbi)) 3711 sbi->kbytes_written = 3712 le64_to_cpu(seg_i->journal->info.kbytes_written); 3713 3714 f2fs_build_gc_manager(sbi); 3715 3716 err = f2fs_build_stats(sbi); 3717 if (err) 3718 goto free_nm; 3719 3720 /* get an inode for node space */ 3721 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 3722 if (IS_ERR(sbi->node_inode)) { 3723 f2fs_err(sbi, "Failed to read node inode"); 3724 err = PTR_ERR(sbi->node_inode); 3725 goto free_stats; 3726 } 3727 3728 /* read root inode and dentry */ 3729 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 3730 if (IS_ERR(root)) { 3731 f2fs_err(sbi, "Failed to read root inode"); 3732 err = PTR_ERR(root); 3733 goto free_node_inode; 3734 } 3735 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 3736 !root->i_size || !root->i_nlink) { 3737 iput(root); 3738 err = -EINVAL; 3739 goto free_node_inode; 3740 } 3741 3742 sb->s_root = d_make_root(root); /* allocate root dentry */ 3743 if (!sb->s_root) { 3744 err = -ENOMEM; 3745 goto free_node_inode; 3746 } 3747 3748 err = f2fs_register_sysfs(sbi); 3749 if (err) 3750 goto free_root_inode; 3751 3752 #ifdef CONFIG_QUOTA 3753 /* Enable quota usage during mount */ 3754 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 3755 err = f2fs_enable_quotas(sb); 3756 if (err) 3757 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 3758 } 3759 #endif 3760 /* if there are any orphan inodes, free them */ 3761 err = f2fs_recover_orphan_inodes(sbi); 3762 if (err) 3763 goto free_meta; 3764 3765 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 3766 goto reset_checkpoint; 3767 3768 /* recover fsynced data */ 3769 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 3770 !test_opt(sbi, NORECOVERY)) { 3771 /* 3772 * mount should be failed, when device has readonly mode, and 3773 * previous checkpoint was not done by clean system shutdown. 3774 */ 3775 if (f2fs_hw_is_readonly(sbi)) { 3776 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3777 err = -EROFS; 3778 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable"); 3779 goto free_meta; 3780 } 3781 f2fs_info(sbi, "write access unavailable, skipping recovery"); 3782 goto reset_checkpoint; 3783 } 3784 3785 if (need_fsck) 3786 set_sbi_flag(sbi, SBI_NEED_FSCK); 3787 3788 if (skip_recovery) 3789 goto reset_checkpoint; 3790 3791 err = f2fs_recover_fsync_data(sbi, false); 3792 if (err < 0) { 3793 if (err != -ENOMEM) 3794 skip_recovery = true; 3795 need_fsck = true; 3796 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 3797 err); 3798 goto free_meta; 3799 } 3800 } else { 3801 err = f2fs_recover_fsync_data(sbi, true); 3802 3803 if (!f2fs_readonly(sb) && err > 0) { 3804 err = -EINVAL; 3805 f2fs_err(sbi, "Need to recover fsync data"); 3806 goto free_meta; 3807 } 3808 } 3809 3810 /* 3811 * If the f2fs is not readonly and fsync data recovery succeeds, 3812 * check zoned block devices' write pointer consistency. 3813 */ 3814 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 3815 err = f2fs_check_write_pointer(sbi); 3816 if (err) 3817 goto free_meta; 3818 } 3819 3820 reset_checkpoint: 3821 f2fs_init_inmem_curseg(sbi); 3822 3823 /* f2fs_recover_fsync_data() cleared this already */ 3824 clear_sbi_flag(sbi, SBI_POR_DOING); 3825 3826 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 3827 err = f2fs_disable_checkpoint(sbi); 3828 if (err) 3829 goto sync_free_meta; 3830 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 3831 f2fs_enable_checkpoint(sbi); 3832 } 3833 3834 /* 3835 * If filesystem is not mounted as read-only then 3836 * do start the gc_thread. 3837 */ 3838 if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) { 3839 /* After POR, we can run background GC thread.*/ 3840 err = f2fs_start_gc_thread(sbi); 3841 if (err) 3842 goto sync_free_meta; 3843 } 3844 kvfree(options); 3845 3846 /* recover broken superblock */ 3847 if (recovery) { 3848 err = f2fs_commit_super(sbi, true); 3849 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 3850 sbi->valid_super_block ? 