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 #include <linux/zstd.h> 29 #include <linux/lz4.h> 30 31 #include "f2fs.h" 32 #include "node.h" 33 #include "segment.h" 34 #include "xattr.h" 35 #include "gc.h" 36 #include "iostat.h" 37 38 #define CREATE_TRACE_POINTS 39 #include <trace/events/f2fs.h> 40 41 static struct kmem_cache *f2fs_inode_cachep; 42 43 #ifdef CONFIG_F2FS_FAULT_INJECTION 44 45 const char *f2fs_fault_name[FAULT_MAX] = { 46 [FAULT_KMALLOC] = "kmalloc", 47 [FAULT_KVMALLOC] = "kvmalloc", 48 [FAULT_PAGE_ALLOC] = "page alloc", 49 [FAULT_PAGE_GET] = "page get", 50 [FAULT_ALLOC_NID] = "alloc nid", 51 [FAULT_ORPHAN] = "orphan", 52 [FAULT_BLOCK] = "no more block", 53 [FAULT_DIR_DEPTH] = "too big dir depth", 54 [FAULT_EVICT_INODE] = "evict_inode fail", 55 [FAULT_TRUNCATE] = "truncate fail", 56 [FAULT_READ_IO] = "read IO error", 57 [FAULT_CHECKPOINT] = "checkpoint error", 58 [FAULT_DISCARD] = "discard error", 59 [FAULT_WRITE_IO] = "write IO error", 60 [FAULT_SLAB_ALLOC] = "slab alloc", 61 }; 62 63 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, 64 unsigned int type) 65 { 66 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; 67 68 if (rate) { 69 atomic_set(&ffi->inject_ops, 0); 70 ffi->inject_rate = rate; 71 } 72 73 if (type) 74 ffi->inject_type = type; 75 76 if (!rate && !type) 77 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 78 } 79 #endif 80 81 /* f2fs-wide shrinker description */ 82 static struct shrinker f2fs_shrinker_info = { 83 .scan_objects = f2fs_shrink_scan, 84 .count_objects = f2fs_shrink_count, 85 .seeks = DEFAULT_SEEKS, 86 }; 87 88 enum { 89 Opt_gc_background, 90 Opt_disable_roll_forward, 91 Opt_norecovery, 92 Opt_discard, 93 Opt_nodiscard, 94 Opt_noheap, 95 Opt_heap, 96 Opt_user_xattr, 97 Opt_nouser_xattr, 98 Opt_acl, 99 Opt_noacl, 100 Opt_active_logs, 101 Opt_disable_ext_identify, 102 Opt_inline_xattr, 103 Opt_noinline_xattr, 104 Opt_inline_xattr_size, 105 Opt_inline_data, 106 Opt_inline_dentry, 107 Opt_noinline_dentry, 108 Opt_flush_merge, 109 Opt_noflush_merge, 110 Opt_nobarrier, 111 Opt_fastboot, 112 Opt_extent_cache, 113 Opt_noextent_cache, 114 Opt_noinline_data, 115 Opt_data_flush, 116 Opt_reserve_root, 117 Opt_resgid, 118 Opt_resuid, 119 Opt_mode, 120 Opt_io_size_bits, 121 Opt_fault_injection, 122 Opt_fault_type, 123 Opt_lazytime, 124 Opt_nolazytime, 125 Opt_quota, 126 Opt_noquota, 127 Opt_usrquota, 128 Opt_grpquota, 129 Opt_prjquota, 130 Opt_usrjquota, 131 Opt_grpjquota, 132 Opt_prjjquota, 133 Opt_offusrjquota, 134 Opt_offgrpjquota, 135 Opt_offprjjquota, 136 Opt_jqfmt_vfsold, 137 Opt_jqfmt_vfsv0, 138 Opt_jqfmt_vfsv1, 139 Opt_whint, 140 Opt_alloc, 141 Opt_fsync, 142 Opt_test_dummy_encryption, 143 Opt_inlinecrypt, 144 Opt_checkpoint_disable, 145 Opt_checkpoint_disable_cap, 146 Opt_checkpoint_disable_cap_perc, 147 Opt_checkpoint_enable, 148 Opt_checkpoint_merge, 149 Opt_nocheckpoint_merge, 150 Opt_compress_algorithm, 151 Opt_compress_log_size, 152 Opt_compress_extension, 153 Opt_nocompress_extension, 154 Opt_compress_chksum, 155 Opt_compress_mode, 156 Opt_compress_cache, 157 Opt_atgc, 158 Opt_gc_merge, 159 Opt_nogc_merge, 160 Opt_discard_unit, 161 Opt_err, 162 }; 163 164 static match_table_t f2fs_tokens = { 165 {Opt_gc_background, "background_gc=%s"}, 166 {Opt_disable_roll_forward, "disable_roll_forward"}, 167 {Opt_norecovery, "norecovery"}, 168 {Opt_discard, "discard"}, 169 {Opt_nodiscard, "nodiscard"}, 170 {Opt_noheap, "no_heap"}, 171 {Opt_heap, "heap"}, 172 {Opt_user_xattr, "user_xattr"}, 173 {Opt_nouser_xattr, "nouser_xattr"}, 174 {Opt_acl, "acl"}, 175 {Opt_noacl, "noacl"}, 176 {Opt_active_logs, "active_logs=%u"}, 177 {Opt_disable_ext_identify, "disable_ext_identify"}, 178 {Opt_inline_xattr, "inline_xattr"}, 179 {Opt_noinline_xattr, "noinline_xattr"}, 180 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 181 {Opt_inline_data, "inline_data"}, 182 {Opt_inline_dentry, "inline_dentry"}, 183 {Opt_noinline_dentry, "noinline_dentry"}, 184 {Opt_flush_merge, "flush_merge"}, 185 {Opt_noflush_merge, "noflush_merge"}, 186 {Opt_nobarrier, "nobarrier"}, 187 {Opt_fastboot, "fastboot"}, 188 {Opt_extent_cache, "extent_cache"}, 189 {Opt_noextent_cache, "noextent_cache"}, 190 {Opt_noinline_data, "noinline_data"}, 191 {Opt_data_flush, "data_flush"}, 192 {Opt_reserve_root, "reserve_root=%u"}, 193 {Opt_resgid, "resgid=%u"}, 194 {Opt_resuid, "resuid=%u"}, 195 {Opt_mode, "mode=%s"}, 196 {Opt_io_size_bits, "io_bits=%u"}, 197 {Opt_fault_injection, "fault_injection=%u"}, 198 {Opt_fault_type, "fault_type=%u"}, 199 {Opt_lazytime, "lazytime"}, 200 {Opt_nolazytime, "nolazytime"}, 201 {Opt_quota, "quota"}, 202 {Opt_noquota, "noquota"}, 203 {Opt_usrquota, "usrquota"}, 204 {Opt_grpquota, "grpquota"}, 205 {Opt_prjquota, "prjquota"}, 206 {Opt_usrjquota, "usrjquota=%s"}, 207 {Opt_grpjquota, "grpjquota=%s"}, 208 {Opt_prjjquota, "prjjquota=%s"}, 209 {Opt_offusrjquota, "usrjquota="}, 210 {Opt_offgrpjquota, "grpjquota="}, 211 {Opt_offprjjquota, "prjjquota="}, 212 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 213 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 214 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 215 {Opt_whint, "whint_mode=%s"}, 216 {Opt_alloc, "alloc_mode=%s"}, 217 {Opt_fsync, "fsync_mode=%s"}, 218 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 219 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 220 {Opt_inlinecrypt, "inlinecrypt"}, 221 {Opt_checkpoint_disable, "checkpoint=disable"}, 222 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 223 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 224 {Opt_checkpoint_enable, "checkpoint=enable"}, 225 {Opt_checkpoint_merge, "checkpoint_merge"}, 226 {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, 227 {Opt_compress_algorithm, "compress_algorithm=%s"}, 228 {Opt_compress_log_size, "compress_log_size=%u"}, 229 {Opt_compress_extension, "compress_extension=%s"}, 230 {Opt_nocompress_extension, "nocompress_extension=%s"}, 231 {Opt_compress_chksum, "compress_chksum"}, 232 {Opt_compress_mode, "compress_mode=%s"}, 233 {Opt_compress_cache, "compress_cache"}, 234 {Opt_atgc, "atgc"}, 235 {Opt_gc_merge, "gc_merge"}, 236 {Opt_nogc_merge, "nogc_merge"}, 237 {Opt_discard_unit, "discard_unit=%s"}, 238 {Opt_err, NULL}, 239 }; 240 241 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 242 { 243 struct va_format vaf; 244 va_list args; 245 int level; 246 247 va_start(args, fmt); 248 249 level = printk_get_level(fmt); 250 vaf.fmt = printk_skip_level(fmt); 251 vaf.va = &args; 252 printk("%c%cF2FS-fs (%s): %pV\n", 253 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 254 255 va_end(args); 256 } 257 258 #ifdef CONFIG_UNICODE 259 static const struct f2fs_sb_encodings { 260 __u16 magic; 261 char *name; 262 char *version; 263 } f2fs_sb_encoding_map[] = { 264 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"}, 265 }; 266 267 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb, 268 const struct f2fs_sb_encodings **encoding, 269 __u16 *flags) 270 { 271 __u16 magic = le16_to_cpu(sb->s_encoding); 272 int i; 273 274 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 275 if (magic == f2fs_sb_encoding_map[i].magic) 276 break; 277 278 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map)) 279 return -EINVAL; 280 281 *encoding = &f2fs_sb_encoding_map[i]; 282 *flags = le16_to_cpu(sb->s_encoding_flags); 283 284 return 0; 285 } 286 287 struct kmem_cache *f2fs_cf_name_slab; 288 static int __init f2fs_create_casefold_cache(void) 289 { 290 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 291 F2FS_NAME_LEN); 292 if (!f2fs_cf_name_slab) 293 return -ENOMEM; 294 return 0; 295 } 296 297 static void f2fs_destroy_casefold_cache(void) 298 { 299 kmem_cache_destroy(f2fs_cf_name_slab); 300 } 301 #else 302 static int __init f2fs_create_casefold_cache(void) { return 0; } 303 static void f2fs_destroy_casefold_cache(void) { } 304 #endif 305 306 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 307 { 308 block_t limit = min((sbi->user_block_count << 1) / 1000, 309 sbi->user_block_count - sbi->reserved_blocks); 310 311 /* limit is 0.2% */ 312 if (test_opt(sbi, RESERVE_ROOT) && 313 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 314 F2FS_OPTION(sbi).root_reserved_blocks = limit; 315 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 316 F2FS_OPTION(sbi).root_reserved_blocks); 317 } 318 if (!test_opt(sbi, RESERVE_ROOT) && 319 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 320 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 321 !gid_eq(F2FS_OPTION(sbi).s_resgid, 322 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 323 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 324 from_kuid_munged(&init_user_ns, 325 F2FS_OPTION(sbi).s_resuid), 326 from_kgid_munged(&init_user_ns, 327 F2FS_OPTION(sbi).s_resgid)); 328 } 329 330 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 331 { 332 if (!F2FS_OPTION(sbi).unusable_cap_perc) 333 return; 334 335 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 336 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 337 else 338 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 339 F2FS_OPTION(sbi).unusable_cap_perc; 340 341 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 342 F2FS_OPTION(sbi).unusable_cap, 343 F2FS_OPTION(sbi).unusable_cap_perc); 344 } 345 346 static void init_once(void *foo) 347 { 348 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 349 350 inode_init_once(&fi->vfs_inode); 351 } 352 353 #ifdef CONFIG_QUOTA 354 static const char * const quotatypes[] = INITQFNAMES; 355 #define QTYPE2NAME(t) (quotatypes[t]) 356 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 357 substring_t *args) 358 { 359 struct f2fs_sb_info *sbi = F2FS_SB(sb); 360 char *qname; 361 int ret = -EINVAL; 362 363 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 364 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 365 return -EINVAL; 366 } 367 if (f2fs_sb_has_quota_ino(sbi)) { 368 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 369 return 0; 370 } 371 372 qname = match_strdup(args); 373 if (!qname) { 374 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 375 return -ENOMEM; 376 } 377 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 378 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 379 ret = 0; 380 else 381 f2fs_err(sbi, "%s quota file already specified", 382 QTYPE2NAME(qtype)); 383 goto errout; 384 } 385 if (strchr(qname, '/')) { 386 f2fs_err(sbi, "quotafile must be on filesystem root"); 387 goto errout; 388 } 389 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 390 set_opt(sbi, QUOTA); 391 return 0; 392 errout: 393 kfree(qname); 394 return ret; 395 } 396 397 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 398 { 399 struct f2fs_sb_info *sbi = F2FS_SB(sb); 400 401 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 402 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 403 return -EINVAL; 404 } 405 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 406 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 407 return 0; 408 } 409 410 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 411 { 412 /* 413 * We do the test below only for project quotas. 'usrquota' and 414 * 'grpquota' mount options are allowed even without quota feature 415 * to support legacy quotas in quota files. 416 */ 417 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 418 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 419 return -1; 420 } 421 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 422 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 423 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 424 if (test_opt(sbi, USRQUOTA) && 425 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 426 clear_opt(sbi, USRQUOTA); 427 428 if (test_opt(sbi, GRPQUOTA) && 429 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 430 clear_opt(sbi, GRPQUOTA); 431 432 if (test_opt(sbi, PRJQUOTA) && 433 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 434 clear_opt(sbi, PRJQUOTA); 435 436 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 437 test_opt(sbi, PRJQUOTA)) { 438 f2fs_err(sbi, "old and new quota format mixing"); 439 return -1; 440 } 441 442 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 443 f2fs_err(sbi, "journaled quota format not specified"); 444 return -1; 445 } 446 } 447 448 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 449 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 450 F2FS_OPTION(sbi).s_jquota_fmt = 0; 451 } 452 return 0; 453 } 454 #endif 455 456 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 457 const char *opt, 458 const substring_t *arg, 459 bool is_remount) 460 { 461 struct f2fs_sb_info *sbi = F2FS_SB(sb); 462 #ifdef CONFIG_FS_ENCRYPTION 463 int err; 464 465 if (!f2fs_sb_has_encrypt(sbi)) { 466 f2fs_err(sbi, "Encrypt feature is off"); 467 return -EINVAL; 468 } 469 470 /* 471 * This mount option is just for testing, and it's not worthwhile to 472 * implement the extra complexity (e.g. RCU protection) that would be 473 * needed to allow it to be set or changed during remount. We do allow 474 * it to be specified during remount, but only if there is no change. 475 */ 476 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) { 477 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 478 return -EINVAL; 479 } 480 err = fscrypt_set_test_dummy_encryption( 481 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy); 482 if (err) { 483 if (err == -EEXIST) 484 f2fs_warn(sbi, 485 "Can't change test_dummy_encryption on remount"); 486 else if (err == -EINVAL) 487 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 488 opt); 489 else 490 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 491 opt, err); 492 return -EINVAL; 493 } 494 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 495 #else 496 f2fs_warn(sbi, "Test dummy encryption mount option ignored"); 497 #endif 498 return 0; 499 } 500 501 #ifdef CONFIG_F2FS_FS_COMPRESSION 502 /* 503 * 1. The same extension name cannot not appear in both compress and non-compress extension 504 * at the same time. 505 * 2. If the compress extension specifies all files, the types specified by the non-compress 506 * extension will be treated as special cases and will not be compressed. 507 * 3. Don't allow the non-compress extension specifies all files. 508 */ 509 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi) 510 { 511 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 512 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 513 int ext_cnt, noext_cnt, index = 0, no_index = 0; 514 515 ext = F2FS_OPTION(sbi).extensions; 516 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 517 noext = F2FS_OPTION(sbi).noextensions; 518 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 519 520 if (!noext_cnt) 521 return 0; 522 523 for (no_index = 0; no_index < noext_cnt; no_index++) { 524 if (!strcasecmp("*", noext[no_index])) { 525 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files"); 526 return -EINVAL; 527 } 528 for (index = 0; index < ext_cnt; index++) { 529 if (!strcasecmp(ext[index], noext[no_index])) { 530 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension", 531 ext[index]); 532 return -EINVAL; 533 } 534 } 535 } 536 return 0; 537 } 538 539 #ifdef CONFIG_F2FS_FS_LZ4 540 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) 541 { 542 #ifdef CONFIG_F2FS_FS_LZ4HC 543 unsigned int level; 544 #endif 545 546 if (strlen(str) == 3) { 547 F2FS_OPTION(sbi).compress_level = 0; 548 return 0; 549 } 550 551 #ifdef CONFIG_F2FS_FS_LZ4HC 552 str += 3; 553 554 if (str[0] != ':') { 555 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 556 return -EINVAL; 557 } 558 if (kstrtouint(str + 1, 10, &level)) 559 return -EINVAL; 560 561 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) { 562 f2fs_info(sbi, "invalid lz4hc compress level: %d", level); 563 return -EINVAL; 564 } 565 566 F2FS_OPTION(sbi).compress_level = level; 567 return 0; 568 #else 569 f2fs_info(sbi, "kernel doesn't support lz4hc compression"); 570 return -EINVAL; 571 #endif 572 } 573 #endif 574 575 #ifdef CONFIG_F2FS_FS_ZSTD 576 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) 577 { 578 unsigned int level; 579 int len = 4; 580 581 if (strlen(str) == len) { 582 F2FS_OPTION(sbi).compress_level = 0; 583 return 0; 584 } 585 586 str += len; 587 588 if (str[0] != ':') { 589 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 590 return -EINVAL; 591 } 592 if (kstrtouint(str + 1, 10, &level)) 593 return -EINVAL; 594 595 if (!level || level > ZSTD_maxCLevel()) { 596 f2fs_info(sbi, "invalid zstd compress level: %d", level); 597 return -EINVAL; 598 } 599 600 F2FS_OPTION(sbi).compress_level = level; 601 return 0; 602 } 603 #endif 604 #endif 605 606 static int parse_options(struct super_block *sb, char *options, bool is_remount) 607 { 608 struct f2fs_sb_info *sbi = F2FS_SB(sb); 609 substring_t args[MAX_OPT_ARGS]; 610 #ifdef CONFIG_F2FS_FS_COMPRESSION 611 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 612 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 613 int ext_cnt, noext_cnt; 614 #endif 615 char *p, *name; 616 int arg = 0; 617 kuid_t uid; 618 kgid_t gid; 619 int ret; 620 621 if (!options) 622 goto default_check; 623 624 while ((p = strsep(&options, ",")) != NULL) { 625 int token; 626 627 if (!