1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * fs/f2fs/super.c 4 * 5 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 6 * http://www.samsung.com/ 7 */ 8 #include <linux/module.h> 9 #include <linux/init.h> 10 #include <linux/fs.h> 11 #include <linux/statfs.h> 12 #include <linux/buffer_head.h> 13 #include <linux/backing-dev.h> 14 #include <linux/kthread.h> 15 #include <linux/parser.h> 16 #include <linux/mount.h> 17 #include <linux/seq_file.h> 18 #include <linux/proc_fs.h> 19 #include <linux/random.h> 20 #include <linux/exportfs.h> 21 #include <linux/blkdev.h> 22 #include <linux/quotaops.h> 23 #include <linux/f2fs_fs.h> 24 #include <linux/sysfs.h> 25 #include <linux/quota.h> 26 #include <linux/unicode.h> 27 #include <linux/part_stat.h> 28 29 #include "f2fs.h" 30 #include "node.h" 31 #include "segment.h" 32 #include "xattr.h" 33 #include "gc.h" 34 #include "trace.h" 35 36 #define CREATE_TRACE_POINTS 37 #include <trace/events/f2fs.h> 38 39 static struct kmem_cache *f2fs_inode_cachep; 40 41 #ifdef CONFIG_F2FS_FAULT_INJECTION 42 43 const char *f2fs_fault_name[FAULT_MAX] = { 44 [FAULT_KMALLOC] = "kmalloc", 45 [FAULT_KVMALLOC] = "kvmalloc", 46 [FAULT_PAGE_ALLOC] = "page alloc", 47 [FAULT_PAGE_GET] = "page get", 48 [FAULT_ALLOC_BIO] = "alloc bio", 49 [FAULT_ALLOC_NID] = "alloc nid", 50 [FAULT_ORPHAN] = "orphan", 51 [FAULT_BLOCK] = "no more block", 52 [FAULT_DIR_DEPTH] = "too big dir depth", 53 [FAULT_EVICT_INODE] = "evict_inode fail", 54 [FAULT_TRUNCATE] = "truncate fail", 55 [FAULT_READ_IO] = "read IO error", 56 [FAULT_CHECKPOINT] = "checkpoint error", 57 [FAULT_DISCARD] = "discard error", 58 [FAULT_WRITE_IO] = "write IO error", 59 }; 60 61 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, 62 unsigned int type) 63 { 64 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; 65 66 if (rate) { 67 atomic_set(&ffi->inject_ops, 0); 68 ffi->inject_rate = rate; 69 } 70 71 if (type) 72 ffi->inject_type = type; 73 74 if (!rate && !type) 75 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 76 } 77 #endif 78 79 /* f2fs-wide shrinker description */ 80 static struct shrinker f2fs_shrinker_info = { 81 .scan_objects = f2fs_shrink_scan, 82 .count_objects = f2fs_shrink_count, 83 .seeks = DEFAULT_SEEKS, 84 }; 85 86 enum { 87 Opt_gc_background, 88 Opt_disable_roll_forward, 89 Opt_norecovery, 90 Opt_discard, 91 Opt_nodiscard, 92 Opt_noheap, 93 Opt_heap, 94 Opt_user_xattr, 95 Opt_nouser_xattr, 96 Opt_acl, 97 Opt_noacl, 98 Opt_active_logs, 99 Opt_disable_ext_identify, 100 Opt_inline_xattr, 101 Opt_noinline_xattr, 102 Opt_inline_xattr_size, 103 Opt_inline_data, 104 Opt_inline_dentry, 105 Opt_noinline_dentry, 106 Opt_flush_merge, 107 Opt_noflush_merge, 108 Opt_nobarrier, 109 Opt_fastboot, 110 Opt_extent_cache, 111 Opt_noextent_cache, 112 Opt_noinline_data, 113 Opt_data_flush, 114 Opt_reserve_root, 115 Opt_resgid, 116 Opt_resuid, 117 Opt_mode, 118 Opt_io_size_bits, 119 Opt_fault_injection, 120 Opt_fault_type, 121 Opt_lazytime, 122 Opt_nolazytime, 123 Opt_quota, 124 Opt_noquota, 125 Opt_usrquota, 126 Opt_grpquota, 127 Opt_prjquota, 128 Opt_usrjquota, 129 Opt_grpjquota, 130 Opt_prjjquota, 131 Opt_offusrjquota, 132 Opt_offgrpjquota, 133 Opt_offprjjquota, 134 Opt_jqfmt_vfsold, 135 Opt_jqfmt_vfsv0, 136 Opt_jqfmt_vfsv1, 137 Opt_whint, 138 Opt_alloc, 139 Opt_fsync, 140 Opt_test_dummy_encryption, 141 Opt_inlinecrypt, 142 Opt_checkpoint_disable, 143 Opt_checkpoint_disable_cap, 144 Opt_checkpoint_disable_cap_perc, 145 Opt_checkpoint_enable, 146 Opt_compress_algorithm, 147 Opt_compress_log_size, 148 Opt_compress_extension, 149 Opt_err, 150 }; 151 152 static match_table_t f2fs_tokens = { 153 {Opt_gc_background, "background_gc=%s"}, 154 {Opt_disable_roll_forward, "disable_roll_forward"}, 155 {Opt_norecovery, "norecovery"}, 156 {Opt_discard, "discard"}, 157 {Opt_nodiscard, "nodiscard"}, 158 {Opt_noheap, "no_heap"}, 159 {Opt_heap, "heap"}, 160 {Opt_user_xattr, "user_xattr"}, 161 {Opt_nouser_xattr, "nouser_xattr"}, 162 {Opt_acl, "acl"}, 163 {Opt_noacl, "noacl"}, 164 {Opt_active_logs, "active_logs=%u"}, 165 {Opt_disable_ext_identify, "disable_ext_identify"}, 166 {Opt_inline_xattr, "inline_xattr"}, 167 {Opt_noinline_xattr, "noinline_xattr"}, 168 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 169 {Opt_inline_data, "inline_data"}, 170 {Opt_inline_dentry, "inline_dentry"}, 171 {Opt_noinline_dentry, "noinline_dentry"}, 172 {Opt_flush_merge, "flush_merge"}, 173 {Opt_noflush_merge, "noflush_merge"}, 174 {Opt_nobarrier, "nobarrier"}, 175 {Opt_fastboot, "fastboot"}, 176 {Opt_extent_cache, "extent_cache"}, 177 {Opt_noextent_cache, "noextent_cache"}, 178 {Opt_noinline_data, "noinline_data"}, 179 {Opt_data_flush, "data_flush"}, 180 {Opt_reserve_root, "reserve_root=%u"}, 181 {Opt_resgid, "resgid=%u"}, 182 {Opt_resuid, "resuid=%u"}, 183 {Opt_mode, "mode=%s"}, 184 {Opt_io_size_bits, "io_bits=%u"}, 185 {Opt_fault_injection, "fault_injection=%u"}, 186 {Opt_fault_type, "fault_type=%u"}, 187 {Opt_lazytime, "lazytime"}, 188 {Opt_nolazytime, "nolazytime"}, 189 {Opt_quota, "quota"}, 190 {Opt_noquota, "noquota"}, 191 {Opt_usrquota, "usrquota"}, 192 {Opt_grpquota, "grpquota"}, 193 {Opt_prjquota, "prjquota"}, 194 {Opt_usrjquota, "usrjquota=%s"}, 195 {Opt_grpjquota, "grpjquota=%s"}, 196 {Opt_prjjquota, "prjjquota=%s"}, 197 {Opt_offusrjquota, "usrjquota="}, 198 {Opt_offgrpjquota, "grpjquota="}, 199 {Opt_offprjjquota, "prjjquota="}, 200 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 201 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 202 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 203 {Opt_whint, "whint_mode=%s"}, 204 {Opt_alloc, "alloc_mode=%s"}, 205 {Opt_fsync, "fsync_mode=%s"}, 206 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 207 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 208 {Opt_inlinecrypt, "inlinecrypt"}, 209 {Opt_checkpoint_disable, "checkpoint=disable"}, 210 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 211 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 212 {Opt_checkpoint_enable, "checkpoint=enable"}, 213 {Opt_compress_algorithm, "compress_algorithm=%s"}, 214 {Opt_compress_log_size, "compress_log_size=%u"}, 215 {Opt_compress_extension, "compress_extension=%s"}, 216 {Opt_err, NULL}, 217 }; 218 219 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 220 { 221 struct va_format vaf; 222 va_list args; 223 int level; 224 225 va_start(args, fmt); 226 227 level = printk_get_level(fmt); 228 vaf.fmt = printk_skip_level(fmt); 229 vaf.va = &args; 230 printk("%c%cF2FS-fs (%s): %pV\n", 231 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 232 233 va_end(args); 234 } 235 236 #ifdef CONFIG_UNICODE 237 static const struct f2fs_sb_encodings { 238 __u16 magic; 239 char *name; 240 char *version; 241 } f2fs_sb_encoding_map[] = { 242 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"}, 243 }; 244 245 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb, 246 const struct f2fs_sb_encodings **encoding, 247 __u16 *flags) 248 { 249 __u16 magic = le16_to_cpu(sb->s_encoding); 250 int i; 251 252 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 253 if (magic == f2fs_sb_encoding_map[i].magic) 254 break; 255 256 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map)) 257 return -EINVAL; 258 259 *encoding = &f2fs_sb_encoding_map[i]; 260 *flags = le16_to_cpu(sb->s_encoding_flags); 261 262 return 0; 263 } 264 #endif 265 266 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 267 { 268 block_t limit = min((sbi->user_block_count << 1) / 1000, 269 sbi->user_block_count - sbi->reserved_blocks); 270 271 /* limit is 0.2% */ 272 if (test_opt(sbi, RESERVE_ROOT) && 273 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 274 F2FS_OPTION(sbi).root_reserved_blocks = limit; 275 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 276 F2FS_OPTION(sbi).root_reserved_blocks); 277 } 278 if (!test_opt(sbi, RESERVE_ROOT) && 279 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 280 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 281 !gid_eq(F2FS_OPTION(sbi).s_resgid, 282 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 283 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 284 from_kuid_munged(&init_user_ns, 285 F2FS_OPTION(sbi).s_resuid), 286 from_kgid_munged(&init_user_ns, 287 F2FS_OPTION(sbi).s_resgid)); 288 } 289 290 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 291 { 292 if (!F2FS_OPTION(sbi).unusable_cap_perc) 293 return; 294 295 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 296 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 297 else 298 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 299 F2FS_OPTION(sbi).unusable_cap_perc; 300 301 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 302 F2FS_OPTION(sbi).unusable_cap, 303 F2FS_OPTION(sbi).unusable_cap_perc); 304 } 305 306 static void init_once(void *foo) 307 { 308 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 309 310 inode_init_once(&fi->vfs_inode); 311 } 312 313 #ifdef CONFIG_QUOTA 314 static const char * const quotatypes[] = INITQFNAMES; 315 #define QTYPE2NAME(t) (quotatypes[t]) 316 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 317 substring_t *args) 318 { 319 struct f2fs_sb_info *sbi = F2FS_SB(sb); 320 char *qname; 321 int ret = -EINVAL; 322 323 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 324 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 325 return -EINVAL; 326 } 327 if (f2fs_sb_has_quota_ino(sbi)) { 328 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 329 return 0; 330 } 331 332 qname = match_strdup(args); 333 if (!qname) { 334 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 335 return -ENOMEM; 336 } 337 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 338 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 339 ret = 0; 340 else 341 f2fs_err(sbi, "%s quota file already specified", 342 QTYPE2NAME(qtype)); 343 goto errout; 344 } 345 if (strchr(qname, '/')) { 346 f2fs_err(sbi, "quotafile must be on filesystem root"); 347 goto errout; 348 } 349 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 350 set_opt(sbi, QUOTA); 351 return 0; 352 errout: 353 kvfree(qname); 354 return ret; 355 } 356 357 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 358 { 359 struct f2fs_sb_info *sbi = F2FS_SB(sb); 360 361 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 362 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 363 return -EINVAL; 364 } 365 kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 366 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 367 return 0; 368 } 369 370 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 371 { 372 /* 373 * We do the test below only for project quotas. 'usrquota' and 374 * 'grpquota' mount options are allowed even without quota feature 375 * to support legacy quotas in quota files. 376 */ 377 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 378 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 379 return -1; 380 } 381 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 382 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 383 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 384 if (test_opt(sbi, USRQUOTA) && 385 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 386 clear_opt(sbi, USRQUOTA); 387 388 if (test_opt(sbi, GRPQUOTA) && 389 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 390 clear_opt(sbi, GRPQUOTA); 391 392 if (test_opt(sbi, PRJQUOTA) && 393 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 394 clear_opt(sbi, PRJQUOTA); 395 396 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 397 test_opt(sbi, PRJQUOTA)) { 398 f2fs_err(sbi, "old and new quota format mixing"); 399 return -1; 400 } 401 402 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 403 f2fs_err(sbi, "journaled quota format not specified"); 404 return -1; 405 } 406 } 407 408 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 409 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 410 F2FS_OPTION(sbi).s_jquota_fmt = 0; 411 } 412 return 0; 413 } 414 #endif 415 416 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 417 const char *opt, 418 const substring_t *arg, 419 bool is_remount) 420 { 421 struct f2fs_sb_info *sbi = F2FS_SB(sb); 422 #ifdef CONFIG_FS_ENCRYPTION 423 int err; 424 425 if (!f2fs_sb_has_encrypt(sbi)) { 426 f2fs_err(sbi, "Encrypt feature is off"); 427 return -EINVAL; 428 } 429 430 /* 431 * This mount option is just for testing, and it's not worthwhile to 432 * implement the extra complexity (e.g. RCU protection) that would be 433 * needed to allow it to be set or changed during remount. We do allow 434 * it to be specified during remount, but only if there is no change. 435 */ 436 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_ctx.ctx) { 437 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 438 return -EINVAL; 439 } 440 err = fscrypt_set_test_dummy_encryption( 441 sb, arg, &F2FS_OPTION(sbi).dummy_enc_ctx); 442 if (err) { 443 if (err == -EEXIST) 444 f2fs_warn(sbi, 445 "Can't change test_dummy_encryption on remount"); 446 else if (err == -EINVAL) 447 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 448 opt); 449 else 450 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 451 opt, err); 452 return -EINVAL; 453 } 454 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 455 #else 456 f2fs_warn(sbi, "Test dummy encryption mount option ignored"); 457 #endif 458 return 0; 459 } 460 461 static int parse_options(struct super_block *sb, char *options, bool is_remount) 462 { 463 struct f2fs_sb_info *sbi = F2FS_SB(sb); 464 substring_t args[MAX_OPT_ARGS]; 465 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 466 char *p, *name; 467 int arg = 0, ext_cnt; 468 kuid_t uid; 469 kgid_t gid; 470 int ret; 471 472 if (!options) 473 return 0; 474 475 while ((p = strsep(&options, ",")) != NULL) { 476 int token; 477 if (!*p) 478 continue; 479 /* 480 * Initialize args struct so we know whether arg was 481 * found; some options take optional arguments. 