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.val[0] = (u32)id; 1446 buf->f_fsid.val[1] = (u32)(id >> 32); 1447 1448 #ifdef CONFIG_QUOTA 1449 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1450 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1451 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1452 } 1453 #endif 1454 return 0; 1455 } 1456 1457 static inline void f2fs_show_quota_options(struct seq_file *seq, 1458 struct super_block *sb) 1459 { 1460 #ifdef CONFIG_QUOTA 1461 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1462 1463 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1464 char *fmtname = ""; 1465 1466 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1467 case QFMT_VFS_OLD: 1468 fmtname = "vfsold"; 1469 break; 1470 case QFMT_VFS_V0: 1471 fmtname = "vfsv0"; 1472 break; 1473 case QFMT_VFS_V1: 1474 fmtname = "vfsv1"; 1475 break; 1476 } 1477 seq_printf(seq, ",jqfmt=%s", fmtname); 1478 } 1479 1480 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1481 seq_show_option(seq, "usrjquota", 1482 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1483 1484 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1485 seq_show_option(seq, "grpjquota", 1486 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1487 1488 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1489 seq_show_option(seq, "prjjquota", 1490 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1491 #endif 1492 } 1493 1494 static inline void f2fs_show_compress_options(struct seq_file *seq, 1495 struct super_block *sb) 1496 { 1497 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1498 char *algtype = ""; 1499 int i; 1500 1501 if (!f2fs_sb_has_compression(sbi)) 1502 return; 1503 1504 switch (F2FS_OPTION(sbi).compress_algorithm) { 1505 case COMPRESS_LZO: 1506 algtype = "lzo"; 1507 break; 1508 case COMPRESS_LZ4: 1509 algtype = "lz4"; 1510 break; 1511 case COMPRESS_ZSTD: 1512 algtype = "zstd"; 1513 break; 1514 case COMPRESS_LZORLE: 1515 algtype = "lzo-rle"; 1516 break; 1517 } 1518 seq_printf(seq, ",compress_algorithm=%s", algtype); 1519 1520 seq_printf(seq, ",compress_log_size=%u", 1521 F2FS_OPTION(sbi).compress_log_size); 1522 1523 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1524 seq_printf(seq, ",compress_extension=%s", 1525 F2FS_OPTION(sbi).extensions[i]); 1526 } 1527 } 1528 1529 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1530 { 1531 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1532 1533 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1534 seq_printf(seq, ",background_gc=%s", "sync"); 1535 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1536 seq_printf(seq, ",background_gc=%s", "on"); 1537 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1538 seq_printf(seq, ",background_gc=%s", "off"); 1539 1540 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1541 seq_puts(seq, ",disable_roll_forward"); 1542 if (test_opt(sbi, NORECOVERY)) 1543 seq_puts(seq, ",norecovery"); 1544 if (test_opt(sbi, DISCARD)) 1545 seq_puts(seq, ",discard"); 1546 else 1547 seq_puts(seq, ",nodiscard"); 1548 if (test_opt(sbi, NOHEAP)) 1549 seq_puts(seq, ",no_heap"); 1550 else 1551 seq_puts(seq, ",heap"); 1552 #ifdef CONFIG_F2FS_FS_XATTR 1553 if (test_opt(sbi, XATTR_USER)) 1554 seq_puts(seq, ",user_xattr"); 1555 else 1556 seq_puts(seq, ",nouser_xattr"); 1557 if (test_opt(sbi, INLINE_XATTR)) 1558 seq_puts(seq, ",inline_xattr"); 1559 else 1560 seq_puts(seq, ",noinline_xattr"); 1561 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1562 seq_printf(seq, ",inline_xattr_size=%u", 1563 F2FS_OPTION(sbi).inline_xattr_size); 1564 #endif 1565 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1566 if (test_opt(sbi, POSIX_ACL)) 1567 seq_puts(seq, ",acl"); 1568 else 1569 seq_puts(seq, ",noacl"); 1570 #endif 1571 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1572 seq_puts(seq, ",disable_ext_identify"); 1573 if (test_opt(sbi, INLINE_DATA)) 1574 seq_puts(seq, ",inline_data"); 1575 else 1576 seq_puts(seq, ",noinline_data"); 1577 if (test_opt(sbi, INLINE_DENTRY)) 1578 seq_puts(seq, ",inline_dentry"); 1579 else 1580 seq_puts(seq, ",noinline_dentry"); 1581 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1582 seq_puts(seq, ",flush_merge"); 1583 if (test_opt(sbi, NOBARRIER)) 1584 seq_puts(seq, ",nobarrier"); 1585 if (test_opt(sbi, FASTBOOT)) 1586 seq_puts(seq, ",fastboot"); 1587 if (test_opt(sbi, EXTENT_CACHE)) 1588 seq_puts(seq, ",extent_cache"); 1589 else 1590 seq_puts(seq, ",noextent_cache"); 1591 if (test_opt(sbi, DATA_FLUSH)) 1592 seq_puts(seq, ",data_flush"); 1593 1594 seq_puts(seq, ",mode="); 1595 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1596 seq_puts(seq, "adaptive"); 1597 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1598 seq_puts(seq, "lfs"); 1599 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1600 if (test_opt(sbi, RESERVE_ROOT)) 1601 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1602 F2FS_OPTION(sbi).root_reserved_blocks, 1603 from_kuid_munged(&init_user_ns, 1604 F2FS_OPTION(sbi).s_resuid), 1605 from_kgid_munged(&init_user_ns, 1606 F2FS_OPTION(sbi).s_resgid)); 1607 if (F2FS_IO_SIZE_BITS(sbi)) 1608 seq_printf(seq, ",io_bits=%u", 1609 F2FS_OPTION(sbi).write_io_size_bits); 1610 #ifdef CONFIG_F2FS_FAULT_INJECTION 1611 if (test_opt(sbi, FAULT_INJECTION)) { 1612 seq_printf(seq, ",fault_injection=%u", 1613 F2FS_OPTION(sbi).fault_info.inject_rate); 1614 seq_printf(seq, ",fault_type=%u", 1615 F2FS_OPTION(sbi).fault_info.inject_type); 1616 } 1617 #endif 1618 #ifdef CONFIG_QUOTA 1619 if (test_opt(sbi, QUOTA)) 1620 seq_puts(seq, ",quota"); 1621 if (test_opt(sbi, USRQUOTA)) 1622 seq_puts(seq, ",usrquota"); 1623 if (test_opt(sbi, GRPQUOTA)) 1624 seq_puts(seq, ",grpquota"); 1625 if (test_opt(sbi, PRJQUOTA)) 1626 seq_puts(seq, ",prjquota"); 1627 #endif 1628 f2fs_show_quota_options(seq, sbi->sb); 1629 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1630 seq_printf(seq, ",whint_mode=%s", "user-based"); 1631 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1632 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1633 1634 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1635 1636 if (sbi->sb->s_flags & SB_INLINECRYPT) 1637 seq_puts(seq, ",inlinecrypt"); 1638 1639 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1640 seq_printf(seq, ",alloc_mode=%s", "default"); 1641 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1642 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1643 1644 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1645 seq_printf(seq, ",checkpoint=disable:%u", 1646 F2FS_OPTION(sbi).unusable_cap); 1647 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1648 seq_printf(seq, ",fsync_mode=%s", "posix"); 1649 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1650 seq_printf(seq, ",fsync_mode=%s", "strict"); 1651 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1652 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 1653 1654 #ifdef CONFIG_F2FS_FS_COMPRESSION 1655 f2fs_show_compress_options(seq, sbi->sb); 1656 #endif 1657 1658 if (test_opt(sbi, ATGC)) 1659 seq_puts(seq, ",atgc"); 1660 return 0; 1661 } 1662 1663 static void default_options(struct f2fs_sb_info *sbi) 1664 { 1665 /* init some FS parameters */ 1666 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 1667 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 1668 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1669 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1670 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1671 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 1672 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 1673 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 1674 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 1675 F2FS_OPTION(sbi).compress_ext_cnt = 0; 1676 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 1677 1678 sbi->sb->s_flags &= ~SB_INLINECRYPT; 1679 1680 set_opt(sbi, INLINE_XATTR); 1681 set_opt(sbi, INLINE_DATA); 1682 set_opt(sbi, INLINE_DENTRY); 1683 set_opt(sbi, EXTENT_CACHE); 1684 set_opt(sbi, NOHEAP); 1685 clear_opt(sbi, DISABLE_CHECKPOINT); 1686 F2FS_OPTION(sbi).unusable_cap = 0; 1687 sbi->sb->s_flags |= SB_LAZYTIME; 1688 set_opt(sbi, FLUSH_MERGE); 1689 set_opt(sbi, DISCARD); 1690 if (f2fs_sb_has_blkzoned(sbi)) 1691 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 1692 else 1693 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 1694 1695 #ifdef CONFIG_F2FS_FS_XATTR 1696 set_opt(sbi, XATTR_USER); 1697 #endif 1698 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1699 set_opt(sbi, POSIX_ACL); 1700 #endif 1701 1702 f2fs_build_fault_attr(sbi, 0, 0); 1703 } 1704 1705 #ifdef CONFIG_QUOTA 1706 static int f2fs_enable_quotas(struct super_block *sb); 1707 #endif 1708 1709 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 1710 { 1711 unsigned int s_flags = sbi->sb->s_flags; 1712 struct cp_control cpc; 1713 int err = 0; 1714 int ret; 1715 block_t unusable; 1716 1717 if (s_flags & SB_RDONLY) { 1718 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 1719 return -EINVAL; 1720 } 1721 sbi->sb->s_flags |= SB_ACTIVE; 1722 1723 f2fs_update_time(sbi, DISABLE_TIME); 1724 1725 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 1726 down_write(&sbi->gc_lock); 1727 err = f2fs_gc(sbi, true, false, NULL_SEGNO); 1728 if (err == -ENODATA) { 1729 err = 0; 1730 break; 1731 } 1732 if (err && err != -EAGAIN) 1733 break; 1734 } 1735 1736 ret = sync_filesystem(sbi->sb); 1737 if (ret || err) { 1738 err = ret ? ret: err; 1739 goto restore_flag; 1740 } 1741 1742 unusable = f2fs_get_unusable_blocks(sbi); 1743 if (f2fs_disable_cp_again(sbi, unusable)) { 1744 err = -EAGAIN; 1745 goto restore_flag; 1746 } 1747 1748 down_write(&sbi->gc_lock); 1749 cpc.reason = CP_PAUSE; 1750 set_sbi_flag(sbi, SBI_CP_DISABLED); 1751 err = f2fs_write_checkpoint(sbi, &cpc); 1752 if (err) 1753 goto out_unlock; 1754 1755 spin_lock(&sbi->stat_lock); 1756 sbi->unusable_block_count = unusable; 1757 spin_unlock(&sbi->stat_lock); 1758 1759 out_unlock: 1760 