*p) 628 continue; 629 /* 630 * Initialize args struct so we know whether arg was 631 * found; some options take optional arguments. 632 */ 633 args[0].to = args[0].from = NULL; 634 token = match_token(p, f2fs_tokens, args); 635 636 switch (token) { 637 case Opt_gc_background: 638 name = match_strdup(&args[0]); 639 640 if (!name) 641 return -ENOMEM; 642 if (!strcmp(name, "on")) { 643 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 644 } else if (!strcmp(name, "off")) { 645 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 646 } else if (!strcmp(name, "sync")) { 647 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 648 } else { 649 kfree(name); 650 return -EINVAL; 651 } 652 kfree(name); 653 break; 654 case Opt_disable_roll_forward: 655 set_opt(sbi, DISABLE_ROLL_FORWARD); 656 break; 657 case Opt_norecovery: 658 /* this option mounts f2fs with ro */ 659 set_opt(sbi, NORECOVERY); 660 if (!f2fs_readonly(sb)) 661 return -EINVAL; 662 break; 663 case Opt_discard: 664 if (!f2fs_hw_support_discard(sbi)) { 665 f2fs_warn(sbi, "device does not support discard"); 666 break; 667 } 668 set_opt(sbi, DISCARD); 669 break; 670 case Opt_nodiscard: 671 if (f2fs_hw_should_discard(sbi)) { 672 f2fs_warn(sbi, "discard is required for zoned block devices"); 673 return -EINVAL; 674 } 675 clear_opt(sbi, DISCARD); 676 break; 677 case Opt_noheap: 678 set_opt(sbi, NOHEAP); 679 break; 680 case Opt_heap: 681 clear_opt(sbi, NOHEAP); 682 break; 683 #ifdef CONFIG_F2FS_FS_XATTR 684 case Opt_user_xattr: 685 set_opt(sbi, XATTR_USER); 686 break; 687 case Opt_nouser_xattr: 688 clear_opt(sbi, XATTR_USER); 689 break; 690 case Opt_inline_xattr: 691 set_opt(sbi, INLINE_XATTR); 692 break; 693 case Opt_noinline_xattr: 694 clear_opt(sbi, INLINE_XATTR); 695 break; 696 case Opt_inline_xattr_size: 697 if (args->from && match_int(args, &arg)) 698 return -EINVAL; 699 set_opt(sbi, INLINE_XATTR_SIZE); 700 F2FS_OPTION(sbi).inline_xattr_size = arg; 701 break; 702 #else 703 case Opt_user_xattr: 704 f2fs_info(sbi, "user_xattr options not supported"); 705 break; 706 case Opt_nouser_xattr: 707 f2fs_info(sbi, "nouser_xattr options not supported"); 708 break; 709 case Opt_inline_xattr: 710 f2fs_info(sbi, "inline_xattr options not supported"); 711 break; 712 case Opt_noinline_xattr: 713 f2fs_info(sbi, "noinline_xattr options not supported"); 714 break; 715 #endif 716 #ifdef CONFIG_F2FS_FS_POSIX_ACL 717 case Opt_acl: 718 set_opt(sbi, POSIX_ACL); 719 break; 720 case Opt_noacl: 721 clear_opt(sbi, POSIX_ACL); 722 break; 723 #else 724 case Opt_acl: 725 f2fs_info(sbi, "acl options not supported"); 726 break; 727 case Opt_noacl: 728 f2fs_info(sbi, "noacl options not supported"); 729 break; 730 #endif 731 case Opt_active_logs: 732 if (args->from && match_int(args, &arg)) 733 return -EINVAL; 734 if (arg != 2 && arg != 4 && 735 arg != NR_CURSEG_PERSIST_TYPE) 736 return -EINVAL; 737 F2FS_OPTION(sbi).active_logs = arg; 738 break; 739 case Opt_disable_ext_identify: 740 set_opt(sbi, DISABLE_EXT_IDENTIFY); 741 break; 742 case Opt_inline_data: 743 set_opt(sbi, INLINE_DATA); 744 break; 745 case Opt_inline_dentry: 746 set_opt(sbi, INLINE_DENTRY); 747 break; 748 case Opt_noinline_dentry: 749 clear_opt(sbi, INLINE_DENTRY); 750 break; 751 case Opt_flush_merge: 752 set_opt(sbi, FLUSH_MERGE); 753 break; 754 case Opt_noflush_merge: 755 clear_opt(sbi, FLUSH_MERGE); 756 break; 757 case Opt_nobarrier: 758 set_opt(sbi, NOBARRIER); 759 break; 760 case Opt_fastboot: 761 set_opt(sbi, FASTBOOT); 762 break; 763 case Opt_extent_cache: 764 set_opt(sbi, EXTENT_CACHE); 765 break; 766 case Opt_noextent_cache: 767 clear_opt(sbi, EXTENT_CACHE); 768 break; 769 case Opt_noinline_data: 770 clear_opt(sbi, INLINE_DATA); 771 break; 772 case Opt_data_flush: 773 set_opt(sbi, DATA_FLUSH); 774 break; 775 case Opt_reserve_root: 776 if (args->from && match_int(args, &arg)) 777 return -EINVAL; 778 if (test_opt(sbi, RESERVE_ROOT)) { 779 f2fs_info(sbi, "Preserve previous reserve_root=%u", 780 F2FS_OPTION(sbi).root_reserved_blocks); 781 } else { 782 F2FS_OPTION(sbi).root_reserved_blocks = arg; 783 set_opt(sbi, RESERVE_ROOT); 784 } 785 break; 786 case Opt_resuid: 787 if (args->from && match_int(args, &arg)) 788 return -EINVAL; 789 uid = make_kuid(current_user_ns(), arg); 790 if (!uid_valid(uid)) { 791 f2fs_err(sbi, "Invalid uid value %d", arg); 792 return -EINVAL; 793 } 794 F2FS_OPTION(sbi).s_resuid = uid; 795 break; 796 case Opt_resgid: 797 if (args->from && match_int(args, &arg)) 798 return -EINVAL; 799 gid = make_kgid(current_user_ns(), arg); 800 if (!gid_valid(gid)) { 801 f2fs_err(sbi, "Invalid gid value %d", arg); 802 return -EINVAL; 803 } 804 F2FS_OPTION(sbi).s_resgid = gid; 805 break; 806 case Opt_mode: 807 name = match_strdup(&args[0]); 808 809 if (!name) 810 return -ENOMEM; 811 if (!strcmp(name, "adaptive")) { 812 if (f2fs_sb_has_blkzoned(sbi)) { 813 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 814 kfree(name); 815 return -EINVAL; 816 } 817 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 818 } else if (!strcmp(name, "lfs")) { 819 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 820 } else { 821 kfree(name); 822 return -EINVAL; 823 } 824 kfree(name); 825 break; 826 case Opt_io_size_bits: 827 if (args->from && match_int(args, &arg)) 828 return -EINVAL; 829 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) { 830 f2fs_warn(sbi, "Not support %d, larger than %d", 831 1 << arg, BIO_MAX_VECS); 832 return -EINVAL; 833 } 834 F2FS_OPTION(sbi).write_io_size_bits = arg; 835 break; 836 #ifdef CONFIG_F2FS_FAULT_INJECTION 837 case Opt_fault_injection: 838 if (args->from && match_int(args, &arg)) 839 return -EINVAL; 840 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 841 set_opt(sbi, FAULT_INJECTION); 842 break; 843 844 case Opt_fault_type: 845 if (args->from && match_int(args, &arg)) 846 return -EINVAL; 847 f2fs_build_fault_attr(sbi, 0, arg); 848 set_opt(sbi, FAULT_INJECTION); 849 break; 850 #else 851 case Opt_fault_injection: 852 f2fs_info(sbi, "fault_injection options not supported"); 853 break; 854 855 case Opt_fault_type: 856 f2fs_info(sbi, "fault_type options not supported"); 857 break; 858 #endif 859 case Opt_lazytime: 860 sb->s_flags |= SB_LAZYTIME; 861 break; 862 case Opt_nolazytime: 863 sb->s_flags &= ~SB_LAZYTIME; 864 break; 865 #ifdef CONFIG_QUOTA 866 case Opt_quota: 867 case Opt_usrquota: 868 set_opt(sbi, USRQUOTA); 869 break; 870 case Opt_grpquota: 871 set_opt(sbi, GRPQUOTA); 872 break; 873 case Opt_prjquota: 874 set_opt(sbi, PRJQUOTA); 875 break; 876 case Opt_usrjquota: 877 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 878 if (ret) 879 return ret; 880 break; 881 case Opt_grpjquota: 882 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 883 if (ret) 884 return ret; 885 break; 886 case Opt_prjjquota: 887 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 888 if (ret) 889 return ret; 890 break; 891 case Opt_offusrjquota: 892 ret = f2fs_clear_qf_name(sb, USRQUOTA); 893 if (ret) 894 return ret; 895 break; 896 case Opt_offgrpjquota: 897 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 898 if (ret) 899 return ret; 900 break; 901 case Opt_offprjjquota: 902 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 903 if (ret) 904 return ret; 905 break; 906 case Opt_jqfmt_vfsold: 907 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 908 break; 909 case Opt_jqfmt_vfsv0: 910 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 911 break; 912 case Opt_jqfmt_vfsv1: 913 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 914 break; 915 case Opt_noquota: 916 clear_opt(sbi, QUOTA); 917 clear_opt(sbi, USRQUOTA); 918 clear_opt(sbi, GRPQUOTA); 919 clear_opt(sbi, PRJQUOTA); 920 break; 921 #else 922 case Opt_quota: 923 case Opt_usrquota: 924 case Opt_grpquota: 925 case Opt_prjquota: 926 case Opt_usrjquota: 927 case Opt_grpjquota: 928 case Opt_prjjquota: 929 case Opt_offusrjquota: 930 case Opt_offgrpjquota: 931 case Opt_offprjjquota: 932 case Opt_jqfmt_vfsold: 933 case Opt_jqfmt_vfsv0: 934 case Opt_jqfmt_vfsv1: 935 case Opt_noquota: 936 f2fs_info(sbi, "quota operations not supported"); 937 break; 938 #endif 939 case Opt_whint: 940 name = match_strdup(&args[0]); 941 if (!name) 942 return -ENOMEM; 943 if (!strcmp(name, "user-based")) { 944 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER; 945 } else if (!strcmp(name, "off")) { 946 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 947 } else if (!strcmp(name, "fs-based")) { 948 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS; 949 } else { 950 kfree(name); 951 return -EINVAL; 952 } 953 kfree(name); 954 break; 955 case Opt_alloc: 956 name = match_strdup(&args[0]); 957 if (!name) 958 return -ENOMEM; 959 960 if (!strcmp(name, "default")) { 961 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 962 } else if (!strcmp(name, "reuse")) { 963 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 964 } else { 965 kfree(name); 966 return -EINVAL; 967 } 968 kfree(name); 969 break; 970 case Opt_fsync: 971 name = match_strdup(&args[0]); 972 if (!name) 973 return -ENOMEM; 974 if (!strcmp(name, "posix")) { 975 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 976 } else if (!strcmp(name, "strict")) { 977 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 978 } else if (!strcmp(name, "nobarrier")) { 979 F2FS_OPTION(sbi).fsync_mode = 980 FSYNC_MODE_NOBARRIER; 981 } else { 982 kfree(name); 983 return -EINVAL; 984 } 985 kfree(name); 986 break; 987 case Opt_test_dummy_encryption: 988 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 989 is_remount); 990 if (ret) 991 return ret; 992 break; 993 case Opt_inlinecrypt: 994 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 995 sb->s_flags |= SB_INLINECRYPT; 996 #else 997 f2fs_info(sbi, "inline encryption not supported"); 998 #endif 999 break; 1000 case Opt_checkpoint_disable_cap_perc: 1001 if (args->from && match_int(args, &arg)) 1002 return -EINVAL; 1003 if (arg < 0 || arg > 100) 1004 return -EINVAL; 1005 F2FS_OPTION(sbi).unusable_cap_perc = arg; 1006 set_opt(sbi, DISABLE_CHECKPOINT); 1007 break; 1008 case Opt_checkpoint_disable_cap: 1009 if (args->from && match_int(args, &arg)) 1010 return -EINVAL; 1011 F2FS_OPTION(sbi).unusable_cap = arg; 1012 set_opt(sbi, DISABLE_CHECKPOINT); 1013 break; 1014 case Opt_checkpoint_disable: 1015 set_opt(sbi, DISABLE_CHECKPOINT); 1016 break; 1017 case Opt_checkpoint_enable: 1018 clear_opt(sbi, DISABLE_CHECKPOINT); 1019 break; 1020 case Opt_checkpoint_merge: 1021 set_opt(sbi, MERGE_CHECKPOINT); 1022 break; 1023 case Opt_nocheckpoint_merge: 1024 clear_opt(sbi, MERGE_CHECKPOINT); 1025 break; 1026 #ifdef CONFIG_F2FS_FS_COMPRESSION 1027 case Opt_compress_algorithm: 1028 if (!f2fs_sb_has_compression(sbi)) { 1029 f2fs_info(sbi, "Image doesn't support compression"); 1030 break; 1031 } 1032 name = match_strdup(&args[0]); 1033 if (!name) 1034 return -ENOMEM; 1035 if (!strcmp(name, "lzo")) { 1036 #ifdef CONFIG_F2FS_FS_LZO 1037 F2FS_OPTION(sbi).compress_level = 0; 1038 F2FS_OPTION(sbi).compress_algorithm = 1039 COMPRESS_LZO; 1040 #else 1041 f2fs_info(sbi, "kernel doesn't support lzo compression"); 1042 #endif 1043 } else if (!strncmp(name, "lz4", 3)) { 1044 #ifdef CONFIG_F2FS_FS_LZ4 1045 ret = f2fs_set_lz4hc_level(sbi, name); 1046 if (ret) { 1047 kfree(name); 1048 return -EINVAL; 1049 } 1050 F2FS_OPTION(sbi).compress_algorithm = 1051 COMPRESS_LZ4; 1052 #else 1053 f2fs_info(sbi, "kernel doesn't support lz4 compression"); 1054 #endif 1055 } else if (!strncmp(name, "zstd", 4)) { 1056 #ifdef CONFIG_F2FS_FS_ZSTD 1057 ret = f2fs_set_zstd_level(sbi, name); 1058 if (ret) { 1059 kfree(name); 1060 return -EINVAL; 1061 } 1062 F2FS_OPTION(sbi).compress_algorithm = 1063 COMPRESS_ZSTD; 1064 #else 1065 f2fs_info(sbi, "kernel doesn't support zstd compression"); 1066 #endif 1067 } else if (!strcmp(name, "lzo-rle")) { 1068 #ifdef CONFIG_F2FS_FS_LZORLE 1069 F2FS_OPTION(sbi).compress_level = 0; 1070 F2FS_OPTION(sbi).compress_algorithm = 1071 COMPRESS_LZORLE; 1072 #else 1073 f2fs_info(sbi, "kernel doesn't support lzorle compression"); 1074 #endif 1075 } else { 1076 kfree(name); 1077 return -EINVAL; 1078 } 1079 kfree(name); 1080 break; 1081 case Opt_compress_log_size: 1082 if (!f2fs_sb_has_compression(sbi)) { 1083 f2fs_info(sbi, "Image doesn't support compression"); 1084 break; 1085 } 1086 if (args->from && match_int(args, &arg)) 1087 return -EINVAL; 1088 if (arg < MIN_COMPRESS_LOG_SIZE || 1089 arg > MAX_COMPRESS_LOG_SIZE) { 1090 f2fs_err(sbi, 1091 "Compress cluster log size is out of range"); 1092 return -EINVAL; 1093 } 1094 F2FS_OPTION(sbi).compress_log_size = arg; 1095 break; 1096 case Opt_compress_extension: 1097 if (!f2fs_sb_has_compression(sbi)) { 1098 f2fs_info(sbi, "Image doesn't support compression"); 1099 break; 1100 } 1101 name = match_strdup(&args[0]); 1102 if (!name) 1103 return -ENOMEM; 1104 1105 ext = F2FS_OPTION(sbi).extensions; 1106 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1107 1108 if (strlen(name) >= F2FS_EXTENSION_LEN || 1109 ext_cnt >= COMPRESS_EXT_NUM) { 1110 f2fs_err(sbi, 1111 "invalid extension length/number"); 1112 kfree(name); 1113 return -EINVAL; 1114 } 1115 1116 strcpy(ext[ext_cnt], name); 1117 F2FS_OPTION(sbi).compress_ext_cnt++; 1118 kfree(name); 1119 break; 1120 case Opt_nocompress_extension: 1121 if (!f2fs_sb_has_compression(sbi)) { 1122 f2fs_info(sbi, "Image doesn't support compression"); 1123 break; 1124 } 1125 name = match_strdup(&args[0]); 1126 if (!name) 1127 return -ENOMEM; 1128 1129 noext = F2FS_OPTION(sbi).noextensions; 1130 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1131 1132 if (strlen(name) >= F2FS_EXTENSION_LEN || 1133 noext_cnt >= COMPRESS_EXT_NUM) { 1134 f2fs_err(sbi, 1135 "invalid extension length/number"); 1136 kfree(name); 1137 return -EINVAL; 1138 } 1139 1140 strcpy(noext[noext_cnt], name); 1141 F2FS_OPTION(sbi).nocompress_ext_cnt++; 1142 kfree(name); 1143 break; 1144 case Opt_compress_chksum: 1145 F2FS_OPTION(sbi).compress_chksum = true; 1146 break; 1147 case Opt_compress_mode: 1148 name = match_strdup(&args[0]); 1149 if (!name) 1150 return -ENOMEM; 1151 if (!strcmp(name, "fs")) { 1152 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1153 } else if (!strcmp(name, "user")) { 1154 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 1155 } else { 1156 kfree(name); 1157 return -EINVAL; 1158 } 1159 kfree(name); 1160 break; 1161 case Opt_compress_cache: 1162 set_opt(sbi, COMPRESS_CACHE); 1163 break; 1164 #else 1165 case Opt_compress_algorithm: 1166 case Opt_compress_log_size: 1167 case Opt_compress_extension: 1168 case Opt_nocompress_extension: 1169 case Opt_compress_chksum: 1170 case Opt_compress_mode: 1171 case Opt_compress_cache: 1172 f2fs_info(sbi, "compression options not supported"); 1173 break; 1174 #endif 1175 case Opt_atgc: 1176 set_opt(sbi, ATGC); 1177 break; 1178 case Opt_gc_merge: 1179 set_opt(sbi, GC_MERGE); 1180 break; 1181 case Opt_nogc_merge: 1182 clear_opt(sbi, GC_MERGE); 1183 break; 1184 case Opt_discard_unit: 1185 name = match_strdup(&args[0]); 1186 if (!name) 1187 return -ENOMEM; 1188 if (!strcmp(name, "block")) { 1189 F2FS_OPTION(sbi).discard_unit = 1190 DISCARD_UNIT_BLOCK; 1191 } else if (!strcmp(name, "segment")) { 1192 F2FS_OPTION(sbi).discard_unit = 1193 DISCARD_UNIT_SEGMENT; 1194 } else if (!strcmp(name, "section")) { 1195 F2FS_OPTION(sbi).discard_unit = 1196 DISCARD_UNIT_SECTION; 1197 } else { 1198 kfree(name); 1199 return -EINVAL; 1200 } 1201 kfree(name); 1202 break; 1203 default: 1204 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 1205 p); 1206 return -EINVAL; 1207 } 1208 } 1209 default_check: 1210 #ifdef CONFIG_QUOTA 1211 if (f2fs_check_quota_options(sbi)) 1212 return -EINVAL; 1213 #else 1214 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 1215 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1216 return -EINVAL; 1217 } 1218 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 1219 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1220 return -EINVAL; 1221 } 1222 #endif 1223 #ifndef CONFIG_UNICODE 1224 if (f2fs_sb_has_casefold(sbi)) { 1225 f2fs_err(sbi, 1226 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1227 return -EINVAL; 1228 } 1229 #endif 1230 /* 1231 * The BLKZONED feature indicates that the drive was formatted with 1232 * zone alignment optimization. This is optional for host-aware 1233 * devices, but mandatory for host-managed zoned block devices. 