482 */ 483 args[0].to = args[0].from = NULL; 484 token = match_token(p, f2fs_tokens, args); 485 486 switch (token) { 487 case Opt_gc_background: 488 name = match_strdup(&args[0]); 489 490 if (!name) 491 return -ENOMEM; 492 if (!strcmp(name, "on")) { 493 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 494 } else if (!strcmp(name, "off")) { 495 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 496 } else if (!strcmp(name, "sync")) { 497 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 498 } else { 499 kvfree(name); 500 return -EINVAL; 501 } 502 kvfree(name); 503 break; 504 case Opt_disable_roll_forward: 505 set_opt(sbi, DISABLE_ROLL_FORWARD); 506 break; 507 case Opt_norecovery: 508 /* this option mounts f2fs with ro */ 509 set_opt(sbi, NORECOVERY); 510 if (!f2fs_readonly(sb)) 511 return -EINVAL; 512 break; 513 case Opt_discard: 514 set_opt(sbi, DISCARD); 515 break; 516 case Opt_nodiscard: 517 if (f2fs_sb_has_blkzoned(sbi)) { 518 f2fs_warn(sbi, "discard is required for zoned block devices"); 519 return -EINVAL; 520 } 521 clear_opt(sbi, DISCARD); 522 break; 523 case Opt_noheap: 524 set_opt(sbi, NOHEAP); 525 break; 526 case Opt_heap: 527 clear_opt(sbi, NOHEAP); 528 break; 529 #ifdef CONFIG_F2FS_FS_XATTR 530 case Opt_user_xattr: 531 set_opt(sbi, XATTR_USER); 532 break; 533 case Opt_nouser_xattr: 534 clear_opt(sbi, XATTR_USER); 535 break; 536 case Opt_inline_xattr: 537 set_opt(sbi, INLINE_XATTR); 538 break; 539 case Opt_noinline_xattr: 540 clear_opt(sbi, INLINE_XATTR); 541 break; 542 case Opt_inline_xattr_size: 543 if (args->from && match_int(args, &arg)) 544 return -EINVAL; 545 set_opt(sbi, INLINE_XATTR_SIZE); 546 F2FS_OPTION(sbi).inline_xattr_size = arg; 547 break; 548 #else 549 case Opt_user_xattr: 550 f2fs_info(sbi, "user_xattr options not supported"); 551 break; 552 case Opt_nouser_xattr: 553 f2fs_info(sbi, "nouser_xattr options not supported"); 554 break; 555 case Opt_inline_xattr: 556 f2fs_info(sbi, "inline_xattr options not supported"); 557 break; 558 case Opt_noinline_xattr: 559 f2fs_info(sbi, "noinline_xattr options not supported"); 560 break; 561 #endif 562 #ifdef CONFIG_F2FS_FS_POSIX_ACL 563 case Opt_acl: 564 set_opt(sbi, POSIX_ACL); 565 break; 566 case Opt_noacl: 567 clear_opt(sbi, POSIX_ACL); 568 break; 569 #else 570 case Opt_acl: 571 f2fs_info(sbi, "acl options not supported"); 572 break; 573 case Opt_noacl: 574 f2fs_info(sbi, "noacl options not supported"); 575 break; 576 #endif 577 case Opt_active_logs: 578 if (args->from && match_int(args, &arg)) 579 return -EINVAL; 580 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE) 581 return -EINVAL; 582 F2FS_OPTION(sbi).active_logs = arg; 583 break; 584 case Opt_disable_ext_identify: 585 set_opt(sbi, DISABLE_EXT_IDENTIFY); 586 break; 587 case Opt_inline_data: 588 set_opt(sbi, INLINE_DATA); 589 break; 590 case Opt_inline_dentry: 591 set_opt(sbi, INLINE_DENTRY); 592 break; 593 case Opt_noinline_dentry: 594 clear_opt(sbi, INLINE_DENTRY); 595 break; 596 case Opt_flush_merge: 597 set_opt(sbi, FLUSH_MERGE); 598 break; 599 case Opt_noflush_merge: 600 clear_opt(sbi, FLUSH_MERGE); 601 break; 602 case Opt_nobarrier: 603 set_opt(sbi, NOBARRIER); 604 break; 605 case Opt_fastboot: 606 set_opt(sbi, FASTBOOT); 607 break; 608 case Opt_extent_cache: 609 set_opt(sbi, EXTENT_CACHE); 610 break; 611 case Opt_noextent_cache: 612 clear_opt(sbi, EXTENT_CACHE); 613 break; 614 case Opt_noinline_data: 615 clear_opt(sbi, INLINE_DATA); 616 break; 617 case Opt_data_flush: 618 set_opt(sbi, DATA_FLUSH); 619 break; 620 case Opt_reserve_root: 621 if (args->from && match_int(args, &arg)) 622 return -EINVAL; 623 if (test_opt(sbi, RESERVE_ROOT)) { 624 f2fs_info(sbi, "Preserve previous reserve_root=%u", 625 F2FS_OPTION(sbi).root_reserved_blocks); 626 } else { 627 F2FS_OPTION(sbi).root_reserved_blocks = arg; 628 set_opt(sbi, RESERVE_ROOT); 629 } 630 break; 631 case Opt_resuid: 632 if (args->from && match_int(args, &arg)) 633 return -EINVAL; 634 uid = make_kuid(current_user_ns(), arg); 635 if (!uid_valid(uid)) { 636 f2fs_err(sbi, "Invalid uid value %d", arg); 637 return -EINVAL; 638 } 639 F2FS_OPTION(sbi).s_resuid = uid; 640 break; 641 case Opt_resgid: 642 if (args->from && match_int(args, &arg)) 643 return -EINVAL; 644 gid = make_kgid(current_user_ns(), arg); 645 if (!gid_valid(gid)) { 646 f2fs_err(sbi, "Invalid gid value %d", arg); 647 return -EINVAL; 648 } 649 F2FS_OPTION(sbi).s_resgid = gid; 650 break; 651 case Opt_mode: 652 name = match_strdup(&args[0]); 653 654 if (!name) 655 return -ENOMEM; 656 if (!strcmp(name, "adaptive")) { 657 if (f2fs_sb_has_blkzoned(sbi)) { 658 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 659 kvfree(name); 660 return -EINVAL; 661 } 662 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 663 } else if (!strcmp(name, "lfs")) { 664 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 665 } else { 666 kvfree(name); 667 return -EINVAL; 668 } 669 kvfree(name); 670 break; 671 case Opt_io_size_bits: 672 if (args->from && match_int(args, &arg)) 673 return -EINVAL; 674 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) { 675 f2fs_warn(sbi, "Not support %d, larger than %d", 676 1 << arg, BIO_MAX_PAGES); 677 return -EINVAL; 678 } 679 F2FS_OPTION(sbi).write_io_size_bits = arg; 680 break; 681 #ifdef CONFIG_F2FS_FAULT_INJECTION 682 case Opt_fault_injection: 683 if (args->from && match_int(args, &arg)) 684 return -EINVAL; 685 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 686 set_opt(sbi, FAULT_INJECTION); 687 break; 688 689 case Opt_fault_type: 690 if (args->from && match_int(args, &arg)) 691 return -EINVAL; 692 f2fs_build_fault_attr(sbi, 0, arg); 693 set_opt(sbi, FAULT_INJECTION); 694 break; 695 #else 696 case Opt_fault_injection: 697 f2fs_info(sbi, "fault_injection options not supported"); 698 break; 699 700 case Opt_fault_type: 701 f2fs_info(sbi, "fault_type options not supported"); 702 break; 703 #endif 704 case Opt_lazytime: 705 sb->s_flags |= SB_LAZYTIME; 706 break; 707 case Opt_nolazytime: 708 sb->s_flags &= ~SB_LAZYTIME; 709 break; 710 #ifdef CONFIG_QUOTA 711 case Opt_quota: 712 case Opt_usrquota: 713 set_opt(sbi, USRQUOTA); 714 break; 715 case Opt_grpquota: 716 set_opt(sbi, GRPQUOTA); 717 break; 718 case Opt_prjquota: 719 set_opt(sbi, PRJQUOTA); 720 break; 721 case Opt_usrjquota: 722 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 723 if (ret) 724 return ret; 725 break; 726 case Opt_grpjquota: 727 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 728 if (ret) 729 return ret; 730 break; 731 case Opt_prjjquota: 732 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 733 if (ret) 734 return ret; 735 break; 736 case Opt_offusrjquota: 737 ret = f2fs_clear_qf_name(sb, USRQUOTA); 738 if (ret) 739 return ret; 740 break; 741 case Opt_offgrpjquota: 742 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 743 if (ret) 744 return ret; 745 break; 746 case Opt_offprjjquota: 747 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 748 if (ret) 749 return ret; 750 break; 751 case Opt_jqfmt_vfsold: 752 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 753 break; 754 case Opt_jqfmt_vfsv0: 755 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 756 break; 757 case Opt_jqfmt_vfsv1: 758 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 759 break; 760 case Opt_noquota: 761 clear_opt(sbi, QUOTA); 762 clear_opt(sbi, USRQUOTA); 763 clear_opt(sbi, GRPQUOTA); 764 clear_opt(sbi, PRJQUOTA); 765 break; 766 #else 767 case Opt_quota: 768 case Opt_usrquota: 769 case Opt_grpquota: 770 case Opt_prjquota: 771 case Opt_usrjquota: 772 case Opt_grpjquota: 773 case Opt_prjjquota: 774 case Opt_offusrjquota: 775 case Opt_offgrpjquota: 776 case Opt_offprjjquota: 777 case Opt_jqfmt_vfsold: 778 case Opt_jqfmt_vfsv0: 779 case Opt_jqfmt_vfsv1: 780 case Opt_noquota: 781 f2fs_info(sbi, "quota operations not supported"); 782 break; 783 #endif 784 case Opt_whint: 785 name = match_strdup(&args[0]); 786 if (!name) 787 return -ENOMEM; 788 if (!strcmp(name, "user-based")) { 789 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER; 790 } else if (!strcmp(name, "off")) { 791 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 792 } else if (!strcmp(name, "fs-based")) { 793 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS; 794 } else { 795 kvfree(name); 796 return -EINVAL; 797 } 798 kvfree(name); 799 break; 800 case Opt_alloc: 801 name = match_strdup(&args[0]); 802 if (!name) 803 return -ENOMEM; 804 805 if (!strcmp(name, "default")) { 806 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 807 } else if (!strcmp(name, "reuse")) { 808 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 809 } else { 810 kvfree(name); 811 return -EINVAL; 812 } 813 kvfree(name); 814 break; 815 case Opt_fsync: 816 name = match_strdup(&args[0]); 817 if (!name) 818 return -ENOMEM; 819 if (!strcmp(name, "posix")) { 820 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 821 } else if (!strcmp(name, "strict")) { 822 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 823 } else if (!strcmp(name, "nobarrier")) { 824 F2FS_OPTION(sbi).fsync_mode = 825 FSYNC_MODE_NOBARRIER; 826 } else { 827 kvfree(name); 828 return -EINVAL; 829 } 830 kvfree(name); 831 break; 832 case Opt_test_dummy_encryption: 833 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 834 is_remount); 835 if (ret) 836 return ret; 837 break; 838 case Opt_inlinecrypt: 839 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 840 sb->s_flags |= SB_INLINECRYPT; 841 #else 842 f2fs_info(sbi, "inline encryption not supported"); 843 #endif 844 break; 845 case Opt_checkpoint_disable_cap_perc: 846 if (args->from && match_int(args, &arg)) 847 return -EINVAL; 848 if (arg < 0 || arg > 100) 849 return -EINVAL; 850 F2FS_OPTION(sbi).unusable_cap_perc = arg; 851 set_opt(sbi, DISABLE_CHECKPOINT); 852 break; 853 case Opt_checkpoint_disable_cap: 854 if (args->from && match_int(args, &arg)) 855 return -EINVAL; 856 F2FS_OPTION(sbi).unusable_cap = arg; 857 set_opt(sbi, DISABLE_CHECKPOINT); 858 break; 859 case Opt_checkpoint_disable: 860 set_opt(sbi, DISABLE_CHECKPOINT); 861 break; 862 case Opt_checkpoint_enable: 863 clear_opt(sbi, DISABLE_CHECKPOINT); 864 break; 865 case Opt_compress_algorithm: 866 if (!f2fs_sb_has_compression(sbi)) { 867 f2fs_err(sbi, "Compression feature if off"); 868 return -EINVAL; 869 } 870 name = match_strdup(&args[0]); 871 if (!name) 872 return -ENOMEM; 873 if (!strcmp(name, "lzo")) { 874 F2FS_OPTION(sbi).compress_algorithm = 875 COMPRESS_LZO; 876 } else if (!strcmp(name, "lz4")) { 877 F2FS_OPTION(sbi).compress_algorithm = 878 COMPRESS_LZ4; 879 } else if (!strcmp(name, "zstd")) { 880 F2FS_OPTION(sbi).compress_algorithm = 881 COMPRESS_ZSTD; 882 } else if (!strcmp(name, "lzo-rle")) { 883 F2FS_OPTION(sbi).compress_algorithm = 884 COMPRESS_LZORLE; 885 } else { 886 kfree(name); 887 return -EINVAL; 888 } 889 kfree(name); 890 break; 891 case Opt_compress_log_size: 892 if (!f2fs_sb_has_compression(sbi)) { 893 f2fs_err(sbi, "Compression feature is off"); 894 return -EINVAL; 895 } 896 if (args->from && match_int(args, &arg)) 897 return -EINVAL; 898 if (arg < MIN_COMPRESS_LOG_SIZE || 899 arg > MAX_COMPRESS_LOG_SIZE) { 900 f2fs_err(sbi, 901 "Compress cluster log size is out of range"); 902 return -EINVAL; 903 } 904 F2FS_OPTION(sbi).compress_log_size = arg; 905 break; 906 case Opt_compress_extension: 907 if (!f2fs_sb_has_compression(sbi)) { 908 f2fs_err(sbi, "Compression feature is off"); 909 return -EINVAL; 910 } 911 name = match_strdup(&args[0]); 912 if (!name) 913 return -ENOMEM; 914 915 ext = F2FS_OPTION(sbi).extensions; 916 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 917 918 if (strlen(name) >= F2FS_EXTENSION_LEN || 919 ext_cnt >= COMPRESS_EXT_NUM) { 920 f2fs_err(sbi, 921 "invalid extension length/number"); 922 kfree(name); 923 return -EINVAL; 924 } 925 926 strcpy(ext[ext_cnt], name); 927 F2FS_OPTION(sbi).compress_ext_cnt++; 928 kfree(name); 929 break; 930 default: 931 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 932 p); 933 return -EINVAL; 934 } 935 } 936 #ifdef CONFIG_QUOTA 937 if (f2fs_check_quota_options(sbi)) 938 return -EINVAL; 939 #else 940 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 941 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 942 return -EINVAL; 943 } 944 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 945 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 946 return -EINVAL; 947 } 948 #endif 949 #ifndef CONFIG_UNICODE 950 if (f2fs_sb_has_casefold(sbi)) { 951 f2fs_err(sbi, 952 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 953 return -EINVAL; 954 } 955 #endif 956 957 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 958 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO", 959 F2FS_IO_SIZE_KB(sbi)); 960 return -EINVAL; 961 } 962 963 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 964 int min_size, max_size; 965 966 if (!f2fs_sb_has_extra_attr(sbi) || 967 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 968 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 969 return -EINVAL; 970 } 971 if (!test_opt(sbi, INLINE_XATTR)) { 972 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 973 return -EINVAL; 974 } 975 976 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 977 max_size = MAX_INLINE_XATTR_SIZE; 978 979 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 980 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 981 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 982 min_size, max_size); 983 return -EINVAL; 984 } 985 } 986 987 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 988 f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n"); 989 return -EINVAL; 990 } 991 992 /* Not pass down write hints if the number of active logs is lesser 993 * than NR_CURSEG_TYPE. 994 */ 995 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE) 996 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 997 return 0; 998 } 999 1000 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1001 { 1002 struct f2fs_inode_info *fi; 1003 1004 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); 1005 if (!fi) 1006 return NULL; 1007 1008 init_once((void *) fi); 1009 1010 /* Initialize f2fs-specific inode info */ 1011 atomic_set(&fi->dirty_pages, 0); 1012 init_rwsem(&fi->i_sem); 1013 spin_lock_init(&fi->i_size_lock); 1014 INIT_LIST_HEAD(&fi->dirty_list); 1015 INIT_LIST_HEAD(&fi->gdirty_list); 1016 INIT_LIST_HEAD(&fi->inmem_ilist); 1017 INIT_LIST_HEAD(&fi->inmem_pages); 1018 mutex_init(&fi->inmem_lock); 1019 init_rwsem(&fi->i_gc_rwsem[READ]); 1020 init_rwsem(&fi->i_gc_rwsem[WRITE]); 1021 init_rwsem(&fi->i_mmap_sem); 1022 init_rwsem(&fi->i_xattr_sem); 1023 1024 /* Will be used by directory only */ 1025 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1026 1027 return &fi->vfs_inode; 1028 } 1029 1030 static int f2fs_drop_inode(struct inode *inode) 1031 { 1032 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1033 int ret; 1034 1035 /* 1036 * during filesystem shutdown, if checkpoint is disabled, 1037 * drop useless meta/node dirty pages. 1038 */ 1039 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1040 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1041 inode->i_ino == F2FS_META_INO(sbi)) { 1042 trace_f2fs_drop_inode(inode, 1); 1043 return 1; 1044 } 1045 } 1046 1047 /* 1048 * This is to avoid a deadlock condition like below. 