1 : 2, err); 3851 } 3852 3853 f2fs_join_shrinker(sbi); 3854 3855 f2fs_tuning_parameters(sbi); 3856 3857 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 3858 cur_cp_version(F2FS_CKPT(sbi))); 3859 f2fs_update_time(sbi, CP_TIME); 3860 f2fs_update_time(sbi, REQ_TIME); 3861 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3862 return 0; 3863 3864 sync_free_meta: 3865 /* safe to flush all the data */ 3866 sync_filesystem(sbi->sb); 3867 retry_cnt = 0; 3868 3869 free_meta: 3870 #ifdef CONFIG_QUOTA 3871 f2fs_truncate_quota_inode_pages(sb); 3872 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 3873 f2fs_quota_off_umount(sbi->sb); 3874 #endif 3875 /* 3876 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 3877 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 3878 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 3879 * falls into an infinite loop in f2fs_sync_meta_pages(). 3880 */ 3881 truncate_inode_pages_final(META_MAPPING(sbi)); 3882 /* evict some inodes being cached by GC */ 3883 evict_inodes(sb); 3884 f2fs_unregister_sysfs(sbi); 3885 free_root_inode: 3886 dput(sb->s_root); 3887 sb->s_root = NULL; 3888 free_node_inode: 3889 f2fs_release_ino_entry(sbi, true); 3890 truncate_inode_pages_final(NODE_MAPPING(sbi)); 3891 iput(sbi->node_inode); 3892 sbi->node_inode = NULL; 3893 free_stats: 3894 f2fs_destroy_stats(sbi); 3895 free_nm: 3896 f2fs_destroy_node_manager(sbi); 3897 free_sm: 3898 f2fs_destroy_segment_manager(sbi); 3899 f2fs_destroy_post_read_wq(sbi); 3900 free_devices: 3901 destroy_device_list(sbi); 3902 kvfree(sbi->ckpt); 3903 free_meta_inode: 3904 make_bad_inode(sbi->meta_inode); 3905 iput(sbi->meta_inode); 3906 sbi->meta_inode = NULL; 3907 free_page_array_cache: 3908 f2fs_destroy_page_array_cache(sbi); 3909 free_xattr_cache: 3910 f2fs_destroy_xattr_caches(sbi); 3911 free_io_dummy: 3912 mempool_destroy(sbi->write_io_dummy); 3913 free_percpu: 3914 destroy_percpu_info(sbi); 3915 free_bio_info: 3916 for (i = 0; i < NR_PAGE_TYPE; i++) 3917 kvfree(sbi->write_io[i]); 3918 3919 #ifdef CONFIG_UNICODE 3920 utf8_unload(sb->s_encoding); 3921 #endif 3922 free_options: 3923 #ifdef CONFIG_QUOTA 3924 for (i = 0; i < MAXQUOTAS; i++) 3925 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 3926 #endif 3927 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 3928 kvfree(options); 3929 free_sb_buf: 3930 kfree(raw_super); 3931 free_sbi: 3932 if (sbi->s_chksum_driver) 3933 crypto_free_shash(sbi->s_chksum_driver); 3934 kfree(sbi); 3935 3936 /* give only one another chance */ 3937 if (retry_cnt > 0 && skip_recovery) { 3938 retry_cnt--; 3939 shrink_dcache_sb(sb); 3940 goto try_onemore; 3941 } 3942 return err; 3943 } 3944 3945 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 3946 const char *dev_name, void *data) 3947 { 3948 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 3949 } 3950 3951 static void kill_f2fs_super(struct super_block *sb) 3952 { 3953 if (sb->s_root) { 3954 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3955 3956 set_sbi_flag(sbi, SBI_IS_CLOSE); 3957 f2fs_stop_gc_thread(sbi); 3958 f2fs_stop_discard_thread(sbi); 3959 3960 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 3961 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3962 struct cp_control cpc = { 3963 .reason = CP_UMOUNT, 3964 }; 3965 f2fs_write_checkpoint(sbi, &cpc); 3966 } 3967 3968 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 3969 sb->s_flags &= ~SB_RDONLY; 3970 } 3971 kill_block_super(sb); 3972 } 3973 3974 static struct file_system_type f2fs_fs_type = { 3975 .owner = THIS_MODULE, 3976 .name = "f2fs", 3977 .mount = f2fs_mount, 3978 .kill_sb = kill_f2fs_super, 3979 .fs_flags = FS_REQUIRES_DEV, 3980 }; 3981 MODULE_ALIAS_FS("f2fs"); 3982 3983 static int __init init_inodecache(void) 3984 { 3985 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 3986 sizeof(struct f2fs_inode_info), 0, 3987 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 3988 if (!f2fs_inode_cachep) 3989 return -ENOMEM; 3990 return 0; 3991 } 3992 3993 static void destroy_inodecache(void) 3994 { 3995 /* 3996 * Make sure all delayed rcu free inodes are flushed before we 3997 * destroy cache. 