1234 */ 1235 #ifndef CONFIG_BLK_DEV_ZONED 1236 if (f2fs_sb_has_blkzoned(sbi)) { 1237 f2fs_err(sbi, "Zoned block device support is not enabled"); 1238 return -EINVAL; 1239 } 1240 #endif 1241 if (f2fs_sb_has_blkzoned(sbi)) { 1242 if (F2FS_OPTION(sbi).discard_unit != 1243 DISCARD_UNIT_SECTION) { 1244 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1245 F2FS_OPTION(sbi).discard_unit = 1246 DISCARD_UNIT_SECTION; 1247 } 1248 } 1249 1250 #ifdef CONFIG_F2FS_FS_COMPRESSION 1251 if (f2fs_test_compress_extension(sbi)) { 1252 f2fs_err(sbi, "invalid compress or nocompress extension"); 1253 return -EINVAL; 1254 } 1255 #endif 1256 1257 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 1258 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO", 1259 F2FS_IO_SIZE_KB(sbi)); 1260 return -EINVAL; 1261 } 1262 1263 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1264 int min_size, max_size; 1265 1266 if (!f2fs_sb_has_extra_attr(sbi) || 1267 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1268 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1269 return -EINVAL; 1270 } 1271 if (!test_opt(sbi, INLINE_XATTR)) { 1272 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1273 return -EINVAL; 1274 } 1275 1276 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 1277 max_size = MAX_INLINE_XATTR_SIZE; 1278 1279 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1280 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1281 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1282 min_size, max_size); 1283 return -EINVAL; 1284 } 1285 } 1286 1287 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1288 f2fs_err(sbi, "LFS not compatible with checkpoint=disable"); 1289 return -EINVAL; 1290 } 1291 1292 /* Not pass down write hints if the number of active logs is lesser 1293 * than NR_CURSEG_PERSIST_TYPE. 1294 */ 1295 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE) 1296 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1297 1298 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1299 f2fs_err(sbi, "Allow to mount readonly mode only"); 1300 return -EROFS; 1301 } 1302 return 0; 1303 } 1304 1305 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1306 { 1307 struct f2fs_inode_info *fi; 1308 1309 fi = f2fs_kmem_cache_alloc(f2fs_inode_cachep, 1310 GFP_F2FS_ZERO, false, F2FS_SB(sb)); 1311 if (!fi) 1312 return NULL; 1313 1314 init_once((void *) fi); 1315 1316 /* Initialize f2fs-specific inode info */ 1317 atomic_set(&fi->dirty_pages, 0); 1318 atomic_set(&fi->i_compr_blocks, 0); 1319 init_rwsem(&fi->i_sem); 1320 spin_lock_init(&fi->i_size_lock); 1321 INIT_LIST_HEAD(&fi->dirty_list); 1322 INIT_LIST_HEAD(&fi->gdirty_list); 1323 INIT_LIST_HEAD(&fi->inmem_ilist); 1324 INIT_LIST_HEAD(&fi->inmem_pages); 1325 mutex_init(&fi->inmem_lock); 1326 init_rwsem(&fi->i_gc_rwsem[READ]); 1327 init_rwsem(&fi->i_gc_rwsem[WRITE]); 1328 init_rwsem(&fi->i_xattr_sem); 1329 1330 /* Will be used by directory only */ 1331 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1332 1333 return &fi->vfs_inode; 1334 } 1335 1336 static int f2fs_drop_inode(struct inode *inode) 1337 { 1338 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1339 int ret; 1340 1341 /* 1342 * during filesystem shutdown, if checkpoint is disabled, 1343 * drop useless meta/node dirty pages. 1344 */ 1345 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1346 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1347 inode->i_ino == F2FS_META_INO(sbi)) { 1348 trace_f2fs_drop_inode(inode, 1); 1349 return 1; 1350 } 1351 } 1352 1353 /* 1354 * This is to avoid a deadlock condition like below. 1355 * writeback_single_inode(inode) 1356 * - f2fs_write_data_page 1357 * - f2fs_gc -> iput -> evict 1358 * - inode_wait_for_writeback(inode) 1359 */ 1360 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1361 if (!inode->i_nlink && !is_bad_inode(inode)) { 1362 /* to avoid evict_inode call simultaneously */ 1363 atomic_inc(&inode->i_count); 1364 spin_unlock(&inode->i_lock); 1365 1366 /* some remained atomic pages should discarded */ 1367 if (f2fs_is_atomic_file(inode)) 1368 f2fs_drop_inmem_pages(inode); 1369 1370 /* should remain fi->extent_tree for writepage */ 1371 f2fs_destroy_extent_node(inode); 1372 1373 sb_start_intwrite(inode->i_sb); 1374 f2fs_i_size_write(inode, 0); 1375 1376 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1377 inode, NULL, 0, DATA); 1378 truncate_inode_pages_final(inode->i_mapping); 1379 1380 if (F2FS_HAS_BLOCKS(inode)) 1381 f2fs_truncate(inode); 1382 1383 sb_end_intwrite(inode->i_sb); 1384 1385 spin_lock(&inode->i_lock); 1386 atomic_dec(&inode->i_count); 1387 } 1388 trace_f2fs_drop_inode(inode, 0); 1389 return 0; 1390 } 1391 ret = generic_drop_inode(inode); 1392 if (!ret) 1393 ret = fscrypt_drop_inode(inode); 1394 trace_f2fs_drop_inode(inode, ret); 1395 return ret; 1396 } 1397 1398 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1399 { 1400 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1401 int ret = 0; 1402 1403 spin_lock(&sbi->inode_lock[DIRTY_META]); 1404 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1405 ret = 1; 1406 } else { 1407 set_inode_flag(inode, FI_DIRTY_INODE); 1408 stat_inc_dirty_inode(sbi, DIRTY_META); 1409 } 1410 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1411 list_add_tail(&F2FS_I(inode)->gdirty_list, 1412 &sbi->inode_list[DIRTY_META]); 1413 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1414 } 1415 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1416 return ret; 1417 } 1418 1419 void f2fs_inode_synced(struct inode *inode) 1420 { 1421 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1422 1423 spin_lock(&sbi->inode_lock[DIRTY_META]); 1424 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1425 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1426 return; 1427 } 1428 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1429 list_del_init(&F2FS_I(inode)->gdirty_list); 1430 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1431 } 1432 clear_inode_flag(inode, FI_DIRTY_INODE); 1433 clear_inode_flag(inode, FI_AUTO_RECOVER); 1434 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1435 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1436 } 1437 1438 /* 1439 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1440 * 1441 * We should call set_dirty_inode to write the dirty inode through write_inode. 1442 */ 1443 static void f2fs_dirty_inode(struct inode *inode, int flags) 1444 { 1445 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1446 1447 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1448 inode->i_ino == F2FS_META_INO(sbi)) 1449 return; 1450 1451 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1452 clear_inode_flag(inode, FI_AUTO_RECOVER); 1453 1454 f2fs_inode_dirtied(inode, false); 1455 } 1456 1457 static void f2fs_free_inode(struct inode *inode) 1458 { 1459 fscrypt_free_inode(inode); 1460 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1461 } 1462 1463 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1464 { 1465 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1466 percpu_counter_destroy(&sbi->total_valid_inode_count); 1467 } 1468 1469 static void destroy_device_list(struct f2fs_sb_info *sbi) 1470 { 1471 int i; 1472 1473 for (i = 0; i < sbi->s_ndevs; i++) { 1474 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1475 #ifdef CONFIG_BLK_DEV_ZONED 1476 kvfree(FDEV(i).blkz_seq); 1477 kfree(FDEV(i).zone_capacity_blocks); 1478 #endif 1479 } 1480 kvfree(sbi->devs); 1481 } 1482 1483 static void f2fs_put_super(struct super_block *sb) 1484 { 1485 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1486 int i; 1487 bool dropped; 1488 1489 /* unregister procfs/sysfs entries in advance to avoid race case */ 1490 f2fs_unregister_sysfs(sbi); 1491 1492 f2fs_quota_off_umount(sb); 1493 1494 /* prevent remaining shrinker jobs */ 1495 mutex_lock(&sbi->umount_mutex); 1496 1497 /* 1498 * flush all issued checkpoints and stop checkpoint issue thread. 1499 * after then, all checkpoints should be done by each process context. 1500 */ 1501 f2fs_stop_ckpt_thread(sbi); 1502 1503 /* 1504 * We don't need to do checkpoint when superblock is clean. 1505 * But, the previous checkpoint was not done by umount, it needs to do 1506 * clean checkpoint again. 1507 */ 1508 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1509 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1510 struct cp_control cpc = { 1511 .reason = CP_UMOUNT, 1512 }; 1513 f2fs_write_checkpoint(sbi, &cpc); 1514 } 1515 1516 /* be sure to wait for any on-going discard commands */ 1517 dropped = f2fs_issue_discard_timeout(sbi); 1518 1519 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && 1520 !sbi->discard_blks && !dropped) { 1521 struct cp_control cpc = { 1522 .reason = CP_UMOUNT | CP_TRIMMED, 1523 }; 1524 f2fs_write_checkpoint(sbi, &cpc); 1525 } 1526 1527 /* 1528 * normally superblock is clean, so we need to release this. 1529 * In addition, EIO will skip do checkpoint, we need this as well. 1530 */ 1531 f2fs_release_ino_entry(sbi, true); 1532 1533 f2fs_leave_shrinker(sbi); 1534 mutex_unlock(&sbi->umount_mutex); 1535 1536 /* our cp_error case, we can wait for any writeback page */ 1537 f2fs_flush_merged_writes(sbi); 1538 1539 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1540 1541 f2fs_bug_on(sbi, sbi->fsync_node_num); 1542 1543 f2fs_destroy_compress_inode(sbi); 1544 1545 iput(sbi->node_inode); 1546 sbi->node_inode = NULL; 1547 1548 iput(sbi->meta_inode); 1549 sbi->meta_inode = NULL; 1550 1551 /* 1552 * iput() can update stat information, if f2fs_write_checkpoint() 1553 * above failed with error. 1554 */ 1555 f2fs_destroy_stats(sbi); 1556 1557 /* destroy f2fs internal modules */ 1558 f2fs_destroy_node_manager(sbi); 1559 f2fs_destroy_segment_manager(sbi); 1560 1561 f2fs_destroy_post_read_wq(sbi); 1562 1563 kvfree(sbi->ckpt); 1564 1565 sb->s_fs_info = NULL; 1566 if (sbi->s_chksum_driver) 1567 crypto_free_shash(sbi->s_chksum_driver); 1568 kfree(sbi->raw_super); 1569 1570 destroy_device_list(sbi); 1571 f2fs_destroy_page_array_cache(sbi); 1572 f2fs_destroy_xattr_caches(sbi); 1573 mempool_destroy(sbi->write_io_dummy); 1574 #ifdef CONFIG_QUOTA 1575 for (i = 0; i < MAXQUOTAS; i++) 1576 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1577 #endif 1578 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1579 destroy_percpu_info(sbi); 1580 f2fs_destroy_iostat(sbi); 1581 for (i = 0; i < NR_PAGE_TYPE; i++) 1582 kvfree(sbi->write_io[i]); 1583 #ifdef CONFIG_UNICODE 1584 utf8_unload(sb->s_encoding); 1585 #endif 1586 kfree(sbi); 1587 } 1588 1589 int f2fs_sync_fs(struct super_block *sb, int sync) 1590 { 1591 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1592 int err = 0; 1593 1594 if (unlikely(f2fs_cp_error(sbi))) 1595 return 0; 1596 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1597 return 0; 1598 1599 trace_f2fs_sync_fs(sb, sync); 1600 1601 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1602 return -EAGAIN; 1603 1604 if (sync) 1605 err = f2fs_issue_checkpoint(sbi); 1606 1607 return err; 1608 } 1609 1610 static int f2fs_freeze(struct super_block *sb) 1611 { 1612 if (f2fs_readonly(sb)) 1613 return 0; 1614 1615 /* IO error happened before */ 1616 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1617 return -EIO; 1618 1619 /* must be clean, since sync_filesystem() was already called */ 1620 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1621 return -EINVAL; 1622 1623 /* ensure no checkpoint required */ 1624 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list)) 1625 return -EINVAL; 1626 return 0; 1627 } 1628 1629 static int f2fs_unfreeze(struct super_block *sb) 1630 { 1631 return 0; 1632 } 1633 1634 #ifdef CONFIG_QUOTA 1635 static int f2fs_statfs_project(struct super_block *sb, 1636 kprojid_t projid, struct kstatfs *buf) 1637 { 1638 struct kqid qid; 1639 struct dquot *dquot; 1640 u64 limit; 1641 u64 curblock; 1642 1643 qid = make_kqid_projid(projid); 1644 dquot = dqget(sb, qid); 1645 if (IS_ERR(dquot)) 1646 return PTR_ERR(dquot); 1647 spin_lock(&dquot->dq_dqb_lock); 1648 1649 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1650 dquot->dq_dqb.dqb_bhardlimit); 1651 if (limit) 1652 limit >>= sb->s_blocksize_bits; 1653 1654 if (limit && buf->f_blocks > limit) { 1655 curblock = (dquot->dq_dqb.dqb_curspace + 1656 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1657 buf->f_blocks = limit; 1658 buf->f_bfree = buf->f_bavail = 1659 (buf->f_blocks > curblock) ? 1660 (buf->f_blocks - curblock) : 0; 1661 } 1662 1663 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1664 dquot->dq_dqb.dqb_ihardlimit); 1665 1666 if (limit && buf->f_files > limit) { 1667 buf->f_files = limit; 1668 buf->f_ffree = 1669 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1670 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1671 } 1672 1673 spin_unlock(&dquot->dq_dqb_lock); 1674 dqput(dquot); 1675 return 0; 1676 } 1677 #endif 1678 1679 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1680 { 1681 struct super_block *sb = dentry->d_sb; 1682 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1683 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1684 block_t total_count, user_block_count, start_count; 1685 u64 avail_node_count; 1686 1687 total_count = le64_to_cpu(sbi->raw_super->block_count); 1688 user_block_count = sbi->user_block_count; 1689 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1690 buf->f_type = F2FS_SUPER_MAGIC; 1691 buf->f_bsize = sbi->blocksize; 1692 1693 buf->f_blocks = total_count - start_count; 1694 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1695 sbi->current_reserved_blocks; 1696 1697 spin_lock(&sbi->stat_lock); 1698 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1699 buf->f_bfree = 0; 1700 else 1701 buf->f_bfree -= sbi->unusable_block_count; 1702 spin_unlock(&sbi->stat_lock); 1703 1704 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1705 buf->f_bavail = buf->f_bfree - 1706 F2FS_OPTION(sbi).root_reserved_blocks; 1707 else 1708 buf->f_bavail = 0; 1709 1710 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1711 1712 if (avail_node_count > user_block_count) { 1713 buf->f_files = user_block_count; 1714 buf->f_ffree = buf->f_bavail; 1715 } else { 1716 buf->f_files = avail_node_count; 1717 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1718 buf->f_bavail); 1719 } 1720 1721 buf->f_namelen = F2FS_NAME_LEN; 1722 buf->f_fsid = u64_to_fsid(id); 1723 1724 #ifdef CONFIG_QUOTA 1725 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1726 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1727 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1728 } 1729 #endif 1730 return 0; 1731 } 1732 1733 static inline void f2fs_show_quota_options(struct seq_file *seq, 1734 struct super_block *sb) 1735 { 1736 #ifdef CONFIG_QUOTA 1737 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1738 1739 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1740 char *fmtname = ""; 1741 1742 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1743 case QFMT_VFS_OLD: 1744 fmtname = "vfsold"; 1745 break; 1746 case QFMT_VFS_V0: 1747 fmtname = "vfsv0"; 1748 break; 1749 case QFMT_VFS_V1: 1750 fmtname = "vfsv1"; 1751 break; 1752 } 1753 seq_printf(seq, ",jqfmt=%s", fmtname); 1754 } 1755 1756 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1757 seq_show_option(seq, "usrjquota", 1758 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1759 1760 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1761 seq_show_option(seq, "grpjquota", 1762 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1763 1764 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1765 seq_show_option(seq, "prjjquota", 1766 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1767 #endif 1768 } 1769 1770 #ifdef CONFIG_F2FS_FS_COMPRESSION 1771 static inline void f2fs_show_compress_options(struct seq_file *seq, 1772 struct super_block *sb) 1773 { 1774 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1775 char *algtype = ""; 1776 int i; 1777 1778 if (!f2fs_sb_has_compression(sbi)) 1779 return; 1780 1781 switch (F2FS_OPTION(sbi).compress_algorithm) { 1782 case COMPRESS_LZO: 1783 algtype = "lzo"; 1784 break; 1785 case COMPRESS_LZ4: 1786 algtype = "lz4"; 1787 break; 1788 case COMPRESS_ZSTD: 1789 algtype = "zstd"; 1790 break; 1791 case COMPRESS_LZORLE: 1792 algtype = "lzo-rle"; 1793 break; 1794 } 1795 seq_printf(seq, ",compress_algorithm=%s", algtype); 1796 1797 if (F2FS_OPTION(sbi).compress_level) 1798 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1799 1800 seq_printf(seq, ",compress_log_size=%u", 1801 F2FS_OPTION(sbi).compress_log_size); 1802 1803 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1804 seq_printf(seq, ",compress_extension=%s", 1805 F2FS_OPTION(sbi).extensions[i]); 1806 } 1807 1808 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1809 seq_printf(seq, ",nocompress_extension=%s", 1810 