1049 * writeback_single_inode(inode) 1050 * - f2fs_write_data_page 1051 * - f2fs_gc -> iput -> evict 1052 * - inode_wait_for_writeback(inode) 1053 */ 1054 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1055 if (!inode->i_nlink && !is_bad_inode(inode)) { 1056 /* to avoid evict_inode call simultaneously */ 1057 atomic_inc(&inode->i_count); 1058 spin_unlock(&inode->i_lock); 1059 1060 /* some remained atomic pages should discarded */ 1061 if (f2fs_is_atomic_file(inode)) 1062 f2fs_drop_inmem_pages(inode); 1063 1064 /* should remain fi->extent_tree for writepage */ 1065 f2fs_destroy_extent_node(inode); 1066 1067 sb_start_intwrite(inode->i_sb); 1068 f2fs_i_size_write(inode, 0); 1069 1070 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1071 inode, NULL, 0, DATA); 1072 truncate_inode_pages_final(inode->i_mapping); 1073 1074 if (F2FS_HAS_BLOCKS(inode)) 1075 f2fs_truncate(inode); 1076 1077 sb_end_intwrite(inode->i_sb); 1078 1079 spin_lock(&inode->i_lock); 1080 atomic_dec(&inode->i_count); 1081 } 1082 trace_f2fs_drop_inode(inode, 0); 1083 return 0; 1084 } 1085 ret = generic_drop_inode(inode); 1086 if (!ret) 1087 ret = fscrypt_drop_inode(inode); 1088 trace_f2fs_drop_inode(inode, ret); 1089 return ret; 1090 } 1091 1092 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1093 { 1094 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1095 int ret = 0; 1096 1097 spin_lock(&sbi->inode_lock[DIRTY_META]); 1098 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1099 ret = 1; 1100 } else { 1101 set_inode_flag(inode, FI_DIRTY_INODE); 1102 stat_inc_dirty_inode(sbi, DIRTY_META); 1103 } 1104 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1105 list_add_tail(&F2FS_I(inode)->gdirty_list, 1106 &sbi->inode_list[DIRTY_META]); 1107 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1108 } 1109 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1110 return ret; 1111 } 1112 1113 void f2fs_inode_synced(struct inode *inode) 1114 { 1115 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1116 1117 spin_lock(&sbi->inode_lock[DIRTY_META]); 1118 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1119 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1120 return; 1121 } 1122 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1123 list_del_init(&F2FS_I(inode)->gdirty_list); 1124 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1125 } 1126 clear_inode_flag(inode, FI_DIRTY_INODE); 1127 clear_inode_flag(inode, FI_AUTO_RECOVER); 1128 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1129 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1130 } 1131 1132 /* 1133 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1134 * 1135 * We should call set_dirty_inode to write the dirty inode through write_inode. 1136 */ 1137 static void f2fs_dirty_inode(struct inode *inode, int flags) 1138 { 1139 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1140 1141 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1142 inode->i_ino == F2FS_META_INO(sbi)) 1143 return; 1144 1145 if (flags == I_DIRTY_TIME) 1146 return; 1147 1148 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1149 clear_inode_flag(inode, FI_AUTO_RECOVER); 1150 1151 f2fs_inode_dirtied(inode, false); 1152 } 1153 1154 static void f2fs_free_inode(struct inode *inode) 1155 { 1156 fscrypt_free_inode(inode); 1157 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1158 } 1159 1160 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1161 { 1162 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1163 percpu_counter_destroy(&sbi->total_valid_inode_count); 1164 } 1165 1166 static void destroy_device_list(struct f2fs_sb_info *sbi) 1167 { 1168 int i; 1169 1170 for (i = 0; i < sbi->s_ndevs; i++) { 1171 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1172 #ifdef CONFIG_BLK_DEV_ZONED 1173 kvfree(FDEV(i).blkz_seq); 1174 #endif 1175 } 1176 kvfree(sbi->devs); 1177 } 1178 1179 static void f2fs_put_super(struct super_block *sb) 1180 { 1181 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1182 int i; 1183 bool dropped; 1184 1185 f2fs_quota_off_umount(sb); 1186 1187 /* prevent remaining shrinker jobs */ 1188 mutex_lock(&sbi->umount_mutex); 1189 1190 /* 1191 * We don't need to do checkpoint when superblock is clean. 1192 * But, the previous checkpoint was not done by umount, it needs to do 1193 * clean checkpoint again. 1194 */ 1195 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1196 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1197 struct cp_control cpc = { 1198 .reason = CP_UMOUNT, 1199 }; 1200 f2fs_write_checkpoint(sbi, &cpc); 1201 } 1202 1203 /* be sure to wait for any on-going discard commands */ 1204 dropped = f2fs_issue_discard_timeout(sbi); 1205 1206 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && 1207 !sbi->discard_blks && !dropped) { 1208 struct cp_control cpc = { 1209 .reason = CP_UMOUNT | CP_TRIMMED, 1210 }; 1211 f2fs_write_checkpoint(sbi, &cpc); 1212 } 1213 1214 /* 1215 * normally superblock is clean, so we need to release this. 1216 * In addition, EIO will skip do checkpoint, we need this as well. 1217 */ 1218 f2fs_release_ino_entry(sbi, true); 1219 1220 f2fs_leave_shrinker(sbi); 1221 mutex_unlock(&sbi->umount_mutex); 1222 1223 /* our cp_error case, we can wait for any writeback page */ 1224 f2fs_flush_merged_writes(sbi); 1225 1226 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1227 1228 f2fs_bug_on(sbi, sbi->fsync_node_num); 1229 1230 iput(sbi->node_inode); 1231 sbi->node_inode = NULL; 1232 1233 iput(sbi->meta_inode); 1234 sbi->meta_inode = NULL; 1235 1236 /* 1237 * iput() can update stat information, if f2fs_write_checkpoint() 1238 * above failed with error. 1239 */ 1240 f2fs_destroy_stats(sbi); 1241 1242 /* destroy f2fs internal modules */ 1243 f2fs_destroy_node_manager(sbi); 1244 f2fs_destroy_segment_manager(sbi); 1245 1246 f2fs_destroy_post_read_wq(sbi); 1247 1248 kvfree(sbi->ckpt); 1249 1250 f2fs_unregister_sysfs(sbi); 1251 1252 sb->s_fs_info = NULL; 1253 if (sbi->s_chksum_driver) 1254 crypto_free_shash(sbi->s_chksum_driver); 1255 kvfree(sbi->raw_super); 1256 1257 destroy_device_list(sbi); 1258 f2fs_destroy_xattr_caches(sbi); 1259 mempool_destroy(sbi->write_io_dummy); 1260 #ifdef CONFIG_QUOTA 1261 for (i = 0; i < MAXQUOTAS; i++) 1262 kvfree(F2FS_OPTION(sbi).s_qf_names[i]); 1263 #endif 1264 fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx); 1265 destroy_percpu_info(sbi); 1266 for (i = 0; i < NR_PAGE_TYPE; i++) 1267 kvfree(sbi->write_io[i]); 1268 #ifdef CONFIG_UNICODE 1269 utf8_unload(sbi->s_encoding); 1270 #endif 1271 kvfree(sbi); 1272 } 1273 1274 int f2fs_sync_fs(struct super_block *sb, int sync) 1275 { 1276 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1277 int err = 0; 1278 1279 if (unlikely(f2fs_cp_error(sbi))) 1280 return 0; 1281 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1282 return 0; 1283 1284 trace_f2fs_sync_fs(sb, sync); 1285 1286 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1287 return -EAGAIN; 1288 1289 if (sync) { 1290 struct cp_control cpc; 1291 1292 cpc.reason = __get_cp_reason(sbi); 1293 1294 down_write(&sbi->gc_lock); 1295 err = f2fs_write_checkpoint(sbi, &cpc); 1296 up_write(&sbi->gc_lock); 1297 } 1298 f2fs_trace_ios(NULL, 1); 1299 1300 return err; 1301 } 1302 1303 static int f2fs_freeze(struct super_block *sb) 1304 { 1305 if (f2fs_readonly(sb)) 1306 return 0; 1307 1308 /* IO error happened before */ 1309 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1310 return -EIO; 1311 1312 /* must be clean, since sync_filesystem() was already called */ 1313 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1314 return -EINVAL; 1315 return 0; 1316 } 1317 1318 static int f2fs_unfreeze(struct super_block *sb) 1319 { 1320 return 0; 1321 } 1322 1323 #ifdef CONFIG_QUOTA 1324 static int f2fs_statfs_project(struct super_block *sb, 1325 kprojid_t projid, struct kstatfs *buf) 1326 { 1327 struct kqid qid; 1328 struct dquot *dquot; 1329 u64 limit; 1330 u64 curblock; 1331 1332 qid = make_kqid_projid(projid); 1333 dquot = dqget(sb, qid); 1334 if (IS_ERR(dquot)) 1335 return PTR_ERR(dquot); 1336 spin_lock(&dquot->dq_dqb_lock); 1337 1338 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1339 dquot->dq_dqb.dqb_bhardlimit); 1340 if (limit) 1341 limit >>= sb->s_blocksize_bits; 1342 1343 if (limit && buf->f_blocks > limit) { 1344 curblock = (dquot->dq_dqb.dqb_curspace + 1345 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1346 buf->f_blocks = limit; 1347 buf->f_bfree = buf->f_bavail = 1348 (buf->f_blocks > curblock) ? 1349 (buf->f_blocks - curblock) : 0; 1350 } 1351 1352 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1353 dquot->dq_dqb.dqb_ihardlimit); 1354 1355 if (limit && buf->f_files > limit) { 1356 buf->f_files = limit; 1357 buf->f_ffree = 1358 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1359 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1360 } 1361 1362 spin_unlock(&dquot->dq_dqb_lock); 1363 dqput(dquot); 1364 return 0; 1365 } 1366 #endif 1367 1368 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1369 { 1370 struct super_block *sb = dentry->d_sb; 1371 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1372 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1373 block_t total_count, user_block_count, start_count; 1374 u64 avail_node_count; 1375 1376 total_count = le64_to_cpu(sbi->raw_super->block_count); 1377 user_block_count = sbi->user_block_count; 1378 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1379 buf->f_type = F2FS_SUPER_MAGIC; 1380 buf->f_bsize = sbi->blocksize; 1381 1382 buf->f_blocks = total_count - start_count; 1383 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1384 sbi->current_reserved_blocks; 1385 1386 spin_lock(&sbi->stat_lock); 1387 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1388 buf->f_bfree = 0; 1389 else 1390 buf->f_bfree -= sbi->unusable_block_count; 1391 spin_unlock(&sbi->stat_lock); 1392 1393 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1394 buf->f_bavail = buf->f_bfree - 1395 F2FS_OPTION(sbi).root_reserved_blocks; 1396 else 1397 buf->f_bavail = 0; 1398 1399 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1400 1401 if (avail_node_count > user_block_count) { 1402 buf->f_files = user_block_count; 1403 buf->f_ffree = buf->f_bavail; 1404 } else { 1405 buf->f_files = avail_node_count; 1406 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1407 buf->f_bavail); 1408 } 1409 1410 buf->f_namelen = F2FS_NAME_LEN; 1411 buf->f_fsid.val[0] = (u32)id; 1412 buf->f_fsid.val[1] = (u32)(id >> 32); 1413 1414 #ifdef CONFIG_QUOTA 1415 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1416 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1417 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1418 } 1419 #endif 1420 return 0; 1421 } 1422 1423 static inline void f2fs_show_quota_options(struct seq_file *seq, 1424 struct super_block *sb) 1425 { 1426 #ifdef CONFIG_QUOTA 1427 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1428 1429 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1430 char *fmtname = ""; 1431 1432 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1433 case QFMT_VFS_OLD: 1434 fmtname = "vfsold"; 1435 break; 1436 case QFMT_VFS_V0: 1437 fmtname = "vfsv0"; 1438 break; 1439 case QFMT_VFS_V1: 1440 fmtname = "vfsv1"; 1441 break; 1442 } 1443 seq_printf(seq, ",jqfmt=%s", fmtname); 1444 } 1445 1446 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1447 seq_show_option(seq, "usrjquota", 1448 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1449 1450 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1451 seq_show_option(seq, "grpjquota", 1452 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1453 1454 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1455 seq_show_option(seq, "prjjquota", 1456 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1457 #endif 1458 } 1459 1460 static inline void f2fs_show_compress_options(struct seq_file *seq, 1461 struct super_block *sb) 1462 { 1463 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1464 char *algtype = ""; 1465 int i; 1466 1467 if (!f2fs_sb_has_compression(sbi)) 1468 return; 1469 1470 switch (F2FS_OPTION(sbi).compress_algorithm) { 1471 case COMPRESS_LZO: 1472 algtype = "lzo"; 1473 break; 1474 case COMPRESS_LZ4: 1475 algtype = "lz4"; 1476 break; 1477 case COMPRESS_ZSTD: 1478 algtype = "zstd"; 1479 break; 1480 case COMPRESS_LZORLE: 1481 algtype = "lzo-rle"; 1482 break; 1483 } 1484 seq_printf(seq, ",compress_algorithm=%s", algtype); 1485 1486 seq_printf(seq, ",compress_log_size=%u", 1487 F2FS_OPTION(sbi).compress_log_size); 1488 1489 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1490 seq_printf(seq, ",compress_extension=%s", 1491 F2FS_OPTION(sbi).extensions[i]); 1492 } 1493 } 1494 1495 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1496 { 1497 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1498 1499 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1500 seq_printf(seq, ",background_gc=%s", "sync"); 1501 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1502 seq_printf(seq, ",background_gc=%s", "on"); 1503 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1504 seq_printf(seq, ",background_gc=%s", "off"); 1505 1506 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1507 seq_puts(seq, ",disable_roll_forward"); 1508 if (test_opt(sbi, NORECOVERY)) 1509 seq_puts(seq, ",norecovery"); 1510 if (test_opt(sbi, DISCARD)) 1511 seq_puts(seq, ",discard"); 1512 else 1513 seq_puts(seq, ",nodiscard"); 1514 if (test_opt(sbi, NOHEAP)) 1515 seq_puts(seq, ",no_heap"); 1516 else 1517 seq_puts(seq, ",heap"); 1518 #ifdef CONFIG_F2FS_FS_XATTR 1519 if (test_opt(sbi, XATTR_USER)) 1520 seq_puts(seq, ",user_xattr"); 1521 else 1522 seq_puts(seq, ",nouser_xattr"); 1523 if (test_opt(sbi, INLINE_XATTR)) 1524 seq_puts(seq, ",inline_xattr"); 1525 else 1526 seq_puts(seq, ",noinline_xattr"); 1527 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1528 seq_printf(seq, ",inline_xattr_size=%u", 1529 F2FS_OPTION(sbi).inline_xattr_size); 1530 #endif 1531 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1532 if (test_opt(sbi, POSIX_ACL)) 1533 seq_puts(seq, ",acl"); 1534 else 1535 seq_puts(seq, ",noacl"); 1536 #endif 1537 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1538 seq_puts(seq, ",disable_ext_identify"); 1539 if (test_opt(sbi, INLINE_DATA)) 1540 seq_puts(seq, ",inline_data"); 1541 else 1542 seq_puts(seq, ",noinline_data"); 1543 if (test_opt(sbi, INLINE_DENTRY)) 1544 seq_puts(seq, ",inline_dentry"); 1545 else 1546 seq_puts(seq, ",noinline_dentry"); 1547 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1548 seq_puts(seq, ",flush_merge"); 1549 if (test_opt(sbi, NOBARRIER)) 1550 seq_puts(seq, ",nobarrier"); 1551 if (test_opt(sbi, FASTBOOT)) 1552 seq_puts(seq, ",fastboot"); 1553 if (test_opt(sbi, EXTENT_CACHE)) 1554 seq_puts(seq, ",extent_cache"); 1555 else 1556 seq_puts(seq, ",noextent_cache"); 1557 if (test_opt(sbi, DATA_FLUSH)) 1558 seq_puts(seq, ",data_flush"); 1559 1560 seq_puts(seq, ",mode="); 1561 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1562 