3998 */ 3999 rcu_barrier(); 4000 kmem_cache_destroy(f2fs_inode_cachep); 4001 } 4002 4003 static int __init init_f2fs_fs(void) 4004 { 4005 int err; 4006 4007 if (PAGE_SIZE != F2FS_BLKSIZE) { 4008 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4009 PAGE_SIZE, F2FS_BLKSIZE); 4010 return -EINVAL; 4011 } 4012 4013 f2fs_build_trace_ios(); 4014 4015 err = init_inodecache(); 4016 if (err) 4017 goto fail; 4018 err = f2fs_create_node_manager_caches(); 4019 if (err) 4020 goto free_inodecache; 4021 err = f2fs_create_segment_manager_caches(); 4022 if (err) 4023 goto free_node_manager_caches; 4024 err = f2fs_create_checkpoint_caches(); 4025 if (err) 4026 goto free_segment_manager_caches; 4027 err = f2fs_create_extent_cache(); 4028 if (err) 4029 goto free_checkpoint_caches; 4030 err = f2fs_create_garbage_collection_cache(); 4031 if (err) 4032 goto free_extent_cache; 4033 err = f2fs_init_sysfs(); 4034 if (err) 4035 goto free_garbage_collection_cache; 4036 err = register_shrinker(&f2fs_shrinker_info); 4037 if (err) 4038 goto free_sysfs; 4039 err = register_filesystem(&f2fs_fs_type); 4040 if (err) 4041 goto free_shrinker; 4042 f2fs_create_root_stats(); 4043 err = f2fs_init_post_read_processing(); 4044 if (err) 4045 goto free_root_stats; 4046 err = f2fs_init_bio_entry_cache(); 4047 if (err) 4048 goto free_post_read; 4049 err = f2fs_init_bioset(); 4050 if (err) 4051 goto free_bio_enrty_cache; 4052 err = f2fs_init_compress_mempool(); 4053 if (err) 4054 goto free_bioset; 4055 err = f2fs_init_compress_cache(); 4056 if (err) 4057 goto free_compress_mempool; 4058 return 0; 4059 free_compress_mempool: 4060 f2fs_destroy_compress_mempool(); 4061 free_bioset: 4062 f2fs_destroy_bioset(); 4063 free_bio_enrty_cache: 4064 f2fs_destroy_bio_entry_cache(); 4065 free_post_read: 4066 f2fs_destroy_post_read_processing(); 4067 free_root_stats: 4068 f2fs_destroy_root_stats(); 4069 unregister_filesystem(&f2fs_fs_type); 4070 free_shrinker: 4071 unregister_shrinker(&f2fs_shrinker_info); 4072 free_sysfs: 4073 f2fs_exit_sysfs(); 4074 free_garbage_collection_cache: 4075 f2fs_destroy_garbage_collection_cache(); 4076 free_extent_cache: 4077 f2fs_destroy_extent_cache(); 4078 free_checkpoint_caches: 4079 f2fs_destroy_checkpoint_caches(); 4080 free_segment_manager_caches: 4081 f2fs_destroy_segment_manager_caches(); 4082 free_node_manager_caches: 4083 f2fs_destroy_node_manager_caches(); 4084 free_inodecache: 4085 destroy_inodecache(); 4086 fail: 4087 return err; 4088 } 4089 4090 static void __exit exit_f2fs_fs(void) 4091 { 4092 f2fs_destroy_compress_cache(); 4093 f2fs_destroy_compress_mempool(); 4094 f2fs_destroy_bioset(); 4095 f2fs_destroy_bio_entry_cache(); 4096 f2fs_destroy_post_read_processing(); 4097 f2fs_destroy_root_stats(); 4098 unregister_filesystem(&f2fs_fs_type); 4099 unregister_shrinker(&f2fs_shrinker_info); 4100 f2fs_exit_sysfs(); 4101 f2fs_destroy_garbage_collection_cache(); 4102 f2fs_destroy_extent_cache(); 4103 f2fs_destroy_checkpoint_caches(); 4104 f2fs_destroy_segment_manager_caches(); 4105 f2fs_destroy_node_manager_caches(); 4106 destroy_inodecache(); 4107 f2fs_destroy_trace_ios(); 4108 } 4109 4110 module_init(init_f2fs_fs) 4111 module_exit(exit_f2fs_fs) 4112 4113 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4114 MODULE_DESCRIPTION("Flash Friendly File System"); 4115 MODULE_LICENSE("GPL"); 4116 4117