F2FS_OPTION(sbi).noextensions[i]); 1811 } 1812 1813 if (F2FS_OPTION(sbi).compress_chksum) 1814 seq_puts(seq, ",compress_chksum"); 1815 1816 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1817 seq_printf(seq, ",compress_mode=%s", "fs"); 1818 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1819 seq_printf(seq, ",compress_mode=%s", "user"); 1820 1821 if (test_opt(sbi, COMPRESS_CACHE)) 1822 seq_puts(seq, ",compress_cache"); 1823 } 1824 #endif 1825 1826 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1827 { 1828 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1829 1830 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1831 seq_printf(seq, ",background_gc=%s", "sync"); 1832 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1833 seq_printf(seq, ",background_gc=%s", "on"); 1834 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1835 seq_printf(seq, ",background_gc=%s", "off"); 1836 1837 if (test_opt(sbi, GC_MERGE)) 1838 seq_puts(seq, ",gc_merge"); 1839 1840 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1841 seq_puts(seq, ",disable_roll_forward"); 1842 if (test_opt(sbi, NORECOVERY)) 1843 seq_puts(seq, ",norecovery"); 1844 if (test_opt(sbi, DISCARD)) 1845 seq_puts(seq, ",discard"); 1846 else 1847 seq_puts(seq, ",nodiscard"); 1848 if (test_opt(sbi, NOHEAP)) 1849 seq_puts(seq, ",no_heap"); 1850 else 1851 seq_puts(seq, ",heap"); 1852 #ifdef CONFIG_F2FS_FS_XATTR 1853 if (test_opt(sbi, XATTR_USER)) 1854 seq_puts(seq, ",user_xattr"); 1855 else 1856 seq_puts(seq, ",nouser_xattr"); 1857 if (test_opt(sbi, INLINE_XATTR)) 1858 seq_puts(seq, ",inline_xattr"); 1859 else 1860 seq_puts(seq, ",noinline_xattr"); 1861 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1862 seq_printf(seq, ",inline_xattr_size=%u", 1863 F2FS_OPTION(sbi).inline_xattr_size); 1864 #endif 1865 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1866 if (test_opt(sbi, POSIX_ACL)) 1867 seq_puts(seq, ",acl"); 1868 else 1869 seq_puts(seq, ",noacl"); 1870 #endif 1871 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1872 seq_puts(seq, ",disable_ext_identify"); 1873 if (test_opt(sbi, INLINE_DATA)) 1874 seq_puts(seq, ",inline_data"); 1875 else 1876 seq_puts(seq, ",noinline_data"); 1877 if (test_opt(sbi, INLINE_DENTRY)) 1878 seq_puts(seq, ",inline_dentry"); 1879 else 1880 seq_puts(seq, ",noinline_dentry"); 1881 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1882 seq_puts(seq, ",flush_merge"); 1883 if (test_opt(sbi, NOBARRIER)) 1884 seq_puts(seq, ",nobarrier"); 1885 if (test_opt(sbi, FASTBOOT)) 1886 seq_puts(seq, ",fastboot"); 1887 if (test_opt(sbi, EXTENT_CACHE)) 1888 seq_puts(seq, ",extent_cache"); 1889 else 1890 seq_puts(seq, ",noextent_cache"); 1891 if (test_opt(sbi, DATA_FLUSH)) 1892 seq_puts(seq, ",data_flush"); 1893 1894 seq_puts(seq, ",mode="); 1895 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1896 seq_puts(seq, "adaptive"); 1897 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1898 seq_puts(seq, "lfs"); 1899 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1900 if (test_opt(sbi, RESERVE_ROOT)) 1901 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1902 F2FS_OPTION(sbi).root_reserved_blocks, 1903 from_kuid_munged(&init_user_ns, 1904 F2FS_OPTION(sbi).s_resuid), 1905 from_kgid_munged(&init_user_ns, 1906 F2FS_OPTION(sbi).s_resgid)); 1907 if (F2FS_IO_SIZE_BITS(sbi)) 1908 seq_printf(seq, ",io_bits=%u", 1909 F2FS_OPTION(sbi).write_io_size_bits); 1910 #ifdef CONFIG_F2FS_FAULT_INJECTION 1911 if (test_opt(sbi, FAULT_INJECTION)) { 1912 seq_printf(seq, ",fault_injection=%u", 1913 F2FS_OPTION(sbi).fault_info.inject_rate); 1914 seq_printf(seq, ",fault_type=%u", 1915 F2FS_OPTION(sbi).fault_info.inject_type); 1916 } 1917 #endif 1918 #ifdef CONFIG_QUOTA 1919 if (test_opt(sbi, QUOTA)) 1920 seq_puts(seq, ",quota"); 1921 if (test_opt(sbi, USRQUOTA)) 1922 seq_puts(seq, ",usrquota"); 1923 if (test_opt(sbi, GRPQUOTA)) 1924 seq_puts(seq, ",grpquota"); 1925 if (test_opt(sbi, PRJQUOTA)) 1926 seq_puts(seq, ",prjquota"); 1927 #endif 1928 f2fs_show_quota_options(seq, sbi->sb); 1929 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1930 seq_printf(seq, ",whint_mode=%s", "user-based"); 1931 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1932 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1933 1934 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1935 1936 if (sbi->sb->s_flags & SB_INLINECRYPT) 1937 seq_puts(seq, ",inlinecrypt"); 1938 1939 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1940 seq_printf(seq, ",alloc_mode=%s", "default"); 1941 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1942 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1943 1944 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1945 seq_printf(seq, ",checkpoint=disable:%u", 1946 F2FS_OPTION(sbi).unusable_cap); 1947 if (test_opt(sbi, MERGE_CHECKPOINT)) 1948 seq_puts(seq, ",checkpoint_merge"); 1949 else 1950 seq_puts(seq, ",nocheckpoint_merge"); 1951 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1952 seq_printf(seq, ",fsync_mode=%s", "posix"); 1953 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1954 seq_printf(seq, ",fsync_mode=%s", "strict"); 1955 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1956 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 1957 1958 #ifdef CONFIG_F2FS_FS_COMPRESSION 1959 f2fs_show_compress_options(seq, sbi->sb); 1960 #endif 1961 1962 if (test_opt(sbi, ATGC)) 1963 seq_puts(seq, ",atgc"); 1964 1965 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 1966 seq_printf(seq, ",discard_unit=%s", "block"); 1967 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 1968 seq_printf(seq, ",discard_unit=%s", "segment"); 1969 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 1970 seq_printf(seq, ",discard_unit=%s", "section"); 1971 1972 return 0; 1973 } 1974 1975 static void default_options(struct f2fs_sb_info *sbi) 1976 { 1977 /* init some FS parameters */ 1978 if (f2fs_sb_has_readonly(sbi)) 1979 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 1980 else 1981 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 1982 1983 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 1984 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1985 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1986 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1987 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 1988 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 1989 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 1990 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 1991 F2FS_OPTION(sbi).compress_ext_cnt = 0; 1992 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1993 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 1994 1995 sbi->sb->s_flags &= ~SB_INLINECRYPT; 1996 1997 set_opt(sbi, INLINE_XATTR); 1998 set_opt(sbi, INLINE_DATA); 1999 set_opt(sbi, INLINE_DENTRY); 2000 set_opt(sbi, EXTENT_CACHE); 2001 set_opt(sbi, NOHEAP); 2002 clear_opt(sbi, DISABLE_CHECKPOINT); 2003 set_opt(sbi, MERGE_CHECKPOINT); 2004 F2FS_OPTION(sbi).unusable_cap = 0; 2005 sbi->sb->s_flags |= SB_LAZYTIME; 2006 set_opt(sbi, FLUSH_MERGE); 2007 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2008 set_opt(sbi, DISCARD); 2009 if (f2fs_sb_has_blkzoned(sbi)) { 2010 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2011 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2012 } else { 2013 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2014 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2015 } 2016 2017 #ifdef CONFIG_F2FS_FS_XATTR 2018 set_opt(sbi, XATTR_USER); 2019 #endif 2020 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2021 set_opt(sbi, POSIX_ACL); 2022 #endif 2023 2024 f2fs_build_fault_attr(sbi, 0, 0); 2025 } 2026 2027 #ifdef CONFIG_QUOTA 2028 static int f2fs_enable_quotas(struct super_block *sb); 2029 #endif 2030 2031 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2032 { 2033 unsigned int s_flags = sbi->sb->s_flags; 2034 struct cp_control cpc; 2035 int err = 0; 2036 int ret; 2037 block_t unusable; 2038 2039 if (s_flags & SB_RDONLY) { 2040 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2041 return -EINVAL; 2042 } 2043 sbi->sb->s_flags |= SB_ACTIVE; 2044 2045 f2fs_update_time(sbi, DISABLE_TIME); 2046 2047 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2048 down_write(&sbi->gc_lock); 2049 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO); 2050 if (err == -ENODATA) { 2051 err = 0; 2052 break; 2053 } 2054 if (err && err != -EAGAIN) 2055 break; 2056 } 2057 2058 ret = sync_filesystem(sbi->sb); 2059 if (ret || err) { 2060 err = ret ? ret : err; 2061 goto restore_flag; 2062 } 2063 2064 unusable = f2fs_get_unusable_blocks(sbi); 2065 if (f2fs_disable_cp_again(sbi, unusable)) { 2066 err = -EAGAIN; 2067 goto restore_flag; 2068 } 2069 2070 down_write(&sbi->gc_lock); 2071 cpc.reason = CP_PAUSE; 2072 set_sbi_flag(sbi, SBI_CP_DISABLED); 2073 err = f2fs_write_checkpoint(sbi, &cpc); 2074 if (err) 2075 goto out_unlock; 2076 2077 spin_lock(&sbi->stat_lock); 2078 sbi->unusable_block_count = unusable; 2079 spin_unlock(&sbi->stat_lock); 2080 2081 out_unlock: 2082 up_write(&sbi->gc_lock); 2083 restore_flag: 2084 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2085 return err; 2086 } 2087 2088 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2089 { 2090 int retry = DEFAULT_RETRY_IO_COUNT; 2091 2092 /* we should flush all the data to keep data consistency */ 2093 do { 2094 sync_inodes_sb(sbi->sb); 2095 cond_resched(); 2096 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT); 2097 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2098 2099 if (unlikely(retry < 0)) 2100 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2101 2102 down_write(&sbi->gc_lock); 2103 f2fs_dirty_to_prefree(sbi); 2104 2105 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2106 set_sbi_flag(sbi, SBI_IS_DIRTY); 2107 up_write(&sbi->gc_lock); 2108 2109 f2fs_sync_fs(sbi->sb, 1); 2110 } 2111 2112 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2113 { 2114 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2115 struct f2fs_mount_info org_mount_opt; 2116 unsigned long old_sb_flags; 2117 int err; 2118 bool need_restart_gc = false, need_stop_gc = false; 2119 bool need_restart_ckpt = false, need_stop_ckpt = false; 2120 bool need_restart_flush = false, need_stop_flush = false; 2121 bool need_restart_discard = false, need_stop_discard = false; 2122 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 2123 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2124 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 2125 bool no_atgc = !test_opt(sbi, ATGC); 2126 bool no_discard = !test_opt(sbi, DISCARD); 2127 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2128 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2129 struct discard_cmd_control *dcc; 2130 #ifdef CONFIG_QUOTA 2131 int i, j; 2132 #endif 2133 2134 /* 2135 * Save the old mount options in case we 2136 * need to restore them. 2137 */ 2138 org_mount_opt = sbi->mount_opt; 2139 old_sb_flags = sb->s_flags; 2140 2141 #ifdef CONFIG_QUOTA 2142 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2143 for (i = 0; i < MAXQUOTAS; i++) { 2144 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2145 org_mount_opt.s_qf_names[i] = 2146 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2147 GFP_KERNEL); 2148 if (!org_mount_opt.s_qf_names[i]) { 2149 for (j = 0; j < i; j++) 2150 kfree(org_mount_opt.s_qf_names[j]); 2151 return -ENOMEM; 2152 } 2153 } else { 2154 org_mount_opt.s_qf_names[i] = NULL; 2155 } 2156 } 2157 #endif 2158 2159 /* recover superblocks we couldn't write due to previous RO mount */ 2160 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2161 err = f2fs_commit_super(sbi, false); 2162 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2163 err); 2164 if (!err) 2165 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2166 } 2167 2168 default_options(sbi); 2169 2170 /* parse mount options */ 2171 err = parse_options(sb, data, true); 2172 if (err) 2173 goto restore_opts; 2174 2175 /* 2176 * Previous and new state of filesystem is RO, 2177 * so skip checking GC and FLUSH_MERGE conditions. 2178 */ 2179 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2180 goto skip; 2181 2182 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) { 2183 err = -EROFS; 2184 goto restore_opts; 2185 } 2186 2187 #ifdef CONFIG_QUOTA 2188 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2189 err = dquot_suspend(sb, -1); 2190 if (err < 0) 2191 goto restore_opts; 2192 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2193 /* dquot_resume needs RW */ 2194 sb->s_flags &= ~SB_RDONLY; 2195 if (sb_any_quota_suspended(sb)) { 2196 dquot_resume(sb, -1); 2197 } else if (f2fs_sb_has_quota_ino(sbi)) { 2198 err = f2fs_enable_quotas(sb); 2199 if (err) 2200 goto restore_opts; 2201 } 2202 } 2203 #endif 2204 /* disallow enable atgc dynamically */ 2205 if (no_atgc == !!test_opt(sbi, ATGC)) { 2206 err = -EINVAL; 2207 f2fs_warn(sbi, "switch atgc option is not allowed"); 2208 goto restore_opts; 2209 } 2210 2211 /* disallow enable/disable extent_cache dynamically */ 2212 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 2213 err = -EINVAL; 2214 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2215 goto restore_opts; 2216 } 2217 2218 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 2219 err = -EINVAL; 2220 f2fs_warn(sbi, "switch io_bits option is not allowed"); 2221 goto restore_opts; 2222 } 2223 2224 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2225 err = -EINVAL; 2226 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2227 goto restore_opts; 2228 } 2229 2230 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2231 err = -EINVAL; 2232 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2233 goto restore_opts; 2234 } 2235 2236 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2237 err = -EINVAL; 2238 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2239 goto restore_opts; 2240 } 2241 2242 /* 2243 * We stop the GC thread if FS is mounted as RO 2244 * or if background_gc = off is passed in mount 2245 * option. Also sync the filesystem. 2246 */ 2247 if ((*flags & SB_RDONLY) || 2248 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2249 !test_opt(sbi, GC_MERGE))) { 2250 if (sbi->gc_thread) { 2251 f2fs_stop_gc_thread(sbi); 2252 need_restart_gc = true; 2253 } 2254 } else if (!sbi->gc_thread) { 2255 err = f2fs_start_gc_thread(sbi); 2256 if (err) 2257 goto restore_opts; 2258 need_stop_gc = true; 2259 } 2260 2261 if (*flags & SB_RDONLY || 2262 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 2263 sync_inodes_sb(sb); 2264 2265 set_sbi_flag(sbi, SBI_IS_DIRTY); 2266 set_sbi_flag(sbi, SBI_IS_CLOSE); 2267 f2fs_sync_fs(sb, 1); 2268 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2269 } 2270 2271 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2272 !test_opt(sbi, MERGE_CHECKPOINT)) { 2273 f2fs_stop_ckpt_thread(sbi); 2274 need_restart_ckpt = true; 2275 } else { 2276 err = f2fs_start_ckpt_thread(sbi); 2277 if (err) { 2278 f2fs_err(sbi, 2279 "Failed to start F2FS issue_checkpoint_thread (%d)", 2280 err); 2281 goto restore_gc; 2282 } 2283 need_stop_ckpt = true; 2284 } 2285 2286 /* 2287 * We stop issue flush thread if FS is mounted as RO 2288 * or if flush_merge is not passed in mount option. 