seq_puts(seq, "adaptive"); 1563 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1564 seq_puts(seq, "lfs"); 1565 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1566 if (test_opt(sbi, RESERVE_ROOT)) 1567 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1568 F2FS_OPTION(sbi).root_reserved_blocks, 1569 from_kuid_munged(&init_user_ns, 1570 F2FS_OPTION(sbi).s_resuid), 1571 from_kgid_munged(&init_user_ns, 1572 F2FS_OPTION(sbi).s_resgid)); 1573 if (F2FS_IO_SIZE_BITS(sbi)) 1574 seq_printf(seq, ",io_bits=%u", 1575 F2FS_OPTION(sbi).write_io_size_bits); 1576 #ifdef CONFIG_F2FS_FAULT_INJECTION 1577 if (test_opt(sbi, FAULT_INJECTION)) { 1578 seq_printf(seq, ",fault_injection=%u", 1579 F2FS_OPTION(sbi).fault_info.inject_rate); 1580 seq_printf(seq, ",fault_type=%u", 1581 F2FS_OPTION(sbi).fault_info.inject_type); 1582 } 1583 #endif 1584 #ifdef CONFIG_QUOTA 1585 if (test_opt(sbi, QUOTA)) 1586 seq_puts(seq, ",quota"); 1587 if (test_opt(sbi, USRQUOTA)) 1588 seq_puts(seq, ",usrquota"); 1589 if (test_opt(sbi, GRPQUOTA)) 1590 seq_puts(seq, ",grpquota"); 1591 if (test_opt(sbi, PRJQUOTA)) 1592 seq_puts(seq, ",prjquota"); 1593 #endif 1594 f2fs_show_quota_options(seq, sbi->sb); 1595 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1596 seq_printf(seq, ",whint_mode=%s", "user-based"); 1597 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1598 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1599 1600 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1601 1602 if (sbi->sb->s_flags & SB_INLINECRYPT) 1603 seq_puts(seq, ",inlinecrypt"); 1604 1605 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1606 seq_printf(seq, ",alloc_mode=%s", "default"); 1607 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1608 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1609 1610 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1611 seq_printf(seq, ",checkpoint=disable:%u", 1612 F2FS_OPTION(sbi).unusable_cap); 1613 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1614 seq_printf(seq, ",fsync_mode=%s", "posix"); 1615 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1616 seq_printf(seq, ",fsync_mode=%s", "strict"); 1617 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1618 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 1619 1620 f2fs_show_compress_options(seq, sbi->sb); 1621 return 0; 1622 } 1623 1624 static void default_options(struct f2fs_sb_info *sbi) 1625 { 1626 /* init some FS parameters */ 1627 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE; 1628 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 1629 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1630 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1631 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1632 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 1633 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 1634 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 1635 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 1636 F2FS_OPTION(sbi).compress_ext_cnt = 0; 1637 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 1638 1639 sbi->sb->s_flags &= ~SB_INLINECRYPT; 1640 1641 set_opt(sbi, INLINE_XATTR); 1642 set_opt(sbi, INLINE_DATA); 1643 set_opt(sbi, INLINE_DENTRY); 1644 set_opt(sbi, EXTENT_CACHE); 1645 set_opt(sbi, NOHEAP); 1646 clear_opt(sbi, DISABLE_CHECKPOINT); 1647 F2FS_OPTION(sbi).unusable_cap = 0; 1648 sbi->sb->s_flags |= SB_LAZYTIME; 1649 set_opt(sbi, FLUSH_MERGE); 1650 set_opt(sbi, DISCARD); 1651 if (f2fs_sb_has_blkzoned(sbi)) 1652 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 1653 else 1654 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 1655 1656 #ifdef CONFIG_F2FS_FS_XATTR 1657 set_opt(sbi, XATTR_USER); 1658 #endif 1659 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1660 set_opt(sbi, POSIX_ACL); 1661 #endif 1662 1663 f2fs_build_fault_attr(sbi, 0, 0); 1664 } 1665 1666 #ifdef CONFIG_QUOTA 1667 static int f2fs_enable_quotas(struct super_block *sb); 1668 #endif 1669 1670 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 1671 { 1672 unsigned int s_flags = sbi->sb->s_flags; 1673 struct cp_control cpc; 1674 int err = 0; 1675 int ret; 1676 block_t unusable; 1677 1678 if (s_flags & SB_RDONLY) { 1679 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 1680 return -EINVAL; 1681 } 1682 sbi->sb->s_flags |= SB_ACTIVE; 1683 1684 f2fs_update_time(sbi, DISABLE_TIME); 1685 1686 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 1687 down_write(&sbi->gc_lock); 1688 err = f2fs_gc(sbi, true, false, NULL_SEGNO); 1689 if (err == -ENODATA) { 1690 err = 0; 1691 break; 1692 } 1693 if (err && err != -EAGAIN) 1694 break; 1695 } 1696 1697 ret = sync_filesystem(sbi->sb); 1698 if (ret || err) { 1699 err = ret ? ret: err; 1700 goto restore_flag; 1701 } 1702 1703 unusable = f2fs_get_unusable_blocks(sbi); 1704 if (f2fs_disable_cp_again(sbi, unusable)) { 1705 err = -EAGAIN; 1706 goto restore_flag; 1707 } 1708 1709 down_write(&sbi->gc_lock); 1710 cpc.reason = CP_PAUSE; 1711 set_sbi_flag(sbi, SBI_CP_DISABLED); 1712 err = f2fs_write_checkpoint(sbi, &cpc); 1713 if (err) 1714 goto out_unlock; 1715 1716 spin_lock(&sbi->stat_lock); 1717 sbi->unusable_block_count = unusable; 1718 spin_unlock(&sbi->stat_lock); 1719 1720 out_unlock: 1721 up_write(&sbi->gc_lock); 1722 restore_flag: 1723 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 1724 return err; 1725 } 1726 1727 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 1728 { 1729 down_write(&sbi->gc_lock); 1730 f2fs_dirty_to_prefree(sbi); 1731 1732 clear_sbi_flag(sbi, SBI_CP_DISABLED); 1733 set_sbi_flag(sbi, SBI_IS_DIRTY); 1734 up_write(&sbi->gc_lock); 1735 1736 f2fs_sync_fs(sbi->sb, 1); 1737 } 1738 1739 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 1740 { 1741 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1742 struct f2fs_mount_info org_mount_opt; 1743 unsigned long old_sb_flags; 1744 int err; 1745 bool need_restart_gc = false; 1746 bool need_stop_gc = false; 1747 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 1748 bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT); 1749 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 1750 bool checkpoint_changed; 1751 #ifdef CONFIG_QUOTA 1752 int i, j; 1753 #endif 1754 1755 /* 1756 * Save the old mount options in case we 1757 * need to restore them. 1758 */ 1759 org_mount_opt = sbi->mount_opt; 1760 old_sb_flags = sb->s_flags; 1761 1762 #ifdef CONFIG_QUOTA 1763 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 1764 for (i = 0; i < MAXQUOTAS; i++) { 1765 if (F2FS_OPTION(sbi).s_qf_names[i]) { 1766 org_mount_opt.s_qf_names[i] = 1767 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 1768 GFP_KERNEL); 1769 if (!org_mount_opt.s_qf_names[i]) { 1770 for (j = 0; j < i; j++) 1771 kvfree(org_mount_opt.s_qf_names[j]); 1772 return -ENOMEM; 1773 } 1774 } else { 1775 org_mount_opt.s_qf_names[i] = NULL; 1776 } 1777 } 1778 #endif 1779 1780 /* recover superblocks we couldn't write due to previous RO mount */ 1781 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 1782 err = f2fs_commit_super(sbi, false); 1783 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 1784 err); 1785 if (!err) 1786 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1787 } 1788 1789 default_options(sbi); 1790 1791 /* parse mount options */ 1792 err = parse_options(sb, data, true); 1793 if (err) 1794 goto restore_opts; 1795 checkpoint_changed = 1796 disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT); 1797 1798 /* 1799 * Previous and new state of filesystem is RO, 1800 * so skip checking GC and FLUSH_MERGE conditions. 1801 */ 1802 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 1803 goto skip; 1804 1805 #ifdef CONFIG_QUOTA 1806 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 1807 err = dquot_suspend(sb, -1); 1808 if (err < 0) 1809 goto restore_opts; 1810 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 1811 /* dquot_resume needs RW */ 1812 sb->s_flags &= ~SB_RDONLY; 1813 if (sb_any_quota_suspended(sb)) { 1814 dquot_resume(sb, -1); 1815 } else if (f2fs_sb_has_quota_ino(sbi)) { 1816 err = f2fs_enable_quotas(sb); 1817 if (err) 1818 goto restore_opts; 1819 } 1820 } 1821 #endif 1822 /* disallow enable/disable extent_cache dynamically */ 1823 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 1824 err = -EINVAL; 1825 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 1826 goto restore_opts; 1827 } 1828 1829 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 1830 err = -EINVAL; 1831 f2fs_warn(sbi, "switch io_bits option is not allowed"); 1832 goto restore_opts; 1833 } 1834 1835 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 1836 err = -EINVAL; 1837 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 1838 goto restore_opts; 1839 } 1840 1841 /* 1842 * We stop the GC thread if FS is mounted as RO 1843 * or if background_gc = off is passed in mount 1844 * option. Also sync the filesystem. 1845 */ 1846 if ((*flags & SB_RDONLY) || 1847 F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) { 1848 if (sbi->gc_thread) { 1849 f2fs_stop_gc_thread(sbi); 1850 need_restart_gc = true; 1851 } 1852 } else if (!sbi->gc_thread) { 1853 err = f2fs_start_gc_thread(sbi); 1854 if (err) 1855 goto restore_opts; 1856 need_stop_gc = true; 1857 } 1858 1859 if (*flags & SB_RDONLY || 1860 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 1861 writeback_inodes_sb(sb, WB_REASON_SYNC); 1862 sync_inodes_sb(sb); 1863 1864 set_sbi_flag(sbi, SBI_IS_DIRTY); 1865 set_sbi_flag(sbi, SBI_IS_CLOSE); 1866 f2fs_sync_fs(sb, 1); 1867 clear_sbi_flag(sbi, SBI_IS_CLOSE); 1868 } 1869 1870 if (checkpoint_changed) { 1871 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 1872 err = f2fs_disable_checkpoint(sbi); 1873 if (err) 1874 goto restore_gc; 1875 } else { 1876 f2fs_enable_checkpoint(sbi); 1877 } 1878 } 1879 1880 /* 1881 * We stop issue flush thread if FS is mounted as RO 1882 * or if flush_merge is not passed in mount option. 1883 */ 1884 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 1885 clear_opt(sbi, FLUSH_MERGE); 1886 f2fs_destroy_flush_cmd_control(sbi, false); 1887 } else { 1888 err = f2fs_create_flush_cmd_control(sbi); 1889 if (err) 1890 goto restore_gc; 1891 } 1892 skip: 1893 #ifdef CONFIG_QUOTA 1894 /* Release old quota file names */ 1895 for (i = 0; i < MAXQUOTAS; i++) 1896 kvfree(org_mount_opt.s_qf_names[i]); 1897 #endif 1898 /* Update the POSIXACL Flag */ 1899 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 1900 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 1901 1902 limit_reserve_root(sbi); 1903 adjust_unusable_cap_perc(sbi); 1904 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 1905 return 0; 1906 restore_gc: 1907 if (need_restart_gc) { 1908 if (f2fs_start_gc_thread(sbi)) 1909 f2fs_warn(sbi, "background gc thread has stopped"); 1910 } else if (need_stop_gc) { 1911 f2fs_stop_gc_thread(sbi); 1912 } 1913 restore_opts: 1914 #ifdef CONFIG_QUOTA 1915 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 1916 for (i = 0; i < MAXQUOTAS; i++) { 1917 kvfree(F2FS_OPTION(sbi).s_qf_names[i]); 1918 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 1919 } 1920 #endif 1921 sbi->mount_opt = org_mount_opt; 1922 sb->s_flags = old_sb_flags; 1923 return err; 1924 } 1925 1926 #ifdef CONFIG_QUOTA 1927 /* Read data from quotafile */ 1928 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 1929 size_t len, loff_t off) 1930 { 1931 struct inode *inode = sb_dqopt(sb)->files[type]; 1932 struct address_space *mapping = inode->i_mapping; 1933 block_t blkidx = F2FS_BYTES_TO_BLK(off); 1934 int offset = off & (sb->s_blocksize - 1); 1935 int tocopy; 1936 size_t toread; 1937 loff_t i_size = i_size_read(inode); 1938 struct page *page; 1939 char *kaddr; 1940 1941 if (off > i_size) 1942 return 0; 1943 1944 if (off + len > i_size) 1945 len = i_size - off; 1946 toread = len; 1947 while (toread > 0) { 1948 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 1949 repeat: 1950 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 1951 if (IS_ERR(page)) { 1952 if (PTR_ERR(page) == -ENOMEM) { 1953 congestion_wait(BLK_RW_ASYNC, 1954 DEFAULT_IO_TIMEOUT); 1955 goto repeat; 1956 } 1957 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 1958 return PTR_ERR(page); 1959 } 1960 1961 lock_page(page); 1962 1963 if (unlikely(page->mapping != mapping)) { 1964 f2fs_put_page(page, 1); 1965 goto repeat; 1966 } 1967 if (unlikely(!PageUptodate(page))) { 1968 f2fs_put_page(page, 1); 1969 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 1970 return -EIO; 1971 } 1972 1973 kaddr = kmap_atomic(page); 1974 memcpy(data, kaddr + offset, tocopy); 1975 kunmap_atomic(kaddr); 1976 f2fs_put_page(page, 1); 1977 1978 offset = 0; 1979 toread -= tocopy; 1980 data += tocopy; 1981 blkidx++; 1982 } 1983 return len; 1984 } 1985 1986 /* Write to quotafile */ 1987 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 1988 const char *data, size_t len, loff_t off) 1989 { 1990 struct inode *inode = sb_dqopt(sb)->files[type]; 1991 struct address_space *mapping = inode->i_mapping; 1992 const struct address_space_operations *a_ops = mapping->a_ops; 1993 int offset = off & (sb->s_blocksize - 1); 1994 size_t towrite = len; 1995 struct page *page; 1996 void *fsdata = NULL; 1997 char *kaddr; 1998 int err = 0; 1999 int tocopy; 2000 2001 while (towrite > 0) { 2002 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2003 towrite); 2004 retry: 2005 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2006 &page, &fsdata); 2007 if (unlikely(err)) { 2008 if (err == -ENOMEM) { 2009 congestion_wait(BLK_RW_ASYNC, 2010 DEFAULT_IO_TIMEOUT); 2011 goto retry; 2012 } 2013 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2014 break; 2015 } 2016 2017 kaddr = kmap_atomic(page); 2018 memcpy(kaddr + offset, data, tocopy); 2019 kunmap_atomic(kaddr); 2020 flush_dcache_page(page); 2021 2022 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2023 page, fsdata); 2024 offset = 0; 2025 towrite -= tocopy; 2026 off += tocopy; 2027 data += tocopy; 2028 cond_resched(); 2029 } 2030 2031 if (len == towrite) 2032 return err; 2033 inode->i_mtime = inode->i_ctime = current_time(inode); 2034 f2fs_mark_inode_dirty_sync(inode, false); 2035 return len - towrite; 2036 } 2037 2038 static struct dquot **f2fs_get_dquots(struct inode *inode) 2039 { 2040 return F2FS_I(inode)->i_dquot; 2041 } 2042 2043 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2044 { 2045 return &F2FS_I(inode)->i_reserved_quota; 2046 } 2047 2048 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2049 { 2050 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2051 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2052 return 0; 2053 } 2054 2055 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2056 F2FS_OPTION(sbi).s_jquota_fmt, type); 2057 } 2058 