up_write(&sbi->gc_lock); 1761 restore_flag: 1762 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 1763 return err; 1764 } 1765 1766 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 1767 { 1768 down_write(&sbi->gc_lock); 1769 f2fs_dirty_to_prefree(sbi); 1770 1771 clear_sbi_flag(sbi, SBI_CP_DISABLED); 1772 set_sbi_flag(sbi, SBI_IS_DIRTY); 1773 up_write(&sbi->gc_lock); 1774 1775 f2fs_sync_fs(sbi->sb, 1); 1776 } 1777 1778 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 1779 { 1780 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1781 struct f2fs_mount_info org_mount_opt; 1782 unsigned long old_sb_flags; 1783 int err; 1784 bool need_restart_gc = false; 1785 bool need_stop_gc = false; 1786 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 1787 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT); 1788 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 1789 bool no_atgc = !test_opt(sbi, ATGC); 1790 bool checkpoint_changed; 1791 #ifdef CONFIG_QUOTA 1792 int i, j; 1793 #endif 1794 1795 /* 1796 * Save the old mount options in case we 1797 * need to restore them. 1798 */ 1799 org_mount_opt = sbi->mount_opt; 1800 old_sb_flags = sb->s_flags; 1801 1802 #ifdef CONFIG_QUOTA 1803 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 1804 for (i = 0; i < MAXQUOTAS; i++) { 1805 if (F2FS_OPTION(sbi).s_qf_names[i]) { 1806 org_mount_opt.s_qf_names[i] = 1807 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 1808 GFP_KERNEL); 1809 if (!org_mount_opt.s_qf_names[i]) { 1810 for (j = 0; j < i; j++) 1811 kfree(org_mount_opt.s_qf_names[j]); 1812 return -ENOMEM; 1813 } 1814 } else { 1815 org_mount_opt.s_qf_names[i] = NULL; 1816 } 1817 } 1818 #endif 1819 1820 /* recover superblocks we couldn't write due to previous RO mount */ 1821 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 1822 err = f2fs_commit_super(sbi, false); 1823 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 1824 err); 1825 if (!err) 1826 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1827 } 1828 1829 default_options(sbi); 1830 1831 /* parse mount options */ 1832 err = parse_options(sb, data, true); 1833 if (err) 1834 goto restore_opts; 1835 checkpoint_changed = 1836 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT); 1837 1838 /* 1839 * Previous and new state of filesystem is RO, 1840 * so skip checking GC and FLUSH_MERGE conditions. 1841 */ 1842 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 1843 goto skip; 1844 1845 #ifdef CONFIG_QUOTA 1846 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 1847 err = dquot_suspend(sb, -1); 1848 if (err < 0) 1849 goto restore_opts; 1850 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 1851 /* dquot_resume needs RW */ 1852 sb->s_flags &= ~SB_RDONLY; 1853 if (sb_any_quota_suspended(sb)) { 1854 dquot_resume(sb, -1); 1855 } else if (f2fs_sb_has_quota_ino(sbi)) { 1856 err = f2fs_enable_quotas(sb); 1857 if (err) 1858 goto restore_opts; 1859 } 1860 } 1861 #endif 1862 /* disallow enable atgc dynamically */ 1863 if (no_atgc == !!test_opt(sbi, ATGC)) { 1864 err = -EINVAL; 1865 f2fs_warn(sbi, "switch atgc option is not allowed"); 1866 goto restore_opts; 1867 } 1868 1869 /* disallow enable/disable extent_cache dynamically */ 1870 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 1871 err = -EINVAL; 1872 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 1873 goto restore_opts; 1874 } 1875 1876 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 1877 err = -EINVAL; 1878 f2fs_warn(sbi, "switch io_bits option is not allowed"); 1879 goto restore_opts; 1880 } 1881 1882 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 1883 err = -EINVAL; 1884 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 1885 goto restore_opts; 1886 } 1887 1888 /* 1889 * We stop the GC thread if FS is mounted as RO 1890 * or if background_gc = off is passed in mount 1891 * option. Also sync the filesystem. 1892 */ 1893 if ((*flags & SB_RDONLY) || 1894 F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) { 1895 if (sbi->gc_thread) { 1896 f2fs_stop_gc_thread(sbi); 1897 need_restart_gc = true; 1898 } 1899 } else if (!sbi->gc_thread) { 1900 err = f2fs_start_gc_thread(sbi); 1901 if (err) 1902 goto restore_opts; 1903 need_stop_gc = true; 1904 } 1905 1906 if (*flags & SB_RDONLY || 1907 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 1908 writeback_inodes_sb(sb, WB_REASON_SYNC); 1909 sync_inodes_sb(sb); 1910 1911 set_sbi_flag(sbi, SBI_IS_DIRTY); 1912 set_sbi_flag(sbi, SBI_IS_CLOSE); 1913 f2fs_sync_fs(sb, 1); 1914 clear_sbi_flag(sbi, SBI_IS_CLOSE); 1915 } 1916 1917 if (checkpoint_changed) { 1918 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 1919 err = f2fs_disable_checkpoint(sbi); 1920 if (err) 1921 goto restore_gc; 1922 } else { 1923 f2fs_enable_checkpoint(sbi); 1924 } 1925 } 1926 1927 /* 1928 * We stop issue flush thread if FS is mounted as RO 1929 * or if flush_merge is not passed in mount option. 1930 */ 1931 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 1932 clear_opt(sbi, FLUSH_MERGE); 1933 f2fs_destroy_flush_cmd_control(sbi, false); 1934 } else { 1935 err = f2fs_create_flush_cmd_control(sbi); 1936 if (err) 1937 goto restore_gc; 1938 } 1939 skip: 1940 #ifdef CONFIG_QUOTA 1941 /* Release old quota file names */ 1942 for (i = 0; i < MAXQUOTAS; i++) 1943 kfree(org_mount_opt.s_qf_names[i]); 1944 #endif 1945 /* Update the POSIXACL Flag */ 1946 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 1947 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 1948 1949 limit_reserve_root(sbi); 1950 adjust_unusable_cap_perc(sbi); 1951 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 1952 return 0; 1953 restore_gc: 1954 if (need_restart_gc) { 1955 if (f2fs_start_gc_thread(sbi)) 1956 f2fs_warn(sbi, "background gc thread has stopped"); 1957 } else if (need_stop_gc) { 1958 f2fs_stop_gc_thread(sbi); 1959 } 1960 restore_opts: 1961 #ifdef CONFIG_QUOTA 1962 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 1963 for (i = 0; i < MAXQUOTAS; i++) { 1964 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1965 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 1966 } 1967 #endif 1968 sbi->mount_opt = org_mount_opt; 1969 sb->s_flags = old_sb_flags; 1970 return err; 1971 } 1972 1973 #ifdef CONFIG_QUOTA 1974 /* Read data from quotafile */ 1975 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 1976 size_t len, loff_t off) 1977 { 1978 struct inode *inode = sb_dqopt(sb)->files[type]; 1979 struct address_space *mapping = inode->i_mapping; 1980 block_t blkidx = F2FS_BYTES_TO_BLK(off); 1981 int offset = off & (sb->s_blocksize - 1); 1982 int tocopy; 1983 size_t toread; 1984 loff_t i_size = i_size_read(inode); 1985 struct page *page; 1986 char *kaddr; 1987 1988 if (off > i_size) 1989 return 0; 1990 1991 if (off + len > i_size) 1992 len = i_size - off; 1993 toread = len; 1994 while (toread > 0) { 1995 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 1996 repeat: 1997 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 1998 if (IS_ERR(page)) { 1999 if (PTR_ERR(page) == -ENOMEM) { 2000 congestion_wait(BLK_RW_ASYNC, 2001 DEFAULT_IO_TIMEOUT); 2002 goto repeat; 2003 } 2004 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2005 return PTR_ERR(page); 2006 } 2007 2008 lock_page(page); 2009 2010 if (unlikely(page->mapping != mapping)) { 2011 f2fs_put_page(page, 1); 2012 goto repeat; 2013 } 2014 if (unlikely(!PageUptodate(page))) { 2015 f2fs_put_page(page, 1); 2016 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2017 return -EIO; 2018 } 2019 2020 kaddr = kmap_atomic(page); 2021 memcpy(data, kaddr + offset, tocopy); 2022 kunmap_atomic(kaddr); 2023 f2fs_put_page(page, 1); 2024 2025 offset = 0; 2026 toread -= tocopy; 2027 data += tocopy; 2028 blkidx++; 2029 } 2030 return len; 2031 } 2032 2033 /* Write to quotafile */ 2034 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2035 const char *data, size_t len, loff_t off) 2036 { 2037 struct inode *inode = sb_dqopt(sb)->files[type]; 2038 struct address_space *mapping = inode->i_mapping; 2039 const struct address_space_operations *a_ops = mapping->a_ops; 2040 int offset = off & (sb->s_blocksize - 1); 2041 size_t towrite = len; 2042 struct page *page; 2043 void *fsdata = NULL; 2044 char *kaddr; 2045 int err = 0; 2046 int tocopy; 2047 2048 while (towrite > 0) { 2049 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2050 towrite); 2051 retry: 2052 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2053 &page, &fsdata); 2054 if (unlikely(err)) { 2055 if (err == -ENOMEM) { 2056 congestion_wait(BLK_RW_ASYNC, 2057 DEFAULT_IO_TIMEOUT); 2058 goto retry; 2059 } 2060 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2061 break; 2062 } 2063 2064 kaddr = kmap_atomic(page); 2065 memcpy(kaddr + offset, data, tocopy); 2066 kunmap_atomic(kaddr); 2067 flush_dcache_page(page); 2068 2069 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2070 page, fsdata); 2071 offset = 0; 2072 towrite -= tocopy; 2073 off += tocopy; 2074 data += tocopy; 2075 cond_resched(); 2076 } 2077 2078 if (len == towrite) 2079 return err; 2080 inode->i_mtime = inode->i_ctime = current_time(inode); 2081 f2fs_mark_inode_dirty_sync(inode, false); 2082 return len - towrite; 2083 } 2084 2085 static struct dquot **f2fs_get_dquots(struct inode *inode) 2086 { 2087 return F2FS_I(inode)->i_dquot; 2088 } 2089 2090 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2091 { 2092 return &F2FS_I(inode)->i_reserved_quota; 2093 } 2094 2095 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2096 { 2097 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2098 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2099 return 0; 2100 } 2101 2102 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2103 F2FS_OPTION(sbi).s_jquota_fmt, type); 2104 } 2105 2106 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2107 { 2108 int enabled = 0; 2109 int i, err; 2110 2111 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2112 err = f2fs_enable_quotas(sbi->sb); 2113 if (err) { 2114 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2115 