2289 */ 2290 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2291 clear_opt(sbi, FLUSH_MERGE); 2292 f2fs_destroy_flush_cmd_control(sbi, false); 2293 need_restart_flush = true; 2294 } else { 2295 err = f2fs_create_flush_cmd_control(sbi); 2296 if (err) 2297 goto restore_ckpt; 2298 need_stop_flush = true; 2299 } 2300 2301 if (no_discard == !!test_opt(sbi, DISCARD)) { 2302 if (test_opt(sbi, DISCARD)) { 2303 err = f2fs_start_discard_thread(sbi); 2304 if (err) 2305 goto restore_flush; 2306 need_stop_discard = true; 2307 } else { 2308 dcc = SM_I(sbi)->dcc_info; 2309 f2fs_stop_discard_thread(sbi); 2310 if (atomic_read(&dcc->discard_cmd_cnt)) 2311 f2fs_issue_discard_timeout(sbi); 2312 need_restart_discard = true; 2313 } 2314 } 2315 2316 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2317 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2318 err = f2fs_disable_checkpoint(sbi); 2319 if (err) 2320 goto restore_discard; 2321 } else { 2322 f2fs_enable_checkpoint(sbi); 2323 } 2324 } 2325 2326 skip: 2327 #ifdef CONFIG_QUOTA 2328 /* Release old quota file names */ 2329 for (i = 0; i < MAXQUOTAS; i++) 2330 kfree(org_mount_opt.s_qf_names[i]); 2331 #endif 2332 /* Update the POSIXACL Flag */ 2333 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2334 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2335 2336 limit_reserve_root(sbi); 2337 adjust_unusable_cap_perc(sbi); 2338 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2339 return 0; 2340 restore_discard: 2341 if (need_restart_discard) { 2342 if (f2fs_start_discard_thread(sbi)) 2343 f2fs_warn(sbi, "discard has been stopped"); 2344 } else if (need_stop_discard) { 2345 f2fs_stop_discard_thread(sbi); 2346 } 2347 restore_flush: 2348 if (need_restart_flush) { 2349 if (f2fs_create_flush_cmd_control(sbi)) 2350 f2fs_warn(sbi, "background flush thread has stopped"); 2351 } else if (need_stop_flush) { 2352 clear_opt(sbi, FLUSH_MERGE); 2353 f2fs_destroy_flush_cmd_control(sbi, false); 2354 } 2355 restore_ckpt: 2356 if (need_restart_ckpt) { 2357 if (f2fs_start_ckpt_thread(sbi)) 2358 f2fs_warn(sbi, "background ckpt thread has stopped"); 2359 } else if (need_stop_ckpt) { 2360 f2fs_stop_ckpt_thread(sbi); 2361 } 2362 restore_gc: 2363 if (need_restart_gc) { 2364 if (f2fs_start_gc_thread(sbi)) 2365 f2fs_warn(sbi, "background gc thread has stopped"); 2366 } else if (need_stop_gc) { 2367 f2fs_stop_gc_thread(sbi); 2368 } 2369 restore_opts: 2370 #ifdef CONFIG_QUOTA 2371 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2372 for (i = 0; i < MAXQUOTAS; i++) { 2373 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2374 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2375 } 2376 #endif 2377 sbi->mount_opt = org_mount_opt; 2378 sb->s_flags = old_sb_flags; 2379 return err; 2380 } 2381 2382 #ifdef CONFIG_QUOTA 2383 /* Read data from quotafile */ 2384 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2385 size_t len, loff_t off) 2386 { 2387 struct inode *inode = sb_dqopt(sb)->files[type]; 2388 struct address_space *mapping = inode->i_mapping; 2389 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2390 int offset = off & (sb->s_blocksize - 1); 2391 int tocopy; 2392 size_t toread; 2393 loff_t i_size = i_size_read(inode); 2394 struct page *page; 2395 char *kaddr; 2396 2397 if (off > i_size) 2398 return 0; 2399 2400 if (off + len > i_size) 2401 len = i_size - off; 2402 toread = len; 2403 while (toread > 0) { 2404 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2405 repeat: 2406 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2407 if (IS_ERR(page)) { 2408 if (PTR_ERR(page) == -ENOMEM) { 2409 congestion_wait(BLK_RW_ASYNC, 2410 DEFAULT_IO_TIMEOUT); 2411 goto repeat; 2412 } 2413 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2414 return PTR_ERR(page); 2415 } 2416 2417 lock_page(page); 2418 2419 if (unlikely(page->mapping != mapping)) { 2420 f2fs_put_page(page, 1); 2421 goto repeat; 2422 } 2423 if (unlikely(!PageUptodate(page))) { 2424 f2fs_put_page(page, 1); 2425 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2426 return -EIO; 2427 } 2428 2429 kaddr = kmap_atomic(page); 2430 memcpy(data, kaddr + offset, tocopy); 2431 kunmap_atomic(kaddr); 2432 f2fs_put_page(page, 1); 2433 2434 offset = 0; 2435 toread -= tocopy; 2436 data += tocopy; 2437 blkidx++; 2438 } 2439 return len; 2440 } 2441 2442 /* Write to quotafile */ 2443 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2444 const char *data, size_t len, loff_t off) 2445 { 2446 struct inode *inode = sb_dqopt(sb)->files[type]; 2447 struct address_space *mapping = inode->i_mapping; 2448 const struct address_space_operations *a_ops = mapping->a_ops; 2449 int offset = off & (sb->s_blocksize - 1); 2450 size_t towrite = len; 2451 struct page *page; 2452 void *fsdata = NULL; 2453 char *kaddr; 2454 int err = 0; 2455 int tocopy; 2456 2457 while (towrite > 0) { 2458 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2459 towrite); 2460 retry: 2461 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2462 &page, &fsdata); 2463 if (unlikely(err)) { 2464 if (err == -ENOMEM) { 2465 congestion_wait(BLK_RW_ASYNC, 2466 DEFAULT_IO_TIMEOUT); 2467 goto retry; 2468 } 2469 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2470 break; 2471 } 2472 2473 kaddr = kmap_atomic(page); 2474 memcpy(kaddr + offset, data, tocopy); 2475 kunmap_atomic(kaddr); 2476 flush_dcache_page(page); 2477 2478 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2479 page, fsdata); 2480 offset = 0; 2481 towrite -= tocopy; 2482 off += tocopy; 2483 data += tocopy; 2484 cond_resched(); 2485 } 2486 2487 if (len == towrite) 2488 return err; 2489 inode->i_mtime = inode->i_ctime = current_time(inode); 2490 f2fs_mark_inode_dirty_sync(inode, false); 2491 return len - towrite; 2492 } 2493 2494 static struct dquot **f2fs_get_dquots(struct inode *inode) 2495 { 2496 return F2FS_I(inode)->i_dquot; 2497 } 2498 2499 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2500 { 2501 return &F2FS_I(inode)->i_reserved_quota; 2502 } 2503 2504 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2505 { 2506 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2507 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2508 return 0; 2509 } 2510 2511 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2512 F2FS_OPTION(sbi).s_jquota_fmt, type); 2513 } 2514 2515 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2516 { 2517 int enabled = 0; 2518 int i, err; 2519 2520 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2521 err = f2fs_enable_quotas(sbi->sb); 2522 if (err) { 2523 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2524 return 0; 2525 } 2526 return 1; 2527 } 2528 2529 for (i = 0; i < MAXQUOTAS; i++) { 2530 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2531 err = f2fs_quota_on_mount(sbi, i); 2532 if (!err) { 2533 enabled = 1; 2534 continue; 2535 } 2536 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2537 err, i); 2538 } 2539 } 2540 return enabled; 2541 } 2542 2543 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2544 unsigned int flags) 2545 { 2546 struct inode *qf_inode; 2547 unsigned long qf_inum; 2548 int err; 2549 2550 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2551 2552 qf_inum = f2fs_qf_ino(sb, type); 2553 if (!qf_inum) 2554 return -EPERM; 2555 2556 qf_inode = f2fs_iget(sb, qf_inum); 2557 if (IS_ERR(qf_inode)) { 2558 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2559 return PTR_ERR(qf_inode); 2560 } 2561 2562 /* Don't account quota for quota files to avoid recursion */ 2563 qf_inode->i_flags |= S_NOQUOTA; 2564 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2565 iput(qf_inode); 2566 return err; 2567 } 2568 2569 static int f2fs_enable_quotas(struct super_block *sb) 2570 { 2571 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2572 int type, err = 0; 2573 unsigned long qf_inum; 2574 bool quota_mopt[MAXQUOTAS] = { 2575 test_opt(sbi, USRQUOTA), 2576 test_opt(sbi, GRPQUOTA), 2577 test_opt(sbi, PRJQUOTA), 2578 }; 2579 2580 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2581 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2582 return 0; 2583 } 2584 2585 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2586 2587 for (type = 0; type < MAXQUOTAS; type++) { 2588 qf_inum = f2fs_qf_ino(sb, type); 2589 if (qf_inum) { 2590 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2591 DQUOT_USAGE_ENABLED | 2592 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2593 if (err) { 2594 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2595 type, err); 2596 for (type--; type >= 0; type--) 2597 dquot_quota_off(sb, type); 2598 set_sbi_flag(F2FS_SB(sb), 2599 SBI_QUOTA_NEED_REPAIR); 2600 return err; 2601 } 2602 } 2603 } 2604 return 0; 2605 } 2606 2607 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 2608 { 2609 struct quota_info *dqopt = sb_dqopt(sbi->sb); 2610 struct address_space *mapping = dqopt->files[type]->i_mapping; 2611 int ret = 0; 2612 2613 ret = dquot_writeback_dquots(sbi->sb, type); 2614 if (ret) 2615 goto out; 2616 2617 ret = filemap_fdatawrite(mapping); 2618 if (ret) 2619 goto out; 2620 2621 /* if we are using journalled quota */ 2622 if (is_journalled_quota(sbi)) 2623 goto out; 2624 2625 ret = filemap_fdatawait(mapping); 2626 2627 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 2628 out: 2629 if (ret) 2630 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2631 return ret; 2632 } 2633 2634 int f2fs_quota_sync(struct super_block *sb, int type) 2635 { 2636 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2637 struct quota_info *dqopt = sb_dqopt(sb); 2638 int cnt; 2639 int ret; 2640 2641 /* 2642 * Now when everything is written we can discard the pagecache so 2643 * that userspace sees the changes. 2644 */ 2645 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2646 2647 if (type != -1 && cnt != type) 2648 continue; 2649 2650 if (!sb_has_quota_active(sb, type)) 2651 return 0; 2652 2653 inode_lock(dqopt->files[cnt]); 2654 2655 /* 2656 * do_quotactl 2657 * f2fs_quota_sync 2658 * down_read(quota_sem) 2659 * dquot_writeback_dquots() 2660 * f2fs_dquot_commit 2661 * block_operation 2662 * down_read(quota_sem) 2663 */ 2664 f2fs_lock_op(sbi); 2665 down_read(&sbi->quota_sem); 2666 2667 ret = f2fs_quota_sync_file(sbi, cnt); 2668 2669 up_read(&sbi->quota_sem); 2670 f2fs_unlock_op(sbi); 2671 2672 inode_unlock(dqopt->files[cnt]); 2673 2674 if (ret) 2675 break; 2676 } 2677 return ret; 2678 } 2679 2680 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2681 const struct path *path) 2682 { 2683 struct inode *inode; 2684 int err; 2685 2686 /* if quota sysfile exists, deny enabling quota with specific file */ 2687 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2688 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2689 return -EBUSY; 2690 } 2691 2692 err = f2fs_quota_sync(sb, type); 2693 if (err) 2694 return err; 2695 2696 err = dquot_quota_on(sb, type, format_id, path); 2697 if (err) 2698 return err; 2699 2700 inode = d_inode(path->dentry); 2701 2702 inode_lock(inode); 2703 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2704 f2fs_set_inode_flags(inode); 2705 inode_unlock(inode); 2706 f2fs_mark_inode_dirty_sync(inode, false); 2707 2708 return 0; 2709 } 2710 2711 static int __f2fs_quota_off(struct super_block *sb, int type) 2712 { 2713 struct inode *inode = sb_dqopt(sb)->files[type]; 2714 int err; 2715 2716 if (!inode || !igrab(inode)) 2717 return dquot_quota_off(sb, type); 2718 2719 err = f2fs_quota_sync(sb, type); 2720 if (err) 2721 goto out_put; 2722 2723 err = dquot_quota_off(sb, type); 2724 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2725 goto out_put; 2726 2727 inode_lock(inode); 2728 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2729 f2fs_set_inode_flags(inode); 2730 inode_unlock(inode); 2731 f2fs_mark_inode_dirty_sync(inode, false); 2732 out_put: 2733 iput(inode); 2734 return err; 2735 } 2736 2737 static int f2fs_quota_off(struct super_block *sb, int type) 2738 { 2739 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2740 int err; 2741 2742 err = __f2fs_quota_off(sb, type); 2743 2744 /* 2745 * quotactl can shutdown journalled quota, result in inconsistence 2746 * between quota record and fs data by following updates, tag the 2747 * flag to let fsck be aware of it. 2748 */ 2749 if (is_journalled_quota(sbi)) 2750 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2751 return err; 2752 } 2753 2754 void f2fs_quota_off_umount(struct super_block *sb) 2755 { 2756 int type; 2757 int err; 2758 2759 for (type = 0; type < MAXQUOTAS; type++) { 2760 err = __f2fs_quota_off(sb, type); 2761 if (err) { 2762 int ret = dquot_quota_off(sb, type); 2763 2764 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2765 type, err, ret); 2766 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2767 } 2768 } 2769 /* 2770 * In case of checkpoint=disable, we must flush quota blocks. 2771 * This can cause NULL exception for node_inode in end_io, since 2772 * put_super already dropped it. 2773 */ 2774 sync_filesystem(sb); 2775 } 2776 2777 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2778 { 2779 struct quota_info *dqopt = sb_dqopt(sb); 2780 int type; 2781 2782 for (type = 0; type < MAXQUOTAS; type++) { 2783 if (!dqopt->files[type]) 2784 continue; 2785 f2fs_inode_synced(dqopt->files[type]); 2786 } 2787 } 2788 2789 static int f2fs_dquot_commit(struct dquot *dquot) 2790 { 2791 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2792 int ret; 2793 2794 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2795 ret = dquot_commit(dquot); 2796 if (ret < 0) 2797 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2798 up_read(&sbi->quota_sem); 2799 return ret; 2800 } 2801 2802 static int f2fs_dquot_acquire(struct dquot *dquot) 2803 { 2804 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2805 int ret; 2806 2807 down_read(&sbi->quota_sem); 2808 ret = dquot_acquire(dquot); 2809 if (ret < 0) 2810 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2811 up_read(&sbi->quota_sem); 2812 return ret; 2813 } 2814 2815 static int f2fs_dquot_release(struct dquot *dquot) 2816 { 2817 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2818 int ret = dquot_release(dquot); 2819 2820 if (ret < 0) 2821 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2822 return ret; 2823 } 2824 2825 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2826 { 2827 struct super_block *sb = dquot->dq_sb; 2828 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2829 int ret = dquot_mark_dquot_dirty(dquot); 2830 2831 /* if we are using journalled quota */ 2832 if (is_journalled_quota(sbi)) 2833 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2834 2835 return ret; 2836 } 2837 2838 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2839 { 2840 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2841 int ret = dquot_commit_info(sb, type); 2842 2843 if (ret < 0) 2844 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2845 return ret; 2846 } 2847 2848 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2849 { 2850 *projid = F2FS_I(inode)->i_projid; 2851 return 0; 2852 } 2853 2854 static const struct dquot_operations f2fs_quota_operations = { 2855 .get_reserved_space = f2fs_get_reserved_space, 2856 .write_dquot = f2fs_dquot_commit, 2857 .acquire_dquot = f2fs_dquot_acquire, 2858 .release_dquot = f2fs_dquot_release, 2859 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2860 .write_info = f2fs_dquot_commit_info, 2861 .alloc_dquot = dquot_alloc, 2862 .destroy_dquot = dquot_destroy, 2863 .get_projid = f2fs_get_projid, 2864 .get_next_id = dquot_get_next_id, 2865 }; 2866 2867 static const struct quotactl_ops f2fs_quotactl_ops = { 2868 .quota_on = f2fs_quota_on, 2869 .quota_off = f2fs_quota_off, 2870 .quota_sync = f2fs_quota_sync, 2871 .get_state = dquot_get_state, 2872 .set_info = dquot_set_dqinfo, 2873 .get_dqblk = dquot_get_dqblk, 2874 .set_dqblk = dquot_set_dqblk, 2875 .get_nextdqblk = dquot_get_next_dqblk, 2876 }; 2877 #else 2878 int f2fs_quota_sync(struct super_block *sb, int type) 2879 { 2880 return 0; 2881 } 2882 2883 void f2fs_quota_off_umount(struct super_block *sb) 2884 { 2885 } 2886 #endif 2887 2888 static const struct super_operations f2fs_sops = { 2889 .alloc_inode = f2fs_alloc_inode, 2890 .free_inode = f2fs_free_inode, 2891 .drop_inode = f2fs_drop_inode, 2892 .write_inode = f2fs_write_inode, 2893 .dirty_inode = f2fs_dirty_inode, 2894 .show_options = f2fs_show_options, 2895 #ifdef CONFIG_QUOTA 2896 .quota_read = f2fs_quota_read, 2897 .quota_write = f2fs_quota_write, 2898 .get_dquots = f2fs_get_dquots, 2899 #endif 2900 .evict_inode = f2fs_evict_inode, 2901 .put_super = f2fs_put_super, 2902 .sync_fs = f2fs_sync_fs, 2903 .freeze_fs = f2fs_freeze, 2904 .unfreeze_fs = f2fs_unfreeze, 2905 .statfs = f2fs_statfs, 2906 .remount_fs = f2fs_remount, 2907 }; 2908 2909 #ifdef CONFIG_FS_ENCRYPTION 2910 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2911 { 2912 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2913 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2914 ctx, len, NULL); 2915 } 2916 2917 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2918 void *fs_data) 2919 { 2920 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2921 2922 /* 2923 * Encrypting the root directory is not allowed because fsck 2924 * expects lost+found directory to exist and remain unencrypted 2925 * if LOST_FOUND feature is enabled. 