2059 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2060 { 2061 int enabled = 0; 2062 int i, err; 2063 2064 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2065 err = f2fs_enable_quotas(sbi->sb); 2066 if (err) { 2067 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2068 return 0; 2069 } 2070 return 1; 2071 } 2072 2073 for (i = 0; i < MAXQUOTAS; i++) { 2074 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2075 err = f2fs_quota_on_mount(sbi, i); 2076 if (!err) { 2077 enabled = 1; 2078 continue; 2079 } 2080 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2081 err, i); 2082 } 2083 } 2084 return enabled; 2085 } 2086 2087 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2088 unsigned int flags) 2089 { 2090 struct inode *qf_inode; 2091 unsigned long qf_inum; 2092 int err; 2093 2094 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2095 2096 qf_inum = f2fs_qf_ino(sb, type); 2097 if (!qf_inum) 2098 return -EPERM; 2099 2100 qf_inode = f2fs_iget(sb, qf_inum); 2101 if (IS_ERR(qf_inode)) { 2102 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2103 return PTR_ERR(qf_inode); 2104 } 2105 2106 /* Don't account quota for quota files to avoid recursion */ 2107 qf_inode->i_flags |= S_NOQUOTA; 2108 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2109 iput(qf_inode); 2110 return err; 2111 } 2112 2113 static int f2fs_enable_quotas(struct super_block *sb) 2114 { 2115 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2116 int type, err = 0; 2117 unsigned long qf_inum; 2118 bool quota_mopt[MAXQUOTAS] = { 2119 test_opt(sbi, USRQUOTA), 2120 test_opt(sbi, GRPQUOTA), 2121 test_opt(sbi, PRJQUOTA), 2122 }; 2123 2124 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2125 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2126 return 0; 2127 } 2128 2129 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2130 2131 for (type = 0; type < MAXQUOTAS; type++) { 2132 qf_inum = f2fs_qf_ino(sb, type); 2133 if (qf_inum) { 2134 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2135 DQUOT_USAGE_ENABLED | 2136 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2137 if (err) { 2138 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2139 type, err); 2140 for (type--; type >= 0; type--) 2141 dquot_quota_off(sb, type); 2142 set_sbi_flag(F2FS_SB(sb), 2143 SBI_QUOTA_NEED_REPAIR); 2144 return err; 2145 } 2146 } 2147 } 2148 return 0; 2149 } 2150 2151 int f2fs_quota_sync(struct super_block *sb, int type) 2152 { 2153 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2154 struct quota_info *dqopt = sb_dqopt(sb); 2155 int cnt; 2156 int ret; 2157 2158 /* 2159 * do_quotactl 2160 * f2fs_quota_sync 2161 * down_read(quota_sem) 2162 * dquot_writeback_dquots() 2163 * f2fs_dquot_commit 2164 * block_operation 2165 * down_read(quota_sem) 2166 */ 2167 f2fs_lock_op(sbi); 2168 2169 down_read(&sbi->quota_sem); 2170 ret = dquot_writeback_dquots(sb, type); 2171 if (ret) 2172 goto out; 2173 2174 /* 2175 * Now when everything is written we can discard the pagecache so 2176 * that userspace sees the changes. 2177 */ 2178 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2179 struct address_space *mapping; 2180 2181 if (type != -1 && cnt != type) 2182 continue; 2183 if (!sb_has_quota_active(sb, cnt)) 2184 continue; 2185 2186 mapping = dqopt->files[cnt]->i_mapping; 2187 2188 ret = filemap_fdatawrite(mapping); 2189 if (ret) 2190 goto out; 2191 2192 /* if we are using journalled quota */ 2193 if (is_journalled_quota(sbi)) 2194 continue; 2195 2196 ret = filemap_fdatawait(mapping); 2197 if (ret) 2198 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2199 2200 inode_lock(dqopt->files[cnt]); 2201 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0); 2202 inode_unlock(dqopt->files[cnt]); 2203 } 2204 out: 2205 if (ret) 2206 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2207 up_read(&sbi->quota_sem); 2208 f2fs_unlock_op(sbi); 2209 return ret; 2210 } 2211 2212 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2213 const struct path *path) 2214 { 2215 struct inode *inode; 2216 int err; 2217 2218 /* if quota sysfile exists, deny enabling quota with specific file */ 2219 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2220 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2221 return -EBUSY; 2222 } 2223 2224 err = f2fs_quota_sync(sb, type); 2225 if (err) 2226 return err; 2227 2228 err = dquot_quota_on(sb, type, format_id, path); 2229 if (err) 2230 return err; 2231 2232 inode = d_inode(path->dentry); 2233 2234 inode_lock(inode); 2235 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2236 f2fs_set_inode_flags(inode); 2237 inode_unlock(inode); 2238 f2fs_mark_inode_dirty_sync(inode, false); 2239 2240 return 0; 2241 } 2242 2243 static int __f2fs_quota_off(struct super_block *sb, int type) 2244 { 2245 struct inode *inode = sb_dqopt(sb)->files[type]; 2246 int err; 2247 2248 if (!inode || !igrab(inode)) 2249 return dquot_quota_off(sb, type); 2250 2251 err = f2fs_quota_sync(sb, type); 2252 if (err) 2253 goto out_put; 2254 2255 err = dquot_quota_off(sb, type); 2256 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2257 goto out_put; 2258 2259 inode_lock(inode); 2260 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2261 f2fs_set_inode_flags(inode); 2262 inode_unlock(inode); 2263 f2fs_mark_inode_dirty_sync(inode, false); 2264 out_put: 2265 iput(inode); 2266 return err; 2267 } 2268 2269 static int f2fs_quota_off(struct super_block *sb, int type) 2270 { 2271 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2272 int err; 2273 2274 err = __f2fs_quota_off(sb, type); 2275 2276 /* 2277 * quotactl can shutdown journalled quota, result in inconsistence 2278 * between quota record and fs data by following updates, tag the 2279 * flag to let fsck be aware of it. 2280 */ 2281 if (is_journalled_quota(sbi)) 2282 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2283 return err; 2284 } 2285 2286 void f2fs_quota_off_umount(struct super_block *sb) 2287 { 2288 int type; 2289 int err; 2290 2291 for (type = 0; type < MAXQUOTAS; type++) { 2292 err = __f2fs_quota_off(sb, type); 2293 if (err) { 2294 int ret = dquot_quota_off(sb, type); 2295 2296 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2297 type, err, ret); 2298 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2299 } 2300 } 2301 /* 2302 * In case of checkpoint=disable, we must flush quota blocks. 2303 * This can cause NULL exception for node_inode in end_io, since 2304 * put_super already dropped it. 2305 */ 2306 sync_filesystem(sb); 2307 } 2308 2309 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2310 { 2311 struct quota_info *dqopt = sb_dqopt(sb); 2312 int type; 2313 2314 for (type = 0; type < MAXQUOTAS; type++) { 2315 if (!dqopt->files[type]) 2316 continue; 2317 f2fs_inode_synced(dqopt->files[type]); 2318 } 2319 } 2320 2321 static int f2fs_dquot_commit(struct dquot *dquot) 2322 { 2323 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2324 int ret; 2325 2326 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2327 ret = dquot_commit(dquot); 2328 if (ret < 0) 2329 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2330 up_read(&sbi->quota_sem); 2331 return ret; 2332 } 2333 2334 static int f2fs_dquot_acquire(struct dquot *dquot) 2335 { 2336 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2337 int ret; 2338 2339 down_read(&sbi->quota_sem); 2340 ret = dquot_acquire(dquot); 2341 if (ret < 0) 2342 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2343 up_read(&sbi->quota_sem); 2344 return ret; 2345 } 2346 2347 static int f2fs_dquot_release(struct dquot *dquot) 2348 { 2349 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2350 int ret = dquot_release(dquot); 2351 2352 if (ret < 0) 2353 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2354 return ret; 2355 } 2356 2357 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2358 { 2359 struct super_block *sb = dquot->dq_sb; 2360 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2361 int ret = dquot_mark_dquot_dirty(dquot); 2362 2363 /* if we are using journalled quota */ 2364 if (is_journalled_quota(sbi)) 2365 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2366 2367 return ret; 2368 } 2369 2370 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2371 { 2372 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2373 int ret = dquot_commit_info(sb, type); 2374 2375 if (ret < 0) 2376 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2377 return ret; 2378 } 2379 2380 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2381 { 2382 *projid = F2FS_I(inode)->i_projid; 2383 return 0; 2384 } 2385 2386 static const struct dquot_operations f2fs_quota_operations = { 2387 .get_reserved_space = f2fs_get_reserved_space, 2388 .write_dquot = f2fs_dquot_commit, 2389 .acquire_dquot = f2fs_dquot_acquire, 2390 .release_dquot = f2fs_dquot_release, 2391 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2392 .write_info = f2fs_dquot_commit_info, 2393 .alloc_dquot = dquot_alloc, 2394 .destroy_dquot = dquot_destroy, 2395 .get_projid = f2fs_get_projid, 2396 .get_next_id = dquot_get_next_id, 2397 }; 2398 2399 static const struct quotactl_ops f2fs_quotactl_ops = { 2400 .quota_on = f2fs_quota_on, 2401 .quota_off = f2fs_quota_off, 2402 .quota_sync = f2fs_quota_sync, 2403 .get_state = dquot_get_state, 2404 .set_info = dquot_set_dqinfo, 2405 .get_dqblk = dquot_get_dqblk, 2406 .set_dqblk = dquot_set_dqblk, 2407 .get_nextdqblk = dquot_get_next_dqblk, 2408 }; 2409 #else 2410 int f2fs_quota_sync(struct super_block *sb, int type) 2411 { 2412 return 0; 2413 } 2414 2415 void f2fs_quota_off_umount(struct super_block *sb) 2416 { 2417 } 2418 #endif 2419 2420 static const struct super_operations f2fs_sops = { 2421 .alloc_inode = f2fs_alloc_inode, 2422 .free_inode = f2fs_free_inode, 2423 .drop_inode = f2fs_drop_inode, 2424 .write_inode = f2fs_write_inode, 2425 .dirty_inode = f2fs_dirty_inode, 2426 .show_options = f2fs_show_options, 2427 #ifdef CONFIG_QUOTA 2428 .quota_read = f2fs_quota_read, 2429 .quota_write = f2fs_quota_write, 2430 .get_dquots = f2fs_get_dquots, 2431 #endif 2432 .evict_inode = f2fs_evict_inode, 2433 .put_super = f2fs_put_super, 2434 .sync_fs = f2fs_sync_fs, 2435 .freeze_fs = f2fs_freeze, 2436 .unfreeze_fs = f2fs_unfreeze, 2437 .statfs = f2fs_statfs, 2438 .remount_fs = f2fs_remount, 2439 }; 2440 2441 #ifdef CONFIG_FS_ENCRYPTION 2442 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2443 { 2444 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2445 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2446 ctx, len, NULL); 2447 } 2448 2449 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2450 void *fs_data) 2451 { 2452 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2453 2454 /* 2455 * Encrypting the root directory is not allowed because fsck 2456 * expects lost+found directory to exist and remain unencrypted 2457 * if LOST_FOUND feature is enabled. 2458 * 2459 */ 2460 if (f2fs_sb_has_lost_found(sbi) && 2461 inode->i_ino == F2FS_ROOT_INO(sbi)) 2462 return -EPERM; 2463 2464 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2465 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2466 ctx, len, fs_data, XATTR_CREATE); 2467 } 2468 2469 static const union fscrypt_context * 2470 f2fs_get_dummy_context(struct super_block *sb) 2471 { 2472 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_ctx.ctx; 2473 } 2474 2475 static bool f2fs_has_stable_inodes(struct super_block *sb) 2476 { 2477 return true; 2478 } 2479 2480 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 2481 int *ino_bits_ret, int *lblk_bits_ret) 2482 { 2483 *ino_bits_ret = 8 * sizeof(nid_t); 2484 *lblk_bits_ret = 8 * sizeof(block_t); 2485 } 2486 2487 static int f2fs_get_num_devices(struct super_block *sb) 2488 { 2489 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2490 2491 if (f2fs_is_multi_device(sbi)) 2492 return sbi->s_ndevs; 2493 return 1; 2494 } 2495 2496 static void f2fs_get_devices(struct super_block *sb, 2497 struct request_queue **devs) 2498 { 2499 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2500 int i; 2501 2502 for (i = 0; i < sbi->s_ndevs; i++) 2503 devs[i] = bdev_get_queue(FDEV(i).bdev); 2504 } 2505 2506 static const struct fscrypt_operations f2fs_cryptops = { 2507 .key_prefix = "f2fs:", 2508 .get_context = f2fs_get_context, 2509 .set_context = f2fs_set_context, 2510 .get_dummy_context = f2fs_get_dummy_context, 2511 .empty_dir = f2fs_empty_dir, 2512 .max_namelen = F2FS_NAME_LEN, 2513 .has_stable_inodes = f2fs_has_stable_inodes, 2514 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 2515 .get_num_devices = f2fs_get_num_devices, 2516 .get_devices = f2fs_get_devices, 2517 }; 2518 #endif 2519 2520 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 2521 u64 ino, u32 generation) 2522 { 2523 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2524 struct inode *inode; 2525 2526 if (f2fs_check_nid_range(sbi, ino)) 2527 return ERR_PTR(-ESTALE); 2528 2529 /* 2530 * f2fs_iget isn't quite right if the inode is currently unallocated! 