return 0; 2116 } 2117 return 1; 2118 } 2119 2120 for (i = 0; i < MAXQUOTAS; i++) { 2121 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2122 err = f2fs_quota_on_mount(sbi, i); 2123 if (!err) { 2124 enabled = 1; 2125 continue; 2126 } 2127 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2128 err, i); 2129 } 2130 } 2131 return enabled; 2132 } 2133 2134 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2135 unsigned int flags) 2136 { 2137 struct inode *qf_inode; 2138 unsigned long qf_inum; 2139 int err; 2140 2141 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2142 2143 qf_inum = f2fs_qf_ino(sb, type); 2144 if (!qf_inum) 2145 return -EPERM; 2146 2147 qf_inode = f2fs_iget(sb, qf_inum); 2148 if (IS_ERR(qf_inode)) { 2149 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2150 return PTR_ERR(qf_inode); 2151 } 2152 2153 /* Don't account quota for quota files to avoid recursion */ 2154 qf_inode->i_flags |= S_NOQUOTA; 2155 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2156 iput(qf_inode); 2157 return err; 2158 } 2159 2160 static int f2fs_enable_quotas(struct super_block *sb) 2161 { 2162 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2163 int type, err = 0; 2164 unsigned long qf_inum; 2165 bool quota_mopt[MAXQUOTAS] = { 2166 test_opt(sbi, USRQUOTA), 2167 test_opt(sbi, GRPQUOTA), 2168 test_opt(sbi, PRJQUOTA), 2169 }; 2170 2171 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2172 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2173 return 0; 2174 } 2175 2176 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2177 2178 for (type = 0; type < MAXQUOTAS; type++) { 2179 qf_inum = f2fs_qf_ino(sb, type); 2180 if (qf_inum) { 2181 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2182 DQUOT_USAGE_ENABLED | 2183 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2184 if (err) { 2185 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2186 type, err); 2187 for (type--; type >= 0; type--) 2188 dquot_quota_off(sb, type); 2189 set_sbi_flag(F2FS_SB(sb), 2190 SBI_QUOTA_NEED_REPAIR); 2191 return err; 2192 } 2193 } 2194 } 2195 return 0; 2196 } 2197 2198 int f2fs_quota_sync(struct super_block *sb, int type) 2199 { 2200 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2201 struct quota_info *dqopt = sb_dqopt(sb); 2202 int cnt; 2203 int ret; 2204 2205 /* 2206 * do_quotactl 2207 * f2fs_quota_sync 2208 * down_read(quota_sem) 2209 * dquot_writeback_dquots() 2210 * f2fs_dquot_commit 2211 * block_operation 2212 * down_read(quota_sem) 2213 */ 2214 f2fs_lock_op(sbi); 2215 2216 down_read(&sbi->quota_sem); 2217 ret = dquot_writeback_dquots(sb, type); 2218 if (ret) 2219 goto out; 2220 2221 /* 2222 * Now when everything is written we can discard the pagecache so 2223 * that userspace sees the changes. 2224 */ 2225 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2226 struct address_space *mapping; 2227 2228 if (type != -1 && cnt != type) 2229 continue; 2230 if (!sb_has_quota_active(sb, cnt)) 2231 continue; 2232 2233 mapping = dqopt->files[cnt]->i_mapping; 2234 2235 ret = filemap_fdatawrite(mapping); 2236 if (ret) 2237 goto out; 2238 2239 /* if we are using journalled quota */ 2240 if (is_journalled_quota(sbi)) 2241 continue; 2242 2243 ret = filemap_fdatawait(mapping); 2244 if (ret) 2245 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2246 2247 inode_lock(dqopt->files[cnt]); 2248 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0); 2249 inode_unlock(dqopt->files[cnt]); 2250 } 2251 out: 2252 if (ret) 2253 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2254 up_read(&sbi->quota_sem); 2255 f2fs_unlock_op(sbi); 2256 return ret; 2257 } 2258 2259 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2260 const struct path *path) 2261 { 2262 struct inode *inode; 2263 int err; 2264 2265 /* if quota sysfile exists, deny enabling quota with specific file */ 2266 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2267 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2268 return -EBUSY; 2269 } 2270 2271 err = f2fs_quota_sync(sb, type); 2272 if (err) 2273 return err; 2274 2275 err = dquot_quota_on(sb, type, format_id, path); 2276 if (err) 2277 return err; 2278 2279 inode = d_inode(path->dentry); 2280 2281 inode_lock(inode); 2282 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2283 f2fs_set_inode_flags(inode); 2284 inode_unlock(inode); 2285 f2fs_mark_inode_dirty_sync(inode, false); 2286 2287 return 0; 2288 } 2289 2290 static int __f2fs_quota_off(struct super_block *sb, int type) 2291 { 2292 struct inode *inode = sb_dqopt(sb)->files[type]; 2293 int err; 2294 2295 if (!inode || !igrab(inode)) 2296 return dquot_quota_off(sb, type); 2297 2298 err = f2fs_quota_sync(sb, type); 2299 if (err) 2300 goto out_put; 2301 2302 err = dquot_quota_off(sb, type); 2303 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2304 goto out_put; 2305 2306 inode_lock(inode); 2307 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2308 f2fs_set_inode_flags(inode); 2309 inode_unlock(inode); 2310 f2fs_mark_inode_dirty_sync(inode, false); 2311 out_put: 2312 iput(inode); 2313 return err; 2314 } 2315 2316 static int f2fs_quota_off(struct super_block *sb, int type) 2317 { 2318 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2319 int err; 2320 2321 err = __f2fs_quota_off(sb, type); 2322 2323 /* 2324 * quotactl can shutdown journalled quota, result in inconsistence 2325 * between quota record and fs data by following updates, tag the 2326 * flag to let fsck be aware of it. 2327 */ 2328 if (is_journalled_quota(sbi)) 2329 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2330 return err; 2331 } 2332 2333 void f2fs_quota_off_umount(struct super_block *sb) 2334 { 2335 int type; 2336 int err; 2337 2338 for (type = 0; type < MAXQUOTAS; type++) { 2339 err = __f2fs_quota_off(sb, type); 2340 if (err) { 2341 int ret = dquot_quota_off(sb, type); 2342 2343 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2344 type, err, ret); 2345 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2346 } 2347 } 2348 /* 2349 * In case of checkpoint=disable, we must flush quota blocks. 2350 * This can cause NULL exception for node_inode in end_io, since 2351 * put_super already dropped it. 2352 */ 2353 sync_filesystem(sb); 2354 } 2355 2356 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2357 { 2358 struct quota_info *dqopt = sb_dqopt(sb); 2359 int type; 2360 2361 for (type = 0; type < MAXQUOTAS; type++) { 2362 if (!dqopt->files[type]) 2363 continue; 2364 f2fs_inode_synced(dqopt->files[type]); 2365 } 2366 } 2367 2368 static int f2fs_dquot_commit(struct dquot *dquot) 2369 { 2370 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2371 int ret; 2372 2373 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2374 ret = dquot_commit(dquot); 2375 if (ret < 0) 2376 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2377 up_read(&sbi->quota_sem); 2378 return ret; 2379 } 2380 2381 static int f2fs_dquot_acquire(struct dquot *dquot) 2382 { 2383 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2384 int ret; 2385 2386 down_read(&sbi->quota_sem); 2387 ret = dquot_acquire(dquot); 2388 if (ret < 0) 2389 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2390 up_read(&sbi->quota_sem); 2391 return ret; 2392 } 2393 2394 static int f2fs_dquot_release(struct dquot *dquot) 2395 { 2396 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2397 int ret = dquot_release(dquot); 2398 2399 if (ret < 0) 2400 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2401 return ret; 2402 } 2403 2404 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2405 { 2406 struct super_block *sb = dquot->dq_sb; 2407 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2408 int ret = dquot_mark_dquot_dirty(dquot); 2409 2410 /* if we are using journalled quota */ 2411 if (is_journalled_quota(sbi)) 2412 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2413 2414 return ret; 2415 } 2416 2417 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2418 { 2419 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2420 int ret = dquot_commit_info(sb, type); 2421 2422 if (ret < 0) 2423 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2424 return ret; 2425 } 2426 2427 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2428 { 2429 *projid = F2FS_I(inode)->i_projid; 2430 return 0; 2431 } 2432 2433 static const struct dquot_operations f2fs_quota_operations = { 2434 .get_reserved_space = f2fs_get_reserved_space, 2435 .write_dquot = f2fs_dquot_commit, 2436 .acquire_dquot = f2fs_dquot_acquire, 2437 .release_dquot = f2fs_dquot_release, 2438 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2439 .write_info = f2fs_dquot_commit_info, 2440 .alloc_dquot = dquot_alloc, 2441 .destroy_dquot = dquot_destroy, 2442 .get_projid = f2fs_get_projid, 2443 .get_next_id = dquot_get_next_id, 2444 }; 2445 2446 static const struct quotactl_ops f2fs_quotactl_ops = { 2447 .quota_on = f2fs_quota_on, 2448 .quota_off = f2fs_quota_off, 2449 .quota_sync = f2fs_quota_sync, 2450 .get_state = dquot_get_state, 2451 .set_info = dquot_set_dqinfo, 2452 .get_dqblk = dquot_get_dqblk, 2453 .set_dqblk = dquot_set_dqblk, 2454 .get_nextdqblk = dquot_get_next_dqblk, 2455 }; 2456 #else 2457 int f2fs_quota_sync(struct super_block *sb, int type) 2458 { 2459 return 0; 2460 } 2461 2462 void f2fs_quota_off_umount(struct super_block *sb) 2463 { 2464 } 2465 #endif 2466 2467 static const struct super_operations f2fs_sops = { 2468 .alloc_inode = f2fs_alloc_inode, 2469 .free_inode = f2fs_free_inode, 2470 .drop_inode = f2fs_drop_inode, 2471 .write_inode = f2fs_write_inode, 2472 .dirty_inode = f2fs_dirty_inode, 2473 .show_options = f2fs_show_options, 2474 #ifdef CONFIG_QUOTA 2475 .quota_read = f2fs_quota_read, 2476 .quota_write = f2fs_quota_write, 2477 .get_dquots = f2fs_get_dquots, 2478 #endif 2479 .evict_inode = f2fs_evict_inode, 2480 .put_super = f2fs_put_super, 2481 .sync_fs = f2fs_sync_fs, 2482 .freeze_fs = f2fs_freeze, 2483 .unfreeze_fs = f2fs_unfreeze, 2484 .statfs = f2fs_statfs, 2485 .remount_fs = f2fs_remount, 2486 }; 2487 2488 #ifdef CONFIG_FS_ENCRYPTION 2489 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2490 { 2491 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2492 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2493 ctx, len, NULL); 2494 } 2495 2496 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2497 void *fs_data) 2498 { 2499 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2500 2501 /* 2502 * Encrypting the root directory is not allowed because fsck 2503 * expects lost+found directory to exist and remain unencrypted 2504 * if LOST_FOUND feature is enabled. 