2926 * 2927 */ 2928 if (f2fs_sb_has_lost_found(sbi) && 2929 inode->i_ino == F2FS_ROOT_INO(sbi)) 2930 return -EPERM; 2931 2932 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2933 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2934 ctx, len, fs_data, XATTR_CREATE); 2935 } 2936 2937 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 2938 { 2939 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 2940 } 2941 2942 static bool f2fs_has_stable_inodes(struct super_block *sb) 2943 { 2944 return true; 2945 } 2946 2947 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 2948 int *ino_bits_ret, int *lblk_bits_ret) 2949 { 2950 *ino_bits_ret = 8 * sizeof(nid_t); 2951 *lblk_bits_ret = 8 * sizeof(block_t); 2952 } 2953 2954 static int f2fs_get_num_devices(struct super_block *sb) 2955 { 2956 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2957 2958 if (f2fs_is_multi_device(sbi)) 2959 return sbi->s_ndevs; 2960 return 1; 2961 } 2962 2963 static void f2fs_get_devices(struct super_block *sb, 2964 struct request_queue **devs) 2965 { 2966 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2967 int i; 2968 2969 for (i = 0; i < sbi->s_ndevs; i++) 2970 devs[i] = bdev_get_queue(FDEV(i).bdev); 2971 } 2972 2973 static const struct fscrypt_operations f2fs_cryptops = { 2974 .key_prefix = "f2fs:", 2975 .get_context = f2fs_get_context, 2976 .set_context = f2fs_set_context, 2977 .get_dummy_policy = f2fs_get_dummy_policy, 2978 .empty_dir = f2fs_empty_dir, 2979 .max_namelen = F2FS_NAME_LEN, 2980 .has_stable_inodes = f2fs_has_stable_inodes, 2981 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 2982 .get_num_devices = f2fs_get_num_devices, 2983 .get_devices = f2fs_get_devices, 2984 }; 2985 #endif 2986 2987 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 2988 u64 ino, u32 generation) 2989 { 2990 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2991 struct inode *inode; 2992 2993 if (f2fs_check_nid_range(sbi, ino)) 2994 return ERR_PTR(-ESTALE); 2995 2996 /* 2997 * f2fs_iget isn't quite right if the inode is currently unallocated! 2998 * However f2fs_iget currently does appropriate checks to handle stale 2999 * inodes so everything is OK. 3000 */ 3001 inode = f2fs_iget(sb, ino); 3002 if (IS_ERR(inode)) 3003 return ERR_CAST(inode); 3004 if (unlikely(generation && inode->i_generation != generation)) { 3005 /* we didn't find the right inode.. */ 3006 iput(inode); 3007 return ERR_PTR(-ESTALE); 3008 } 3009 return inode; 3010 } 3011 3012 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3013 int fh_len, int fh_type) 3014 { 3015 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3016 f2fs_nfs_get_inode); 3017 } 3018 3019 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3020 int fh_len, int fh_type) 3021 { 3022 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3023 f2fs_nfs_get_inode); 3024 } 3025 3026 static const struct export_operations f2fs_export_ops = { 3027 .fh_to_dentry = f2fs_fh_to_dentry, 3028 .fh_to_parent = f2fs_fh_to_parent, 3029 .get_parent = f2fs_get_parent, 3030 }; 3031 3032 loff_t max_file_blocks(struct inode *inode) 3033 { 3034 loff_t result = 0; 3035 loff_t leaf_count; 3036 3037 /* 3038 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3039 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3040 * space in inode.i_addr, it will be more safe to reassign 3041 * result as zero. 3042 */ 3043 3044 if (inode && f2fs_compressed_file(inode)) 3045 leaf_count = ADDRS_PER_BLOCK(inode); 3046 else 3047 leaf_count = DEF_ADDRS_PER_BLOCK; 3048 3049 /* two direct node blocks */ 3050 result += (leaf_count * 2); 3051 3052 /* two indirect node blocks */ 3053 leaf_count *= NIDS_PER_BLOCK; 3054 result += (leaf_count * 2); 3055 3056 /* one double indirect node block */ 3057 leaf_count *= NIDS_PER_BLOCK; 3058 result += leaf_count; 3059 3060 return result; 3061 } 3062 3063 static int __f2fs_commit_super(struct buffer_head *bh, 3064 struct f2fs_super_block *super) 3065 { 3066 lock_buffer(bh); 3067 if (super) 3068 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 3069 set_buffer_dirty(bh); 3070 unlock_buffer(bh); 3071 3072 /* it's rare case, we can do fua all the time */ 3073 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 3074 } 3075 3076 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3077 struct buffer_head *bh) 3078 { 3079 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3080 (bh->b_data + F2FS_SUPER_OFFSET); 3081 struct super_block *sb = sbi->sb; 3082 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3083 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3084 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3085 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3086 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3087 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3088 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3089 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3090 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3091 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3092 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3093 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3094 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3095 u64 main_end_blkaddr = main_blkaddr + 3096 (segment_count_main << log_blocks_per_seg); 3097 u64 seg_end_blkaddr = segment0_blkaddr + 3098 (segment_count << log_blocks_per_seg); 3099 3100 if (segment0_blkaddr != cp_blkaddr) { 3101 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3102 segment0_blkaddr, cp_blkaddr); 3103 return true; 3104 } 3105 3106 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3107 sit_blkaddr) { 3108 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3109 cp_blkaddr, sit_blkaddr, 3110 segment_count_ckpt << log_blocks_per_seg); 3111 return true; 3112 } 3113 3114 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3115 nat_blkaddr) { 3116 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3117 sit_blkaddr, nat_blkaddr, 3118 segment_count_sit << log_blocks_per_seg); 3119 return true; 3120 } 3121 3122 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3123 ssa_blkaddr) { 3124 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3125 nat_blkaddr, ssa_blkaddr, 3126 segment_count_nat << log_blocks_per_seg); 3127 return true; 3128 } 3129 3130 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3131 main_blkaddr) { 3132 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3133 ssa_blkaddr, main_blkaddr, 3134 segment_count_ssa << log_blocks_per_seg); 3135 return true; 3136 } 3137 3138 if (main_end_blkaddr > seg_end_blkaddr) { 3139 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3140 main_blkaddr, seg_end_blkaddr, 3141 segment_count_main << log_blocks_per_seg); 3142 return true; 3143 } else if (main_end_blkaddr < seg_end_blkaddr) { 3144 int err = 0; 3145 char *res; 3146 3147 /* fix in-memory information all the time */ 3148 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3149 segment0_blkaddr) >> log_blocks_per_seg); 3150 3151 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 3152 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3153 res = "internally"; 3154 } else { 3155 err = __f2fs_commit_super(bh, NULL); 3156 res = err ? "failed" : "done"; 3157 } 3158 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3159 res, main_blkaddr, seg_end_blkaddr, 3160 segment_count_main << log_blocks_per_seg); 3161 if (err) 3162 return true; 3163 } 3164 return false; 3165 } 3166 3167 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3168 struct buffer_head *bh) 3169 { 3170 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3171 block_t total_sections, blocks_per_seg; 3172 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3173 (bh->b_data + F2FS_SUPER_OFFSET); 3174 size_t crc_offset = 0; 3175 __u32 crc = 0; 3176 3177 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3178 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3179 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3180 return -EINVAL; 3181 } 3182 3183 /* Check checksum_offset and crc in superblock */ 3184 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3185 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3186 if (crc_offset != 3187 offsetof(struct f2fs_super_block, crc)) { 3188 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3189 crc_offset); 3190 return -EFSCORRUPTED; 3191 } 3192 crc = le32_to_cpu(raw_super->crc); 3193 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3194 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3195 return -EFSCORRUPTED; 3196 } 3197 } 3198 3199 /* Currently, support only 4KB block size */ 3200 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3201 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3202 le32_to_cpu(raw_super->log_blocksize), 3203 F2FS_BLKSIZE_BITS); 3204 return -EFSCORRUPTED; 3205 } 3206 3207 /* check log blocks per segment */ 3208 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3209 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3210 le32_to_cpu(raw_super->log_blocks_per_seg)); 3211 return -EFSCORRUPTED; 3212 } 3213 3214 /* Currently, support 512/1024/2048/4096 bytes sector size */ 3215 if (le32_to_cpu(raw_super->log_sectorsize) > 3216 F2FS_MAX_LOG_SECTOR_SIZE || 3217 le32_to_cpu(raw_super->log_sectorsize) < 3218 F2FS_MIN_LOG_SECTOR_SIZE) { 3219 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3220 le32_to_cpu(raw_super->log_sectorsize)); 3221 return -EFSCORRUPTED; 3222 } 3223 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3224 le32_to_cpu(raw_super->log_sectorsize) != 3225 F2FS_MAX_LOG_SECTOR_SIZE) { 3226 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3227 le32_to_cpu(raw_super->log_sectors_per_block), 3228 le32_to_cpu(raw_super->log_sectorsize)); 3229 return -EFSCORRUPTED; 3230 } 3231 3232 segment_count = le32_to_cpu(raw_super->segment_count); 3233 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3234 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3235 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3236 total_sections = le32_to_cpu(raw_super->section_count); 3237 3238 /* blocks_per_seg should be 512, given the above check */ 3239 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 3240 3241 if (segment_count > F2FS_MAX_SEGMENT || 3242 segment_count < F2FS_MIN_SEGMENTS) { 3243 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3244 return -EFSCORRUPTED; 3245 } 3246 3247 if (total_sections > segment_count_main || total_sections < 1 || 3248 segs_per_sec > segment_count || !segs_per_sec) { 3249 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3250 segment_count, total_sections, segs_per_sec); 3251 return -EFSCORRUPTED; 3252 } 3253 3254 if (segment_count_main != total_sections * segs_per_sec) { 3255 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3256 segment_count_main, total_sections, segs_per_sec); 3257 return -EFSCORRUPTED; 3258 } 3259 3260 if ((segment_count / segs_per_sec) < total_sections) { 3261 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3262 segment_count, segs_per_sec, total_sections); 3263 return -EFSCORRUPTED; 3264 } 3265 3266 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3267 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3268 segment_count, le64_to_cpu(raw_super->block_count)); 3269 return -EFSCORRUPTED; 3270 } 3271 3272 if (RDEV(0).path[0]) { 3273 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3274 int i = 1; 3275 3276 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3277 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3278 i++; 3279 } 3280 if (segment_count != dev_seg_count) { 3281 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3282 segment_count, dev_seg_count); 3283 return -EFSCORRUPTED; 3284 } 3285 } else { 3286 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3287 !bdev_is_zoned(sbi->sb->s_bdev)) { 3288 f2fs_info(sbi, "Zoned block device path is missing"); 3289 return -EFSCORRUPTED; 3290 } 3291 } 3292 3293 if (secs_per_zone > total_sections || !secs_per_zone) { 3294 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3295 secs_per_zone, total_sections); 3296 return -EFSCORRUPTED; 3297 } 3298 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3299 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3300 (le32_to_cpu(raw_super->extension_count) + 3301 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3302 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3303 le32_to_cpu(raw_super->extension_count), 3304 raw_super->hot_ext_count, 3305 F2FS_MAX_EXTENSION); 3306 return -EFSCORRUPTED; 3307 } 3308 3309 if (le32_to_cpu(raw_super->cp_payload) >= 3310 (blocks_per_seg - F2FS_CP_PACKS - 3311 NR_CURSEG_PERSIST_TYPE)) { 3312 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3313 le32_to_cpu(raw_super->cp_payload), 3314 blocks_per_seg - F2FS_CP_PACKS - 3315 NR_CURSEG_PERSIST_TYPE); 3316 return -EFSCORRUPTED; 3317 } 3318 3319 /* check reserved ino info */ 3320 if (le32_to_cpu(raw_super->node_ino) != 1 || 3321 le32_to_cpu(raw_super->meta_ino) != 2 || 3322 le32_to_cpu(raw_super->root_ino) != 3) { 3323 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3324 le32_to_cpu(raw_super->node_ino), 3325 le32_to_cpu(raw_super->meta_ino), 3326 le32_to_cpu(raw_super->root_ino)); 3327 return -EFSCORRUPTED; 3328 } 3329 3330 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3331 if (sanity_check_area_boundary(sbi, bh)) 3332 return -EFSCORRUPTED; 3333 3334 return 0; 3335 } 3336 3337 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3338 { 3339 unsigned int total, fsmeta; 3340 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3341 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3342 unsigned int ovp_segments, reserved_segments; 3343 unsigned int main_segs, blocks_per_seg; 3344 unsigned int sit_segs, nat_segs; 3345 unsigned int sit_bitmap_size, nat_bitmap_size; 3346 unsigned int log_blocks_per_seg; 3347 unsigned int segment_count_main; 3348 unsigned int cp_pack_start_sum, cp_payload; 3349 block_t user_block_count, valid_user_blocks; 3350 block_t avail_node_count, valid_node_count; 3351 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3352 int i, j; 3353 3354 total = le32_to_cpu(raw_super->segment_count); 3355 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3356 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3357 fsmeta += sit_segs; 3358 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3359 fsmeta += nat_segs; 3360 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3361 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3362 3363 if (unlikely(fsmeta >= total)) 3364 return 1; 3365 3366 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3367 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3368 3369 if (!f2fs_sb_has_readonly(sbi) && 3370 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3371 ovp_segments == 0 || reserved_segments == 0)) { 3372 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3373 return 1; 3374 } 3375 user_block_count = le64_to_cpu(ckpt->user_block_count); 3376 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3377 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3378 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3379 if (!user_block_count || user_block_count >= 3380 segment_count_main << log_blocks_per_seg) { 3381 f2fs_err(sbi, "Wrong user_block_count: %u", 3382 user_block_count); 3383 return 1; 3384 } 3385 3386 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3387 if (valid_user_blocks > user_block_count) { 3388 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3389 valid_user_blocks, user_block_count); 3390 return 1; 3391 } 3392 3393 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3394 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3395 if (valid_node_count > avail_node_count) { 3396 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3397 valid_node_count, avail_node_count); 3398 return 1; 3399 } 3400 3401 main_segs = le32_to_cpu(raw_super->segment_count_main); 3402 blocks_per_seg = sbi->blocks_per_seg; 3403 3404 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3405 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3406 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3407 return 1; 3408 3409 if (f2fs_sb_has_readonly(sbi)) 3410 goto check_data; 3411 3412 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3413 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3414 le32_to_cpu(ckpt->cur_node_segno[j])) { 3415 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3416 i, j, 3417 le32_to_cpu(ckpt->cur_node_segno[i])); 3418 return 1; 3419 } 3420 } 3421 } 3422 check_data: 3423 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3424 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3425 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3426 return 1; 3427 3428 if (f2fs_sb_has_readonly(sbi)) 3429 goto skip_cross; 3430 3431 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3432 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3433 le32_to_cpu(ckpt->cur_data_segno[j])) { 3434 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3435 i, j, 3436 le32_to_cpu(ckpt->cur_data_segno[i])); 3437 return 1; 3438 } 3439 } 3440 } 3441 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3442 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3443 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3444 le32_to_cpu(ckpt->cur_data_segno[j])) { 3445 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3446 i, j, 3447 le32_to_cpu(ckpt->cur_node_segno[i])); 3448 return 1; 3449 } 3450 } 3451 } 3452 skip_cross: 3453 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3454 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3455 3456 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3457 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3458 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3459 