2531 * However f2fs_iget currently does appropriate checks to handle stale 2532 * inodes so everything is OK. 2533 */ 2534 inode = f2fs_iget(sb, ino); 2535 if (IS_ERR(inode)) 2536 return ERR_CAST(inode); 2537 if (unlikely(generation && inode->i_generation != generation)) { 2538 /* we didn't find the right inode.. */ 2539 iput(inode); 2540 return ERR_PTR(-ESTALE); 2541 } 2542 return inode; 2543 } 2544 2545 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 2546 int fh_len, int fh_type) 2547 { 2548 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 2549 f2fs_nfs_get_inode); 2550 } 2551 2552 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 2553 int fh_len, int fh_type) 2554 { 2555 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 2556 f2fs_nfs_get_inode); 2557 } 2558 2559 static const struct export_operations f2fs_export_ops = { 2560 .fh_to_dentry = f2fs_fh_to_dentry, 2561 .fh_to_parent = f2fs_fh_to_parent, 2562 .get_parent = f2fs_get_parent, 2563 }; 2564 2565 static loff_t max_file_blocks(void) 2566 { 2567 loff_t result = 0; 2568 loff_t leaf_count = DEF_ADDRS_PER_BLOCK; 2569 2570 /* 2571 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 2572 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 2573 * space in inode.i_addr, it will be more safe to reassign 2574 * result as zero. 2575 */ 2576 2577 /* two direct node blocks */ 2578 result += (leaf_count * 2); 2579 2580 /* two indirect node blocks */ 2581 leaf_count *= NIDS_PER_BLOCK; 2582 result += (leaf_count * 2); 2583 2584 /* one double indirect node block */ 2585 leaf_count *= NIDS_PER_BLOCK; 2586 result += leaf_count; 2587 2588 return result; 2589 } 2590 2591 static int __f2fs_commit_super(struct buffer_head *bh, 2592 struct f2fs_super_block *super) 2593 { 2594 lock_buffer(bh); 2595 if (super) 2596 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 2597 set_buffer_dirty(bh); 2598 unlock_buffer(bh); 2599 2600 /* it's rare case, we can do fua all the time */ 2601 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 2602 } 2603 2604 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 2605 struct buffer_head *bh) 2606 { 2607 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2608 (bh->b_data + F2FS_SUPER_OFFSET); 2609 struct super_block *sb = sbi->sb; 2610 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 2611 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 2612 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 2613 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 2614 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 2615 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 2616 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 2617 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 2618 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 2619 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 2620 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2621 u32 segment_count = le32_to_cpu(raw_super->segment_count); 2622 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2623 u64 main_end_blkaddr = main_blkaddr + 2624 (segment_count_main << log_blocks_per_seg); 2625 u64 seg_end_blkaddr = segment0_blkaddr + 2626 (segment_count << log_blocks_per_seg); 2627 2628 if (segment0_blkaddr != cp_blkaddr) { 2629 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 2630 segment0_blkaddr, cp_blkaddr); 2631 return true; 2632 } 2633 2634 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 2635 sit_blkaddr) { 2636 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 2637 cp_blkaddr, sit_blkaddr, 2638 segment_count_ckpt << log_blocks_per_seg); 2639 return true; 2640 } 2641 2642 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 2643 nat_blkaddr) { 2644 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 2645 sit_blkaddr, nat_blkaddr, 2646 segment_count_sit << log_blocks_per_seg); 2647 return true; 2648 } 2649 2650 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 2651 ssa_blkaddr) { 2652 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 2653 nat_blkaddr, ssa_blkaddr, 2654 segment_count_nat << log_blocks_per_seg); 2655 return true; 2656 } 2657 2658 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 2659 main_blkaddr) { 2660 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 2661 ssa_blkaddr, main_blkaddr, 2662 segment_count_ssa << log_blocks_per_seg); 2663 return true; 2664 } 2665 2666 if (main_end_blkaddr > seg_end_blkaddr) { 2667 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)", 2668 main_blkaddr, 2669 segment0_blkaddr + 2670 (segment_count << log_blocks_per_seg), 2671 segment_count_main << log_blocks_per_seg); 2672 return true; 2673 } else if (main_end_blkaddr < seg_end_blkaddr) { 2674 int err = 0; 2675 char *res; 2676 2677 /* fix in-memory information all the time */ 2678 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 2679 segment0_blkaddr) >> log_blocks_per_seg); 2680 2681 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 2682 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2683 res = "internally"; 2684 } else { 2685 err = __f2fs_commit_super(bh, NULL); 2686 res = err ? "failed" : "done"; 2687 } 2688 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)", 2689 res, main_blkaddr, 2690 segment0_blkaddr + 2691 (segment_count << log_blocks_per_seg), 2692 segment_count_main << log_blocks_per_seg); 2693 if (err) 2694 return true; 2695 } 2696 return false; 2697 } 2698 2699 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 2700 struct buffer_head *bh) 2701 { 2702 block_t segment_count, segs_per_sec, secs_per_zone; 2703 block_t total_sections, blocks_per_seg; 2704 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2705 (bh->b_data + F2FS_SUPER_OFFSET); 2706 unsigned int blocksize; 2707 size_t crc_offset = 0; 2708 __u32 crc = 0; 2709 2710 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 2711 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 2712 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 2713 return -EINVAL; 2714 } 2715 2716 /* Check checksum_offset and crc in superblock */ 2717 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 2718 crc_offset = le32_to_cpu(raw_super->checksum_offset); 2719 if (crc_offset != 2720 offsetof(struct f2fs_super_block, crc)) { 2721 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 2722 crc_offset); 2723 return -EFSCORRUPTED; 2724 } 2725 crc = le32_to_cpu(raw_super->crc); 2726 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 2727 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 2728 return -EFSCORRUPTED; 2729 } 2730 } 2731 2732 /* Currently, support only 4KB page cache size */ 2733 if (F2FS_BLKSIZE != PAGE_SIZE) { 2734 f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB", 2735 PAGE_SIZE); 2736 return -EFSCORRUPTED; 2737 } 2738 2739 /* Currently, support only 4KB block size */ 2740 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); 2741 if (blocksize != F2FS_BLKSIZE) { 2742 f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB", 2743 blocksize); 2744 return -EFSCORRUPTED; 2745 } 2746 2747 /* check log blocks per segment */ 2748 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 2749 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 2750 le32_to_cpu(raw_super->log_blocks_per_seg)); 2751 return -EFSCORRUPTED; 2752 } 2753 2754 /* Currently, support 512/1024/2048/4096 bytes sector size */ 2755 if (le32_to_cpu(raw_super->log_sectorsize) > 2756 F2FS_MAX_LOG_SECTOR_SIZE || 2757 le32_to_cpu(raw_super->log_sectorsize) < 2758 F2FS_MIN_LOG_SECTOR_SIZE) { 2759 f2fs_info(sbi, "Invalid log sectorsize (%u)", 2760 le32_to_cpu(raw_super->log_sectorsize)); 2761 return -EFSCORRUPTED; 2762 } 2763 if (le32_to_cpu(raw_super->log_sectors_per_block) + 2764 le32_to_cpu(raw_super->log_sectorsize) != 2765 F2FS_MAX_LOG_SECTOR_SIZE) { 2766 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 2767 le32_to_cpu(raw_super->log_sectors_per_block), 2768 le32_to_cpu(raw_super->log_sectorsize)); 2769 return -EFSCORRUPTED; 2770 } 2771 2772 segment_count = le32_to_cpu(raw_super->segment_count); 2773 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 2774 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 2775 total_sections = le32_to_cpu(raw_super->section_count); 2776 2777 /* blocks_per_seg should be 512, given the above check */ 2778 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 2779 2780 if (segment_count > F2FS_MAX_SEGMENT || 2781 segment_count < F2FS_MIN_SEGMENTS) { 2782 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 2783 return -EFSCORRUPTED; 2784 } 2785 2786 if (total_sections > segment_count || 2787 total_sections < F2FS_MIN_SEGMENTS || 2788 segs_per_sec > segment_count || !segs_per_sec) { 2789 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 2790 segment_count, total_sections, segs_per_sec); 2791 return -EFSCORRUPTED; 2792 } 2793 2794 if ((segment_count / segs_per_sec) < total_sections) { 2795 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 2796 segment_count, segs_per_sec, total_sections); 2797 return -EFSCORRUPTED; 2798 } 2799 2800 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 2801 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 2802 segment_count, le64_to_cpu(raw_super->block_count)); 2803 return -EFSCORRUPTED; 2804 } 2805 2806 if (RDEV(0).path[0]) { 2807 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 2808 int i = 1; 2809 2810 while (i < MAX_DEVICES && RDEV(i).path[0]) { 2811 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 2812 i++; 2813 } 2814 if (segment_count != dev_seg_count) { 2815 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 2816 segment_count, dev_seg_count); 2817 return -EFSCORRUPTED; 2818 } 2819 } 2820 2821 if (secs_per_zone > total_sections || !secs_per_zone) { 2822 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 2823 secs_per_zone, total_sections); 2824 return -EFSCORRUPTED; 2825 } 2826 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 2827 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 2828 (le32_to_cpu(raw_super->extension_count) + 2829 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 2830 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 2831 le32_to_cpu(raw_super->extension_count), 2832 raw_super->hot_ext_count, 2833 F2FS_MAX_EXTENSION); 2834 return -EFSCORRUPTED; 2835 } 2836 2837 if (le32_to_cpu(raw_super->cp_payload) > 2838 (blocks_per_seg - F2FS_CP_PACKS)) { 2839 f2fs_info(sbi, "Insane cp_payload (%u > %u)", 2840 le32_to_cpu(raw_super->cp_payload), 2841 blocks_per_seg - F2FS_CP_PACKS); 2842 return -EFSCORRUPTED; 2843 } 2844 2845 /* check reserved ino info */ 2846 if (le32_to_cpu(raw_super->node_ino) != 1 || 2847 le32_to_cpu(raw_super->meta_ino) != 2 || 2848 le32_to_cpu(raw_super->root_ino) != 3) { 2849 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 2850 le32_to_cpu(raw_super->node_ino), 2851 le32_to_cpu(raw_super->meta_ino), 2852 le32_to_cpu(raw_super->root_ino)); 2853 return -EFSCORRUPTED; 2854 } 2855 2856 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 2857 if (sanity_check_area_boundary(sbi, bh)) 2858 return -EFSCORRUPTED; 2859 2860 return 0; 2861 } 2862 2863 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 2864 { 2865 unsigned int total, fsmeta; 2866 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 2867 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 2868 unsigned int ovp_segments, reserved_segments; 2869 unsigned int main_segs, blocks_per_seg; 2870 unsigned int sit_segs, nat_segs; 2871 unsigned int sit_bitmap_size, nat_bitmap_size; 2872 unsigned int log_blocks_per_seg; 2873 unsigned int segment_count_main; 2874 unsigned int cp_pack_start_sum, cp_payload; 2875 block_t user_block_count, valid_user_blocks; 2876 block_t avail_node_count, valid_node_count; 2877 int i, j; 2878 2879 total = le32_to_cpu(raw_super->segment_count); 2880 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 2881 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 2882 fsmeta += sit_segs; 2883 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 2884 fsmeta += nat_segs; 2885 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 2886 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 2887 2888 if (unlikely(fsmeta >= total)) 2889 return 1; 2890 2891 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 2892 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 2893 2894 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS || 2895 ovp_segments == 0 || reserved_segments == 0)) { 2896 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 2897 return 1; 2898 } 2899 2900 user_block_count = le64_to_cpu(ckpt->user_block_count); 2901 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2902 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2903 if (!user_block_count || user_block_count >= 2904 segment_count_main << log_blocks_per_seg) { 2905 f2fs_err(sbi, "Wrong user_block_count: %u", 2906 user_block_count); 2907 return 1; 2908 } 2909 2910 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 2911 if (valid_user_blocks > user_block_count) { 2912 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 2913 valid_user_blocks, user_block_count); 2914 return 1; 2915 } 2916 2917 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 2918 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 2919 if (valid_node_count > avail_node_count) { 2920 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 2921 valid_node_count, avail_node_count); 2922 return 1; 2923 } 2924 2925 main_segs = le32_to_cpu(raw_super->segment_count_main); 2926 blocks_per_seg = sbi->blocks_per_seg; 2927 2928 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 2929 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 2930 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 2931 return 1; 2932 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 2933 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 2934 le32_to_cpu(ckpt->cur_node_segno[j])) { 2935 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 2936 i, j, 2937 le32_to_cpu(ckpt->cur_node_segno[i])); 2938 return 1; 2939 } 2940 } 2941 } 2942 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 2943 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 2944 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 2945 return 1; 2946 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 2947 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 2948 le32_to_cpu(ckpt->cur_data_segno[j])) { 2949 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 2950 i, j, 2951 le32_to_cpu(ckpt->cur_data_segno[i])); 2952 return 1; 2953 } 2954 } 2955 } 2956 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 2957 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 2958 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 2959 le32_to_cpu(ckpt->cur_data_segno[j])) { 2960 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 2961 i, j, 2962 le32_to_cpu(ckpt->cur_node_segno[i])); 2963 return 1; 2964 } 2965 } 2966 } 2967 2968 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 2969 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 2970 2971 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 2972 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 2973 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 2974 sit_bitmap_size, nat_bitmap_size); 2975 return 1; 2976 } 2977 2978 cp_pack_start_sum = __start_sum_addr(sbi); 2979 cp_payload = __cp_payload(sbi); 2980 if (cp_pack_start_sum < cp_payload + 1 || 2981 cp_pack_start_sum > blocks_per_seg - 1 - 2982 NR_CURSEG_TYPE) { 2983 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 2984 cp_pack_start_sum); 2985 return 1; 2986 } 2987 2988 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 2989 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 2990 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 2991 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 2992 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 2993 le32_to_cpu(ckpt->checksum_offset)); 2994 return 1; 2995 } 2996 2997 if (unlikely(f2fs_cp_error(sbi))) { 2998 f2fs_err(sbi, "A bug case: need to run fsck"); 2999 return 1; 3000 } 3001 return 0; 3002 } 3003 3004 static void init_sb_info(struct f2fs_sb_info *sbi) 3005 { 3006 struct f2fs_super_block *raw_super = sbi->raw_super; 3007 int i; 3008 3009 sbi->log_sectors_per_block = 3010 le32_to_cpu(raw_super->log_sectors_per_block); 3011 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3012 sbi->blocksize = 1 << sbi->log_blocksize; 3013 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3014 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3015 