2505 * 2506 */ 2507 if (f2fs_sb_has_lost_found(sbi) && 2508 inode->i_ino == F2FS_ROOT_INO(sbi)) 2509 return -EPERM; 2510 2511 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2512 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2513 ctx, len, fs_data, XATTR_CREATE); 2514 } 2515 2516 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 2517 { 2518 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 2519 } 2520 2521 static bool f2fs_has_stable_inodes(struct super_block *sb) 2522 { 2523 return true; 2524 } 2525 2526 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 2527 int *ino_bits_ret, int *lblk_bits_ret) 2528 { 2529 *ino_bits_ret = 8 * sizeof(nid_t); 2530 *lblk_bits_ret = 8 * sizeof(block_t); 2531 } 2532 2533 static int f2fs_get_num_devices(struct super_block *sb) 2534 { 2535 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2536 2537 if (f2fs_is_multi_device(sbi)) 2538 return sbi->s_ndevs; 2539 return 1; 2540 } 2541 2542 static void f2fs_get_devices(struct super_block *sb, 2543 struct request_queue **devs) 2544 { 2545 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2546 int i; 2547 2548 for (i = 0; i < sbi->s_ndevs; i++) 2549 devs[i] = bdev_get_queue(FDEV(i).bdev); 2550 } 2551 2552 static const struct fscrypt_operations f2fs_cryptops = { 2553 .key_prefix = "f2fs:", 2554 .get_context = f2fs_get_context, 2555 .set_context = f2fs_set_context, 2556 .get_dummy_policy = f2fs_get_dummy_policy, 2557 .empty_dir = f2fs_empty_dir, 2558 .max_namelen = F2FS_NAME_LEN, 2559 .has_stable_inodes = f2fs_has_stable_inodes, 2560 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 2561 .get_num_devices = f2fs_get_num_devices, 2562 .get_devices = f2fs_get_devices, 2563 }; 2564 #endif 2565 2566 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 2567 u64 ino, u32 generation) 2568 { 2569 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2570 struct inode *inode; 2571 2572 if (f2fs_check_nid_range(sbi, ino)) 2573 return ERR_PTR(-ESTALE); 2574 2575 /* 2576 * f2fs_iget isn't quite right if the inode is currently unallocated! 2577 * However f2fs_iget currently does appropriate checks to handle stale 2578 * inodes so everything is OK. 2579 */ 2580 inode = f2fs_iget(sb, ino); 2581 if (IS_ERR(inode)) 2582 return ERR_CAST(inode); 2583 if (unlikely(generation && inode->i_generation != generation)) { 2584 /* we didn't find the right inode.. */ 2585 iput(inode); 2586 return ERR_PTR(-ESTALE); 2587 } 2588 return inode; 2589 } 2590 2591 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 2592 int fh_len, int fh_type) 2593 { 2594 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 2595 f2fs_nfs_get_inode); 2596 } 2597 2598 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 2599 int fh_len, int fh_type) 2600 { 2601 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 2602 f2fs_nfs_get_inode); 2603 } 2604 2605 static const struct export_operations f2fs_export_ops = { 2606 .fh_to_dentry = f2fs_fh_to_dentry, 2607 .fh_to_parent = f2fs_fh_to_parent, 2608 .get_parent = f2fs_get_parent, 2609 }; 2610 2611 static loff_t max_file_blocks(void) 2612 { 2613 loff_t result = 0; 2614 loff_t leaf_count = DEF_ADDRS_PER_BLOCK; 2615 2616 /* 2617 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 2618 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 2619 * space in inode.i_addr, it will be more safe to reassign 2620 * result as zero. 2621 */ 2622 2623 /* two direct node blocks */ 2624 result += (leaf_count * 2); 2625 2626 /* two indirect node blocks */ 2627 leaf_count *= NIDS_PER_BLOCK; 2628 result += (leaf_count * 2); 2629 2630 /* one double indirect node block */ 2631 leaf_count *= NIDS_PER_BLOCK; 2632 result += leaf_count; 2633 2634 return result; 2635 } 2636 2637 static int __f2fs_commit_super(struct buffer_head *bh, 2638 struct f2fs_super_block *super) 2639 { 2640 lock_buffer(bh); 2641 if (super) 2642 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 2643 set_buffer_dirty(bh); 2644 unlock_buffer(bh); 2645 2646 /* it's rare case, we can do fua all the time */ 2647 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 2648 } 2649 2650 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 2651 struct buffer_head *bh) 2652 { 2653 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2654 (bh->b_data + F2FS_SUPER_OFFSET); 2655 struct super_block *sb = sbi->sb; 2656 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 2657 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 2658 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 2659 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 2660 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 2661 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 2662 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 2663 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 2664 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 2665 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 2666 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2667 u32 segment_count = le32_to_cpu(raw_super->segment_count); 2668 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2669 u64 main_end_blkaddr = main_blkaddr + 2670 (segment_count_main << log_blocks_per_seg); 2671 u64 seg_end_blkaddr = segment0_blkaddr + 2672 (segment_count << log_blocks_per_seg); 2673 2674 if (segment0_blkaddr != cp_blkaddr) { 2675 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 2676 segment0_blkaddr, cp_blkaddr); 2677 return true; 2678 } 2679 2680 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 2681 sit_blkaddr) { 2682 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 2683 cp_blkaddr, sit_blkaddr, 2684 segment_count_ckpt << log_blocks_per_seg); 2685 return true; 2686 } 2687 2688 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 2689 nat_blkaddr) { 2690 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 2691 sit_blkaddr, nat_blkaddr, 2692 segment_count_sit << log_blocks_per_seg); 2693 return true; 2694 } 2695 2696 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 2697 ssa_blkaddr) { 2698 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 2699 nat_blkaddr, ssa_blkaddr, 2700 segment_count_nat << log_blocks_per_seg); 2701 return true; 2702 } 2703 2704 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 2705 main_blkaddr) { 2706 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 2707 ssa_blkaddr, main_blkaddr, 2708 segment_count_ssa << log_blocks_per_seg); 2709 return true; 2710 } 2711 2712 if (main_end_blkaddr > seg_end_blkaddr) { 2713 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 2714 main_blkaddr, seg_end_blkaddr, 2715 segment_count_main << log_blocks_per_seg); 2716 return true; 2717 } else if (main_end_blkaddr < seg_end_blkaddr) { 2718 int err = 0; 2719 char *res; 2720 2721 /* fix in-memory information all the time */ 2722 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 2723 segment0_blkaddr) >> log_blocks_per_seg); 2724 2725 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 2726 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2727 res = "internally"; 2728 } else { 2729 err = __f2fs_commit_super(bh, NULL); 2730 res = err ? "failed" : "done"; 2731 } 2732 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 2733 res, main_blkaddr, seg_end_blkaddr, 2734 segment_count_main << log_blocks_per_seg); 2735 if (err) 2736 return true; 2737 } 2738 return false; 2739 } 2740 2741 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 2742 struct buffer_head *bh) 2743 { 2744 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 2745 block_t total_sections, blocks_per_seg; 2746 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2747 (bh->b_data + F2FS_SUPER_OFFSET); 2748 unsigned int blocksize; 2749 size_t crc_offset = 0; 2750 __u32 crc = 0; 2751 2752 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 2753 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 2754 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 2755 return -EINVAL; 2756 } 2757 2758 /* Check checksum_offset and crc in superblock */ 2759 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 2760 crc_offset = le32_to_cpu(raw_super->checksum_offset); 2761 if (crc_offset != 2762 offsetof(struct f2fs_super_block, crc)) { 2763 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 2764 crc_offset); 2765 return -EFSCORRUPTED; 2766 } 2767 crc = le32_to_cpu(raw_super->crc); 2768 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 2769 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 2770 return -EFSCORRUPTED; 2771 } 2772 } 2773 2774 /* Currently, support only 4KB page cache size */ 2775 if (F2FS_BLKSIZE != PAGE_SIZE) { 2776 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB", 2777 PAGE_SIZE); 2778 return -EFSCORRUPTED; 2779 } 2780 2781 /* Currently, support only 4KB block size */ 2782 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); 2783 if (blocksize != F2FS_BLKSIZE) { 2784 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB", 2785 blocksize); 2786 return -EFSCORRUPTED; 2787 } 2788 2789 /* check log blocks per segment */ 2790 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 2791 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 2792 le32_to_cpu(raw_super->log_blocks_per_seg)); 2793 return -EFSCORRUPTED; 2794 } 2795 2796 /* Currently, support 512/1024/2048/4096 bytes sector size */ 2797 if (le32_to_cpu(raw_super->log_sectorsize) > 2798 F2FS_MAX_LOG_SECTOR_SIZE || 2799 le32_to_cpu(raw_super->log_sectorsize) < 