sit_bitmap_size, nat_bitmap_size); 3460 return 1; 3461 } 3462 3463 cp_pack_start_sum = __start_sum_addr(sbi); 3464 cp_payload = __cp_payload(sbi); 3465 if (cp_pack_start_sum < cp_payload + 1 || 3466 cp_pack_start_sum > blocks_per_seg - 1 - 3467 NR_CURSEG_PERSIST_TYPE) { 3468 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3469 cp_pack_start_sum); 3470 return 1; 3471 } 3472 3473 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3474 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3475 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3476 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3477 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3478 le32_to_cpu(ckpt->checksum_offset)); 3479 return 1; 3480 } 3481 3482 nat_blocks = nat_segs << log_blocks_per_seg; 3483 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 3484 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 3485 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 3486 (cp_payload + F2FS_CP_PACKS + 3487 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 3488 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 3489 cp_payload, nat_bits_blocks); 3490 return -EFSCORRUPTED; 3491 } 3492 3493 if (unlikely(f2fs_cp_error(sbi))) { 3494 f2fs_err(sbi, "A bug case: need to run fsck"); 3495 return 1; 3496 } 3497 return 0; 3498 } 3499 3500 static void init_sb_info(struct f2fs_sb_info *sbi) 3501 { 3502 struct f2fs_super_block *raw_super = sbi->raw_super; 3503 int i; 3504 3505 sbi->log_sectors_per_block = 3506 le32_to_cpu(raw_super->log_sectors_per_block); 3507 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3508 sbi->blocksize = 1 << sbi->log_blocksize; 3509 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3510 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3511 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3512 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3513 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3514 sbi->total_node_count = 3515 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3516 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3517 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3518 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3519 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3520 sbi->cur_victim_sec = NULL_SECNO; 3521 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3522 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3523 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3524 sbi->migration_granularity = sbi->segs_per_sec; 3525 sbi->seq_file_ra_mul = MIN_RA_MUL; 3526 3527 sbi->dir_level = DEF_DIR_LEVEL; 3528 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3529 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3530 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3531 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3532 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3533 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3534 DEF_UMOUNT_DISCARD_TIMEOUT; 3535 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3536 3537 for (i = 0; i < NR_COUNT_TYPE; i++) 3538 atomic_set(&sbi->nr_pages[i], 0); 3539 3540 for (i = 0; i < META; i++) 3541 atomic_set(&sbi->wb_sync_req[i], 0); 3542 3543 INIT_LIST_HEAD(&sbi->s_list); 3544 mutex_init(&sbi->umount_mutex); 3545 init_rwsem(&sbi->io_order_lock); 3546 spin_lock_init(&sbi->cp_lock); 3547 3548 sbi->dirty_device = 0; 3549 spin_lock_init(&sbi->dev_lock); 3550 3551 init_rwsem(&sbi->sb_lock); 3552 init_rwsem(&sbi->pin_sem); 3553 } 3554 3555 static int init_percpu_info(struct f2fs_sb_info *sbi) 3556 { 3557 int err; 3558 3559 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3560 if (err) 3561 return err; 3562 3563 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3564 GFP_KERNEL); 3565 if (err) 3566 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3567 3568 return err; 3569 } 3570 3571 #ifdef CONFIG_BLK_DEV_ZONED 3572 3573 struct f2fs_report_zones_args { 3574 struct f2fs_dev_info *dev; 3575 bool zone_cap_mismatch; 3576 }; 3577 3578 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3579 void *data) 3580 { 3581 struct f2fs_report_zones_args *rz_args = data; 3582 3583 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3584 return 0; 3585 3586 set_bit(idx, rz_args->dev->blkz_seq); 3587 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >> 3588 F2FS_LOG_SECTORS_PER_BLOCK; 3589 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch) 3590 rz_args->zone_cap_mismatch = true; 3591 3592 return 0; 3593 } 3594 3595 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3596 { 3597 struct block_device *bdev = FDEV(devi).bdev; 3598 sector_t nr_sectors = bdev_nr_sectors(bdev); 3599 struct f2fs_report_zones_args rep_zone_arg; 3600 int ret; 3601 3602 if (!f2fs_sb_has_blkzoned(sbi)) 3603 return 0; 3604 3605 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3606 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 3607 return -EINVAL; 3608 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 3609 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3610 __ilog2_u32(sbi->blocks_per_blkz)) 3611 return -EINVAL; 3612 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3613 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3614 sbi->log_blocks_per_blkz; 3615 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 3616 FDEV(devi).nr_blkz++; 3617 3618 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3619 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3620 * sizeof(unsigned long), 3621 GFP_KERNEL); 3622 if (!FDEV(devi).blkz_seq) 3623 return -ENOMEM; 3624 3625 /* Get block zones type and zone-capacity */ 3626 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi, 3627 FDEV(devi).nr_blkz * sizeof(block_t), 3628 GFP_KERNEL); 3629 if (!FDEV(devi).zone_capacity_blocks) 3630 return -ENOMEM; 3631 3632 rep_zone_arg.dev = &FDEV(devi); 3633 rep_zone_arg.zone_cap_mismatch = false; 3634 3635 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3636 &rep_zone_arg); 3637 if (ret < 0) 3638 return ret; 3639 3640 if (!rep_zone_arg.zone_cap_mismatch) { 3641 kfree(FDEV(devi).zone_capacity_blocks); 3642 FDEV(devi).zone_capacity_blocks = NULL; 3643 } 3644 3645 return 0; 3646 } 3647 #endif 3648 3649 /* 3650 * Read f2fs raw super block. 3651 * Because we have two copies of super block, so read both of them 3652 * to get the first valid one. If any one of them is broken, we pass 3653 * them recovery flag back to the caller. 3654 */ 3655 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3656 struct f2fs_super_block **raw_super, 3657 int *valid_super_block, int *recovery) 3658 { 3659 struct super_block *sb = sbi->sb; 3660 int block; 3661 struct buffer_head *bh; 3662 struct f2fs_super_block *super; 3663 int err = 0; 3664 3665 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3666 if (!super) 3667 return -ENOMEM; 3668 3669 for (block = 0; block < 2; block++) { 3670 bh = sb_bread(sb, block); 3671 if (!bh) { 3672 f2fs_err(sbi, "Unable to read %dth superblock", 3673 block + 1); 3674 err = -EIO; 3675 *recovery = 1; 3676 continue; 3677 } 3678 3679 /* sanity checking of raw super */ 3680 err = sanity_check_raw_super(sbi, bh); 3681 if (err) { 3682 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3683 block + 1); 3684 brelse(bh); 3685 *recovery = 1; 3686 continue; 3687 } 3688 3689 if (!*raw_super) { 3690 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3691 sizeof(*super)); 3692 *valid_super_block = block; 3693 *raw_super = super; 3694 } 3695 brelse(bh); 3696 } 3697 3698 /* No valid superblock */ 3699 if (!*raw_super) 3700 kfree(super); 3701 else 3702 err = 0; 3703 3704 return err; 3705 } 3706 3707 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3708 { 3709 struct buffer_head *bh; 3710 __u32 crc = 0; 3711 int err; 3712 3713 if ((recover && f2fs_readonly(sbi->sb)) || 3714 bdev_read_only(sbi->sb->s_bdev)) { 3715 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3716 return -EROFS; 3717 } 3718 3719 /* we should update superblock crc here */ 3720 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3721 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3722 offsetof(struct f2fs_super_block, crc)); 3723 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3724 } 3725 3726 /* write back-up superblock first */ 3727 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3728 if (!bh) 3729 return -EIO; 3730 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3731 brelse(bh); 3732 3733 /* if we are in recovery path, skip writing valid superblock */ 3734 if (recover || err) 3735 return err; 3736 3737 /* write current valid superblock */ 3738 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3739 if (!bh) 3740 return -EIO; 3741 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3742 brelse(bh); 3743 return err; 3744 } 3745 3746 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3747 { 3748 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3749 unsigned int max_devices = MAX_DEVICES; 3750 int i; 3751 3752 /* Initialize single device information */ 3753 if (!RDEV(0).path[0]) { 3754 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3755 return 0; 3756 max_devices = 1; 3757 } 3758 3759 /* 3760 * Initialize multiple devices information, or single 3761 * zoned block device information. 3762 */ 3763 sbi->devs = f2fs_kzalloc(sbi, 3764 array_size(max_devices, 3765 sizeof(struct f2fs_dev_info)), 3766 GFP_KERNEL); 3767 if (!sbi->devs) 3768 return -ENOMEM; 3769 3770 for (i = 0; i < max_devices; i++) { 3771 3772 if (i > 0 && !RDEV(i).path[0]) 3773 break; 3774 3775 if (max_devices == 1) { 3776 /* Single zoned block device mount */ 3777 FDEV(0).bdev = 3778 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3779 sbi->sb->s_mode, sbi->sb->s_type); 3780 } else { 3781 /* Multi-device mount */ 3782 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3783 FDEV(i).total_segments = 3784 le32_to_cpu(RDEV(i).total_segments); 3785 if (i == 0) { 3786 FDEV(i).start_blk = 0; 3787 FDEV(i).end_blk = FDEV(i).start_blk + 3788 (FDEV(i).total_segments << 3789 sbi->log_blocks_per_seg) - 1 + 3790 le32_to_cpu(raw_super->segment0_blkaddr); 3791 } else { 3792 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3793 FDEV(i).end_blk = FDEV(i).start_blk + 3794 (FDEV(i).total_segments << 3795 sbi->log_blocks_per_seg) - 1; 3796 } 3797 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3798 sbi->sb->s_mode, sbi->sb->s_type); 3799 } 3800 if (IS_ERR(FDEV(i).bdev)) 3801 return PTR_ERR(FDEV(i).bdev); 3802 3803 /* to release errored devices */ 3804 sbi->s_ndevs = i + 1; 3805 3806 #ifdef CONFIG_BLK_DEV_ZONED 3807 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3808 !f2fs_sb_has_blkzoned(sbi)) { 3809 f2fs_err(sbi, "Zoned block device feature not enabled"); 3810 return -EINVAL; 3811 } 3812 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3813 if (init_blkz_info(sbi, i)) { 3814 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3815 return -EINVAL; 3816 } 3817 if (max_devices == 1) 3818 break; 3819 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3820 i, FDEV(i).path, 3821 FDEV(i).total_segments, 3822 FDEV(i).start_blk, FDEV(i).end_blk, 3823 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3824 "Host-aware" : "Host-managed"); 3825 continue; 3826 } 3827 #endif 3828 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3829 i, FDEV(i).path, 3830 FDEV(i).total_segments, 3831 FDEV(i).start_blk, FDEV(i).end_blk); 3832 } 3833 f2fs_info(sbi, 3834 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3835 return 0; 3836 } 3837 3838 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3839 { 3840 #ifdef CONFIG_UNICODE 3841 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 3842 const struct f2fs_sb_encodings *encoding_info; 3843 struct unicode_map *encoding; 3844 __u16 encoding_flags; 3845 3846 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info, 3847 &encoding_flags)) { 3848 f2fs_err(sbi, 3849 "Encoding requested by superblock is unknown"); 3850 return -EINVAL; 3851 } 3852 3853 encoding = utf8_load(encoding_info->version); 3854 if (IS_ERR(encoding)) { 3855 f2fs_err(sbi, 3856 "can't mount with superblock charset: %s-%s " 3857 "not supported by the kernel. flags: 0x%x.", 3858 encoding_info->name, encoding_info->version, 3859 encoding_flags); 3860 return PTR_ERR(encoding); 3861 } 3862 f2fs_info(sbi, "Using encoding defined by superblock: " 3863 "%s-%s with flags 0x%hx", encoding_info->name, 3864 encoding_info->version?:"\b", encoding_flags); 3865 3866 sbi->sb->s_encoding = encoding; 3867 sbi->sb->s_encoding_flags = encoding_flags; 3868 } 3869 #else 3870 if (f2fs_sb_has_casefold(sbi)) { 3871 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3872 return -EINVAL; 3873 } 3874 #endif 3875 return 0; 3876 } 3877 3878 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3879 { 3880 struct f2fs_sm_info *sm_i = SM_I(sbi); 3881 3882 /* adjust parameters according to the volume size */ 3883 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3884 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3885 if (f2fs_block_unit_discard(sbi)) 3886 sm_i->dcc_info->discard_granularity = 1; 3887 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 3888 } 3889 3890 sbi->readdir_ra = 1; 3891 } 3892 3893 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3894 { 3895 struct f2fs_sb_info *sbi; 3896 struct f2fs_super_block *raw_super; 3897 struct inode *root; 3898 int err; 3899 bool skip_recovery = false, need_fsck = false; 3900 char *options = NULL; 3901 int recovery, i, valid_super_block; 3902 struct curseg_info *seg_i; 3903 int retry_cnt = 1; 3904 3905 try_onemore: 3906 err = -EINVAL; 3907 raw_super = NULL; 3908 valid_super_block = -1; 3909 recovery = 0; 3910 3911 /* allocate memory for f2fs-specific super block info */ 3912 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 3913 if (!sbi) 3914 return -ENOMEM; 3915 3916 sbi->sb = sb; 3917 3918 /* Load the checksum driver */ 3919 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 3920 if (IS_ERR(sbi->s_chksum_driver)) { 3921 f2fs_err(sbi, "Cannot load crc32 driver."); 3922 err = PTR_ERR(sbi->s_chksum_driver); 3923 sbi->s_chksum_driver = NULL; 3924 goto free_sbi; 3925 } 3926 3927 /* set a block size */ 3928 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 3929 f2fs_err(sbi, "unable to set blocksize"); 3930 goto free_sbi; 3931 } 3932 3933 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 3934 &recovery); 3935 if (err) 3936 goto free_sbi; 3937 3938 sb->s_fs_info = sbi; 3939 sbi->raw_super = raw_super; 3940 3941 /* precompute checksum seed for metadata */ 3942 if (f2fs_sb_has_inode_chksum(sbi)) 3943 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 3944 sizeof(raw_super->uuid)); 3945 3946 default_options(sbi); 3947 /* parse mount options */ 3948 options = kstrdup((const char *)data, GFP_KERNEL); 3949 if (data && !options) { 3950 err = -ENOMEM; 3951 goto free_sb_buf; 3952 } 3953 3954 err = parse_options(sb, options, false); 3955 if (err) 3956 goto free_options; 3957 3958 sb->s_maxbytes = max_file_blocks(NULL) << 3959 le32_to_cpu(raw_super->log_blocksize); 3960 sb->s_max_links = F2FS_LINK_MAX; 3961 3962 err = f2fs_setup_casefold(sbi); 3963 if (err) 3964 goto free_options; 3965 3966 #ifdef CONFIG_QUOTA 3967 sb->dq_op = &f2fs_quota_operations; 3968 sb->s_qcop = &f2fs_quotactl_ops; 3969 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 3970 3971 if (f2fs_sb_has_quota_ino(sbi)) { 3972 for (i = 0; i < MAXQUOTAS; i++) { 3973 if (f2fs_qf_ino(sbi->sb, i)) 3974 sbi->nquota_files++; 3975 } 3976 } 3977 #endif 3978 3979 sb->s_op = &f2fs_sops; 3980 #ifdef CONFIG_FS_ENCRYPTION 3981 sb->s_cop = &f2fs_cryptops; 3982 #endif 3983 #ifdef CONFIG_FS_VERITY 3984 sb->s_vop = &f2fs_verityops; 3985 #endif 3986 sb->s_xattr = f2fs_xattr_handlers; 3987 sb->s_export_op = &f2fs_export_ops; 3988 sb->s_magic = F2FS_SUPER_MAGIC; 3989 sb->s_time_gran = 1; 3990 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 3991 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 3992 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 3993 sb->s_iflags |= SB_I_CGROUPWB; 3994 3995 /* init f2fs-specific super block info */ 3996 sbi->valid_super_block = valid_super_block; 3997 init_rwsem(&sbi->gc_lock); 3998 mutex_init(&sbi->writepages); 3999 init_rwsem(&sbi->cp_global_sem); 4000 init_rwsem(&sbi->node_write); 4001 init_rwsem(&sbi->node_change); 4002 4003 /* disallow all the data/node/meta page writes */ 4004 set_sbi_flag(sbi, SBI_POR_DOING); 4005 spin_lock_init(&sbi->stat_lock); 4006 4007 for (i = 0; i < NR_PAGE_TYPE; i++) { 4008 int n = (i == META) ? 