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3016 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3017 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3018 sbi->total_node_count = 3019 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3020 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3021 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); 3022 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); 3023 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); 3024 sbi->cur_victim_sec = NULL_SECNO; 3025 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3026 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3027 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3028 sbi->migration_granularity = sbi->segs_per_sec; 3029 3030 sbi->dir_level = DEF_DIR_LEVEL; 3031 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3032 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3033 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3034 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3035 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3036 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3037 DEF_UMOUNT_DISCARD_TIMEOUT; 3038 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3039 3040 for (i = 0; i < NR_COUNT_TYPE; i++) 3041 atomic_set(&sbi->nr_pages[i], 0); 3042 3043 for (i = 0; i < META; i++) 3044 atomic_set(&sbi->wb_sync_req[i], 0); 3045 3046 INIT_LIST_HEAD(&sbi->s_list); 3047 mutex_init(&sbi->umount_mutex); 3048 init_rwsem(&sbi->io_order_lock); 3049 spin_lock_init(&sbi->cp_lock); 3050 3051 sbi->dirty_device = 0; 3052 spin_lock_init(&sbi->dev_lock); 3053 3054 init_rwsem(&sbi->sb_lock); 3055 init_rwsem(&sbi->pin_sem); 3056 } 3057 3058 static int init_percpu_info(struct f2fs_sb_info *sbi) 3059 { 3060 int err; 3061 3062 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3063 if (err) 3064 return err; 3065 3066 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3067 GFP_KERNEL); 3068 if (err) 3069 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3070 3071 return err; 3072 } 3073 3074 #ifdef CONFIG_BLK_DEV_ZONED 3075 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3076 void *data) 3077 { 3078 struct f2fs_dev_info *dev = data; 3079 3080 if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL) 3081 set_bit(idx, dev->blkz_seq); 3082 return 0; 3083 } 3084 3085 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3086 { 3087 struct block_device *bdev = FDEV(devi).bdev; 3088 sector_t nr_sectors = bdev->bd_part->nr_sects; 3089 int ret; 3090 3091 if (!f2fs_sb_has_blkzoned(sbi)) 3092 return 0; 3093 3094 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3095 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 3096 return -EINVAL; 3097 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 3098 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3099 __ilog2_u32(sbi->blocks_per_blkz)) 3100 return -EINVAL; 3101 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3102 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3103 sbi->log_blocks_per_blkz; 3104 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 3105 FDEV(devi).nr_blkz++; 3106 3107 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3108 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3109 * sizeof(unsigned long), 3110 GFP_KERNEL); 3111 if (!FDEV(devi).blkz_seq) 3112 return -ENOMEM; 3113 3114 /* Get block zones type */ 3115 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3116 &FDEV(devi)); 3117 if (ret < 0) 3118 return ret; 3119 3120 return 0; 3121 } 3122 #endif 3123 3124 /* 3125 * Read f2fs raw super block. 3126 * Because we have two copies of super block, so read both of them 3127 * to get the first valid one. If any one of them is broken, we pass 3128 * them recovery flag back to the caller. 3129 */ 3130 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3131 struct f2fs_super_block **raw_super, 3132 int *valid_super_block, int *recovery) 3133 { 3134 struct super_block *sb = sbi->sb; 3135 int block; 3136 struct buffer_head *bh; 3137 struct f2fs_super_block *super; 3138 int err = 0; 3139 3140 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3141 if (!super) 3142 return -ENOMEM; 3143 3144 for (block = 0; block < 2; block++) { 3145 bh = sb_bread(sb, block); 3146 if (!bh) { 3147 f2fs_err(sbi, "Unable to read %dth superblock", 3148 block + 1); 3149 err = -EIO; 3150 *recovery = 1; 3151 continue; 3152 } 3153 3154 /* sanity checking of raw super */ 3155 err = sanity_check_raw_super(sbi, bh); 3156 if (err) { 3157 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3158 block + 1); 3159 brelse(bh); 3160 *recovery = 1; 3161 continue; 3162 } 3163 3164 if (!*raw_super) { 3165 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3166 sizeof(*super)); 3167 *valid_super_block = block; 3168 *raw_super = super; 3169 } 3170 brelse(bh); 3171 } 3172 3173 /* No valid superblock */ 3174 if (!*raw_super) 3175 kvfree(super); 3176 else 3177 err = 0; 3178 3179 return err; 3180 } 3181 3182 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3183 { 3184 struct buffer_head *bh; 3185 __u32 crc = 0; 3186 int err; 3187 3188 if ((recover && f2fs_readonly(sbi->sb)) || 3189 bdev_read_only(sbi->sb->s_bdev)) { 3190 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3191 return -EROFS; 3192 } 3193 3194 /* we should update superblock crc here */ 3195 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3196 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3197 offsetof(struct f2fs_super_block, crc)); 3198 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3199 } 3200 3201 /* write back-up superblock first */ 3202 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3203 if (!bh) 3204 return -EIO; 3205 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3206 brelse(bh); 3207 3208 /* if we are in recovery path, skip writing valid superblock */ 3209 if (recover || err) 3210 return err; 3211 3212 /* write current valid superblock */ 3213 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3214 if (!bh) 3215 return -EIO; 3216 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3217 brelse(bh); 3218 return err; 3219 } 3220 3221 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3222 { 3223 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3224 unsigned int max_devices = MAX_DEVICES; 3225 int i; 3226 3227 /* Initialize single device information */ 3228 if (!RDEV(0).path[0]) { 3229 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3230 return 0; 3231 max_devices = 1; 3232 } 3233 3234 /* 3235 * Initialize multiple devices information, or single 3236 * zoned block device information. 3237 */ 3238 sbi->devs = f2fs_kzalloc(sbi, 3239 array_size(max_devices, 3240 sizeof(struct f2fs_dev_info)), 3241 GFP_KERNEL); 3242 if (!sbi->devs) 3243 return -ENOMEM; 3244 3245 for (i = 0; i < max_devices; i++) { 3246 3247 if (i > 0 && !RDEV(i).path[0]) 3248 break; 3249 3250 if (max_devices == 1) { 3251 /* Single zoned block device mount */ 3252 FDEV(0).bdev = 3253 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3254 sbi->sb->s_mode, sbi->sb->s_type); 3255 } else { 3256 /* Multi-device mount */ 3257 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3258 FDEV(i).total_segments = 3259 le32_to_cpu(RDEV(i).total_segments); 3260 if (i == 0) { 3261 FDEV(i).start_blk = 0; 3262 FDEV(i).end_blk = FDEV(i).start_blk + 3263 (FDEV(i).total_segments << 3264 sbi->log_blocks_per_seg) - 1 + 3265 le32_to_cpu(raw_super->segment0_blkaddr); 3266 } else { 3267 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3268 FDEV(i).end_blk = FDEV(i).start_blk + 3269 (FDEV(i).total_segments << 3270 sbi->log_blocks_per_seg) - 1; 3271 } 3272 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3273 sbi->sb->s_mode, sbi->sb->s_type); 3274 } 3275 if (IS_ERR(FDEV(i).bdev)) 3276 return PTR_ERR(FDEV(i).bdev); 3277 3278 /* to release errored devices */ 3279 sbi->s_ndevs = i + 1; 3280 3281 #ifdef CONFIG_BLK_DEV_ZONED 3282 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3283 !f2fs_sb_has_blkzoned(sbi)) { 3284 f2fs_err(sbi, "Zoned block device feature not enabled\n"); 3285 return -EINVAL; 3286 } 3287 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3288 if (init_blkz_info(sbi, i)) { 3289 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3290 return -EINVAL; 3291 } 3292 if (max_devices == 1) 3293 break; 3294 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3295 i, FDEV(i).path, 3296 FDEV(i).total_segments, 3297 FDEV(i).start_blk, FDEV(i).end_blk, 3298 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3299 "Host-aware" : "Host-managed"); 3300 continue; 3301 } 3302 #endif 3303 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3304 i, FDEV(i).path, 3305 FDEV(i).total_segments, 3306 FDEV(i).start_blk, FDEV(i).end_blk); 3307 } 3308 f2fs_info(sbi, 3309 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3310 return 0; 3311 } 3312 3313 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3314 { 3315 #ifdef CONFIG_UNICODE 3316 if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) { 3317 const struct f2fs_sb_encodings *encoding_info; 3318 struct unicode_map *encoding; 3319 __u16 encoding_flags; 3320 3321 if (f2fs_sb_has_encrypt(sbi)) { 3322 f2fs_err(sbi, 3323 "Can't mount with encoding and encryption"); 3324 return -EINVAL; 3325 } 3326 3327 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info, 3328 &encoding_flags)) { 3329 f2fs_err(sbi, 3330 "Encoding requested by superblock is unknown"); 3331 return -EINVAL; 3332 } 3333 3334 encoding = utf8_load(encoding_info->version); 3335 if (IS_ERR(encoding)) { 3336 f2fs_err(sbi, 3337 "can't mount with superblock charset: %s-%s " 3338 "not supported by the kernel. flags: 0x%x.", 3339 encoding_info->name, encoding_info->version, 3340 encoding_flags); 3341 return PTR_ERR(encoding); 3342 } 3343 f2fs_info(sbi, "Using encoding defined by superblock: " 3344 "%s-%s with flags 0x%hx", encoding_info->name, 3345 encoding_info->version?:"\b", encoding_flags); 3346 3347 sbi->s_encoding = encoding; 3348 sbi->s_encoding_flags = encoding_flags; 3349 sbi->sb->s_d_op = &f2fs_dentry_ops; 3350 } 3351 #else 3352 if (f2fs_sb_has_casefold(sbi)) { 3353 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3354 return -EINVAL; 3355 } 3356 #endif 3357 return 0; 3358 } 3359 3360 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3361 { 3362 struct f2fs_sm_info *sm_i = SM_I(sbi); 3363 3364 /* adjust parameters according to the volume size */ 3365 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3366 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3367 sm_i->dcc_info->discard_granularity = 1; 3368 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 3369 } 3370 3371 sbi->readdir_ra = 1; 3372 } 3373 3374 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3375 { 3376 struct f2fs_sb_info *sbi; 3377 struct f2fs_super_block *raw_super; 3378 struct inode *root; 3379 int err; 3380 bool skip_recovery = false, need_fsck = false; 3381 char *options = NULL; 3382 int recovery, i, valid_super_block; 3383 struct curseg_info *seg_i; 3384 int retry_cnt = 1; 3385 3386 try_onemore: 3387 err = -EINVAL; 3388 raw_super = NULL; 3389 valid_super_block = -1; 3390 recovery = 0; 3391 3392 /* allocate memory for f2fs-specific super block info */ 3393 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 3394 if (!sbi) 3395 return -ENOMEM; 3396 3397 sbi->sb = sb; 3398 3399 /* Load the checksum driver */ 3400 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 3401 if (IS_ERR(sbi->s_chksum_driver)) { 3402 f2fs_err(sbi, "Cannot load crc32 driver."); 3403 err = PTR_ERR(sbi->s_chksum_driver); 3404 sbi->s_chksum_driver = NULL; 3405 goto free_sbi; 3406 } 3407 3408 /* set a block size */ 3409 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 3410 f2fs_err(sbi, "unable to set blocksize"); 3411 goto free_sbi; 3412 } 3413 3414 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 3415 &recovery); 3416 if (err) 3417 goto free_sbi; 3418 3419 sb->s_fs_info = sbi; 3420 sbi->raw_super = raw_super; 3421 3422 /* precompute checksum seed for metadata */ 3423 if (f2fs_sb_has_inode_chksum(sbi)) 3424 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 3425 sizeof(raw_super->uuid)); 3426 3427 /* 3428 * The BLKZONED feature indicates that the drive was formatted with 3429 * zone alignment optimization. This is optional for host-aware 3430 * devices, but mandatory for host-managed zoned block devices. 3431 */ 3432 #ifndef CONFIG_BLK_DEV_ZONED 3433 if (f2fs_sb_has_blkzoned(sbi)) { 3434 f2fs_err(sbi, "Zoned block device support is not enabled"); 3435 err = -EOPNOTSUPP; 3436 goto free_sb_buf; 3437 } 3438 #endif 3439 default_options(sbi); 3440 /* parse mount options */ 3441 options = kstrdup((const char *)data, GFP_KERNEL); 3442 if (data && !options) { 3443 err = -ENOMEM; 3444 goto free_sb_buf; 3445 } 3446 3447 err = parse_options(sb, options, false); 3448 if (err) 3449 goto free_options; 3450 3451 sbi->max_file_blocks = max_file_blocks(); 3452 sb->s_maxbytes = sbi->max_file_blocks << 3453 le32_to_cpu(raw_super->log_blocksize); 3454 sb->s_max_links = F2FS_LINK_MAX; 3455 3456 err = f2fs_setup_casefold(sbi); 3457 if (err) 3458 goto free_options; 3459 3460 #ifdef CONFIG_QUOTA 3461 sb->dq_op = &f2fs_quota_operations; 3462 sb->s_qcop = &f2fs_quotactl_ops; 3463 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 3464 3465 if (f2fs_sb_has_quota_ino(sbi)) { 3466 for (i = 0; i < MAXQUOTAS; i++) { 3467 if (f2fs_qf_ino(sbi->sb, i)) 3468 sbi->nquota_files++; 3469 } 3470 } 3471 #endif 3472 3473 sb->s_op = &f2fs_sops; 3474 #ifdef CONFIG_FS_ENCRYPTION 3475 sb->s_cop = &f2fs_cryptops; 3476 #endif 3477 #ifdef CONFIG_FS_VERITY 3478 sb->s_vop = &f2fs_verityops; 3479 #endif 3480 sb->s_xattr = f2fs_xattr_handlers; 3481 sb->s_export_op = &f2fs_export_ops; 3482 sb->s_magic = F2FS_SUPER_MAGIC; 3483 sb->s_time_gran = 1; 3484 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3485 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 3486 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 3487 sb->s_iflags |= SB_I_CGROUPWB; 3488 3489 /* init f2fs-specific super block info */ 3490 sbi->valid_super_block = valid_super_block; 3491 init_rwsem(&sbi->gc_lock); 3492 mutex_init(&sbi->writepages); 3493 mutex_init(&sbi->cp_mutex); 3494 init_rwsem(&sbi->node_write); 3495 init_rwsem(&sbi->node_change); 3496 3497 /* disallow all the data/node/meta page writes */ 3498 set_sbi_flag(sbi, SBI_POR_DOING); 3499 spin_lock_init(&sbi->stat_lock); 3500 3501 /* init iostat info */ 3502 spin_lock_init(&sbi->iostat_lock); 3503 sbi->iostat_enable = false; 3504 sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS; 3505 3506 for (i = 0; i < NR_PAGE_TYPE; i++) { 3507 int n = (i == META) ? 