2800 F2FS_MIN_LOG_SECTOR_SIZE) { 2801 f2fs_info(sbi, "Invalid log sectorsize (%u)", 2802 le32_to_cpu(raw_super->log_sectorsize)); 2803 return -EFSCORRUPTED; 2804 } 2805 if (le32_to_cpu(raw_super->log_sectors_per_block) + 2806 le32_to_cpu(raw_super->log_sectorsize) != 2807 F2FS_MAX_LOG_SECTOR_SIZE) { 2808 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 2809 le32_to_cpu(raw_super->log_sectors_per_block), 2810 le32_to_cpu(raw_super->log_sectorsize)); 2811 return -EFSCORRUPTED; 2812 } 2813 2814 segment_count = le32_to_cpu(raw_super->segment_count); 2815 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2816 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 2817 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 2818 total_sections = le32_to_cpu(raw_super->section_count); 2819 2820 /* blocks_per_seg should be 512, given the above check */ 2821 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 2822 2823 if (segment_count > F2FS_MAX_SEGMENT || 2824 segment_count < F2FS_MIN_SEGMENTS) { 2825 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 2826 return -EFSCORRUPTED; 2827 } 2828 2829 if (total_sections > segment_count_main || total_sections < 1 || 2830 segs_per_sec > segment_count || !segs_per_sec) { 2831 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 2832 segment_count, total_sections, segs_per_sec); 2833 return -EFSCORRUPTED; 2834 } 2835 2836 if (segment_count_main != total_sections * segs_per_sec) { 2837 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 2838 segment_count_main, total_sections, segs_per_sec); 2839 return -EFSCORRUPTED; 2840 } 2841 2842 if ((segment_count / segs_per_sec) < total_sections) { 2843 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 2844 segment_count, segs_per_sec, total_sections); 2845 return -EFSCORRUPTED; 2846 } 2847 2848 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 2849 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 2850 segment_count, le64_to_cpu(raw_super->block_count)); 2851 return -EFSCORRUPTED; 2852 } 2853 2854 if (RDEV(0).path[0]) { 2855 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 2856 int i = 1; 2857 2858 while (i < MAX_DEVICES && RDEV(i).path[0]) { 2859 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 2860 i++; 2861 } 2862 if (segment_count != dev_seg_count) { 2863 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 2864 segment_count, dev_seg_count); 2865 return -EFSCORRUPTED; 2866 } 2867 } else { 2868 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 2869 !bdev_is_zoned(sbi->sb->s_bdev)) { 2870 f2fs_info(sbi, "Zoned block device path is missing"); 2871 return -EFSCORRUPTED; 2872 } 2873 } 2874 2875 if (secs_per_zone > total_sections || !secs_per_zone) { 2876 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 2877 secs_per_zone, total_sections); 2878 return -EFSCORRUPTED; 2879 } 2880 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 2881 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 2882 (le32_to_cpu(raw_super->extension_count) + 2883 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 2884 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 2885 le32_to_cpu(raw_super->extension_count), 2886 raw_super->hot_ext_count, 2887 F2FS_MAX_EXTENSION); 2888 return -EFSCORRUPTED; 2889 } 2890 2891 if (le32_to_cpu(raw_super->cp_payload) > 2892 (blocks_per_seg - F2FS_CP_PACKS)) { 2893 f2fs_info(sbi, "Insane cp_payload (%u > %u)", 2894 le32_to_cpu(raw_super->cp_payload), 2895 blocks_per_seg - F2FS_CP_PACKS); 2896 return -EFSCORRUPTED; 2897 } 2898 2899 /* check reserved ino info */ 2900 if (le32_to_cpu(raw_super->node_ino) != 1 || 2901 le32_to_cpu(raw_super->meta_ino) != 2 || 2902 le32_to_cpu(raw_super->root_ino) != 3) { 2903 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 2904 le32_to_cpu(raw_super->node_ino), 2905 le32_to_cpu(raw_super->meta_ino), 2906 le32_to_cpu(raw_super->root_ino)); 2907 return -EFSCORRUPTED; 2908 } 2909 2910 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 2911 if (sanity_check_area_boundary(sbi, bh)) 2912 return -EFSCORRUPTED; 2913 2914 return 0; 2915 } 2916 2917 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 2918 { 2919 unsigned int total, fsmeta; 2920 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 2921 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 2922 unsigned int ovp_segments, reserved_segments; 2923 unsigned int main_segs, blocks_per_seg; 2924 unsigned int sit_segs, nat_segs; 2925 unsigned int sit_bitmap_size, nat_bitmap_size; 2926 unsigned int log_blocks_per_seg; 2927 unsigned int segment_count_main; 2928 unsigned int cp_pack_start_sum, cp_payload; 2929 block_t user_block_count, valid_user_blocks; 2930 block_t avail_node_count, valid_node_count; 2931 int i, j; 2932 2933 total = le32_to_cpu(raw_super->segment_count); 2934 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 2935 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 2936 fsmeta += sit_segs; 2937 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 2938 fsmeta += nat_segs; 2939 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 2940 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 2941 2942 if (unlikely(fsmeta >= total)) 2943 return 1; 2944 2945 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 2946 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 2947 2948 if (unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 2949 ovp_segments == 0 || reserved_segments == 0)) { 2950 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 2951 return 1; 2952 } 2953 2954 user_block_count = le64_to_cpu(ckpt->user_block_count); 2955 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2956 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2957 if (!user_block_count || user_block_count >= 2958 segment_count_main << log_blocks_per_seg) { 2959 f2fs_err(sbi, "Wrong user_block_count: %u", 2960 user_block_count); 2961 return 1; 2962 } 2963 2964 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 2965 if (valid_user_blocks > user_block_count) { 2966 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 2967 valid_user_blocks, user_block_count); 2968 return 1; 2969 } 2970 2971 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 2972 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 2973 if (valid_node_count > avail_node_count) { 2974 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 2975 valid_node_count, avail_node_count); 2976 return 1; 2977 } 2978 2979 main_segs = le32_to_cpu(raw_super->segment_count_main); 2980 blocks_per_seg = sbi->blocks_per_seg; 2981 2982 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 2983 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 2984 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 2985 return 1; 2986 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 2987 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 2988 le32_to_cpu(ckpt->cur_node_segno[j])) { 2989 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 2990 i, j, 2991 le32_to_cpu(ckpt->cur_node_segno[i])); 2992 return 1; 2993 } 2994 } 2995 } 2996 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 2997 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 2998 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 2999 return 1; 3000 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3001 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3002 le32_to_cpu(ckpt->cur_data_segno[j])) { 3003 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3004 i, j, 3005 le32_to_cpu(ckpt->cur_data_segno[i])); 3006 return 1; 3007 } 3008 } 3009 } 3010 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3011 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3012 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3013 le32_to_cpu(ckpt->cur_data_segno[j])) { 3014 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3015 i, j, 3016 le32_to_cpu(ckpt->cur_node_segno[i])); 3017 return 1; 3018 } 3019 } 3020 } 3021 3022 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3023 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3024 3025 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3026 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3027 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3028 sit_bitmap_size, nat_bitmap_size); 3029 return 1; 3030 } 3031 3032 cp_pack_start_sum = __start_sum_addr(sbi); 3033 cp_payload = __cp_payload(sbi); 3034 if (cp_pack_start_sum < cp_payload + 1 || 3035 cp_pack_start_sum > blocks_per_seg - 1 - 3036 NR_CURSEG_PERSIST_TYPE) { 3037 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3038 cp_pack_start_sum); 3039 return 1; 3040 } 3041 3042 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3043 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3044 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3045 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3046 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3047 le32_to_cpu(ckpt->checksum_offset)); 3048 return 1; 3049 } 3050 3051 if (unlikely(f2fs_cp_error(sbi))) { 3052 f2fs_err(sbi, "A bug case: need to run fsck"); 3053 return 1; 3054 } 3055 return 0; 3056 } 3057 3058 static void init_sb_info(struct f2fs_sb_info *sbi) 3059 { 3060 struct f2fs_super_block *raw_super = sbi->raw_super; 3061 int i; 3062 3063 sbi->log_sectors_per_block = 3064 le32_to_cpu(raw_super->log_sectors_per_block); 3065 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3066 sbi->blocksize = 1 << sbi->log_blocksize; 3067 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3068 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3069 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3070 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3071 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3072 sbi->total_node_count = 3073 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3074 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3075 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); 3076 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); 3077 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); 