1 : NR_TEMP_TYPE; 4009 int j; 4010 4011 sbi->write_io[i] = 4012 f2fs_kmalloc(sbi, 4013 array_size(n, 4014 sizeof(struct f2fs_bio_info)), 4015 GFP_KERNEL); 4016 if (!sbi->write_io[i]) { 4017 err = -ENOMEM; 4018 goto free_bio_info; 4019 } 4020 4021 for (j = HOT; j < n; j++) { 4022 init_rwsem(&sbi->write_io[i][j].io_rwsem); 4023 sbi->write_io[i][j].sbi = sbi; 4024 sbi->write_io[i][j].bio = NULL; 4025 spin_lock_init(&sbi->write_io[i][j].io_lock); 4026 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 4027 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); 4028 init_rwsem(&sbi->write_io[i][j].bio_list_lock); 4029 } 4030 } 4031 4032 init_rwsem(&sbi->cp_rwsem); 4033 init_rwsem(&sbi->quota_sem); 4034 init_waitqueue_head(&sbi->cp_wait); 4035 init_sb_info(sbi); 4036 4037 err = f2fs_init_iostat(sbi); 4038 if (err) 4039 goto free_bio_info; 4040 4041 err = init_percpu_info(sbi); 4042 if (err) 4043 goto free_iostat; 4044 4045 if (F2FS_IO_ALIGNED(sbi)) { 4046 sbi->write_io_dummy = 4047 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 4048 if (!sbi->write_io_dummy) { 4049 err = -ENOMEM; 4050 goto free_percpu; 4051 } 4052 } 4053 4054 /* init per sbi slab cache */ 4055 err = f2fs_init_xattr_caches(sbi); 4056 if (err) 4057 goto free_io_dummy; 4058 err = f2fs_init_page_array_cache(sbi); 4059 if (err) 4060 goto free_xattr_cache; 4061 4062 /* get an inode for meta space */ 4063 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4064 if (IS_ERR(sbi->meta_inode)) { 4065 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4066 err = PTR_ERR(sbi->meta_inode); 4067 goto free_page_array_cache; 4068 } 4069 4070 err = f2fs_get_valid_checkpoint(sbi); 4071 if (err) { 4072 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4073 goto free_meta_inode; 4074 } 4075 4076 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4077 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4078 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4079 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4080 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4081 } 4082 4083 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4084 set_sbi_flag(sbi, SBI_NEED_FSCK); 4085 4086 /* Initialize device list */ 4087 err = f2fs_scan_devices(sbi); 4088 if (err) { 4089 f2fs_err(sbi, "Failed to find devices"); 4090 goto free_devices; 4091 } 4092 4093 err = f2fs_init_post_read_wq(sbi); 4094 if (err) { 4095 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4096 goto free_devices; 4097 } 4098 4099 sbi->total_valid_node_count = 4100 le32_to_cpu(sbi->ckpt->valid_node_count); 4101 percpu_counter_set(&sbi->total_valid_inode_count, 4102 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4103 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4104 sbi->total_valid_block_count = 4105 le64_to_cpu(sbi->ckpt->valid_block_count); 4106 sbi->last_valid_block_count = sbi->total_valid_block_count; 4107 sbi->reserved_blocks = 0; 4108 sbi->current_reserved_blocks = 0; 4109 limit_reserve_root(sbi); 4110 adjust_unusable_cap_perc(sbi); 4111 4112 for (i = 0; i < NR_INODE_TYPE; i++) { 4113 INIT_LIST_HEAD(&sbi->inode_list[i]); 4114 spin_lock_init(&sbi->inode_lock[i]); 4115 } 4116 mutex_init(&sbi->flush_lock); 4117 4118 f2fs_init_extent_cache_info(sbi); 4119 4120 f2fs_init_ino_entry_info(sbi); 4121 4122 f2fs_init_fsync_node_info(sbi); 4123 4124 /* setup checkpoint request control and start checkpoint issue thread */ 4125 f2fs_init_ckpt_req_control(sbi); 4126 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4127 test_opt(sbi, MERGE_CHECKPOINT)) { 4128 err = f2fs_start_ckpt_thread(sbi); 4129 if (err) { 4130 f2fs_err(sbi, 4131 "Failed to start F2FS issue_checkpoint_thread (%d)", 4132 err); 4133 goto stop_ckpt_thread; 4134 } 4135 } 4136 4137 /* setup f2fs internal modules */ 4138 err = f2fs_build_segment_manager(sbi); 4139 if (err) { 4140 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4141 err); 4142 goto free_sm; 4143 } 4144 err = f2fs_build_node_manager(sbi); 4145 if (err) { 4146 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4147 err); 4148 goto free_nm; 4149 } 4150 4151 /* For write statistics */ 4152 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4153 4154 /* Read accumulated write IO statistics if exists */ 4155 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4156 if (__exist_node_summaries(sbi)) 4157 sbi->kbytes_written = 4158 le64_to_cpu(seg_i->journal->info.kbytes_written); 4159 4160 f2fs_build_gc_manager(sbi); 4161 4162 err = f2fs_build_stats(sbi); 4163 if (err) 4164 goto free_nm; 4165 4166 /* get an inode for node space */ 4167 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4168 if (IS_ERR(sbi->node_inode)) { 4169 f2fs_err(sbi, "Failed to read node inode"); 4170 err = PTR_ERR(sbi->node_inode); 4171 goto free_stats; 4172 } 4173 4174 /* read root inode and dentry */ 4175 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4176 if (IS_ERR(root)) { 4177 f2fs_err(sbi, "Failed to read root inode"); 4178 err = PTR_ERR(root); 4179 goto free_node_inode; 4180 } 4181 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4182 !root->i_size || !root->i_nlink) { 4183 iput(root); 4184 err = -EINVAL; 4185 goto free_node_inode; 4186 } 4187 4188 sb->s_root = d_make_root(root); /* allocate root dentry */ 4189 if (!sb->s_root) { 4190 err = -ENOMEM; 4191 goto free_node_inode; 4192 } 4193 4194 err = f2fs_init_compress_inode(sbi); 4195 if (err) 4196 goto free_root_inode; 4197 4198 err = f2fs_register_sysfs(sbi); 4199 if (err) 4200 goto free_compress_inode; 4201 4202 #ifdef CONFIG_QUOTA 4203 /* Enable quota usage during mount */ 4204 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4205 err = f2fs_enable_quotas(sb); 4206 if (err) 4207 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4208 } 4209 #endif 4210 /* if there are any orphan inodes, free them */ 4211 err = f2fs_recover_orphan_inodes(sbi); 4212 if (err) 4213 goto free_meta; 4214 4215 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4216 goto reset_checkpoint; 4217 4218 /* recover fsynced data */ 4219 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4220 !test_opt(sbi, NORECOVERY)) { 4221 /* 4222 * mount should be failed, when device has readonly mode, and 4223 * previous checkpoint was not done by clean system shutdown. 4224 */ 4225 if (f2fs_hw_is_readonly(sbi)) { 4226 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4227 err = f2fs_recover_fsync_data(sbi, true); 4228 if (err > 0) { 4229 err = -EROFS; 4230 f2fs_err(sbi, "Need to recover fsync data, but " 4231 "write access unavailable, please try " 4232 "mount w/ disable_roll_forward or norecovery"); 4233 } 4234 if (err < 0) 4235 goto free_meta; 4236 } 4237 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4238 goto reset_checkpoint; 4239 } 4240 4241 if (need_fsck) 4242 set_sbi_flag(sbi, SBI_NEED_FSCK); 4243 4244 if (skip_recovery) 4245 goto reset_checkpoint; 4246 4247 err = f2fs_recover_fsync_data(sbi, false); 4248 if (err < 0) { 4249 if (err != -ENOMEM) 4250 skip_recovery = true; 4251 need_fsck = true; 4252 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4253 err); 4254 goto free_meta; 4255 } 4256 } else { 4257 err = f2fs_recover_fsync_data(sbi, true); 4258 4259 if (!f2fs_readonly(sb) && err > 0) { 4260 err = -EINVAL; 4261 f2fs_err(sbi, "Need to recover fsync data"); 4262 goto free_meta; 4263 } 4264 } 4265 4266 /* 4267 * If the f2fs is not readonly and fsync data recovery succeeds, 4268 * check zoned block devices' write pointer consistency. 4269 */ 4270 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 4271 err = f2fs_check_write_pointer(sbi); 4272 if (err) 4273 goto free_meta; 4274 } 4275 4276 reset_checkpoint: 4277 f2fs_init_inmem_curseg(sbi); 4278 4279 /* f2fs_recover_fsync_data() cleared this already */ 4280 clear_sbi_flag(sbi, SBI_POR_DOING); 4281 4282 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4283 err = f2fs_disable_checkpoint(sbi); 4284 if (err) 4285 goto sync_free_meta; 4286 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4287 f2fs_enable_checkpoint(sbi); 4288 } 4289 4290 /* 4291 * If filesystem is not mounted as read-only then 4292 * do start the gc_thread. 4293 */ 4294 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4295 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4296 /* After POR, we can run background GC thread.*/ 4297 err = f2fs_start_gc_thread(sbi); 4298 if (err) 4299 goto sync_free_meta; 4300 } 4301 kvfree(options); 4302 4303 /* recover broken superblock */ 4304 if (recovery) { 4305 err = f2fs_commit_super(sbi, true); 4306 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4307 sbi->valid_super_block ? 1 : 2, err); 4308 } 4309 4310 f2fs_join_shrinker(sbi); 4311 4312 f2fs_tuning_parameters(sbi); 4313 4314 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4315 cur_cp_version(F2FS_CKPT(sbi))); 4316 f2fs_update_time(sbi, CP_TIME); 4317 f2fs_update_time(sbi, REQ_TIME); 4318 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4319 return 0; 4320 4321 sync_free_meta: 4322 /* safe to flush all the data */ 4323 sync_filesystem(sbi->sb); 4324 retry_cnt = 0; 4325 4326 free_meta: 4327 #ifdef CONFIG_QUOTA 4328 f2fs_truncate_quota_inode_pages(sb); 4329 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4330 f2fs_quota_off_umount(sbi->sb); 4331 #endif 4332 /* 4333 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4334 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4335 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4336 * falls into an infinite loop in f2fs_sync_meta_pages(). 4337 */ 4338 truncate_inode_pages_final(META_MAPPING(sbi)); 4339 /* evict some inodes being cached by GC */ 4340 evict_inodes(sb); 4341 f2fs_unregister_sysfs(sbi); 4342 free_compress_inode: 4343 f2fs_destroy_compress_inode(sbi); 4344 free_root_inode: 4345 dput(sb->s_root); 4346 sb->s_root = NULL; 4347 free_node_inode: 4348 f2fs_release_ino_entry(sbi, true); 4349 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4350 iput(sbi->node_inode); 4351 sbi->node_inode = NULL; 4352 free_stats: 4353 f2fs_destroy_stats(sbi); 4354 free_nm: 4355 f2fs_destroy_node_manager(sbi); 4356 free_sm: 4357 f2fs_destroy_segment_manager(sbi); 4358 f2fs_destroy_post_read_wq(sbi); 4359 stop_ckpt_thread: 4360 f2fs_stop_ckpt_thread(sbi); 4361 free_devices: 4362 destroy_device_list(sbi); 4363 kvfree(sbi->ckpt); 4364 free_meta_inode: 4365 make_bad_inode(sbi->meta_inode); 4366 iput(sbi->meta_inode); 4367 sbi->meta_inode = NULL; 4368 free_page_array_cache: 4369 f2fs_destroy_page_array_cache(sbi); 4370 free_xattr_cache: 4371 f2fs_destroy_xattr_caches(sbi); 4372 free_io_dummy: 4373 mempool_destroy(sbi->write_io_dummy); 4374 free_percpu: 4375 destroy_percpu_info(sbi); 4376 free_iostat: 4377 f2fs_destroy_iostat(sbi); 4378 free_bio_info: 4379 for (i = 0; i < NR_PAGE_TYPE; i++) 4380 kvfree(sbi->write_io[i]); 4381 4382 #ifdef CONFIG_UNICODE 4383 utf8_unload(sb->s_encoding); 4384 sb->s_encoding = NULL; 4385 #endif 4386 free_options: 4387 #ifdef CONFIG_QUOTA 4388 for (i = 0; i < MAXQUOTAS; i++) 4389 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4390 #endif 4391 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4392 kvfree(options); 4393 free_sb_buf: 4394 kfree(raw_super); 4395 free_sbi: 4396 if (sbi->s_chksum_driver) 4397 crypto_free_shash(sbi->s_chksum_driver); 4398 kfree(sbi); 4399 4400 /* give only one another chance */ 4401 if (retry_cnt > 0 && skip_recovery) { 4402 retry_cnt--; 4403 shrink_dcache_sb(sb); 4404 goto try_onemore; 4405 } 4406 return err; 4407 } 4408 4409 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4410 const char *dev_name, void *data) 4411 { 4412 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4413 } 4414 4415 static void kill_f2fs_super(struct super_block *sb) 4416 { 4417 if (sb->s_root) { 4418 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4419 4420 set_sbi_flag(sbi, SBI_IS_CLOSE); 4421 f2fs_stop_gc_thread(sbi); 4422 f2fs_stop_discard_thread(sbi); 4423 4424 #ifdef CONFIG_F2FS_FS_COMPRESSION 4425 /* 4426 * latter evict_inode() can bypass checking and invalidating 4427 * compress inode cache. 4428 */ 4429 if (test_opt(sbi, COMPRESS_CACHE)) 4430 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4431 #endif 4432 4433 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4434 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4435 struct cp_control cpc = { 4436 .reason = CP_UMOUNT, 4437 }; 4438 f2fs_write_checkpoint(sbi, &cpc); 4439 } 4440 4441 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4442 sb->s_flags &= ~SB_RDONLY; 4443 } 4444 kill_block_super(sb); 4445 } 4446 4447 static struct file_system_type f2fs_fs_type = { 4448 .owner = THIS_MODULE, 4449 .name = "f2fs", 4450 .mount = f2fs_mount, 4451 .kill_sb = kill_f2fs_super, 4452 .fs_flags = FS_REQUIRES_DEV, 4453 }; 4454 MODULE_ALIAS_FS("f2fs"); 4455 4456 static int __init init_inodecache(void) 4457 { 4458 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4459 sizeof(struct f2fs_inode_info), 0, 4460 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4461 if (!f2fs_inode_cachep) 4462 return -ENOMEM; 4463 return 0; 4464 } 4465 4466 static void destroy_inodecache(void) 4467 { 4468 /* 4469 * Make sure all delayed rcu free inodes are flushed before we 4470 * destroy cache. 4471 */ 4472 rcu_barrier(); 4473 kmem_cache_destroy(f2fs_inode_cachep); 4474 } 4475 4476 static int __init init_f2fs_fs(void) 4477 { 4478 int err; 4479 4480 if (PAGE_SIZE != F2FS_BLKSIZE) { 4481 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4482 PAGE_SIZE, F2FS_BLKSIZE); 4483 return -EINVAL; 4484 } 4485 4486 err = init_inodecache(); 4487 if (err) 4488 goto fail; 4489 err = f2fs_create_node_manager_caches(); 4490 if (err) 4491 goto free_inodecache; 4492 err = f2fs_create_segment_manager_caches(); 4493 if (err) 4494 goto free_node_manager_caches; 4495 err = f2fs_create_checkpoint_caches(); 4496 if (err) 4497 goto free_segment_manager_caches; 4498 err = f2fs_create_recovery_cache(); 4499 if (err) 4500 goto free_checkpoint_caches; 4501 err = f2fs_create_extent_cache(); 4502 if (err) 4503 goto free_recovery_cache; 4504 err = f2fs_create_garbage_collection_cache(); 4505 if (err) 4506 goto free_extent_cache; 4507 err = f2fs_init_sysfs(); 4508 if (err) 4509 goto free_garbage_collection_cache; 4510 err = register_shrinker(&f2fs_shrinker_info); 4511 if (err) 4512 goto free_sysfs; 4513 err = register_filesystem(&f2fs_fs_type); 4514 if (err) 4515 goto free_shrinker; 4516 f2fs_create_root_stats(); 4517 err = f2fs_init_post_read_processing(); 4518 if (err) 4519 goto free_root_stats; 4520 err = f2fs_init_iostat_processing(); 4521 if (err) 4522 goto free_post_read; 4523 err = f2fs_init_bio_entry_cache(); 4524 if (err) 4525 goto free_iostat; 4526 err = f2fs_init_bioset(); 4527 if (err) 4528 goto free_bio_enrty_cache; 4529 err = f2fs_init_compress_mempool(); 4530 if (err) 4531 goto free_bioset; 4532 err = f2fs_init_compress_cache(); 4533 if (err) 4534 goto free_compress_mempool; 4535 err = f2fs_create_casefold_cache(); 4536 if (err) 4537 goto free_compress_cache; 4538 return 0; 4539 free_compress_cache: 4540 f2fs_destroy_compress_cache(); 4541 free_compress_mempool: 4542 f2fs_destroy_compress_mempool(); 4543 free_bioset: 4544 f2fs_destroy_bioset(); 4545 free_bio_enrty_cache: 4546 f2fs_destroy_bio_entry_cache(); 4547 free_iostat: 4548 f2fs_destroy_iostat_processing(); 4549 free_post_read: 4550 f2fs_destroy_post_read_processing(); 4551 free_root_stats: 4552 f2fs_destroy_root_stats(); 4553 unregister_filesystem(&f2fs_fs_type); 4554 free_shrinker: 4555 unregister_shrinker(&f2fs_shrinker_info); 4556 free_sysfs: 4557 f2fs_exit_sysfs(); 4558 free_garbage_collection_cache: 4559 f2fs_destroy_garbage_collection_cache(); 4560 free_extent_cache: 4561 f2fs_destroy_extent_cache(); 4562 free_recovery_cache: 4563 f2fs_destroy_recovery_cache(); 4564 free_checkpoint_caches: 4565 f2fs_destroy_checkpoint_caches(); 4566 free_segment_manager_caches: 4567 f2fs_destroy_segment_manager_caches(); 4568 free_node_manager_caches: 4569 f2fs_destroy_node_manager_caches(); 4570 free_inodecache: 4571 destroy_inodecache(); 4572 fail: 4573 return err; 4574 } 4575 4576 static void __exit exit_f2fs_fs(void) 4577 { 4578 f2fs_destroy_casefold_cache(); 4579 f2fs_destroy_compress_cache(); 4580 f2fs_destroy_compress_mempool(); 4581 f2fs_destroy_bioset(); 4582 f2fs_destroy_bio_entry_cache(); 4583 f2fs_destroy_iostat_processing(); 4584 f2fs_destroy_post_read_processing(); 4585 f2fs_destroy_root_stats(); 4586 unregister_filesystem(&f2fs_fs_type); 4587 unregister_shrinker(&f2fs_shrinker_info); 4588 f2fs_exit_sysfs(); 4589 f2fs_destroy_garbage_collection_cache(); 4590 f2fs_destroy_extent_cache(); 4591 f2fs_destroy_recovery_cache(); 4592 f2fs_destroy_checkpoint_caches(); 4593 f2fs_destroy_segment_manager_caches(); 4594 f2fs_destroy_node_manager_caches(); 4595 destroy_inodecache(); 4596 } 4597 4598 module_init(init_f2fs_fs) 4599 module_exit(exit_f2fs_fs) 4600 4601 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4602 MODULE_DESCRIPTION("Flash Friendly File System"); 4603 MODULE_LICENSE("GPL"); 4604 MODULE_SOFTDEP("pre: crc32"); 4605 4606