1: NR_TEMP_TYPE; 3508 int j; 3509 3510 sbi->write_io[i] = 3511 f2fs_kmalloc(sbi, 3512 array_size(n, 3513 sizeof(struct f2fs_bio_info)), 3514 GFP_KERNEL); 3515 if (!sbi->write_io[i]) { 3516 err = -ENOMEM; 3517 goto free_bio_info; 3518 } 3519 3520 for (j = HOT; j < n; j++) { 3521 init_rwsem(&sbi->write_io[i][j].io_rwsem); 3522 sbi->write_io[i][j].sbi = sbi; 3523 sbi->write_io[i][j].bio = NULL; 3524 spin_lock_init(&sbi->write_io[i][j].io_lock); 3525 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 3526 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); 3527 init_rwsem(&sbi->write_io[i][j].bio_list_lock); 3528 } 3529 } 3530 3531 init_rwsem(&sbi->cp_rwsem); 3532 init_rwsem(&sbi->quota_sem); 3533 init_waitqueue_head(&sbi->cp_wait); 3534 init_sb_info(sbi); 3535 3536 err = init_percpu_info(sbi); 3537 if (err) 3538 goto free_bio_info; 3539 3540 if (F2FS_IO_ALIGNED(sbi)) { 3541 sbi->write_io_dummy = 3542 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 3543 if (!sbi->write_io_dummy) { 3544 err = -ENOMEM; 3545 goto free_percpu; 3546 } 3547 } 3548 3549 /* init per sbi slab cache */ 3550 err = f2fs_init_xattr_caches(sbi); 3551 if (err) 3552 goto free_io_dummy; 3553 3554 /* get an inode for meta space */ 3555 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 3556 if (IS_ERR(sbi->meta_inode)) { 3557 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 3558 err = PTR_ERR(sbi->meta_inode); 3559 goto free_xattr_cache; 3560 } 3561 3562 err = f2fs_get_valid_checkpoint(sbi); 3563 if (err) { 3564 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 3565 goto free_meta_inode; 3566 } 3567 3568 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 3569 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3570 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 3571 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3572 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 3573 } 3574 3575 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 3576 set_sbi_flag(sbi, SBI_NEED_FSCK); 3577 3578 /* Initialize device list */ 3579 err = f2fs_scan_devices(sbi); 3580 if (err) { 3581 f2fs_err(sbi, "Failed to find devices"); 3582 goto free_devices; 3583 } 3584 3585 err = f2fs_init_post_read_wq(sbi); 3586 if (err) { 3587 f2fs_err(sbi, "Failed to initialize post read workqueue"); 3588 goto free_devices; 3589 } 3590 3591 sbi->total_valid_node_count = 3592 le32_to_cpu(sbi->ckpt->valid_node_count); 3593 percpu_counter_set(&sbi->total_valid_inode_count, 3594 le32_to_cpu(sbi->ckpt->valid_inode_count)); 3595 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 3596 sbi->total_valid_block_count = 3597 le64_to_cpu(sbi->ckpt->valid_block_count); 3598 sbi->last_valid_block_count = sbi->total_valid_block_count; 3599 sbi->reserved_blocks = 0; 3600 sbi->current_reserved_blocks = 0; 3601 limit_reserve_root(sbi); 3602 adjust_unusable_cap_perc(sbi); 3603 3604 for (i = 0; i < NR_INODE_TYPE; i++) { 3605 INIT_LIST_HEAD(&sbi->inode_list[i]); 3606 spin_lock_init(&sbi->inode_lock[i]); 3607 } 3608 mutex_init(&sbi->flush_lock); 3609 3610 f2fs_init_extent_cache_info(sbi); 3611 3612 f2fs_init_ino_entry_info(sbi); 3613 3614 f2fs_init_fsync_node_info(sbi); 3615 3616 /* setup f2fs internal modules */ 3617 err = f2fs_build_segment_manager(sbi); 3618 if (err) { 3619 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 3620 err); 3621 goto free_sm; 3622 } 3623 err = f2fs_build_node_manager(sbi); 3624 if (err) { 3625 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 3626 err); 3627 goto free_nm; 3628 } 3629 3630 /* For write statistics */ 3631 if (sb->s_bdev->bd_part) 3632 sbi->sectors_written_start = 3633 (u64)part_stat_read(sb->s_bdev->bd_part, 3634 sectors[STAT_WRITE]); 3635 3636 /* Read accumulated write IO statistics if exists */ 3637 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 3638 if (__exist_node_summaries(sbi)) 3639 sbi->kbytes_written = 3640 le64_to_cpu(seg_i->journal->info.kbytes_written); 3641 3642 f2fs_build_gc_manager(sbi); 3643 3644 err = f2fs_build_stats(sbi); 3645 if (err) 3646 goto free_nm; 3647 3648 /* get an inode for node space */ 3649 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 3650 if (IS_ERR(sbi->node_inode)) { 3651 f2fs_err(sbi, "Failed to read node inode"); 3652 err = PTR_ERR(sbi->node_inode); 3653 goto free_stats; 3654 } 3655 3656 /* read root inode and dentry */ 3657 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 3658 if (IS_ERR(root)) { 3659 f2fs_err(sbi, "Failed to read root inode"); 3660 err = PTR_ERR(root); 3661 goto free_node_inode; 3662 } 3663 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 3664 !root->i_size || !root->i_nlink) { 3665 iput(root); 3666 err = -EINVAL; 3667 goto free_node_inode; 3668 } 3669 3670 sb->s_root = d_make_root(root); /* allocate root dentry */ 3671 if (!sb->s_root) { 3672 err = -ENOMEM; 3673 goto free_node_inode; 3674 } 3675 3676 err = f2fs_register_sysfs(sbi); 3677 if (err) 3678 goto free_root_inode; 3679 3680 #ifdef CONFIG_QUOTA 3681 /* Enable quota usage during mount */ 3682 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 3683 err = f2fs_enable_quotas(sb); 3684 if (err) 3685 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 3686 } 3687 #endif 3688 /* if there are any orphan inodes, free them */ 3689 err = f2fs_recover_orphan_inodes(sbi); 3690 if (err) 3691 goto free_meta; 3692 3693 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 3694 goto reset_checkpoint; 3695 3696 /* recover fsynced data */ 3697 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 3698 !test_opt(sbi, NORECOVERY)) { 3699 /* 3700 * mount should be failed, when device has readonly mode, and 3701 * previous checkpoint was not done by clean system shutdown. 3702 */ 3703 if (f2fs_hw_is_readonly(sbi)) { 3704 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3705 err = -EROFS; 3706 f2fs_err(sbi, "Need to recover fsync data, but write access unavailable"); 3707 goto free_meta; 3708 } 3709 f2fs_info(sbi, "write access unavailable, skipping recovery"); 3710 goto reset_checkpoint; 3711 } 3712 3713 if (need_fsck) 3714 set_sbi_flag(sbi, SBI_NEED_FSCK); 3715 3716 if (skip_recovery) 3717 goto reset_checkpoint; 3718 3719 err = f2fs_recover_fsync_data(sbi, false); 3720 if (err < 0) { 3721 if (err != -ENOMEM) 3722 skip_recovery = true; 3723 need_fsck = true; 3724 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 3725 err); 3726 goto free_meta; 3727 } 3728 } else { 3729 err = f2fs_recover_fsync_data(sbi, true); 3730 3731 if (!f2fs_readonly(sb) && err > 0) { 3732 err = -EINVAL; 3733 f2fs_err(sbi, "Need to recover fsync data"); 3734 goto free_meta; 3735 } 3736 } 3737 3738 /* 3739 * If the f2fs is not readonly and fsync data recovery succeeds, 3740 * check zoned block devices' write pointer consistency. 3741 */ 3742 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 3743 err = f2fs_check_write_pointer(sbi); 3744 if (err) 3745 goto free_meta; 3746 } 3747 3748 reset_checkpoint: 3749 /* f2fs_recover_fsync_data() cleared this already */ 3750 clear_sbi_flag(sbi, SBI_POR_DOING); 3751 3752 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 3753 err = f2fs_disable_checkpoint(sbi); 3754 if (err) 3755 goto sync_free_meta; 3756 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 3757 f2fs_enable_checkpoint(sbi); 3758 } 3759 3760 /* 3761 * If filesystem is not mounted as read-only then 3762 * do start the gc_thread. 3763 */ 3764 if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) { 3765 /* After POR, we can run background GC thread.*/ 3766 err = f2fs_start_gc_thread(sbi); 3767 if (err) 3768 goto sync_free_meta; 3769 } 3770 kvfree(options); 3771 3772 /* recover broken superblock */ 3773 if (recovery) { 3774 err = f2fs_commit_super(sbi, true); 3775 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 3776 sbi->valid_super_block ? 1 : 2, err); 3777 } 3778 3779 f2fs_join_shrinker(sbi); 3780 3781 f2fs_tuning_parameters(sbi); 3782 3783 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 3784 cur_cp_version(F2FS_CKPT(sbi))); 3785 f2fs_update_time(sbi, CP_TIME); 3786 f2fs_update_time(sbi, REQ_TIME); 3787 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 3788 return 0; 3789 3790 sync_free_meta: 3791 /* safe to flush all the data */ 3792 sync_filesystem(sbi->sb); 3793 retry_cnt = 0; 3794 3795 free_meta: 3796 #ifdef CONFIG_QUOTA 3797 f2fs_truncate_quota_inode_pages(sb); 3798 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 3799 f2fs_quota_off_umount(sbi->sb); 3800 #endif 3801 /* 3802 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 3803 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 3804 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 3805 * falls into an infinite loop in f2fs_sync_meta_pages(). 3806 */ 3807 truncate_inode_pages_final(META_MAPPING(sbi)); 3808 /* evict some inodes being cached by GC */ 3809 evict_inodes(sb); 3810 f2fs_unregister_sysfs(sbi); 3811 free_root_inode: 3812 dput(sb->s_root); 3813 sb->s_root = NULL; 3814 free_node_inode: 3815 f2fs_release_ino_entry(sbi, true); 3816 truncate_inode_pages_final(NODE_MAPPING(sbi)); 3817 iput(sbi->node_inode); 3818 sbi->node_inode = NULL; 3819 free_stats: 3820 f2fs_destroy_stats(sbi); 3821 free_nm: 3822 f2fs_destroy_node_manager(sbi); 3823 free_sm: 3824 f2fs_destroy_segment_manager(sbi); 3825 f2fs_destroy_post_read_wq(sbi); 3826 free_devices: 3827 destroy_device_list(sbi); 3828 kvfree(sbi->ckpt); 3829 free_meta_inode: 3830 make_bad_inode(sbi->meta_inode); 3831 iput(sbi->meta_inode); 3832 sbi->meta_inode = NULL; 3833 free_xattr_cache: 3834 f2fs_destroy_xattr_caches(sbi); 3835 free_io_dummy: 3836 mempool_destroy(sbi->write_io_dummy); 3837 free_percpu: 3838 destroy_percpu_info(sbi); 3839 free_bio_info: 3840 for (i = 0; i < NR_PAGE_TYPE; i++) 3841 kvfree(sbi->write_io[i]); 3842 3843 #ifdef CONFIG_UNICODE 3844 utf8_unload(sbi->s_encoding); 3845 #endif 3846 free_options: 3847 #ifdef CONFIG_QUOTA 3848 for (i = 0; i < MAXQUOTAS; i++) 3849 kvfree(F2FS_OPTION(sbi).s_qf_names[i]); 3850 #endif 3851 fscrypt_free_dummy_context(&F2FS_OPTION(sbi).dummy_enc_ctx); 3852 kvfree(options); 3853 free_sb_buf: 3854 kvfree(raw_super); 3855 free_sbi: 3856 if (sbi->s_chksum_driver) 3857 crypto_free_shash(sbi->s_chksum_driver); 3858 kvfree(sbi); 3859 3860 /* give only one another chance */ 3861 if (retry_cnt > 0 && skip_recovery) { 3862 retry_cnt--; 3863 shrink_dcache_sb(sb); 3864 goto try_onemore; 3865 } 3866 return err; 3867 } 3868 3869 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 3870 const char *dev_name, void *data) 3871 { 3872 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 3873 } 3874 3875 static void kill_f2fs_super(struct super_block *sb) 3876 { 3877 if (sb->s_root) { 3878 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3879 3880 set_sbi_flag(sbi, SBI_IS_CLOSE); 3881 f2fs_stop_gc_thread(sbi); 3882 f2fs_stop_discard_thread(sbi); 3883 3884 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 3885 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3886 struct cp_control cpc = { 3887 .reason = CP_UMOUNT, 3888 }; 3889 f2fs_write_checkpoint(sbi, &cpc); 3890 } 3891 3892 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 3893 sb->s_flags &= ~SB_RDONLY; 3894 } 3895 kill_block_super(sb); 3896 } 3897 3898 static struct file_system_type f2fs_fs_type = { 3899 .owner = THIS_MODULE, 3900 .name = "f2fs", 3901 .mount = f2fs_mount, 3902 .kill_sb = kill_f2fs_super, 3903 .fs_flags = FS_REQUIRES_DEV, 3904 }; 3905 MODULE_ALIAS_FS("f2fs"); 3906 3907 static int __init init_inodecache(void) 3908 { 3909 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 3910 sizeof(struct f2fs_inode_info), 0, 3911 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 3912 if (!f2fs_inode_cachep) 3913 return -ENOMEM; 3914 return 0; 3915 } 3916 3917 static void destroy_inodecache(void) 3918 { 3919 /* 3920 * Make sure all delayed rcu free inodes are flushed before we 3921 * destroy cache. 3922 */ 3923 rcu_barrier(); 3924 kmem_cache_destroy(f2fs_inode_cachep); 3925 } 3926 3927 static int __init init_f2fs_fs(void) 3928 { 3929 int err; 3930 3931 if (PAGE_SIZE != F2FS_BLKSIZE) { 3932 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 3933 PAGE_SIZE, F2FS_BLKSIZE); 3934 return -EINVAL; 3935 } 3936 3937 f2fs_build_trace_ios(); 3938 3939 err = init_inodecache(); 3940 if (err) 3941 goto fail; 3942 err = f2fs_create_node_manager_caches(); 3943 if (err) 3944 goto free_inodecache; 3945 err = f2fs_create_segment_manager_caches(); 3946 if (err) 3947 goto free_node_manager_caches; 3948 err = f2fs_create_checkpoint_caches(); 3949 if (err) 3950 goto free_segment_manager_caches; 3951 err = f2fs_create_extent_cache(); 3952 if (err) 3953 goto free_checkpoint_caches; 3954 err = f2fs_init_sysfs(); 3955 if (err) 3956 goto free_extent_cache; 3957 err = register_shrinker(&f2fs_shrinker_info); 3958 if (err) 3959 goto free_sysfs; 3960 err = register_filesystem(&f2fs_fs_type); 3961 if (err) 3962 goto free_shrinker; 3963 f2fs_create_root_stats(); 3964 err = f2fs_init_post_read_processing(); 3965 if (err) 3966 goto free_root_stats; 3967 err = f2fs_init_bio_entry_cache(); 3968 if (err) 3969 goto free_post_read; 3970 err = f2fs_init_bioset(); 3971 if (err) 3972 goto free_bio_enrty_cache; 3973 err = f2fs_init_compress_mempool(); 3974 if (err) 3975 goto free_bioset; 3976 return 0; 3977 free_bioset: 3978 f2fs_destroy_bioset(); 3979 free_bio_enrty_cache: 3980 f2fs_destroy_bio_entry_cache(); 3981 free_post_read: 3982 f2fs_destroy_post_read_processing(); 3983 free_root_stats: 3984 f2fs_destroy_root_stats(); 3985 unregister_filesystem(&f2fs_fs_type); 3986 free_shrinker: 3987 unregister_shrinker(&f2fs_shrinker_info); 3988 free_sysfs: 3989 f2fs_exit_sysfs(); 3990 free_extent_cache: 3991 f2fs_destroy_extent_cache(); 3992 free_checkpoint_caches: 3993 f2fs_destroy_checkpoint_caches(); 3994 free_segment_manager_caches: 3995 f2fs_destroy_segment_manager_caches(); 3996 free_node_manager_caches: 3997 f2fs_destroy_node_manager_caches(); 3998 free_inodecache: 3999 destroy_inodecache(); 4000 fail: 4001 return err; 4002 } 4003 4004 static void __exit exit_f2fs_fs(void) 4005 { 4006 f2fs_destroy_compress_mempool(); 4007 f2fs_destroy_bioset(); 4008 f2fs_destroy_bio_entry_cache(); 4009 f2fs_destroy_post_read_processing(); 4010 f2fs_destroy_root_stats(); 4011 unregister_filesystem(&f2fs_fs_type); 4012 unregister_shrinker(&f2fs_shrinker_info); 4013 f2fs_exit_sysfs(); 4014 f2fs_destroy_extent_cache(); 4015 f2fs_destroy_checkpoint_caches(); 4016 f2fs_destroy_segment_manager_caches(); 4017 f2fs_destroy_node_manager_caches(); 4018 destroy_inodecache(); 4019 f2fs_destroy_trace_ios(); 4020 } 4021 4022 module_init(init_f2fs_fs) 4023 module_exit(exit_f2fs_fs) 4024 4025 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4026 MODULE_DESCRIPTION("Flash Friendly File System"); 4027 MODULE_LICENSE("GPL"); 4028 4029