3078 sbi->cur_victim_sec = NULL_SECNO; 3079 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3080 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3081 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3082 sbi->migration_granularity = sbi->segs_per_sec; 3083 3084 sbi->dir_level = DEF_DIR_LEVEL; 3085 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3086 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3087 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3088 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3089 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3090 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3091 DEF_UMOUNT_DISCARD_TIMEOUT; 3092 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3093 3094 for (i = 0; i < NR_COUNT_TYPE; i++) 3095 atomic_set(&sbi->nr_pages[i], 0); 3096 3097 for (i = 0; i < META; i++) 3098 atomic_set(&sbi->wb_sync_req[i], 0); 3099 3100 INIT_LIST_HEAD(&sbi->s_list); 3101 mutex_init(&sbi->umount_mutex); 3102 init_rwsem(&sbi->io_order_lock); 3103 spin_lock_init(&sbi->cp_lock); 3104 3105 sbi->dirty_device = 0; 3106 spin_lock_init(&sbi->dev_lock); 3107 3108 init_rwsem(&sbi->sb_lock); 3109 init_rwsem(&sbi->pin_sem); 3110 } 3111 3112 static int init_percpu_info(struct f2fs_sb_info *sbi) 3113 { 3114 int err; 3115 3116 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3117 if (err) 3118 return err; 3119 3120 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3121 GFP_KERNEL); 3122 if (err) 3123 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3124 3125 return err; 3126 } 3127 3128 #ifdef CONFIG_BLK_DEV_ZONED 3129 3130 struct f2fs_report_zones_args { 3131 struct f2fs_dev_info *dev; 3132 bool zone_cap_mismatch; 3133 }; 3134 3135 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3136 void *data) 3137 { 3138 struct f2fs_report_zones_args *rz_args = data; 3139 3140 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3141 return 0; 3142 3143 set_bit(idx, rz_args->dev->blkz_seq); 3144 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >> 3145 F2FS_LOG_SECTORS_PER_BLOCK; 3146 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch) 3147 rz_args->zone_cap_mismatch = true; 3148 3149 return 0; 3150 } 3151 3152 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3153 { 3154 struct block_device *bdev = FDEV(devi).bdev; 3155 sector_t nr_sectors = bdev->bd_part->nr_sects; 3156 struct f2fs_report_zones_args rep_zone_arg; 3157 int ret; 3158 3159 if (!f2fs_sb_has_blkzoned(sbi)) 3160 return 0; 3161 3162 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3163 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 3164 return -EINVAL; 3165 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 3166 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3167 __ilog2_u32(sbi->blocks_per_blkz)) 3168 return -EINVAL; 3169 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3170 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3171 sbi->log_blocks_per_blkz; 3172 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 3173 FDEV(devi).nr_blkz++; 3174 3175 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3176 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3177 * sizeof(unsigned long), 3178 GFP_KERNEL); 3179 if (!FDEV(devi).blkz_seq) 3180 return -ENOMEM; 3181 3182 /* Get block zones type and zone-capacity */ 3183 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi, 3184 FDEV(devi).nr_blkz * sizeof(block_t), 3185 GFP_KERNEL); 3186 if (!FDEV(devi).zone_capacity_blocks) 3187 return -ENOMEM; 3188 3189 rep_zone_arg.dev = &FDEV(devi); 3190 rep_zone_arg.zone_cap_mismatch = false; 3191 3192 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3193 &rep_zone_arg); 3194 if (ret < 0) 3195 return ret; 3196 3197 if (!rep_zone_arg.zone_cap_mismatch) { 3198 kfree(FDEV(devi).zone_capacity_blocks); 3199 FDEV(devi).zone_capacity_blocks = NULL; 3200 } 3201 3202 return 0; 3203 } 3204 #endif 3205 3206 /* 3207 * Read f2fs raw super block. 3208 * Because we have two copies of super block, so read both of them 3209 * to get the first valid one. If any one of them is broken, we pass 3210 * them recovery flag back to the caller. 3211 */ 3212 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3213 struct f2fs_super_block **raw_super, 3214 int *valid_super_block, int *recovery) 3215 { 3216 struct super_block *sb = sbi->sb; 3217 int block; 3218 struct buffer_head *bh; 3219 struct f2fs_super_block *super; 3220 int err = 0; 3221 3222 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3223 if (!super) 3224 return -ENOMEM; 3225 3226 for (block = 0; block < 2; block++) { 3227 bh = sb_bread(sb, block); 3228 if (!bh) { 3229 f2fs_err(sbi, "Unable to read %dth superblock", 3230 block + 1); 3231 err = -EIO; 3232 *recovery = 1; 3233 continue; 3234 } 3235 3236 /* sanity checking of raw super */ 3237 err = sanity_check_raw_super(sbi, bh); 3238 if (err) { 3239 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3240 block + 1); 3241 brelse(bh); 3242 *recovery = 1; 3243 continue; 3244 } 3245 3246 if (!*raw_super) { 3247 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3248 sizeof(*super)); 3249 *valid_super_block = block; 3250 *raw_super = super; 3251 } 3252 brelse(bh); 3253 } 3254 3255 /* No valid superblock */ 3256 if (!*raw_super) 3257 kfree(super); 3258 else 3259 err = 0; 3260 3261 return err; 3262 } 3263 3264 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3265 { 3266 struct buffer_head *bh; 3267 __u32 crc = 0; 3268 int err; 3269 3270 if ((recover && f2fs_readonly(sbi->sb)) || 3271 bdev_read_only(sbi->sb->s_bdev)) { 3272 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3273 return -EROFS; 3274 } 3275 3276 /* we should update superblock crc here */ 3277 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3278 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3279 offsetof(struct f2fs_super_block, crc)); 3280 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3281 } 3282 3283 /* write back-up superblock first */ 3284 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3285 if (!bh) 3286 return -EIO; 3287 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3288 brelse(bh); 3289 3290 /* if we are in recovery path, skip writing valid superblock */ 3291 if (recover || err) 3292 return err; 3293 3294 /* write current valid superblock */ 3295 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3296 if (!bh) 3297 return -EIO; 3298 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3299 brelse(bh); 3300 return err; 3301 } 3302 3303 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3304 { 3305 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3306 unsigned int max_devices = MAX_DEVICES; 3307 int i; 3308 3309 /* Initialize single device information */ 3310 if (!RDEV(0).path[0]) { 3311 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3312 return 0; 3313 max_devices = 1; 3314 } 3315 3316 /* 3317 * Initialize multiple devices information, or single 3318 * zoned block device information. 3319 */ 3320 sbi->devs = f2fs_kzalloc(sbi, 3321 array_size(max_devices, 3322 sizeof(struct f2fs_dev_info)), 3323 GFP_KERNEL); 3324 if (!sbi->devs) 3325 return -ENOMEM; 3326 3327 for (i = 0; i < max_devices; i++) { 3328 3329 if (i > 0 && !RDEV(i).path[0]) 3330 break; 3331 3332 if (max_devices == 1) { 3333 /* Single zoned block device mount */ 3334 FDEV(0).bdev = 3335 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3336 sbi->sb->s_mode, sbi->sb->s_type); 3337 } else { 3338 /* Multi-device mount */ 3339 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3340 FDEV(i).total_segments = 3341 le32_to_cpu(RDEV(i).total_segments); 3342 if (i == 0) { 3343 FDEV(i).start_blk = 0; 3344 FDEV(i).end_blk = FDEV(i).start_blk + 3345 (FDEV(i).total_segments << 3346 sbi->log_blocks_per_seg) - 1 + 3347 le32_to_cpu(raw_super->segment0_blkaddr); 3348 } else { 3349 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3350 FDEV(i).end_blk = FDEV(i).start_blk + 3351 (FDEV(i).total_segments << 3352 sbi->log_blocks_per_seg) - 1; 3353 } 3354 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3355 sbi->sb->s_mode, sbi->sb->s_type); 3356 } 3357 if (IS_ERR(FDEV(i).bdev)) 3358 return PTR_ERR(FDEV(i).bdev); 3359 3360 /* to release errored devices */ 3361 sbi->s_ndevs = i + 1; 3362 3363 #ifdef CONFIG_BLK_DEV_ZONED 3364 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3365 !f2fs_sb_has_blkzoned(sbi)) { 3366 f2fs_err(sbi, "Zoned block device feature not enabled\n"); 3367 return -EINVAL; 3368 } 3369 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3370 if (init_blkz_info(sbi, i)) { 3371 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3372 return -EINVAL; 3373 } 3374 if (max_devices == 1) 3375 break; 3376 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3377 i, FDEV(i).path, 3378 FDEV(i).total_segments, 3379 FDEV(i).start_blk, FDEV(i).end_blk, 3380 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3381 "Host-aware" : "Host-managed"); 3382 continue; 3383 } 3384 #endif 3385 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3386 i, FDEV(i).path, 3387 FDEV(i).total_segments, 3388 FDEV(i).start_blk, FDEV(i).end_blk); 3389 } 3390 f2fs_info(sbi, 3391 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3392 return 0; 3393 } 3394 3395 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3396 { 3397 #ifdef CONFIG_UNICODE 3398 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 3399 const struct f2fs_sb_encodings *encoding_info; 3400 struct unicode_map *encoding; 3401 __u16 encoding_flags; 3402 3403 if (f2fs_sb_has_encrypt(sbi)) { 3404 f2fs_err(sbi, 3405 "Can't mount with encoding and encryption"); 3406 return -EINVAL; 3407 } 3408 3409 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info, 3410 &encoding_flags)) { 3411 f2fs_err(sbi, 3412 "Encoding requested by superblock is unknown"); 3413 return -EINVAL; 3414 } 3415 3416 encoding = utf8_load(encoding_info->version); 3417 if (IS_ERR(encoding)) { 3418 f2fs_err(sbi, 3419 "can't mount with superblock charset: %s-%s " 3420 "not supported by the kernel. flags: 0x%x.", 3421 encoding_info->name, encoding_info->version, 3422 encoding_flags); 3423 return PTR_ERR(encoding); 3424 } 3425 f2fs_info(sbi, "Using encoding defined by superblock: " 3426 "%s-%s with flags 0x%hx", encoding_info->name, 3427 encoding_info->version?:"\b", encoding_flags); 3428 3429 sbi->sb->s_encoding = encoding; 3430 sbi->sb->s_encoding_flags = encoding_flags; 3431 sbi->sb->s_d_op = &f2fs_dentry_ops; 3432 } 3433 #else 3434 if (f2fs_sb_has_casefold(sbi)) { 3435 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3436 return -EINVAL; 3437 } 3438 #endif 3439 return 0; 3440 } 3441 3442 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3443 { 3444 struct f2fs_sm_info *sm_i = SM_I(sbi); 3445 3446 /* adjust parameters according to the volume size */ 3447 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3448 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3449 sm_i->dcc_info->discard_granularity = 1; 3450 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 3451 } 3452 3453 sbi->readdir_ra = 1; 3454 } 3455 3456 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3457 { 3458 struct f2fs_sb_info *sbi; 3459 struct f2fs_super_block *raw_super; 3460 struct inode *root; 3461 int err; 3462 bool skip_recovery = false, need_fsck = false; 3463 char *options = NULL; 3464 int recovery, i, valid_super_block; 3465 struct curseg_info *seg_i; 3466 int retry_cnt = 1; 3467 3468 try_onemore: 3469 err = -EINVAL; 3470 raw_super = NULL; 3471 valid_super_block = -1; 3472 recovery = 0; 3473 3474 /* allocate memory for f2fs-specific super block info */ 3475 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 3476 if (!sbi) 3477 return -ENOMEM; 3478 3479 sbi->sb = sb; 3480 3481 /* Load the checksum driver */ 3482 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 3483 if (IS_ERR(sbi->s_chksum_driver)) { 3484 f2fs_err(sbi, "Cannot load crc32 driver."); 3485 err = PTR_ERR(sbi->s_chksum_driver); 3486 sbi->s_chksum_driver = NULL; 3487 goto free_sbi; 3488 } 3489 3490 /* set a block size */ 3491 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 3492 f2fs_err(sbi, "unable to set blocksize"); 3493 goto free_sbi; 3494 } 3495 3496 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 3497 &recovery); 3498 if (err) 3499 goto free_sbi; 3500 3501 sb->s_fs_info = sbi; 3502 sbi->raw_super = raw_super; 3503 3504 /* precompute checksum seed for metadata */ 3505 if (f2fs_sb_has_inode_chksum(sbi)) 3506 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 3507 sizeof(raw_super->uuid)); 3508 3509 default_options(sbi); 3510 /* parse mount options */ 3511 options = kstrdup((const char *)data, GFP_KERNEL); 3512 if (data && !options) { 3513 err = -ENOMEM; 3514 goto free_sb_buf; 3515 } 3516 3517 err = parse_options(sb, options, false); 3518 if (err) 3519 goto free_options; 3520 3521 sbi->max_file_blocks = max_file_blocks(); 3522 sb->s_maxbytes = sbi->max_file_blocks << 3523 le32_to_cpu(raw_super->log_blocksize); 3524 sb->s_max_links = F2FS_LINK_MAX; 3525 3526 err = f2fs_setup_casefold(sbi); 3527 if (err) 3528 goto free_options; 3529 3530 #ifdef CONFIG_QUOTA 3531 sb->dq_op = &f2fs_quota_operations; 3532 sb->s_qcop = &f2fs_quotactl_ops; 3533 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 3534 3535 if (f2fs_sb_has_quota_ino(sbi)) { 3536 for (i = 0; i < MAXQUOTAS; i++) { 3537 if (f2fs_qf_ino(sbi->sb, i)) 3538 sbi->nquota_files++; 3539 } 3540 } 3541 #endif 3542 3543 sb->s_op = &f2fs_sops; 3544 #ifdef CONFIG_FS_ENCRYPTION 3545 sb->s_cop = &f2fs_cryptops; 3546 #endif 3547 #ifdef CONFIG_FS_VERITY 3548 sb->s_vop = &f2fs_verityops; 3549 #endif 3550 sb->s_xattr = f2fs_xattr_handlers; 3551 sb->s_export_op = &f2fs_export_ops; 3552 sb->s_magic = F2FS_SUPER_MAGIC; 3553 sb->s_time_gran = 1; 3554 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3555 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 3556 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 3557 sb->s_iflags |= SB_I_CGROUPWB; 3558 3559 /* init f2fs-specific super block info */ 3560 sbi->valid_super_block = valid_super_block; 3561 init_rwsem(&sbi->gc_lock); 3562 mutex_init(&sbi->writepages); 3563 mutex_init(&sbi->cp_mutex); 3564 init_rwsem(&sbi->node_write); 3565 init_rwsem(&sbi->node_change); 3566 3567 /* disallow all the data/node/meta page writes */ 3568 set_sbi_flag(sbi, SBI_POR_DOING); 3569 spin_lock_init(&sbi->stat_lock); 3570 3571 /* init iostat info */ 3572 spin_lock_init(&sbi->iostat_lock); 3573 sbi->iostat_enable = false; 3574 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS; 3575 3576 for (i = 0; i < NR_PAGE_TYPE; i++) { 3577 int n = (i == META) ? 1: NR_TEMP_TYPE; 3578 int j; 3579 3580 sbi->write_io[i] = 3581 f2fs_kmalloc(sbi, 3582 array_size(n, 3583 sizeof(struct f2fs_bio_info)), 3584 GFP_KERNEL); 3585 if (!sbi->write_io[i]) { 3586 err = -ENOMEM; 3587 goto free_bio_info; 3588 } 3589 3590 for (j = HOT; j < n; j++) { 3591 init_rwsem(&sbi->write_io[i][j].io_rwsem); 3592 sbi->write_io[i][j].sbi = sbi; 3593 sbi->write_io[i][j].bio = NULL; 3594 spin_lock_init(&sbi->write_io[i][j].io_lock); 3595 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 3596 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); 3597 init_rwsem(&sbi->write_io[i][j].bio_list_lock); 3598 } 3599 } 3600 3601 init_rwsem(&sbi->cp_rwsem); 3602 init_rwsem(&sbi->quota_sem); 3603 init_waitqueue_head(&sbi->cp_wait); 3604 init_sb_info(sbi); 3605 3606 err = init_percpu_info(sbi); 3607 if (err) 3608 goto free_bio_info; 3609 3610 if (F2FS_IO_ALIGNED(sbi)) { 3611 sbi->write_io_dummy = 3612 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 3613 if (!sbi->write_io_dummy) { 3614 err = -ENOMEM; 3615 goto free_percpu; 3616 } 3617 } 3618 3619 /* init per sbi slab cache */ 3620 err = f2fs_init_xattr_caches(sbi); 3621 if (err) 3622 goto free_io_dummy; 3623 err = f2fs_init_page_array_cache(sbi); 3624 if (err) 3625 goto free_xattr_cache; 3626 3627 /* get an inode for meta space */ 3628 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 3629 if (IS_ERR(sbi->meta_inode)) { 3630 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 3631 err = PTR_ERR(sbi->meta_inode); 3632 goto free_page_array_cache; 3633 } 3634 3635 err = f2fs_get_valid_checkpoint(sbi); 3636 if (err) { 3637 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 3638 goto free_meta_inode; 3639 } 3640 3641 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 3642 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3643 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 3644 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3645 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 3646 } 3647 3648 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 3649 set_sbi_flag(sbi, SBI_NEED_FSCK); 3650 3651 /* Initialize device list */ 3652 err = f2fs_scan_devices(sbi); 3653 if (err) { 3654 f2fs_err(sbi, "Failed to find devices"); 3655 goto free_devices; 3656 } 3657 3658 err = f2fs_init_post_read_wq(sbi); 3659 if (err) { 3660 f2fs_err(sbi, "Failed to initialize post read workqueue"); 3661 goto free_devices; 3662 } 3663 3664 sbi->total_valid_node_count = 3665 le32_to_cpu(sbi->ckpt->valid_node_count); 3666 percpu_counter_set(&sbi->total_valid_inode_count, 3667 le32_to_cpu(sbi->ckpt->valid_inode_count)); 3668 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 3669 sbi->total_valid_block_count = 3670 le64_to_cpu(sbi->ckpt->valid_block_count); 3671 sbi->last_valid_block_count = sbi->total_valid_block_count; 3672 sbi->reserved_blocks = 0; 3673 sbi->current_reserved_blocks = 0; 3674 limit_reserve_root(sbi); 3675 adjust_unusable_cap_perc(sbi); 3676 3677 for (i = 0; i < NR_INODE_TYPE; i++) { 3678 INIT_LIST_HEAD(&sbi->inode_list[i]); 3679 spin_lock_init(&sbi->inode_lock[i]); 3680 } 3681 mutex_init(&sbi->flush_lock); 3682 3683 f2fs_init_extent_cache_info(sbi); 3684 3685 f2fs_init_ino_entry_info(sbi); 3686 3687 f2fs_init_fsync_node_info(sbi); 3688 3689 /* setup f2fs internal modules */ 3690 err = f2fs_build_segment_manager(sbi); 3691 if (err) { 3692 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 3693 err); 3694 goto free_sm; 3695 } 3696 err = f2fs_build_node_manager(sbi); 3697 if (err) { 3698 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 3699 err); 3700 goto free_nm; 3701 } 3702 3703 /* For write statistics */ 3704 if (sb->s_bdev->bd_part) 3705 sbi->sectors_written_start = 3706 (u64)part_stat_read(sb->s_bdev->bd_part, 3707 sectors[STAT_WRITE]); 3708 3709 /* Read accumulated write IO statistics if exists */ 3710 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 3711 if (__exist_node_summaries(sbi)) 3712 sbi->kbytes_written = 3713 le64_to_cpu(seg_i->journal->info.kbytes_written); 3714 3715 f2fs_build_gc_manager(sbi); 3716 3717 err = f2fs_build_stats(sbi); 3718 if (err) 3719 goto free_nm; 3720 3721 /* get an inode for node space */ 3722 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 3723 if (IS_ERR(sbi->node_inode)) { 3724 f2fs_err(sbi, "Failed to read node inode"); 3725 err = PTR_ERR(sbi->node_inode); 3726 goto free_stats; 3727 } 3728 3729 /* read root inode and dentry */ 3730 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 3731 if (IS_ERR(root)) { 3732 f2fs_err(sbi, "Failed to read root inode"); 3733 err = PTR_ERR(root); 3734 goto free_node_inode; 3735 } 3736 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 3737 !root->i_size || !root->i_nlink) { 3738 iput(root); 3739 err = -EINVAL; 3740 goto free_node_inode; 3741 } 3742 3743 sb->s_root = d_make_root(root); /* allocate root dentry */ 3744 if (!sb->s_root) { 3745 err = -ENOMEM; 3746 goto free_node_inode; 3747 } 3748 3749 err = f2fs_register_sysfs(sbi); 3750 if (err) 3751 goto free_root_inode; 3752 3753 #ifdef CONFIG_QUOTA 3754 /* Enable quota usage during mount */ 3755 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 3756 err = f2fs_enable_quotas(sb); 3757 if (err) 3758 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 3759 } 3760 #endif 3761 /* if there are any orphan inodes, free them */ 3762 err = f2fs_recover_orphan_inodes(sbi); 3763 if (err) 3764 goto free_meta; 3765 3766 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 3767 goto reset_checkpoint; 3768 3769 /* recover fsynced data */ 3770 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 3771 !test_opt(sbi, NORECOVERY)) { 3772 /* 3773 * mount should be failed, when device has readonly mode, and 3774 * previous checkpoint was not done by clean system shutdown. 3775 */ 3776 if (f2fs_hw_is_readonly(sbi)) { 3777 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3778 err = -EROFS; 3779 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable"); 3780 goto free_meta; 3781 } 3782 f2fs_info(sbi, "write access unavailable, skipping recovery"); 3783 goto reset_checkpoint; 3784 } 3785 3786 if (need_fsck) 3787 set_sbi_flag(sbi, SBI_NEED_FSCK); 3788 3789 if (skip_recovery) 3790 goto reset_checkpoint; 3791 3792 err = f2fs_recover_fsync_data(sbi, false); 3793 if (err < 0) { 3794 if (err != -ENOMEM) 3795 skip_recovery = true; 3796 need_fsck = true; 3797 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 3798 err); 3799 goto free_meta; 3800 } 3801 } else { 3802 err = f2fs_recover_fsync_data(sbi, true); 3803 3804 if (!f2fs_readonly(sb) && err > 0) { 3805 err = -EINVAL; 3806 f2fs_err(sbi, "Need to recover fsync data"); 3807 goto free_meta; 3808 } 3809 } 3810 3811 /* 3812 * If the f2fs is not readonly and fsync data recovery succeeds, 3813 * check zoned block devices' write pointer consistency. 3814 */ 3815 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 3816 err = f2fs_check_write_pointer(sbi); 3817 if (err) 3818 goto free_meta; 3819 } 3820 3821 reset_checkpoint: 3822 f2fs_init_inmem_curseg(sbi); 3823 3824 /* f2fs_recover_fsync_data() cleared this already */ 3825 clear_sbi_flag(sbi, SBI_POR_DOING); 3826 3827 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 3828 err = f2fs_disable_checkpoint(sbi); 3829 if (err) 3830 goto sync_free_meta; 3831 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 3832 f2fs_enable_checkpoint(sbi); 3833 } 3834 3835 /* 3836 * If filesystem is not mounted as read-only then 3837 * do start the gc_thread. 3838 */ 3839 if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) { 3840 /* After POR, we can run background GC thread.*/ 3841 err = f2fs_start_gc_thread(sbi); 3842 if (err) 3843 goto sync_free_meta; 3844 } 3845 kvfree(options); 3846 3847 /* recover broken superblock */ 3848 if (recovery) { 3849 err = f2fs_commit_super(sbi, true); 3850 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 3851 sbi->valid_super_block ? 1 : 2, err); 3852 } 3853 3854 f2fs_join_shrinker(sbi); 3855 3856 f2fs_tuning_parameters(sbi); 3857 3858 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 3859 cur_cp_version(F2FS_CKPT(sbi))); 3860 f2fs_update_time(sbi, CP_TIME); 3861 f2fs_update_time(sbi, REQ_TIME); 3862 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3863 return 0; 3864 3865 sync_free_meta: 3866 /* safe to flush all the data */ 3867 sync_filesystem(sbi->sb); 3868 retry_cnt = 0; 3869 3870 free_meta: 3871 #ifdef CONFIG_QUOTA 3872 f2fs_truncate_quota_inode_pages(sb); 3873 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 3874 f2fs_quota_off_umount(sbi->sb); 3875 #endif 3876 /* 3877 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 3878 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 3879 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 3880 * falls into an infinite loop in f2fs_sync_meta_pages(). 3881 */ 3882 truncate_inode_pages_final(META_MAPPING(sbi)); 3883 /* evict some inodes being cached by GC */ 3884 evict_inodes(sb); 3885 f2fs_unregister_sysfs(sbi); 3886 free_root_inode: 3887 dput(sb->s_root); 3888 sb->s_root = NULL; 3889 free_node_inode: 3890 f2fs_release_ino_entry(sbi, true); 3891 truncate_inode_pages_final(NODE_MAPPING(sbi)); 3892 iput(sbi->node_inode); 3893 sbi->node_inode = NULL; 3894 free_stats: 3895 f2fs_destroy_stats(sbi); 3896 free_nm: 3897 f2fs_destroy_node_manager(sbi); 3898 free_sm: 3899 f2fs_destroy_segment_manager(sbi); 3900 f2fs_destroy_post_read_wq(sbi); 3901 free_devices: 3902 destroy_device_list(sbi); 3903 kvfree(sbi->ckpt); 3904 free_meta_inode: 3905 make_bad_inode(sbi->meta_inode); 3906 iput(sbi->meta_inode); 3907 sbi->meta_inode = NULL; 3908 free_page_array_cache: 3909 f2fs_destroy_page_array_cache(sbi); 3910 free_xattr_cache: 3911 f2fs_destroy_xattr_caches(sbi); 3912 free_io_dummy: 3913 mempool_destroy(sbi->write_io_dummy); 3914 free_percpu: 3915 destroy_percpu_info(sbi); 3916 free_bio_info: 3917 for (i = 0; i < NR_PAGE_TYPE; i++) 3918 kvfree(sbi->write_io[i]); 3919 3920 #ifdef CONFIG_UNICODE 3921 utf8_unload(sb->s_encoding); 3922 #endif 3923 free_options: 3924 #ifdef CONFIG_QUOTA 3925 for (i = 0; i < MAXQUOTAS; i++) 3926 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 3927 #endif 3928 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 3929 kvfree(options); 3930 free_sb_buf: 3931 kfree(raw_super); 3932 free_sbi: 3933 if (sbi->s_chksum_driver) 3934 crypto_free_shash(sbi->s_chksum_driver); 3935 kfree(sbi); 3936 3937 /* give only one another chance */ 3938 if (retry_cnt > 0 && skip_recovery) { 3939 retry_cnt--; 3940 shrink_dcache_sb(sb); 3941 goto try_onemore; 3942 } 3943 return err; 3944 } 3945 3946 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 3947 const char *dev_name, void *data) 3948 { 3949 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 3950 } 3951 3952 static void kill_f2fs_super(struct super_block *sb) 3953 { 3954 if (sb->s_root) { 3955 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3956 3957 set_sbi_flag(sbi, SBI_IS_CLOSE); 3958 f2fs_stop_gc_thread(sbi); 3959 f2fs_stop_discard_thread(sbi); 3960 3961 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 3962 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3963 struct cp_control cpc = { 3964 .reason = CP_UMOUNT, 3965 }; 3966 f2fs_write_checkpoint(sbi, &cpc); 3967 } 3968 3969 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 3970 sb->s_flags &= ~SB_RDONLY; 3971 } 3972 kill_block_super(sb); 3973 } 3974 3975 static struct file_system_type f2fs_fs_type = { 3976 .owner = THIS_MODULE, 3977 .name = "f2fs", 3978 .mount = f2fs_mount, 3979 .kill_sb = kill_f2fs_super, 3980 .fs_flags = FS_REQUIRES_DEV, 3981 }; 3982 MODULE_ALIAS_FS("f2fs"); 3983 3984 static int __init init_inodecache(void) 3985 { 3986 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 3987 sizeof(struct f2fs_inode_info), 0, 3988 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 3989 if (!f2fs_inode_cachep) 3990 return -ENOMEM; 3991 return 0; 3992 } 3993 3994 static void destroy_inodecache(void) 3995 { 3996 /* 3997 * Make sure all delayed rcu free inodes are flushed before we 3998 * destroy cache. 3999 */ 4000 rcu_barrier(); 4001 kmem_cache_destroy(f2fs_inode_cachep); 4002 } 4003 4004 static int __init init_f2fs_fs(void) 4005 { 4006 int err; 4007 4008 if (PAGE_SIZE != F2FS_BLKSIZE) { 4009 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4010 PAGE_SIZE, F2FS_BLKSIZE); 4011 return -EINVAL; 4012 } 4013 4014 f2fs_build_trace_ios(); 4015 4016 err = init_inodecache(); 4017 if (err) 4018 goto fail; 4019 err = f2fs_create_node_manager_caches(); 4020 if (err) 4021 goto free_inodecache; 4022 err = f2fs_create_segment_manager_caches(); 4023 if (err) 4024 goto free_node_manager_caches; 4025 err = f2fs_create_checkpoint_caches(); 4026 if (err) 4027 goto free_segment_manager_caches; 4028 err = f2fs_create_extent_cache(); 4029 if (err) 4030 goto free_checkpoint_caches; 4031 err = f2fs_create_garbage_collection_cache(); 4032 if (err) 4033 goto free_extent_cache; 4034 err = f2fs_init_sysfs(); 4035 if (err) 4036 goto free_garbage_collection_cache; 4037 err = register_shrinker(&f2fs_shrinker_info); 4038 if (err) 4039 goto free_sysfs; 4040 err = register_filesystem(&f2fs_fs_type); 4041 if (err) 4042 goto free_shrinker; 4043 f2fs_create_root_stats(); 4044 err = f2fs_init_post_read_processing(); 4045 if (err) 4046 goto free_root_stats; 4047 err = f2fs_init_bio_entry_cache(); 4048 if (err) 4049 goto free_post_read; 4050 err = f2fs_init_bioset(); 4051 if (err) 4052 goto free_bio_enrty_cache; 4053 err = f2fs_init_compress_mempool(); 4054 if (err) 4055 goto free_bioset; 4056 err = f2fs_init_compress_cache(); 4057 if (err) 4058 goto free_compress_mempool; 4059 return 0; 4060 free_compress_mempool: 4061 f2fs_destroy_compress_mempool(); 4062 free_bioset: 4063 f2fs_destroy_bioset(); 4064 free_bio_enrty_cache: 4065 f2fs_destroy_bio_entry_cache(); 4066 free_post_read: 4067 f2fs_destroy_post_read_processing(); 4068 free_root_stats: 4069 f2fs_destroy_root_stats(); 4070 unregister_filesystem(&f2fs_fs_type); 4071 free_shrinker: 4072 unregister_shrinker(&f2fs_shrinker_info); 4073 free_sysfs: 4074 f2fs_exit_sysfs(); 4075 free_garbage_collection_cache: 4076 f2fs_destroy_garbage_collection_cache(); 4077 free_extent_cache: 4078 f2fs_destroy_extent_cache(); 4079 free_checkpoint_caches: 4080 f2fs_destroy_checkpoint_caches(); 4081 free_segment_manager_caches: 4082 f2fs_destroy_segment_manager_caches(); 4083 free_node_manager_caches: 4084 f2fs_destroy_node_manager_caches(); 4085 free_inodecache: 4086 destroy_inodecache(); 4087 fail: 4088 return err; 4089 } 4090 4091 static void __exit exit_f2fs_fs(void) 4092 { 4093 f2fs_destroy_compress_cache(); 4094 f2fs_destroy_compress_mempool(); 4095 f2fs_destroy_bioset(); 4096 f2fs_destroy_bio_entry_cache(); 4097 f2fs_destroy_post_read_processing(); 4098 f2fs_destroy_root_stats(); 4099 unregister_filesystem(&f2fs_fs_type); 4100 unregister_shrinker(&f2fs_shrinker_info); 4101 f2fs_exit_sysfs(); 4102 f2fs_destroy_garbage_collection_cache(); 4103 f2fs_destroy_extent_cache(); 4104 f2fs_destroy_checkpoint_caches(); 4105 f2fs_destroy_segment_manager_caches(); 4106 f2fs_destroy_node_manager_caches(); 4107 destroy_inodecache(); 4108 f2fs_destroy_trace_ios(); 4109 } 4110 4111 module_init(init_f2fs_fs) 4112 module_exit(exit_f2fs_fs) 4113 4114 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4115 MODULE_DESCRIPTION("Flash Friendly File System"); 4116 MODULE_LICENSE("GPL"); 4117 4118